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 |
|---|---|---|---|---|---|
The correct establishment and maintenance of unidirectional Notch signaling are critical for the homeostasis of various stem cell lineages . However , the molecular mechanisms that prevent cell-autonomous ectopic Notch signaling activation and deleterious cell fate decisions remain unclear . Here we show that the retromer complex directly and specifically regulates Notch receptor retrograde trafficking in Drosophila neuroblast lineages to ensure the unidirectional Notch signaling from neural progenitors to neuroblasts . Notch polyubiquitination mediated by E3 ubiquitin ligase Itch/Su ( dx ) is inherently inefficient within neural progenitors , relying on retromer-mediated trafficking to avoid aberrant endosomal accumulation of Notch and cell-autonomous signaling activation . Upon retromer dysfunction , hypo-ubiquitinated Notch accumulates in Rab7+ enlarged endosomes , where it is ectopically processed and activated in a ligand-dependent manner , causing progenitor-originated tumorigenesis . Our results therefore unveil a safeguard mechanism whereby retromer retrieves potentially harmful Notch receptors in a timely manner to prevent aberrant Notch activation-induced neural progenitor dedifferentiation and brain tumor formation .
The correct establishment and maintenance of unidirectional Notch signaling are critical for the homeostasis of various stem cell lineages ( Bertet et al . , 2014; Blanpain et al . , 2006; Bowman et al . , 2008; Conboy and Rando , 2002; Fre et al . , 2005; Guo and Ohlstein , 2015; Lin et al . , 2012; Liu et al . , 2017; Ohlstein and Spradling , 2007; Ren et al . , 2018; Song and Lu , 2011; Williams et al . , 2011 ) . The canonical Notch signaling , which requires two adjacent cells to present transmembrane ligands and transmembrane receptors respectively ( Bray , 2006; Kopan and Ilagan , 2009 ) and involves intercellular or intracellular amplification step ( s ) to establish its unidirectionality ( Artavanis-Tsakonas et al . , 1999; Liu et al . , 2017; Losick and Desplan , 2008 ) , is an ideal signaling pathway for binary cell fate specification . Accordingly , Notch signaling has been implicated in cell fate decision-making events in diverse stem cell lineages ( Bertet et al . , 2014; Chen et al . , 2016; Demitrack et al . , 2015; Dong et al . , 2012; Hilton et al . , 2008; Homem and Knoblich , 2012; Liu et al . , 2010; Ohlstein and Spradling , 2007; Pinto-Teixeira et al . , 2018; Watt et al . , 2008 ) . An important strategy utilized by dividing stem cells or progenitors to ensure binary cell fate decisions is asymmetric segregation of the endocytic protein Numb , an evolutionarily conserved Notch signaling antagonist , to one of the daughter cells ( Bultje et al . , 2009; Conboy and Rando , 2002; Gunage et al . , 2014; Lu et al . , 1998; Luo et al . , 2005; Rhyu et al . , 1994; Sallé et al . , 2017; Shen et al . , 2002; Wang et al . , 2006; Wu et al . , 2017; Zhong et al . , 1996 ) . Numb acts as an adaptor to bridge the Notch receptor and its cofactor ( s ) with the endocytic machinery and reduces the surface pool of Notch by promoting its endocytosis ( Hutterer and Knoblich , 2005; Song and Lu , 2012 ) . Endocytosed Notch receptors are often poly-ubiquitinated by E3 ubiquitin ligases , such as Itch/Su ( dx ) ( Suppressor of deltex ) and Nedd4 ( Cornell et al . , 1999; Le Bras et al . , 2011; Qiu et al . , 2000; Sakata et al . , 2004; Wilkin et al . , 2004 ) , and sorted through the ESCRT ( Endosomal Sorting Complex Required for Transport ) pathway for lysosomal degradation ( Horner et al . , 2018; Thompson et al . , 2005; Vaccari et al . , 2009 ) . As a consequence , the daughter cell inheriting relatively more Numb protein becomes the Notch signaling sending cell , unambiguously establishing signaling directionality . Not surprisingly , dysregulation in the asymmetric segregation of Numb has been implicated in a wide range of developmental defects and diseases ( Bowman et al . , 2008; Bu et al . , 2016; Caussinus and Gonzalez , 2005; George et al . , 2013; Li et al . , 2003; Pece et al . , 2004 ) . However , the plasma membrane is not the only location where the Notch receptor can be processed and activated . The proteolytic activity of γ-secretase has been detected in endosomal membranes ( Gupta-Rossi et al . , 2004; Lah and Levey , 2000; Pasternak et al . , 2003; Urra et al . , 2007 ) . Furthermore , it has been postulated that the relatively low pH at the endosomal compartments renders a conformational change in the Notch receptor , allowing for more efficient proteolysis . Indeed , inactivation of the ESCRT complex leads to retention of the Notch receptor in the limiting membrane of multivesicular bodies ( MVBs ) where Notch is ectopically activated via ligand-independent , γ-secretase-dependent proteolysis ( Hori et al . , 2011; Thompson et al . , 2005; Vaccari and Bilder , 2005; Vaccari et al . , 2009; Wilkin et al . , 2008; Zhou et al . , 2016 ) . Other than ESCRT pathway-mediated lysosomal degradation , how protein trafficking machinery prevents deleterious cell-autonomous Notch signaling activation in stem cell lineages remains to be elucidated . Type II neural stem cells , so called neuroblasts , in the Drosophila larval central brain region provide an attractive model system for studying how endosomal trafficking establishes unidirectional Notch signaling and ensures stem cell versus progenitor binary cell fate decisions ( Figure 1A ) ( Liu et al . , 2017; Song and Lu , 2012 ) . Firstly , type II neural stem cell lineages resemble their mammalian counterparts in terms of regulatory molecules and principles , yet with much simpler anatomical structure and lineage composition ( Brand and Livesey , 2011; Homem and Knoblich , 2012; Sousa-Nunes et al . , 2010 ) . Secondly , unidirectional Notch signaling is critical for establishing type II neuroblast versus immature intermediate neural progenitor ( INP ) binary cell fates ( Bowman et al . , 2008; Song and Lu , 2011; Song and Lu , 2012; Wang et al . , 2006; Weng et al . , 2010 ) . Whereas downregulation of Notch signaling in neuroblasts leads to their premature differentiation into INPs and loss of stemness , overactivation of Notch signaling in neural progenitors cause their fate reversion back into neuroblast-like state and tumorigenesis ( Bowman et al . , 2008; Song and Lu , 2011; Song and Lu , 2012; Wang et al . , 2006; Weng et al . , 2010 ) . Thus , the total number of neuroblasts in each brain lobe represents a quantitative and precise readout of Notch signaling strength . Thirdly , Numb is asymmetrically inherited by immature INPs , where it dampens Notch signaling partly by reducing the cell surface pool of mature Notch receptors ( Figure 1B ) ( Bowman et al . , 2008; Lee et al . , 2006b; Song and Lu , 2012; Wang et al . , 2006 ) . In a large-scale unbiased RNAi-based genetic screen for regulators of neuroblast versus progenitor cell fate decision , we identified Vps26 , a subunit of the retromer complex ( Burd and Cullen , 2014; Wang and Bellen , 2015 ) . Specific downregulation of Vps26 in Drosophila central brain neuroblast lineages led to a supernumerous neuroblast phenotype . The retromer complex is an evolutionarily highly conserved endosomal sorting complex , which plays a crucial role in the retrograde trafficking of a specific subset of endocytosed proteins from endosomes back to the trans-Golgi network or the plasma membrane ( Burd and Cullen , 2014; Wang and Bellen , 2015 ) . The core of the retromer complex is a vacuolar protein sorting ( Vps ) trimer composed of Vps35 , Vps26 and Vps29 subunits ( Figure 1C ) . Previous studies have implicated retromer in controlling a wide range of physiological processes , such as regulating fly wing development , maintaining the function of photoreceptors , establishing cell polarity in epithelial cells , controlling LTP ( long-term potential ) in mature hippocampus , modulating fly oogenesis and propagating mitochondrial stress signals ( Belenkaya et al . , 2008; Chen et al . , 2010; Choy et al . , 2014; Coudreuse et al . , 2006; Franch-Marro et al . , 2008; Gomez-Lamarca et al . , 2015; Harterink et al . , 2011; Hesketh et al . , 2014; Pan et al . , 2008; Pocha et al . , 2011; Port et al . , 2008; Starble and Pokrywka , 2018; Temkin et al . , 2011; Temkin et al . , 2017; Wang and Bellen , 2015; Yang et al . , 2008; Zhang et al . , 2018 ) . Dysfunction of retromer-mediated endosomal sorting has been linked to various pathologies , including neurodegenerative diseases such as Alzheimer’s disease and Parkinson’s disease ( McMillan et al . , 2017; Small and Petsko , 2015; Wang and Bellen , 2015 ) . Here our results unveil a safeguard mechanism through which the retromer complex ensures sufficient dampening of Notch signaling in neural progenitors . Upon attenuation of the retromer function , hypo-ubiquitinated Notch that fails to enter the ESCRT-lysosomal pathway accumulates in enlarged Rab7+ endosomes and is ectopically processed and activated . Such cell-autonomous intracellular hyperactivation of Notch signaling causes fate reversion of neural progenitors and the formation of transplantable tumors . These results led us to propose a model whereby retromer serves as ‘bomb squad’ to retrieve and disarm the potentially harmful pool of Notch receptors in a timely manner .
To investigate the function of retromer in neuroblast lineages , we first downregulated Vps26 in all central brain neuroblast lineages using short hairpin microRNAs ( shmiRNAs ) , driven by insc-Gal4 , and observed a supernumerary neuroblast phenotype ( Figure 1D , E ) . Such brain tumor phenotype induced by vps26-RNAi was fully rescued by the coexpression of a shmiRNA-resistant form of the Vps26 transgene , excluding the possibility of an off-target effect of the shmiRNA ( Figure 1D , E ) . Furthermore , homozygous vps35 mutant larval brains exhibited an even more severe supernumerary neuroblast phenotype than vps26-IR , and such phenotype was fully rescued upon specific expression of a Vps35 transgene in all central brain neuroblast lineages ( Figure 1D , E ) . Importantly , human Vps35 also fully rescued the brain tumor phenotype of vps35 mutants back to wild type ( Figure 1D , E ) . Taken together , our results clearly indicated that retromer plays an evolutionarily-conserved role in preventing ectopic neuroblast formation in the central brain area . To investigate the cellular origin of the ectopic neuroblasts formed upon retromer inactivation , we expressed the Vps35 transgene in distinct subset of cells within central brain neuroblast lineages and assessed its ability to rescue the vps35 mutant phenotype . Expression of the Vps35 transgene in type II neuroblast lineages , by PntP1-Gal4 ( Zhu et al . , 2011 ) , fully suppressed the brain tumor phenotype caused by vps35 mutation ( Figure 1E , F and Figure 1—figure supplement 1 ) . By contrast , restoring Vps35 function in type I neuroblast lineages , by ase-Gal4 ( Zhu et al . , 2006 ) , failed to do so ( data not shown ) . These results indicated that the ectopic neuroblasts in retromer mutants are derived from type II neuroblast lineages . Indeed , expression of the Vps35 transgene in Deadpan ( Dpn ) - Asense ( Ase ) - INPs by erm-Gal4 ( II ) but not in Dpn- Ase+ INPs by erm-Gal4 ( III ) ( Pfeiffer et al . , 2008 ) completely rescued the supernumerary neuroblast phenotype caused by Vps35 inactivation ( Figure 1E , F and Figure 1—figure supplement 1 ) . Therefore , the reverting Dpn-Ase- neural progenitors are the cellular origin of brain tumor in vps35 mutants . Supporting this notion , cell polarity remained unaltered in vps35 mutant neuroblasts ( Figure 1G ) , indicating that these ectopic neuroblasts are not resulted from neuroblast symmetric division . Importantly , Numb is normally localized to the basal cortex of vps35 mutant dividing neuroblasts ( Figure 1H ) , arguing against the possibility that defective asymmetric segregation of Numb causes INP dedifferentiation in vps35 mutant brains . Consistently , whereas Vps26 downregulation in type II neuroblast lineages or immature INP lineages , driven by PntP1-Gal4 or erm-Gal4 ( II ) respectively , resulted in supernumerary neuroblast phenotype , its knockdown in mature INP lineages or type I neuroblast lineages , driven by erm-Gal4 ( III ) or ase-Gal4 respectively , failed to induce ectopic neuroblasts ( Figure 1—figure supplement 2 ) . Furthermore , distinct from wild type control type II neuroblast MARCM clones ( Lee and Luo , 1999 ) that contained one and only one Dpn+ Ase- neuroblast ( white bracket in Figure 1I and Figure 1—figure supplement 1B ) , vps35 mutant clones contained multiple ectopic Dpn+ Ase- Pros- neuroblast-like cells ( yellow arrowheads in Figure 1 and Figure 1—figure supplement 3 ) several cell diameters away from the primary neuroblast ( white bracket in Figure 1I ) . These ectopic Dpn+ Ase- Pros- neuroblast-like cells were of intermediate cell sizes between neural progenitors and primary neuroblasts ( yellow arrowheads in Figure 1 and Figure 1—figure supplement 3 ) , indicating that they were undergoing dedifferentiation ( Song and Lu , 2011 ) . In addition , FLP-FRT-based lineage tracing by inducing GFP+ clones exclusively in immature INPs , driven by erm-Gal4 ( II ) , resulted in labeling of INPs ( white arrowhead in Figure 1—figure supplement 4 ) , GMCs , and neurons ( cyan arrowhead in Figure 1—figure supplement 4 ) in wild-type brains . In contrast , in vps35 mutant brains , GFP-labeled ectopic type II neuroblasts of various cellular sizes were found after similar lineage tracing ( yellow arrowheads in Figure 1—figure supplement 4 ) , indicating that immature INPs could indeed dedifferentiate back into neuroblast-like cells upon retromer dysfunction . Taken together , our results clearly indicate that the brain tumor phenotype in vps35 mutants is caused by cell fate reversion of Dpn- Ase- neural progenitors . We next employed transplantation assay to test whether the ectopic neuroblasts in vps35 mutant brains are capable of initiating tumor . Transplantation of vps35 mutant but not wild-type control brain tissues into the abdomens of host flies caused the formation of massive tumors ( yellow bracket in Figure 1J ) that often metastasize to distal organs ( yellow arrowhead in Figure 1J; statistic results in Figure 1K ) . Importantly , the vps35 mutant GFP+ tumor cells extracted from the abdomen of transplanted hosts were Dpn+ Miranda ( Mira ) + neuroblast-like cells ( Figure 1L ) . Thus vps35 mutant cells in the larval brains are indeed tumor-initiating cells . Together , we conclude that retromer acts as a tumor suppressor in the Drosophila brain by preventing neural progenitor dedifferentiation . Since the well-characterized function of retromer is retrograde transport of transmembrane proteins , we next assessed whether the distribution of any subcellular marker ( s ) is altered upon inactivation of retromer function . Compared to wild-type control INPs , vps35 mutant INPs or ectopic neuroblasts displayed dramatically enlarged late endosomes/MVBs ( Figure 2A–C; up to more than 10-fold increase in endosomal vesicle sizes ) . The expression levels of Rab7 remained unchanged in vps35 mutants ( Figure 2D ) , ruling out the likelihood that Vps35 regulates Rab7 gene expression or protein stability . Furthermore , Rab7 primarily colocalized with early endosome marker Rab5 in vps35 but not wild type cells ( Figure 2E ) , demonstrating that the enlarged MVBs in vps35 mutant cells are of early and late endosome hybrid identities . In contrast , other subcellular markers including lysosome ( GFP-LAMP1 ) , recycling endosome ( Rab11 ) , Golgi ( GFP-Golgi ) and mitochondria ( mito-GFP ) remained unchanged in vps35-defective cells ( Figure 2F–2I ) . Therefore , our results strongly suggest that retromer normally functions in neural progenitors to transport cargo proteins away from early and late endosomes . Upon retromer dysfunction , its cargo proteins highly accumulate in MVBs , resulting in enlarged , aberrant endosomal vesicles of hybrid identities . We next sought to identify the critical cargo protein ( s ) of retromer in preventing INP dedifferentiation . Since Notch pathway is both necessary and sufficient to promote self-renewal in type II neuroblast lineages , we first examined the subcellular distribution of transmembrane protein components of Notch signaling pathway . We noted that the Notch receptor and its cofactor Sanpodo ( Couturier et al . , 2012; Hutterer and Knoblich , 2005; O'Connor-Giles and Skeath , 2003; Song and Lu , 2012 ) highly accumulated in enlarged puncta in vps35 mutant cells , mostly colocalizing with Rab7+ enlarged endosomes ( Figure 3A , B and Figure 3—figure supplement 1 ) . In contrast , the distribution of other signaling molecules such as Patched ( Ptc ) and Wnt/Wingless ( Wg ) remained unaltered upon Vps35 depletion ( Figure 3—figure supplement 2A–C ) , indicating that retromer specifically mediates Notch receptor trafficking in neuroblast lineages . Strongly supporting this notion , Notch signaling reporter E ( spl ) mγ-GFP ( Almeida and Bray , 2005; Song and Lu , 2011 ) , which faithfully reflects Notch signaling activity in neuroblast lineages , was undetectable in wild type Dpn- Ase- immature INPs ( white arrowhead in Figure 3C ) but ectopically turned on in Dpn+ Ase- dedifferentiating neural progenitors ( yellow arrowhead in Figure 3C ) upon Vps26 downregulation . In addition , Notch puncta colocalizing with Rab7+ endosomes remained unaltered in vps35 mutant wing imaginal disc epithelia ( arrowheads in Figure 3—figure supplement 2D ) , suggesting a tissue-specific regulation of Notch trafficking by retromer . Collectively , retromer normally suppresses Notch activity through mediating retrograde trafficking of Notch receptors in neural progenitors . We next assessed whether Notch is a crucial cargo of retromer in neuroblast lineages . Neuroblast lineage-specific knockdown of Notch completely suppressed the neuroblast overproliferation phenotype in vps35 mutants ( Figure 3D , E ) , indicating that the dedifferentiation process of vps35 mutant INPs was Notch signaling-dependent . Type II neuroblast lineage-specific or immature INP-specific depletion of the ligand Delta , as well as neuroblast lineage-specific expression of a dominant negative form of Delta ( Dl-DN ) that lacks its intracellular domain ( Baonza et al . , 2000; Flores et al . , 2000; Huppert et al . , 1997 ) , completely or potently suppressed brain tumor phenotypes caused by vps35 mutations ( Figure 3D , E and Figure 3—figure supplement 2E , F ) . Furthermore , type II neuroblast lineage-specific or immature INP-specific expression of a dominant negative form of the metalloprotease Kuzbanian ( Kuz-DN ) , which lacks its protease activity and thereby specifically blocks ligand-induce S2 cleavage of Notch ( Lieber et al . , 2002; Mumm et al . , 2000; Pan and Rubin , 1997 ) , also phenocopied the effect of Notch-RNAi in inhibiting brain tumor formation ( Figure 3D , E ) . These observations indicated that overactivation of Notch signaling in vps35 mutant neural progenitors is largely , if not completely , ligand-dependent . Not surprisingly , a functional γ–secretase is also essential for ectopic activation of Notch signaling in vps35 mutants ( Figure 3D , E ) . In sharp contrast , inactivation of various other signaling pathways , such as Wnt/Wg , Hedgehog or EGFR , or overactivation of Hedgehog signaling showed no effects on the supernumerary neuroblast phenotype in vps35 mutants ( Figure 3D , E and Figure 3—figure supplement 2E , F ) , further demonstrating the high specificity of retromer on Notch signaling pathway in neuroblast lineages . Importantly , Notch colocalized with fly or human Vps35 transgene ( Figure 3F , G ) and endogenous Vps26 ( Figure 3H and Figure 3—figure supplement 3 ) . More remarkably , Notch depletion by RNAi led to a dramatic reduction in Rab7+ endosomal vesicle sizes almost back to normal ( Figure 3I ) , suggesting that Notch receptors constitute the major endosomal contents of these aberrant vps35 mutant vesicles . Taken together , our results strongly suggested that the Notch receptor is a functionally important cargo of retromer in type II neuroblast lineages . To validate that the Notch receptor is a cargo protein of the retromer complex , we assessed their physical interaction by performing coimmunoprecipitation ( coIP ) assays . Vps35 or Vps26 was specifically coimmunoprecipitated with Notch intracellular domain ( NICD ) from HEK293T cell extracts ( Figure 4A ) . Further domain mapping experiments revealed that the ankyrin repeat region but not the C-terminal region of NICD exhibited a strong binding activity to Vps26 ( Figure 4B , C ) . Reciprocal coIP assay showed that Vps26 utilized its middle domain to interact with NICD ( Figure 4D , E ) . Furthermore , Notch-V5 expressed in central brain neuroblast lineages was specifically coimmunoprecipitated with Vps35-FLAG from fly larval brain extracts ( Figure 4F ) , confirming the in vivo protein-protein interaction . Importantly , coIP experiments further revealed interaction between mouse NICD and mouse Vps26 proteins ( Figure 4G , H ) , indicating that the physical association between the retromer cargo-recognition complex and Notch is evolutionarily conserved . Taken together , our results validate that the Notch receptor is a bona fide cargo protein of the retromer complex . Our results presented so far support an intriguing possibility that the retromer complex physically interacts with Notch and transports it away from early and late endosomes in a timely and efficient manner . When retromer is defective , Notch receptors are trapped at early/late aberrant endosomal vesicles and is ectopically processed and activated , causing neural progenitor-derived brain tumor . If this hypothesis is correct , one would expect that blocking the flux of Notch receptors towards its activating compartment or accelerating Notch trafficking away from it might prevent the accumulation and subsequent ectopic activation of Notch in vps35 mutants ( Figure 5A ) . Indeed , overexpression of a dominant negative form of Rab5 GTPase ( Rab5-DN ) , which blocks the fusion of endocytic vesicles with early endosomes , or a constitutively active form of Rab9 GTPase ( Rab9-CA ) , which promotes protein retrograde trafficking from late endosomes to trans-Golgi network ( TGN ) or the plasma membrane ( Figure 5—figure supplement 1 ) , completely suppressed brain tumor formation in vps35 mutant brains ( Figure 5B , C ) . Importantly , both the enlargement of Rab5/Rab7-positive endosomal vesicles and the high accumulation of Notch in these aberrant endosomal compartments in vps35 mutant cells were effectively relieved upon Rab5-DN or Rab9-CA coexpression ( arrowheads in Figure 5D , E ) . On the other hand , overexpression of a constitutively active form of Rab7 ( Rab7-CA ) or the ESCRT-0 complex component Hrs ( Hepatocyte growth factor-regulated tyrosine kinase substrate ) , which accelerates the protein trafficking towards lysosome ( Figure 5—figure supplement 1 ) ( Lloyd et al . , 2002 ) , potently inhibited the neuroblast overproliferation phenotype in vps35 mutant brains ( Figure 5A , B , C ) . Indeed , coexpression of either Rab7-CA or Hrs led to high accumulation of Notch in lysosomes of vps35 mutant cells ( Figure 5F ) . In contrast , overexpression of a constitutively active form of Rab5 GTPase ( Rab5-CA ) , which accelerates the fusion of endocytic vesicles with early endosomes , or a dominant negative form of Rab7 ( Rab7-DN ) or Rab9 ( Rab9-DN ) GTPase , which prevents transport of proteins away from the sorting endosomes , failed to suppress the supernumerary neuroblast phenotype in vps35 mutant brains ( Figure 5—figure supplement 2 ) . In addition , the Delta ligand clearly colocalized with Rab7+ enlarged endosomes in vps35 mutant cells ( Figure 5G ) . Taken together , we concluded that the enlarged , aberrant endosomal vesicles with both early and late endosomal identities are the ligand-dependent activating compartments of the Notch receptor in vps35 mutant neural progenitors . Why Notch needs to be transported away from its activating compartments by retromer under physiological conditions ? Previous studies indicated that the internalized Notch receptors are either sorted through the ESCRT pathway and get degraded in lysosomes or recycled back to the plasma membrane for ligand binding and activation ( Kopan , 2012 ) . Furthermore , ubiquitin is a crucial sorting signal for Notch receptor trafficking . We therefore considered the intriguing possibility that a pool of hypo-ubiquitinated Notch receptors might not be sorted through ESCRT-0 but instead trapped at the limiting membrane of MVBs , where they are retrieved and transported away by retromer in a timely manner . If this hypothesis is correct , one would expect that an elevation in the activity of the E3 ubiquitin ligase ( s ) that promotes Notch polyubiquitination and lysosomal degradation may reduce the pool of hypo-ubiquitinated Notch in retromer mutant neural progenitors and thereby alleviate the brain tumor phenotype . Neuroblast lineage-specific overexpression of HECT domain E3 ubiquitin ligase Itch/Su ( dx ) or Nedd4 , known for mediating Notch receptor polyubiquitination and degradation ( Cornell et al . , 1999; Le Bras et al . , 2011; Qiu et al . , 2000; Sakata et al . , 2004; Wilkin et al . , 2004 ) , showed little inhibitory effect on the supernumerary neuroblast phenotype in vps35 mutants ( Figure 6A , B ) , suggesting that these two E3 ligases are not fully active upon overexpression in neuroblast lineages . Since Ndfip protein ( Nedd4 family interacting protein ) has been reported to recruit and activate Itch/Su ( dx ) or Nedd4 by relieving their autoinhibition caused by intramolecular interaction ( Dalton et al . , 2011; Mund and Pelham , 2009 ) , we coexpressed Ndfip in an attempt to boost the catalytic activity of Itch/Su ( dx ) and Nedd4 . Whereas simultaneous overexpression of Nedd4 and Ndfip barely exhibited any effect on brain tumor phenotype caused by vps35 mutation ( Figure 6—figure supplement 1A–C ) , coexpression of Su ( dx ) and Ndfip indeed led to a complete rescue of the supernumerary neuroblast phenotype in vps35 mutants ( Figure 6A , B ) . Consistent with these observations , the high accumulation of Notch in aberrant endosomal vesicles in vps35 mutant cells was also effectively suppressed by Su ( dx ) and Ndfip coexpression ( Figure 6—figure supplement 1D , E ) . A related and more important prediction of this hypothesis is that the activity of the E3 ubiquitin ligase ( s ) targeting Notch for polyubiquitination and degradation is inherently inefficient in fly neuroblast lineages and depends on retromer-mediated retrieval to avoid ectopic accumulation and processing of Notch in INPs . If this model is correct , we reason that a reduction in the activity of the E3 ubiquitin ligase ( s ) might tilt the balance and lead to a larger pool of hypo-ubiquitinated Notch than normal . If retromer is meanwhile not fully functional , Notch receptors may be stalled in MVBs and eventually result in progenitor-derived tumor . Indeed , we observed a strong synergistic interaction between Su ( dx ) and Vps26 in mediating neuroblast self-renewal . While expression of either vps26-RNAi or Su ( dx ) -C917A , a dominant negative form of Su ( dx ) ( Su ( dx ) -DN ) that lacks its E3 ubiquitin ligase activity ( Wang et al . , 2015 ) , by PntP1-Gal4 , led to a mild neuroblast overproliferation phenotype ( Figure 6C , D ) , simultaneous expression of vps26-RNAi and Su ( dx ) -DN resulted in a severe brain tumor phenotype ( Figure 6C , D ) . More significantly , Notch receptors were highly accumulated in enlarged Rab7-positive endosomal vesicles in neural progenitors expressing both vps26-RNAi and Su ( dx ) -DN , but not in neural progenitors expressing either vps26-RNAi or Su ( dx ) -DN alone ( Figure 6E , F ) . Immunostaining with our newly-raised Su ( dx ) and Ndfip antibodies ( Figure 6—figure supplement 2A , B ) revealed that Su ( dx ) mainly localized to the cell cortex , whereas Ndfip primarily distributed in intracellular vesicles ( Figure 6—figure supplement 2C , D ) . Such largely distinct distribution pattern of Su ( dx ) and Ndfip in INPs might partially explain why Notch polyubiquitination and lysosomal degradation is inherently inefficient in neural progenitors . A third prediction of this hypothesis is that ectopic processing and activation of the Notch receptor in vps35 mutants are independent of the ESCRT pathway . Indeed , blocking the entry to the ESCRT pathway via depletion of Hrs , a key subunit of ESCRT-0 , exhibited no effects on the brain tumor phenotypes in vps35 mutants ( Figure 6G , H ) . Taken together , these findings indicated that retromer prevents neural progenitor dedifferentiation through compensating the insufficient dampening of Notch signaling mediated by the Su ( dx ) /Ndfip-ESCRT-lysosomal pathway . To further confirm this hypothesis , we assessed the cleavage status of the Notch receptor . Our model predicts that , upon retromer dysfunction , the chances for Notch to be transported back to the plasma membrane to access its E3 ubiquitin ligase ( s ) and obtain additional ubiquitin moieties become smaller . As a consequence , a pool of hypo-ubiquitinated Notch might accumulate and ectopically processed . In accordance , our results clearly showed that , a smear of presumably ubiquitinated NICD migrating at approximately 130 kilodaltons ( kDa ) and a un-ubiquitinated NICD band migrating at 100 kDa specifically accumulated in vps35 mutant but not wild type brain extracts , indicating that these Notch fragments are hypo-ubiquitinated NICD ( Figure 6I ) . Importantly , such increased intensity of the smear of these hypo-ubiquitinated Notch fragments in vps35 mutants ( green arrowhead ) was essentially reduced back to normal upon coexpression of Su ( dx ) and Ndfip ( Figure 6I ) . We reason that coexpression of E3 ligase led to a reduction in the pool of hypo-ubiquitinated Notch and a corresponding increase in the pool of Notch harboring sufficient ubiquitin moieties , which was sorted through the ESCRT pathway and degraded in lysosomes . Indeed , blocking the cargo entry into the ESCRT pathway by Hrs downregulation potently inhibited the rescue effects of Su ( dx ) /Ndfip on retromer inactivation-induced brain tumor phenotype ( Figure 6G , H ) . Consistently , un-ubiquitinated NICD fragments also specifically accumulated in larval brain extracts coexpressing vps26-RNAi and Su ( dx ) -DN but not in extracts expressing either vps26-RNAi or Su ( dx ) -DN alone ( Figure 6J ) . Furthermore , in N; NiGFP background , in which a bacterial artificial chromosome ( BAC ) transgene expressing a GFP-tagged Notch ( NiGFP ) functionally replaces endogenous Notch ( Couturier et al . , 2012 ) , accumulation of hypo-ubiquitinated NICD-GFP fragments was also specifically detected in vps35 mutant but not wild type control brain extracts ( Figure 6—figure supplement 3 ) . Collectively , our results supports a safeguard model whereby Notch polyubiquitination mediated by the E3 ubiquitin ligase Itch/Su ( dx ) is inherently inefficient within neural progenitors , relying on retromer-mediated retrograde trafficking to retrieve the pool of hypo-ubiquitinated Notch that fails to enter the ESCRT-lysosomal degradation pathway in a timely manner ( Figure 7 ) . Upon retromer inactivation , hypo-ubiquitinated Notch accumulates in MVBs , ectopically processed in a ligand-dependent fashion , leading to cell-autonomous activation of Notch signaling , neural progenitor dedifferentiation and tumorigenesis ( Figure 7 ) .
Unidirectional Notch signaling is a widely used strategy for initiating and maintaining binary cell fates . However , the molecular mechanisms establishing the unidirectionality of Notch signaling in stem cell lineages remain unclear . Here we reveal that , while asymmetric partition of Numb leads to a biased internalization of the Notch receptor and hence asymmetric dampening of Notch signaling in neural progenitors , it meanwhile poses a high risk of non-canonical endosomal activation of Notch . We find that the retromer complex is the key protein trafficking machinery that resolves this crisis through a timely retrieval of the Notch receptor from its endosomal activation compartments . Upon retromer dysfunction , neural progenitors dedifferentiate into neural stem cell-like status and result in the formation of transplantable tumors . Therefore , retromer acts as a tumor suppressor in Drosophila larval brains . Importantly , mammalian Vps35 physically interacts with Notch , colocalizes with Notch in neural progenitors , and its neuroblast-lineage-specific expression fully rescues neural progenitor-derived brain tumor phenotype in vps35 mutants . Thus , the brain tumor suppressor function of retromer is likely to be conserved in mammals . Intriguingly , downregulation of the retromer complex components has been reported in various human cancers , including glioblastoma ( An et al . , 2012; Bredel et al . , 2005; Lee et al . , 2006b ) . Our studies thus provide a new mechanistic link between the retromer complex and carcinogenesis . Why the E3 ubiquitin ligase system promoting Notch receptor polyubiquitination and degradation is inherently inefficient in neuroblast lineages ? We speculate that Notch is probably not the only substrate of Su ( dx ) and Ndfip in neuroblasts or neural progenitors . Therefore , high levels and/or activity of this E3 ubiquitin ligase system above certain threshold may potentially cause imbalanced homeostasis of its critical substrates and hence perturbed neuroblast lineages . Indeed , co-overexpression of Su ( dx ) and Ndfip led to drastically reduced number of neuroblast lineages and severe tissue atrophy ( Figure 6—figure supplement 4 ) . In this case , a relatively general yet inefficient ubiquitination-degradation system coupled with a highly efficient and selective cargo retrieving system provides a customized regulation of the Notch receptor , ensuring sufficient dampening of Notch signaling in neural progenitors without devastating side effects . Intriguingly , previous studies posited that retromer dysfunction causes increased levels of APP ( β-amyloid precursor protein ) to reside in the endosomes for longer duration than normal , resulting in accelerated processing of APP into amyloid-β , a neurotoxic fragment implicated AD pathogenesis ( Small and Gandy , 2006; Small and Petsko , 2015 ) . Furthermore , retromer maintains the integrity of photoreceptors by avoiding persistent accumulation of rhodopsin in endolysosomal compartments that stresses photoreceptors and causes their degeneration ( Wang et al . , 2014 ) . Taken together with our study here , these findings indicate that retromer serves as bomb squad to retrieve and disarm harmful or toxic protein fragments from endosomes in a timely manner and thereby safeguard the integrity and fitness of the neuronal lineages . How is the Notch receptor ectopically activated in retromer mutants ? We favor the idea that Notch is activated in MVBs in a ligand-dependent , cell-autonomous manner , distinct from the majority of non-canonical Notch activation mechanisms . Most of the endosomal Notch activation events identified before , including ectopic Notch signaling activation in ESCRT mutants , BLOS2 mutants , or Rme8 and Vps26 double knockdown background , as well as Hif-alpha-dependent activation of Notch signaling implicated in crystal cell maintenance and survival , are all ligand-independent ( Baron , 2012; Childress et al . , 2006; Gallagher and Knoblich , 2006; Giebel and Wodarz , 2006; Gomez-Lamarca et al . , 2015; Hori et al . , 2011; Jaekel and Klein , 2006; Mukherjee et al . , 2011; Palmer and Deng , 2015; Schneider et al . , 2013; Thompson et al . , 2005; Vaccari and Bilder , 2005; Vaccari et al . , 2009; Wilkin et al . , 2008; Zhou et al . , 2016 ) . It has been proposed that the proteases within the acidifying environment of MVB lumen are sufficient to remove the extracellular domain of Notch , leading to the S3 cleavage of Notch at the limiting membrane ( Palmer and Deng , 2015; Wilkin et al . , 2008 ) . Strongly supporting this notion , blocking the entry of Notch into the ESCRT pathway but not ligand inactivation potently inhibited ectopic Notch activation induced by ESCRT mutations ( Childress et al . , 2006; Gallagher and Knoblich , 2006; Jaekel and Klein , 2006 ) . In sharp contrast to these previously-revealed mechanisms , attenuating ligand activity but not preventing Notch from entering the ESCRT pathway effectively rescues Notch overactivation phenotype caused by retromer dysfunction ( Figures 3D , E and 6G , H ) . Then how Notch signaling is ectopically activated in a ligand-dependent manner in retromer mutants ? We speculate that , upon retromer dysfunction , both Notch and Delta are entrapped in MVBs , where Notch and Delta are presented by limiting membrane and intravesicular membrane respectively and result in ligand-dependent Notch processing and activation , resembling the scenario presented for ligand-dependent Notch signaling activation in Sara endosome ( Coumailleau et al . , 2009; Kressmann et al . , 2015 ) . The detailed regulatory mechanisms underlying Notch overactivation in retromer mutants warrants future investigation . The ability of vps35 mutant neoplastic neuroblasts to metastasize upon transplantation is intriguing ( Figure 1J , K ) . Metastasis of brain tumor cells derived from neuroblast lineages has never been observed in the developing fly larval brains , likely because the limited time window of fly larval development precludes tumor progression and metastasis . Transplantation assay ( Rossi and Gonzalez , 2015 ) , however , provides the ectopic microenvironment and allows cancer progression in a much longer time scale ( months , or even years upon retransplantation ) . Importantly , mutations that caused metastasis of fly brain tumor cells upon transplantation have also been implicated in various human cancers ( Caussinus and Gonzalez , 2005; Eroglu et al . , 2014; Froldi et al . , 2015; Knoblich , 2010; Landskron et al . , 2018; Liu et al . , 2017; Narbonne-Reveau et al . , 2016 ) . Future studies on the transcriptional profiling of the distal metastatic colonies and stepwise characterization of this long-range metastatic process promise to provide us with fresh mechanistic insights into the enormously complex process of cancer metastasis .
Fly culture and crosses were performed according to standard procedures . Drosophila stocks used in this study include: vps35E42 ( Belenkaya et al . , 2008 ) ( a gift from Dr . Xinhua Lin ) ; vps351 ( Belenkaya et al . , 2008 ) ; vps26G2008 ( BL26623 ) ; UAS-Vps35-FLAG ( this study ) ; UAS-Vps26-Myc ( RR: RNAi resistant form; this study ) ; UAS-vps26-RNAi ( BL38937 ) ; UAS-Notch-RNAi ( VDRC27229 ) ; UAS-Dl-DN ( BL26698 ) ; UAS-Dl-RNAi ( BL34322 ) ; UAS-Psn-DN ( BL8323 ) ; UAS-Kuz-DN ( BL6578 ) ; UAS-Rab5-DN ( BL42704 ) ; UAS-Rab7-CA ( BL42707 ) ; UAS-Hrs ( Lloyd et al . , 2002 ) ; UAS-GFP-LAMP1 ( BL42714 ) ; UAS-mito-GFP ( BL8442 ) ; UASp-YFP-Rab9-WT ( BL9784 ) ; UASp-YFP-Rab9-CA ( BL9785 ) ; UAS-Su ( dx ) ( BL51664 ) ; UAS-Myc-Su ( dx ) ( this study ) ; UAS-FLAG-Ndfip ( this study ) ; UAS-Myc-Su ( dx ) -C917A ( this study ) ; UAS-Myc-Nedd4 ( this study ) ; insc-Gal4 ( Luo et al . , 1994 ) ; PntP1-Gal4 ( Zhu et al . , 2011 ) ; ase-Gal4 ( Zhu et al . , 2006 ) ; erm-Gal4 ( II ) ( Xiao et al . , 2012 ) ; erm-Gal4 ( III ) ( Pfeiffer et al . , 2008; Weng et al . , 2010 ) ; E ( spl ) mγ-GFP ( Almeida and Bray , 2005; Monastirioti et al . , 2010 ) ; UAS-Spdo-GFP ( Song and Lu , 2012 ) ; UAS-wg-RNAi ( BL32994 ) ; UAS-EGFR-DN ( BL5364 ) ; UAS-med-RNAi ( BL52214 ) ; UAS-ptc-DN ( BL31928 ) ; UAS-Hrs-RNAi ( BL34086 , BL33900 ) ; N55e11; NiGFP ( Couturier et al . , 2012 ) and UAS-FLP , Ubi-p63E-FRT > stop > FRT-nlsGFP ( BL28282 ) ( Evans et al . , 2009 ) . All larval brains phenotypes were analyzed at late third instar larval stage . Note that , compared to wild type control , the development of vps35E42 mutant larvae was delayed . Experiments with no special notification were carried out as follows: Eggs were collected for 4–6 hr at 25°C and kept at 25°C until dissection at late third instar larval stage . The experimental conditions shown in Figures 1D , F , 3D , 5B , D , F , 6C , E and I are as follows: Eggs were collected for 4–6 hr at 22°C , kept at 22°C for 24 hr ( Figures 1D , F , 3D , 5B , 6C and E ) or 48 hr ( Figures 5D , F and 6I ) after hatching and shifted to 29°C until dissection at late third instar larval stage . The experimental conditions shown in 3C is as follows: Eggs were collected for 4–6 hr at 25°C , kept at 18°C for 8 days , then shifted to 29°C for 40 hr before dissection . The experimental conditions shown in Figure 1—figure supplement 2 and Figure 6—figure supplement 4 are as follows: Eggs were collected for 4–6 hr at 22°C . Larvae were raised at 29°C immediately after hatching until dissection at late third instar larval stage . Full-length cDNA clones for vps35 , vps26 ( LD29140 ) , and nedd4 were obtained from Drosophila Genomics Resource Center ( DGRC ) . For ndfip and su ( dx ) cDNAs , their respective coding exons were cloned by genomic DNA PCR from w1118 flies and UAS-Su ( dx ) transgenic flies respectively , assembled together by the Gibson Assembly method and fully sequenced . FLAG-Ndfip , Myc-Nedd4 and Myc-Su ( dx ) -WT were constructed by adding a FLAG tag ( DYKDDDDK ) or a Myc tag ( EQKLISEEDL ) respectively to the N-terminus . Vps35-FLAG and Vps26-Myc were constructed by adding a FLAG tag or a Myc tag respectively to the C-terminus . Note that shmiRNA-resistant sequence was introduced into Vps26 before it was cloned into the pUAST vector . A missense mutation ( C917A ) was introduced into Su ( dx ) to generate a ligase-inactivated form . NICD-V5 was generated as described before ( Liu et al . , 2017 ) . All transgenic plasmids were verified by DNA sequencing before germline transformation . For coimmunoprecipitation experiments , Vps26-FLAG and Vps26-Myc were cloned into pcDNA3 . 1 vector respectively ( Invitrogen ) . Vps26 truncated forms Vps26-ΔN ( aa 147–478 ) , Vps26-ΔM ( aa 1–146 and aa 297–478 ) , Vps26-ΔC ( aa 1–296 ) and Vps26-M ( aa 110–357 ) were cloned with a N-terminal FLAG tag into pcDNA3 . 1 vector respectively . Mouse NICD cDNA was obtained from Addgene , while mouse vps26 cDNA were generated by introducing I16V , V17A , E217D to human vps26 cDNA ( Addgene ) . NICD-V5 construct was generated as described before ( Liu et al . , 2017 ) , except that aa 1767–1770 , 1832–1835 , 2202–2205 and 2222–2225 were deleted to remove its nuclear localization sequence . NICD truncated versions NICD-N ( aa 1771–2230 ) and NICD-ANK ( aa 1838–2230 ) were cloned into the vector pcDNA3 . 1 with V5 tag added to C-terminus , and NICD-C-ΔPEST ( aa 2231–2603 ) with a V5 tag inserted between aa 2571 and 2572 . To generate mouse NICD-ΔNLS-V5 , aa 1749–1752 , 1771–1774 , 1811–1814 , 2146–2149 and 2167–2170 were deleted from mouse NICD ( aa 1744–2531 of mouse Notch1 protein ) and a V5 tag was inserted between aa 2396 and 2397 , before cloned into pcDNA3 . 1 vector . mVps26-FLAG and mVps26-Myc were cloned into the pCMV vector respectively . Neuroblast MARCM clones were generated as previously described ( Song and Lu , 2011 ) . Briefly , newly hatched larvae were heat-shocked at 37°C for 90 min and further aged at 25°C for indicated time before dissection . FRTG13 , vps351 was used for neuroblast MARCM clonal analysis , as shown in Figure 1I and Figure 1—figure supplement 2 , with FRTG13 alone serving as a negative control . For larval brain immunostaining , larvae were dissected in Schneider’s Insect Medium ( Sigma-Aldrich ) and proceeded as previously described ( Liu et al . , 2017; Song and Lu , 2011 ) . Briefly , larval brains were fixed with 4% paraformaldehyde in PEM buffer ( 100 mM PIPES at pH 6 . 9 , 1 mM EGTA , 1 mM MgCl2 ) for 22 min at room temperature . Brains were washed several times with PBST buffer ( 1 × PBS plus 0 . 1% Triton X-100 ) and were incubated with appropriate primary antibody overnight at 4°C or for 2 hr at room temperature , labeled with secondary antibodies according to standard procedures , and mounted in Vectashield ( Vector Laboratories ) . For anti-Delta staining , larval brains were fixed with 4% paraformaldehyde/PEM buffer for 20 min at room temperature , blocked in 3% BSA/PBST for 20 min at room temperature , before being incubated with mouse anti-Delta ( 1:200 ) in 0 . 5% BSA/PBST for 12 hr at 4°C . After washing with PBST buffer , brains were incubated with goat anti-mouse secondary antibody ( 1:100 ) in 0 . 5% BSA/PBST for 2 hr at room temperature before being mounted in Vectashield . Antibodies generated in this study were rabbit anti-Vps26 antibody [GST fusion of aa 320–478 , affinity purified ( Abclonal Biotech . ) , used at 1:200] , rabbit anti-Su ( dx ) [GST fusion of aa 350–500 , affinity-purified ( Abclonal Biotech . ) , used at 1:200] and rabbit anti-Ndfip [GST fusion of aa 2–165 , affinity-purified ( Abclonal Biotech . ) , used at 1:100] . To eliminate any non-specific binding , all antibodies were preabsorbed before being used in immunostaining experiments . Images were obtained on a Leica TCS SP8 AOBS confocal microscope and were processed with LAS AF ( Leica ) and Adobe Photoshop CS5 . Other primary antibodies used for immunohistochemistry were chicken anti-GFP ( 1:2000 , Abcam ) , mouse anti-Pros ( 1:100 , Developmental Studies Hybridoma Bank [DSHB] ) , mouse anti-NECD C458 . 2H ( 1:80 , DSHB ) , rat anti-Miranda ( 1:100; Abcam ) , rabbit anti-Dpn ( 1:1000 , Y . N . Jan ) , rabbit anti-Rab7 ( 1:2000 , a generous gift from A . Nakamura ) ( Tanaka and Nakamura , 2008 ) ; guinea pig anti-Numb ( 1:1000 , a generous gift from J . Skeath ) ( O'Connor-Giles and Skeath , 2003 ) , mouse anti-β-galactosidase ( 1:100 , DSHB ) , guinea pig anti-Ase ( 1:400 , Y . N . Jan ) , rabbit anti-aPKC ζ C20 ( 1:1000 , Santa Cruz Biotechnologies ) and mouse anti-DlECD C594 . 9B ( 1:200 , DSHB ) . The outline of individual , dispersed neuroblast lineages was determined by the staining pattern of general cell cortex marker F-actin or CD8-GFP/CD8-RFP and marked by white dashed line . Human embryonic kidney HEK293T cells ( ATCC , RRID: CRL-3216; obtained from Dr . Hong Wu’s laboratory , Peking University , and authenticated by ATCC ) were maintained in DMEM medium ( Invitrogen ) supplemented with 10% FBS at 37˚C and 5% CO2 . DNA transfection was performed using a standard polyethylenimine ( PEI ) protocol . The cell line has been tested for and confirmed to be negative for mycoplasma contamination , using short tandem repeat ( STR ) profiling technique . Coimmunoprecipitation ( CoIP ) assays in HEK 293 T cell extracts were performed as previously described ( Liu et al . , 2017; Song and Lu , 2012 ) . Briefly , 48 hr after transfection , HEK 293 T cells were harvested , washed and resuspended in lysis buffer [50 mM Tris-HCl ( pH 8 . 0 ) ; 120 mM NaCl; 5 mM EDTA; 1% NP-40; 10% glycerol; protease inhibitor cocktail ( Sigma-Aldrich ) ; 2 mM Na3VO4] and kept on ice for 20 min . Cell extracts were sonicated with Bioruptor Plus ( Biosense ) at 4°C . The cell extracts were clarified by centrifugation , and proteins immobilized by binding to anti-FLAG M2 or anti-V5 ( Sigma-Aldrich ) affinity gel for 4 hr or overnight at 4°C . Beads were washed and proteins recovered directly in SDS-PAGE sample buffer . Rabbit anti-FLAG ( Sigma-Aldrich ) , rabbit anti-V5 ( Sigma-Aldrich ) or mouse anti-c-Myc ( CWBIO ) were used for Western blot analysis . For in vivo coIP , larval brains coexpressing UAS-Vps35-FLAG and UAS-Notch-V5 by insc-Gal4 were used as experimental group , whereas larval brains expressing UAS-Vps35-FLAG alone by insc-Gal4 served as control . Approximately 350 late third instar larval brains of each genotype were dissected and collected in ice-cold 1xPBS solution . Protein samples were prepared by grinding brains in lysis buffer [50 mM Tris-HCl , 120 mM NaCl , 5 mM EDTA , 10% glycerol , 1% NP-40 , protease inhibitor cocktail ( Sigma-Aldrich ) ] with a plastic pestle . Immunoprecipitation was carried out with anti-V5 affinity gels ( Sigma-Aldrich ) . GFP+ larval brain pieces were transplanted into the abdomen of young female adult host flies as previously described ( Caussinus and Gonzalez , 2005; Liu et al . , 2017 ) . After transplantation , host flies were transferred to fresh food every day and were observed under a fluorescent scope every two days to analyze tumor formation and metastasis . Drosophila late third instar larval brains were dissected in PBS buffer , and immediately transferred into Fixation buffer I ( 2% paraformaldehyde/2 . 5% glutaraldehyde in 0 . 1 M phosphate buffer , pH 7 . 4 ) for 2 hr at room temperature , and then overnight at 4°C . The samples were then fixed in the Fixation buffer II ( 1% tannic acid/2 . 5% glutaraldehyde in 0 . 1 M phosphate buffer , pH 7 . 4 ) for 2 hr at room temperature . After rinsing several times in phosphate buffer , the brain samples were post-fixed in 2% OsO4 with 1 . 5% Potassium Ferrocyanide for 1 hr at room temperature and stained with 2% aqueous uranyl acetate overnight at 4°C . Following several washes in distilled water , samples were dehydrated through a graded alcohol series and subsequently embedded in Spurr’s resin ( SPI supplies , PA , USA ) . Ultra-thin sections ( 70 nm ) were cut with a diamond knife using an ultramicrotome ( UC7 , Leica Microsystem ) and mounted on copper grids with a single slot . Sections were stained with uranyl acetate and lead citrate , and observed under a FEI Tecnai G2 Spirit transmission electron microscope at 120 kV . | Most cells in the animal body are tailored to perform particular tasks , but stem cells have not yet made their choice . Instead , they have unlimited capacity to divide and , with the right signals , they can start to specialize to become a given type of cells . In the brain , this process starts with a stem cell dividing . One of the daughters will remain a stem cell , while the other , the neural progenitor , will differentiate to form a mature cell such as a neuron . Keeping this tight balance is crucial for the health of the organ: if the progenitor reverts back to being a stem cell , there will be a surplus of undifferentiated cells that can lead to a tumor . A one-way signal driven by the protein Notch partly controls the distinct fates of the two daughter cells . While the neural progenitor carries Notch at its surface , its neural stem cell sister has a Notch receptor on its membrane instead . This ensures that the Notch signaling goes in one direction , from the cell with Notch to the one sporting the receptor . When a stem cell divides , one daughter gets more of a protein called Numb than the other . Numb pulls Notch receptors away from the external membrane and into internal capsules called endosomes . This guarantees that only one of the siblings will be carrying the receptors at its surface . Yet , sometimes the Notch receptors can get activated in the endosomes , which may make neural progenitors revert to being stem cells . It is still unclear what tools the cells have to stop this abnormal activation . Here , Li et al . screened brain cells from fruit fly larvae to find out the genes that might play a role in suppressing the inappropriate Notch signaling . This highlighted a protein complex known as the retromer , which normally helps to transport proteins in the cell . Experiments showed that , in progenitors , the retromer physically interacts with Notch receptors and retrieves them from the endosomes back to the cell surface . If the retromer is inactive , the Notch receptors accumulate in the endosomes , where they can be switched on . It seems that , in fruit flies , the retromer acts as a bomb squad that recognizes and retrieves potentially harmful Notch receptors , thereby preventing brain tumor formation . Several retromer components are less present in patients with various cancers , including glioblastoma , an aggressive form of brain cancer . The results by Li et al . may therefore shed light on the link between the protein complex and the emergence of the disease in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"neuroscience"
] | 2018 | The retromer complex safeguards against neural progenitor-derived tumorigenesis by regulating Notch receptor trafficking |
The mechanisms linking maternal stress in pregnancy with infant neurodevelopment in a sexually dimorphic manner are poorly understood . We tested the hypothesis that maternal hypothalamic-pituitary-adrenal axis activity , measured by hair cortisol concentration ( HCC ) , is associated with microstructure , structural connectivity , and volume of the infant amygdala . In 78 mother-infant dyads , maternal hair was sampled postnatally , and infants underwent magnetic resonance imaging at term-equivalent age . We found a relationship between maternal HCC and amygdala development that differed according to infant sex . Higher HCC was associated with higher left amygdala fractional anisotropy ( β = 0 . 677 , p=0 . 010 ) , lower left amygdala orientation dispersion index ( β = −0 . 597 , p=0 . 034 ) , and higher fractional anisotropy in connections between the right amygdala and putamen ( β = 0 . 475 , p=0 . 007 ) in girls compared to boys . Furthermore , altered amygdala microstructure was only observed in boys , with connectivity changes restricted to girls . Maternal cortisol during pregnancy is related to newborn amygdala architecture and connectivity in a sexually dimorphic manner . Given the fundamental role of the amygdala in the emergence of emotion regulation , these findings offer new insights into mechanisms linking maternal health with neuropsychiatric outcomes of children .
Prenatal exposure to maternal stress is estimated to affect 10–35% of children worldwide , which is a major concern because early life stress may contribute to impaired cognitive development , negative affectivity , autism spectrum disorder ( ASD ) , and psychiatric diagnoses including attention deficit hyperactivity disorder ( ADHD ) , addiction , depression , and schizophrenia ( Van den Bergh et al . , 2017 ) . Neural correlates of prenatally stressed children include alternations in brain structural and functional connectivity , especially in networks involving the amygdala and prefrontal cortex ( Scheinost et al . , 2017 ) . Adaptation of the maternal hypothalamic-pituitary-adrenal ( HPA ) axis is a key mechanism by which maternal stress modulates offspring neurodevelopment ( Moisiadis and Matthews , 2014 ) , and there is evidence that this mechanism operates in a sexually dimorphic manner ( Sutherland and Brunwasser , 2018 ) . For example , higher waking maternal salivary cortisol in pregnancy is associated with increased internalising behaviours in female infants and reduced internalising behaviours in males ( Braithwaite et al . , 2017a; Braithwaite et al . , 2017b ) . Higher maternal salivary cortisol in pregnancy is also associated with stronger amygdala functional connectivity with networks involved in sensory processing and integration in newborn girls , with weaker connectivity to these brain regions in boys ( Graham et al . , 2019 ) ; and in childhood , with larger amygdalae ( Buss et al . , 2012 ) and reduced segregation of structural networks in girls but not boys ( Kim et al . , 2016 ) . The amygdala is further implicated as a neural target of prenatal stress exposure by observations from studies that have characterised maternal stress by symptomatology of depression and/or anxiety , which report alterations in amygdala volume ( Wen et al . , 2017 ) , microstructure ( Rifkin-Graboi et al . , 2013 ) , and functional and structural connectivity among offspring ( Posner et al . , 2016 ) . Candidacy of the amygdala as an important neural target of prenatal stress exposure comes from the following observations in pre-clinical and clinical studies . First , the amygdala develops early in embryonic life ( Humphrey , 1968 ) and contains a high concentration of glucocorticoid receptors ( Wang et al . , 2014 ) ; second , increased maternal glucocorticoids modulate amygdala development and anxiety-like behaviours in experimental models ( Welberg et al . , 2000; Barbazanges et al . , 1996 ) ; third , lesion studies in non-human primates support its critical role in early development of emotion regulation ( Schumann et al . , 2011 ) ; fourth , newborn amygdala functional connectivity is consistently linked with internalising behaviours in children up to the age of 2 years ( Graham et al . , 2019; Rogers et al . , 2017 ) ; fifth , early disruption to cell composition of the amygdala is reported in a model of early life stress ( Kraszpulski et al . , 2006 ) , and in children with autism ( Avino et al . , 2018 ) ; and sixth , in pre-clinical models , stress and glucocorticoid exposure induce dendritic arborisation , amygdala hypertrophy and induce anxiety-like behaviours ( Vyas et al . , 2003; Mitra and Sapolsky , 2008 ) . Neonatal magnetic resonance imaging ( MRI ) serves as an intermediate phenotype for investigating the impact of early life exposures on brain and health because it is distal to the aetiological process , in this case prenatal stress , and is proximal to cognitive , behavioural and disease outcomes . Structural and diffusion MRI ( dMRI ) have been used to characterise brain structural maturation and emerging network connectivity during the perinatal period , and to investigate pathways to atypical development ( Batalle et al . , 2018; Boardman and Counsell , 2020 ) . It is a suitable tool to investigate the impact of prenatal stress exposure on the amygdala because age-specific templates enable accurate parcellation of the amygdala and associated structures ( Alexander et al . , 2017 ) ; and diffusion tensor imaging and neurite orientation and dispersion density imaging ( NODDI ) support inference about tissue microstructure and network connectivity , modelled by fractional anisotropy ( FA ) , mean diffusivity ( MD ) , orientation dispersion index ( ODI ) and neurite density index ( NDI ) ( Galdi et al . , 2020; Zhang et al . , 2012 ) . Hair cortisol concentration ( HCC ) measured in 3 cm hair samples collected from close to the scalp reflects basal HPA axis activity over the 3 months prior to sampling , and in contrast to single measures from saliva or blood , it is not influenced by short-term activation of the HPA axis in response to acute stressors ( Staufenbiel et al . , 2013 ) . Studies in pregnant women have shown HCC to be an efficient method of retrospective assessment of long-term cortisol secretion , and thus long-term HPA axis activity ( Kirschbaum et al . , 2009; D'Anna-Hernandez et al . , 2011 ) . Previous studies have reported sex-specific differences between maternal stress and amygdala functional connectivity and behavioural outcomes among children ( Braithwaite et al . , 2017a; Braithwaite et al . , 2017b; Graham et al . , 2019; Buss et al . , 2012; Kim et al . , 2016 ) , but study designs leave uncertainty about the mechanism linking maternal stress with amygdala development , the potential confounding role of events and environmental exposures during childhood , and the impact of stress on structural connectivity . Resolving these uncertainties is necessary for developing strategies designed to improve socio-emotional development of children born to women who are stressed during pregnancy . Based on studies of the imaging , biochemical , and clinical phenotype of prenatal stress exposure , we hypothesised that higher levels of maternal HPA activity in the final months of pregnancy ascertained from maternal HCC would impact amygdala development and structural connectivity of offspring infants in a sexually-dimorphic manner , and that these effects would be apparent around the time of birth .
The parents of 102 infants consented to take part . Of these , two preterm infants died before term equivalent age , 12 did not complete the MRI protocol or images were not amenable processing due to movement artefact; one had an incidental structural anomaly detected at MRI; and nine withdrew before MRI scan . This left data from 78 mother-infant dyads for analysis , the maternal and infant characteristics for whom are shown in Table 1 . The preterm group ( n = 36 ) had the following co-morbidities: early onset sepsis ( n = 1 ) ; late onset sepsis ( n = 5 ) , necrotising enterocolitis ( n = 3 ) ; histological chorioamnionitis ( n = 14 ) ; mechanical ventilation during NICU stay ( n = 16 , median period of ventilation 2 days [IQR , 1 to 5 days] ) . Maternal hair was sampled at mean 3 . 5 ± 2 . 5 days after delivery , and the median HCC concentration was 5 . 6 pg/mg ( 0 . 5–107 . 1 ) . Maternal HCC was not associated with gestational age ( GA ) at birth ( r = 0 . 200 , p=0 . 094 ) . HCC did not differ between mothers of male and female infants ( p=0 . 997 ) . MRI was carried out at term-equivalent age: median 41 . 9 weeks’ GA ( range 38 . 6–45 . 9 ) . In univariate analysis , there were moderately strong correlations between both FA and MD , and GA at birth and age at scan ( r = 0 . 41–0 . 64 ) , and weak correlations with birth weight z-score and Scottish Index of Multiple Deprivation 2016 ( SIMD2016 ) quintile ( r = 0 . 24–0 . 30 ) . There were no significant correlations between FA and MD in amygdalae and ethnicity or infant sex , or maternal parity , age or BMI . There were moderate-to-strong correlations between NDI in the amygdalae and GA at birth and age at scan ( r = 0 . 43–0 . 74 ) , and a weak correlation with SIMD2016 quintile ( r = 0 . 23–0 . 26 ) . Weak-to-moderate correlations were observed between ODI in amygdalae with GA at birth and age at scan ( r = 0 . 28–0 . 42 ) , Supplementary file 1 . In multiple linear regression models , there was a significant interaction effect between maternal HCC and infant sex in left amygdala FA ( p=0 . 010 ) and ODI ( p=0 . 034 ) , with higher maternal HCC being associated with higher left amygdala FA and lower ODI in girls compared to boys ( Table 2 , Figure 1 ) . When we stratified by sex , there were associations between maternal HCC and infant amygdala microstructure in boys , but not girls . Table 3 shows that in boys , higher maternal HCC was associated with lower left amygdala FA ( β = −0 . 339 ) , lower right amygdala FA ( β = −0 . 287 ) and NDI ( β = −0 . 215 ) , and higher right amygdala MD ( β = 0 . 264 ) and ODI ( β = 0 . 309 ) , after FDR correction . For both hemispheres , the networks with the top 20% number of streamlines were connected to eight structures: thalamus , putamen , insula , superior temporal gyrus , inferior temporal gyrus , middle temporal gyrus , caudate , and lateral orbitofrontal cortex ( Figure 2 and Figure 2—figure supplement 1 ) . Quantification of streamline counts is given in Supplementary file 2 and illustrated in Figure 3 . Maternal HCC was not associated with streamline counts of the left and right amygdala with these regions . In fully adjusted analyses , the interaction between maternal HCC and infant sex was significant for mean FA of connections between the right amygdala and putamen . Higher maternal HCC was associated with higher FA for amygdala-putamen connectivity in girls compared with boys ( p=0 . 007 ) ( Figure 1 ) . The interaction was also seen for connections to left thalamus , putamen , and insula , but the interaction term did not remain after correction for multiple tests ( Supplementary file 3 ) . In sex-stratified analysis , girls had higher FA values in association with high maternal HCC in connections between left amygdala with thalamus , putamen , and inferior temporal gyrus , and the right amygdala with putamen and inferior temporal gyrus , but these were not significant after correction for multiple tests ( Supplementary file 3a ) . Mean volumes of the left and right amygdala were 877 ± 111 mm3 and 823 ± 91 mm3 , respectively . In univariate analysis , there were weak associations ( r = 0 . 24–0 . 3 ) between amygdala volume and GA at birth and birth weight z-score , but not with age at scan , SIMD2016 quintile , sex , ethnicity , or maternal BMI , parity or age ( Supplementary file 1 ) . Maternal HCC was not associated with infant right or left amygdala volume in regression models adjusted for potential covariates , and interaction terms between maternal HCC and infant sex were not significant ( Supplementary file 3b ) . There were seven twin sets in the whole sample . When we repeated analyses including only singletons and the first born of twin pairs , significant associations between maternal HCC , sex and image feature remained , with little change to the value of regression coefficients ( Supplementary file 3c ) . In subgroup analysis of preterm and term infants , the direction and magnitude of interaction effects for both groups were similar to those of the whole sample . Specifically , when tested in term and preterm infants , respectively , higher maternal hair cortisol concentration was associated with higher left amygdala fractional anisotropy ( β = 0 . 735 and 0 . 640 ) , lower left amygdala orientation dispersion index ( β = −0 . 710 and −0 . 614 ) , and higher fractional anisotropy in connections between the right amygdala and putamen ( β = 0 . 733 and 0 . 426 ) in girls compared to boys ( Supplementary file 3c ) . Bronchopulmonary dysplasia , sepsis , and histological chorioamnionitis did not correlate with any image feature tested . The need for mechanical ventilation did correlate with lower FA in the right amygdala and its connections to putamen , but it had little influence on the interactive effect of sex and maternal cortisol on amygdala microstructure or connectivity ( Supplementary file 3d ) .
We report a mechanism that could explain the impact of maternal stress on infant brain development . We found that maternal HCC , a stable marker of chronic maternal HPA axis activity in pregnancy , is associated with microstructure and structural connectivity of the newborn amygdala , a region of functional importance for early social development and emotion regulation . Specifically , HCC interacts with infant sex to modify amygdala FA , ODI , and NDI , which supports the inference that maternal chronic HPA activity has an impact on dendritic structure , axonal configuration , and the packing density of neurites , in a sexually dimorphic manner ( Jespersen et al . , 2012; Grussu et al . , 2017; Sato et al . , 2017; Nazeri et al . , 2020 ) . The findings are consistent with recent reports from the GUSTO ( Growing Up in Singapore Towards Health Outcomes ) cohort that describe associations between maternal depressive symptoms and alterations in offspring amygdala development ( Wen et al . , 2017; Rifkin-Graboi et al . , 2013 ) . That study highlighted the role of maternal mental health on newborn brain development , and focussed attention on the amygdala . Here , we provide mechanistic insights into the relationship between maternal health and wellbeing and amygdala development by using maternal HCC to characterise chronic HPA activity , and the NODDI model for inference about tissue microstructure . We chose to measure NODDI parameters for assessing microstructure because ODI and NDI in grey matter appear to be functionally tractable . For example , diffusion markers of dendritic density and arborisation in grey matter predict differences in intelligence ( Genç et al . , 2018 ) , reduced ODI in grey matter is reported in psychosis and in neurodegenerative disease , and reduced grey matter NDI is reported in Parkinson’s disease , Alzheimer’s disease , autism spectrum disorder , and temporal lobe epilepsy ( for review see Nazeri et al . , 2020 ) . Maternal HCC was also related to structural connectivity of the amygdala in a sex-discordant manner . Higher maternal HCC was associated with higher FA in girls than boys in tracts between right amygdala and putamen . These observations were not explained by differences in streamline counts in relation to maternal HCC . Furthermore , in sex-stratified analysis , there were consistent trends for girls born to women with higher HCC to have higher mean FA between the left amygdala and left thalamus , putamen and inferior temporal gyrus , and between right amygdala and right putamen and right inferior temporal gyrus , although these did not survive FDR correction . During the neonatal period , higher FA in white matter tracts is typically taken to imply microstructural maturation , through increased axon diameter , density , or myelination . Therefore , increased mean FA demonstrated in connections between the amygdala and putamen , in girls exposed to higher cortisol , could be interpreted as increased maturation of these connections . Several plausible biological mechanisms could underlie sex differences in associations between HCC and fetal neurodevelopment . First , the capacity of the placenta to regulate the passage of cortisol from mother to fetus differs according to fetal sex , evidenced by sex-discordant expression of enzymes controlling glucocorticoid metabolism within the placenta ( Carpenter et al . , 2017; Meakin et al . , 2017 ) . Second , evidence from gene expression studies indicate that fetal sex influences the direct actions of cortisol in the placenta ( Rosenfeld , 2015 ) . Third , the same level of cortisol exposure likely holds sex-different actions at the level of the fetal brain , given that male and female fetuses have different glucocorticoid and mineralocorticoid expression across development ( Owen and Matthews , 2003 ) . Finally , there may be sex differences in placental release of corticotropin-releasing hormone ( CRH ) , which influence fetal cortisol exposure and neurodevelopment ( Bangasser and Wiersielis , 2018; Sandman et al . , 2013 ) . To our knowledge , this is the first study to investigate a physiological measure of chronic maternal HPA activity with quantitative biomarkers of brain development , and to include infants born very preterm . The relationships we describe appear to apply across the whole GA range because GA at birth was included as a covariate in regression models , and in sub-group analyses the magnitude and direction of ‘HCC x sex’ interaction effects existed in term and preterm groups . This suggests that maternal health and wellbeing across gestation influences neurodevelopment in preterm and term infants . This is important because preterm birth has previously been associated with both exposure to maternal HPA axis dysregulation ( Duthie and Reynolds , 2013 ) , and an increased risk of inattention and affective disorders ( Johnson , 2007 ) . Furthermore , the finding that maternal HCC is associated with neonatal amygdala development in preterm and term infants is consistent with the observation that maternal cortisol in early gestation predicts amygdala volume at the age of 7 years ( Buss et al . , 2012 ) . Given that mothers’ HPA axis shows considerable trait-stability across pregnancy , whereby women with higher cortisol concentrations in the second trimester also tend to have higher cortisol in the third trimester ( Graham et al . , 2019; Stoye et al . , 2020 ) , the data lend further support to strategies designed to optimise pre- and early pregnancy health for optimising fetal neurodevelopment . Strengths of this study are the use of biophysical tissue modelling ( NODDI ) to enable inference about neurite density and organisation in the amygdala; and use of a data-driven approach to investigate amygdala structural connectivity . A second strength is use of maternal HCC to operationalise stress because it is a quantitative stable marker of cortisol secretion that represents HPA activity over 3 months; as such HCC is unlikely to reflect transient stresses that can occur in pregnancy , and it overcomes the problems of diurnal variation that occur with plasma and saliva measurements . The study has some limitations: first it was not powered to detect both sex and birth gestation interactions , but this should be considered in future study design . Second , follow-up studies that include measures of socio-emotional development are needed to understand functional consequences of these findings . Finally , the newborn amygdalae are relatively small anatomical regions so could be susceptible to partial volume effects influencing microstructural characteristics . To mitigate this risk , we used an age-specific atlas for segmentation , and excluded voxels with a υiso < 0 . 5 . Third , due to the close proximity of the amygdala and putamen the corresponding connectivity FA may to some degree reflect microstructure of the regions themselves . Longitudinal studies that evaluate socio-emotional development are needed to understand the functional consequences of these findings . In addition , as the network architecture underlying socio-emotional function in early life becomes more certain with technological advances , it may be useful to focus image analyses on individual networks . The subcortical-gaze pathway is of specific interest because in adults there is evidence that structural connectivity of the superior colliculus and amygdala is related to processing of facial expressions ( McFadyen , 2019 ) , which is one of the earliest social cognitive abilities to develop and is foundational to other trajectories of cognitive development . In conclusion , dMRI and HCC were used to investigate mechanisms underlying the transmission of prenatal stressors on infant development . Maternal HCC in pregnancy is associated with newborn amygdala microstructure and structural connectivity , in a sex-dimorphic manner . These findings reveal that the amygdala , a structure of known importance for child development , is susceptible to variations in the prenatal stress environment , and that cortisol imparts sex-specific effects on human fetal neurodevelopment .
The ‘Stress Response Systems in Mothers and Infants’ cohort recruited mother-infant dyads from the Royal Infirmary , Edinburgh , between March 2018 and August 2019 . It prospectively tests associations of perinatal glucocorticoid exposure with brain development , and early life exposures including preterm birth with infant HPA axis regulation . We recruited mother-infant dyads who presented to NHS services with threatened preterm labour and delivered at ≤32 completed weeks of gestation , and women who delivered ≥37 weeks’ gestation and received routine care with their infant on postnatal wards . No participants were recruited from high-risk antenatal clinics . Exclusion criteria were congenital fetal abnormality , chromosomal abnormality or regular maternal corticosteroid use . All women gave written informed consent . Ethical approval was granted by South East Scotland 01 Regional Ethics Committee ( 18/SS/0006 ) . Maternal hair was sampled within 10 days of delivery . Hair was cut close to the scalp , at the posterior vertex , and stored in aluminium foil at −20°C . The proximal 3 cm of hair were analysed by liquid chromatography-tandem mass spectrometry ( LC-MS/MS ) , at Dresden Lab Service GmbH ( Dresden , Germany ) , using an established protocol ( Gao et al . , 2013 ) . Adult hair commonly grows at 1 cm/month ( Wennig , 2000 ) and thus hair segments represented maternal HPA axis activity over the last 3 months of pregnancy . Participant demographic information was collected through maternal questionnaire and review of medical records . Collected maternal information included: age at delivery ( years ) , parity ( primiparous/multiparous ) , clinical diagnosis of gestational diabetes , pre-eclampsia , pharmacological treatment for depression during pregnancy , antenatal corticosteroid exposure for threatened preterm birth; body mass index ( BMI ) calculated at antenatal booking; smoking status defined as having smoked any tobacco in pregnancy; SIMD 2016 quintile rank , a score generated by the Scottish government which measures localities’ deprivation according to local income , employment , health , education , geographic access to services , crime and housing . Infant demographics included whether participants were a singleton or twin , ethnicity , GA at birth ( weeks ) , and birth weight z-score calculated according to intergrowth standards ( Villar et al . , 2014 ) . Diffusion MRI processing was performed as follows: for each subject the two dMRI acquisitions were first concatenated and then denoised using a Marchenko-Pastur-PCA-based algorithm ( Veraart et al . , 2016 ) ; the eddy current , head movement and EPI geometric distortions were corrected using outlier replacement and slice-to-volume registration ( Smith et al . , 2004; Andersson et al . , 2003; Andersson and Sotiropoulos , 2016; Andersson et al . , 2017 ) ; bias field inhomogeneity correction was performed by calculating the bias field of the mean b0 volume and applying the correction to all the volumes ( Tustison et al . , 2010 ) . The T2w images were processed using the minimal processing pipeline of the developing human connectome project ( dHCP ) to obtain the bias field corrected T2w , the brain masks and the different tissue probability maps ( Makropoulos et al . , 2018 ) . The mean b0 EPI volume of each subject was co-registered to their structural T2w volume using boundary-based registration ( Greve and Fischl , 2009 ) . The 10 manually labelled subjects of the M-CRIB atlas ( Alexander et al . , 2017 ) were registered to the bias field corrected T2w using rigid , affine and symmetric normalisation ( SyN ) ( Avants et al . , 2008 ) . Next , the registered labels of the 10 atlases were merged using joint label fusion ( Wang et al . , 2013 ) , resulting in a parcellation containing 84 regions of interest ( ROIs ) . Volumes were calculated from ROIs derived in the structural images . ROIs were propagated to the diffusion native space using the previously computed transformation . To calculate the tensor derived metric , only the first shell was used . NODDI metrics were calculated using the recommended values for neonatal grey-matter of the parallel intrinsic diffusivity ( 1 . 25 μm2·ms−1 ) ( Zhang et al . , 2012; Guerrero et al . , 2019 ) . The obtained metrics are: neurite density index ( NDI ) , isotropic volume fraction ( υiso ) and orientation dispersion index ( ODI ) . The mean FA , MD , ODI , and NDI were calculated for the left and right amygdalae M-CRIB ROIs , after exclusion of voxels with a υiso < 0 . 5 . Voxels with a υiso < 0 . 5 were excluded , in order to minimise partial volume effects ( Batalle et al . , 2019 ) . Tractography was performed using constrained spherical deconvolution ( CSD ) and anatomically constrained tractography ( Tournier et al . , 2019; Smith et al . , 2012 ) The required 5-tissue type file , was generated by combining the tissue probability maps obtained from the dHCP pipeline with the subcortical structures derived from the parcellation process . Multi-tissue response function was calculated , with a FA threshold of 0 . 1 . The average response functions were calculated . Then , the multi-tissue fiber orientation distribution ( FOD ) was calculated ( Jeurissen et al . , 2014 ) , and global intensity normalisation on the FODs images was performed . Finally , the tractogram was created , generating 10 million streamlines , with a minimum length of 20 mm and a maximum of 200 mm and a cut-off of 0 . 05 ( default ) , using backtrack and a dynamic seeding ( Smith et al . , 2015a ) . To be able to quantitatively assess connectivity , spherical-deconvolution informed filtering of tractograms two ( SIFT2 ) was applied to the resulting tractograms ( Smith et al . , 2015a ) . The connectivity matrix was constructed using a robust approach , a 2 mm radial search at the end of the streamline was performed to allow the tracts to reach the GM parcellation ( Smith et al . , 2015b ) . The final connectivity matrices were multiplied by the μ coefficient obtained during the SIFT2 process . These connectomes gave a quantification of the SIFT2 weights ( referred to as the streamline counts ) , and the mean FA of connections , between both the left and right amygdala to 41 unilateral regions of interest defined through M-CRIB parcellation . In order to focus analysis on to amygdala’s most structurally connected areas , these 82 ROIs were thresholded according to the number of streamlines connecting them to the left or right amygdala , with the top 20% ( N = 16 ) of connections taken forward for further analysis testing relationships with maternal HCC . Analyses were performed using IBM SPSS Statistics Version 25 Armonk , NY: IBM Corp . Continuous data are summarised as mean ± SD if they had a normal distribution , and median ( range ) if skewed . Maternal HCC was positively skewed , and log-10 transformed for analysis . The relationship between maternal HCC with infant characteristics was tested using independent t-test and Pearson’s correlation for categorical and continuous variables , respectively . Associations between maternal HCC with ( i ) left and right amygdala microstructure ( FA , MD , NDI , ODI ) , ( ii ) structural connectivity ( number of streamlines and mean FA of connections ) , ( iii ) amygdalae volumes were tested using multiple linear regression . In all models , image feature was the dependent variable and maternal HCC was an independent variable . Covariates included infant sex and clinical or demographic factors that were correlated with either left or right amygdala microstructure or volume using Pearson’s correlation . Associations with the following were tested: GA at birth , age at scan , birth weight z-score , SIMD2016 quintile , infant ethnicity , infant sex , and maternal parity , BMI and age . Antenatal corticosteroid treatment for threatened preterm birth was not included as a covariate because it was given to n = 36 ( 100% ) women in the preterm group , was highly correlated with GA at birth ( r = 0 . 958 , p<0 . 001 ) , so its inclusion as a covariate would have introduced multicollinearity in regression analysis . For descriptive purposes , correlations of infant and maternal factors considered as potential covariates are described as weak if r < 0 . 3 , moderate if r = 0 . 3–0 . 7 , and strong if r > 0 . 7 . Sex differences in the relationship between maternal HCC and newborn imaging features were assessed by adding an interaction term between maternal HCC and infant sex in the whole group regression model . If a significant interaction was present , sex stratified analysis was conducted independently in boys and girls . Benjamini and Hochberg false discovery rate ( FDR ) correction was used to adjust p-values for multiple testing . FDR corrections were conducted separately for assessments of left amygdala microstructure ( n = 4 ) , right amygdala microstructure ( n = 4 ) , left amygdala connectivity ( n = 8 ) and right amygdala connectivity ( n = 8 ) . One sensitivity analysis was carried out to assess whether associations between maternal HCC and image features might be enhanced by inclusion of twins . We repeated analysis of features with a significant ‘HCC x sex’ interaction in the whole sample , using only singleton pregnancies and the first-born infant of twin pairs . One sub-group analysis of preterm ( GA at birth ≤32 weeks ) and term infants ( GA at birth ≥37 weeks ) was carried out because the relationship between maternal HCC and infant brain development may be gestation specific . | Stress during pregnancy , for example because of mental or physical disorders , can have long-term effects on child development . Epidemiological studies have shown that individuals exposed to stress in the womb are at higher risk of developmental and mood conditions , such as ADHD and depression . This effect is different between the sexes , and the biological mechanisms that underpin these observations are poorly understood . One possibility is that a baby’s developing amygdala , the part of the brain that processes emotions , is affected by a signal known as cortisol . This hormone is best known for its role in coordinating the stress response , but it also directs the growth of a fetus . Tracking fetal amygdala changes as well as cortisol levels in the pregnant individual could explain how stress during pregnancy affects development . To investigate , Stoye et al . recruited nearly 80 volunteers and their newborn children . MRI scans were used to examine the structure of the amygdala , and how it is connected to other parts of the brain . In parallel , the amount of cortisol was measured in hair samples collected from the volunteers around the time of birth , which reflects stress levels during the final three months of pregnancy . Linking the brain imaging results to the volunteers’ cortisol levels showed that being exposed to higher cortisol levels in the womb affected babies in different ways based on their sex: boys showed alterations in the fine structure of their amygdala , while girls displayed changes in the way that brain region connected to other neural networks . The work by Stoye et al . potentially reveals a biological mechanism by which early exposure to stress could affect brain development differently between the sexes , potentially informing real-world interventions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Maternal cortisol is associated with neonatal amygdala microstructure and connectivity in a sexually dimorphic manner |
The general translation initiation factor eIF2 is a major translational control point . Multiple signaling pathways in the integrated stress response phosphorylate eIF2 serine-51 , inhibiting nucleotide exchange by eIF2B . ISRIB , a potent drug-like small molecule , renders cells insensitive to eIF2α phosphorylation and enhances cognitive function in rodents by blocking long-term depression . ISRIB was identified in a phenotypic cell-based screen , and its mechanism of action remained unknown . We now report that ISRIB is an activator of eIF2B . Our reporter-based shRNA screen revealed an eIF2B requirement for ISRIB activity . Our results define ISRIB as a symmetric molecule , show ISRIB-mediated stabilization of activated eIF2B dimers , and suggest that eIF2B4 ( δ-subunit ) contributes to the ISRIB binding site . We also developed new ISRIB analogs , improving its EC50 to 600 pM in cell culture . By modulating eIF2B function , ISRIB promises to be an invaluable tool in proof-of-principle studies aiming to ameliorate cognitive defects resulting from neurodegenerative diseases .
In the integrated stress response ( ISR ) , phosphorylation of the α-subunit of the eukaryotic translation initiation factor eIF2 ( eIF2α-P ) at serine-51 acts as a major regulatory step that controls the rate of translation initiation . Four distinct eIF2α kinases can catalyze phosphorylation at this single residue , each acting in response to different cellular stress conditions: PERK senses accumulation of unfolded polypeptides in the lumen of the endoplasmic reticulum ( ER ) , GCN2 responds to amino acid starvation and UV-light , PKR responds to viral infection , and HRI responds to heme deficiency . Their convergence on the same molecular event leads to a reduction in overall protein synthesis . Concomitant with a decrease in new protein synthesis , preferential translation of a small subset of mRNAs that contain small upstream open reading frames ( uORFs ) in their 5′ untranslated region is induced ( Harding et al . , 2003; Wek et al . , 2006 ) . ISR-translational targets include the well-known mammalian ATF4 ( Activating Transcription Factor 4 ) and CHOP ( a pro-apoptotic transcription factor ) ( Harding et al . , 2000; Vattem and Wek , 2004; Palam et al . , 2011 ) . ATF4 regulates genes involved in metabolism and nutrient uptake and was shown to have a cytoprotective role upon stress in many cellular contexts ( Ye et al . , 2010 ) . ATF4 is also a negative regulator of ‘memory genes’ and its preferential translation in neurites can transmit a neurodegenerative signal in neurons ( Chen et al . , 2003; Baleriola et al . , 2014 ) . ISR activation leads to preferential translation of key regulatory molecules and thus its level and duration of induction must be tightly regulated . Cells ensure that the effects of eIF2α-P are transient by also activating a negative feedback loop . This is accomplished by GADD34 induction , which encodes the regulatory subunit of the eIF2α phosphatase ( Lee et al . , 2009 ) . GADD34 induction leads to a reduction of eIF2α-P , allowing cells to restore translation ( Novoa et al . , 2001 ) . eIF2 is a trimeric complex ( comprised of α , β and γ subunits ) that binds to both GTP and the initiator methionyl tRNA ( Met-tRNAi ) to form a ternary complex ( eIF2•GTP•Met-tRNAi ) . After engaging the 40S ribosomal subunit at an AUG start codon recognized by Met-tRNAi , GTP is hydrolyzed by the GTPase activating protein ( GAP ) eIF5 , and the 60S ribosomal subunit joins to form a complete 80S ribosome ready for polypeptide elongation . eIF2•GDP is released , and eIF2 must then be reloaded with GTP to enter another round of ternary complex formation ( Hinnebusch and Lorsch , 2012 ) . In addition to being a GAP for eIF2 , eIF5 is also a GDP dissociation inhibitor that prevents GDP release from eIF2 ( Jennings and Pavitt , 2015 ) . The exchange of GDP with GTP in eIF2 is catalyzed by its dedicated guanine nucleotide exchange factor ( GEF ) eIF2B , which has the dual function of catalyzing the release of both eIF5 and GDP ( Jennings et al . , 2013 ) . eIF2B is a complex molecular machine , composed of five different subunits , eIF2B1 through eIF2B5 , also called the α , β , γ , δ , and ε subunits . eIF2B5 catalyzes the GDP/GTP exchange reaction and , together with a partially homologous subunit eIF2B3 , constitutes the ‘catalytic core’ ( Williams et al . , 2001 ) . The three remaining subunits ( eIF2B1 , eIF2B2 , and eIF2B4 ) are also highly homologous to one another and form a ‘regulatory sub-complex’ that provides binding sites for eIF2B's substrate eIF2 ( Dev et al . , 2010 ) . When phosphorylated on Ser-51 , eIF2α-P dissociates more slowly from the eIF2B regulatory sub-complex and locks eIF2B into an inactive state ( Krishnamoorthy et al . , 2001 ) . Phosphorylation thus renders eIF2 an inhibitor of its own GEF . Because eIF2 is more abundant than eIF2B , a small amount of eIF2α-P is sufficient to sequester a large proportion of available eIF2B , leading to a substantial reduction in overall protein synthesis . Using a cell-based high-throughput screen , we recently identified a small molecule , ISRIB ( for integrated stress response inhibitor ) that renders cells resistant to the inhibitory effects of eIF2α-P . ISRIB , the only bona fide ISR inhibitor identified to date , is a highly potent compound ( EC50 = 5 nM in cells ) and has good pharmacokinetic properties ( Sidrauski et al . , 2013 ) . In agreement with the phenotype of genetically modified mice having reduced eIF2α-P , we showed that treatment with ISRIB enhances memory consolidation in rodents . Moreover , ISRIB comprehensively and selectively blocked the effects of eIF2α phosphorylation on mRNA translation and triggered rapid stress granule disassembly ( Sidrauski et al . , 2015 ) . To date , the molecular target of ISRIB is not known . The fast kinetics of action of ISRIB and the remarkable specificity of its effects in response to eIF2α phosphorylation strongly suggested that its target is a factor that closely interacts with the eIF2 translation initiation complex . The existence of eIF2B mutations in yeast that , like ISRIB , render cells resistant to eIF2α-P led us to propose that eIF2B was a likely target of this small molecule ( Sidrauski et al . , 2013 ) . Here , we draw on clues from two independent approaches , an unbiased genetic screen and structure/activity analyses of ISRIB , to converge on the hypothesis that the mammalian eIF2B complex indeed is the molecular target of ISRIB . We demonstrate that a symmetric ISRIB molecule induces or stabilizes eIF2B dimerization , increasing its GEF activity and desensitizing it to inhibition by eIF2-P . Thus ISRIB directly modulates the central regulator in the ISR .
To identify the molecular target of ISRIB , we conducted a genetic screen for genes whose knockdown modulated the sensitivity of cells to the drug . Using this strategy , we were previously able to pinpoint the molecular targets of cytotoxic compounds and to delineate their mechanism of action ( Matheny et al . , 2013; Julien et al . , 2014 ) . Here , we conducted a reporter-based screen using a sub-library of our next-generation shRNA library targeting 2933 genes involved in aspects of proteostasis . This focused library targets each protein-coding gene with ∼25 independent shRNAs and contains a large set ( >1000 ) of negative-control shRNAs . We have previously shown that the use of such libraries and analysis using a rigorous statistical framework generates robust results from forward genetic screens ( Bassik et al . , 2013; Kampmann et al . , 2013 ) . We screened the shRNA library in a K562 cell line expressing an uORF-ATF4-venus reporter ( Figure 1A ) , similar to the translational reporters that we and others previously used to measure activation of the ISR . In cells bearing this reporter , the venus fluorescent protein is translationally induced upon eIF2α phosphorylation . We chose the K562 cell line for the screen because these cells are non-adherent and allow for efficient fluorescence-activated cell sorting ( FACS ) . Treatment with thapsigargin ( Tg ) , an ER stress inducer that inhibits the ER-localized Ca2+-ATPase , resulted in a sixfold increase in mean fluorescence intensity and , as expected , ISRIB substantially reduced induction of the reporter ( Figure 1B ) . As a first step in the screen , we transduced the reporter cell line with the library and selected shRNA-expressing cells . We next divided the population and induced ER stress with Tg in the presence or absence of ISRIB . To optimize the dynamic range of the screen and to focus on early translational effects elicited by eIF2α phosphorylation , we incubated cells for 7 hr , at which time full induction of the reporter was reached . To identify genes whose knockdown resulted in either enhanced or reduced sensitivity to ISRIB , we used a concentration of drug corresponding to the EC50 ( 15 nM ) in this cell type . Cells from each subpopulation ( Tg-treated and Tg + ISRIB-treated ) were then FACS-sorted to isolate the third of the population with the lowest reporter expression and the third of the population with the highest reporter expression ( see schematic in Figure 1C ) . To quantify frequencies of cells expressing each shRNA , we isolated genomic DNA from the sorted populations and then PCR-amplified , purified and analyzed by deep-sequencing the shRNA-encoding cassettes . To determine the enrichment or depletion of each shRNA , we compared its frequency in the Low and High reporter populations . For each gene , we calculated a p value by comparing the distribution of log2 enrichment for the 25 shRNAs targeting the gene to the negative control shRNAs . We then plotted p values for each gene determined in ER stress-induced cells in the absence ( x-axis ) vs the presence ( y-axis ) of ISRIB ( Figure 1D ) . 10 . 7554/eLife . 07314 . 003Figure 1 . Knockdown of eIF2B subunits renders cells more resistant to ISRIB . ( A ) Schematic representation of the ATF4-venus reporter used for the screen . The 5′ end of the human ATF4 mRNA up to the start codon of the ATF4-encoding ORF was fused to venus , followed by the EMCV internal ribosomal entry site ( IRES ) and BFP and inserted into a lentiviral system . ( B ) ISRIB reduces activation of the ATF4-venus reporter . K562 cells were incubated with Tg ( 300 nM ) for 6 hr in the presence of different concentrations of ISRIB . Reporter fluorescence was measured by flow cytometry and median values were plotted ( N = 3 , ± SD ) . ( C ) Schematic of the shRNA screen aimed to identify the target ISRIB . K562 cells expressing the screening reporter were transduced with a pooled shRNA library and transduced cells were selected . The population was then divided into two and either treated with Tg ( ER stress ) or Tg + ISRIB ( ER stress + ISRIB ) for 7 hr . Cells were sorted based on their fluorescence ( venus ) intensity into three bins and the third of the population with the Low and High-reporter levels were collected . Note that the ER stress + ISRIB population had a lower overall fluorescence intensity ( median ) as ISRIB partially blocks induction of the reporter when added at a concentration corresponding to its EC50 in these cells ( 15 nM ) . DNA was extracted from the sorted subpopulations for each treatment and shRNA-encoding cassettes were PCR-amplified and subjected to deep sequencing to determine their frequency . ( D ) Effect of knockdown of individual genes in the proteostasis library on reporter expression upon ISR induction in the presence and absence of ISRIB . Gene p values for enrichment and depletion were compared between the ER stress ( x-axis ) vs the ER stress + ISRIB ( y-axis ) experiments . For each gene , a p value was calculated by comparing the distribution of log2 enrichment values for the 25 shRNAs targeting the gene to the negative control shRNAs . ( E ) The log2 counts for eIF2B5 ( top panel ) or eIF2B4 ( bottom panel ) targeting shRNAs in the High-reporter population ( x-axis ) vs the Low-reporter population ( y-axis ) was plotted and color coded based on the log2 enrichment as depicted in the side bar . Red colors indicate a shift towards higher reporter levels , blue colors shifts towards lower reporter levels . Negative control shRNAs in the library are colored grey . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 00310 . 7554/eLife . 07314 . 004Figure 1—source data 1 . Sequence of the reporter utilized in the shRNA screen . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 00410 . 7554/eLife . 07314 . 005Figure 1—source data 2 . Gene p values for the High and Low reporter populations . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 005 The data shown in Figure 1D revealed that knockdown of the majority of the genes in the library did not change the expression of the reporter upon either treatment and thus congregated in the center of the plot . By contrast , knockdown of genes that changed the expression of the reporter to the same degree in both treatments localized to the diagonal . We focused our analysis on genes that when knocked-down in the presence of ISRIB , affected the expression of the reporter selectively . In this plot these genes are displaced along the y-axis and encode proteins whose reduced expression modulates the cells' sensitivity to ISRIB . Knockdown of genes that confer resistance to ISRIB lie above the diagonal , while knockdown of genes that confer hypersensitivity to ISRIB lie below it . Of particular interest was the pronounced effect of the knockdown of ( i ) two subunits of eIF2B , eIF2B4 and eIF2B5 , that significantly reduced the sensitivity ( p < 1 . 4·10−6 and p < 2 . 4·10−11 , respectively ) and ( ii ) eIF4G1 that significantly enhanced the sensitivity ( p < 3 . 4·10−10 ) of cells to ISRIB , each without affecting induction of the reporter ( i . e . , no displacement along the x-axis ) . Individual shRNAs targeting either eIF2B4 or eIF2B5 were enriched in the High reporter population of the ISRIB-treated sample and stood out from the negative control shRNA population ( Figure 1E ) . Knockdown of other translation initiation factors ( highlighted in Figure 1D ) revealed no effects on ISRIB sensitivity ( locating close to the diagonal of the plot ) . Based on these data and the fact that eIF2α-P is a direct inhibitor of eIF2B , we postulated that eIF2B is a promising candidate target of ISRIB . Moreover , the data suggest that ISRIB acts as an activator of eIF2B: when eIF2B levels are reduced , cells become resistant to the effects of ISRIB when there is a lower supply of molecules that can be activated . Structure-activity studies of synthetic ISRIB analogs provided further clues as to the nature of its molecular target in cells . Of particular note is that the progenitor member of this class ( ISRIB , also denoted herein as ISRIB-A1 , Figure 2A ) exhibits twofold rotational symmetry and is bisected longitudinally by a mirror plane . The molecule is thus achiral but can exist as either cis or trans diastereomers , depending on the relative orientation of the side chains at positions 1 and 4 of the cyclohexane ring ( Figure 2A , ISRIB-A1 and ISRIB-A2 ) . We previously showed in cell-based assays that the trans-isomer ( ISRIB-A1 , EC50 = 5 nM ) is > 100-fold more potent than the cis-isomer ( ISRIB-A2 , EC50 > 600 nM ) . This indicated a preference for an extended binding conformation , with both side chains adopting an equatorial position , as would be expected in the preferred chair conformation of the trans diastereomer ( ISRIB-A1 ) ( Sidrauski et al . , 2013 ) . By contrast , the cis diastereomer ISRIB-A2 would need to adopt a higher-energy boat-like conformation to project both side chains in pseudo-equatorial orientations . Further structure-activity studies revealed that a 1 , 4-phenyl spacer could reasonably substitute for 1 , 4-cyclohexyl , although a 10-fold loss in potency was observed ( ISRIB-A7 , EC50 = 53 nM ) . Replacement of the 1 , 4-cyclohexyl ring with cis or trans-1 , 3-cyclobutyl spacers resulted in a more dramatic loss of potency ( ISRIB-A4 , EC50 = 142 nM; ISRIB-A5 , EC50 = 1000 nM ) , indicating that the distance between the distal aromatic rings in ISRIB analogs is as important as their positioning in space . This distance dependence was also observed in analogs with acyclic spacers ( e . g . , ISRIB-A3 and ISRIB-A6 ) . Thus , the n-butyl linker in ISRIB-A3 ( maintaining the spacing of ISRIB-A1 ) was better tolerated than the shorter n-propyl linker in ISRIB-A6 , an analog without measurable activity . The 60-fold reduction in the potency of ISRIB-A3 as compared to ISRIB-A1 can be explained by the increased flexibility of the n-butyl chain , resulting in a higher entropic cost associated with adopting the conformation required for binding . 10 . 7554/eLife . 07314 . 006Figure 2 . SAR analyses suggest ISRIB interacts with a twofold symmetric target . ( A ) ISRIB analogs bearing various likers ( L ) between the pendant side chains and their corresponding EC50 values . ( B ) Sequential replacement of the para-chloro substituent ( X and Y ) with F , Me , or CN on the distal aromatic rings has unfavorable and additive effects on potency . ( C ) Sequential addition of a meta-substituent ( X and Y ) on the distal aromatic rings had favorable and additive effects on potency . Dose response curves of the different ISRIB analogs are shown in Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 00610 . 7554/eLife . 07314 . 007Figure 2—figure supplement 1 . Activation of the ATF4 luciferase reporter in HEK293T cells was measured . Cells were treated with 1 μg/ml of tunicamycin to induce ER stress and different concentrations of the analogs for 7 hr . Relative luminescence intensity ( RLI ) was plotted as a function of the concentration of the indicated ISRIB analog ( N = 2 , mean ± SD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 007 Extensive structure-activity relationship ( SAR ) studies were also carried out on the distal aryl substituents . Overall , we found that the SAR was consistent with the idea that ISRIB analogs bind across a symmetrical interface . Thus , sequential modification of one and then both side chains in ISRIB analogs was additive , both for favorable modifications and for unfavorable modifications . For example , a para-chloro substituent was found to be optimal in ISRIB analogs . Replacing one or both para-chloro substituents with fluoro , methyl , or cyano groups led to predictable deterioration of potencies , with the doubly modified analogs least potent in every case ( Figure 2B , compare ISRIB-A8 with A9 , ISRIB-A10 with A11 and ISRIB-A12 with A13 ) . Conversely , the addition of a meta-chloro or meta-fluoro substituent enhanced the potency of ISRIB analogs , and introducing such modifications on both side chains produced the most potent analogs ( Figure 2C , compare ISRIB-A14 with A15 , ISRIB-A16 with A17 ) . Among these more potent analogs is ISRIB-A17 , which is nearly 10-fold more potent than ISRIB-A1 , lowering the EC50 into the picomolar range . A full account of our SAR studies will be provided elsewhere but the data presented here demonstrate that the electronics of the phenoxy substituents are important drivers of potency and support the notion that the two halves of ISRIB analogs are engaged in similar recognition events with the target . The most plausible explanation of these findings is that the functional twofold symmetry of ISRIB reflect a target that is likewise twofold symmetric . Taken together , the results obtained by the shRNA screen described above and the recent discovery of eIF2B dimers suggest that ISRIB may act by directly binding to eIF2B at a twofold symmetric interface that stabilizes it as a dimer ( Gordiyenko et al . , 2014; Wortham et al . , 2014 ) . To test directly whether ISRIB induces or stabilizes the dimeric form of eIF2B , we treated cells with or without ISRIB . We prepared extracts in a high-salt buffer to dissociate eIF2B from its substrate eIF2 and analyzed the lysates by velocity sedimentation on sucrose gradients . In the absence of ISRIB , eIF2B ( as detected by immunoblotting with antibodies against eIF2B4 and eIF2B5 ) migrated predominantly in fractions 3–6 in the gradient , consistent a combined molecular mass of four of its subunits ( 225 kDa ) . In the high-salt buffer used , the eIF2B complex lacked the eIF2B1 subunit , which was found predominantly in fractions 1–3 of the gradient . By contrast , when cells were treated with ISRIB , we observed a substantial shift in sedimentation towards a higher molecular mass ( predominantly found in fractions 5–8 ) , demonstrating a substantial increase in complex size . By comparing the relative mobility of eIF2B4 and eIF2B5 to that of a background band ( marked with a red asterisk in the upper panel of Figure 3 ) , the shift in size of eIF2B is easily appreciated . The magnitude of the shift is consistent with a doubling in the molecular mass of the complex . Interestingly , in extracts from ISRIB-treated cells , eIF2B1 also shifted to the heavier fractions , suggesting that its association with the rest of the complex was stabilized . In contrast to the eIF2B subunits , we did not observe a shift in eIF3a or eIF2α . These data strongly support the notion that ISRIB induces the formation of a stable eIF2B dimer . 10 . 7554/eLife . 07314 . 008Figure 3 . ISRIB induces dimerization of eIF2B in cells . ( A ) HEK293T cells were treated with or without 200 nM ISRIB and clarified lysates were loaded on a 5–20% sucrose gradient and subjected to centrifugation . 13 equal-size fractions were collected , protein was precipitated and run on a SDS-PAGE gel and immunoblotted with the indicated antibodies . The red asterisk indicates a background band that cross-reacts with the eIF2B4 antibody . Sedimentation was from left to right . Gradients were calibrated ( in Svedberg units , ‘S’ ) with ovalbumin ( S = 3 . 5; Mr = 44 kD ) ; aldolase ( S = 7 . 3; Mr = 158 kD ) and thyroglobulin ( S = 19; Mr = 669 kD ) . Shown is a representative blot ( N = 3 ) . ( B ) HEK293T cells and lysates were treated with 200 nM ISRIB or 200 nM ISRIBinact ( ISRIB-A18; Figure 3—figure supplement 1 ) and clarified lysates were loaded on a 5–20% sucrose gradient and subjected to centrifugation . 13 equal sized fractions were collected and fractions 6–9 were precipitated , trypsinized and subjected to mass spectrometric analysis . The sum of the normalized peptide intensity of each eIF2B subunit as well as two control proteins , eIF3a and PSMD1 in each fraction was plotted . Two biological replicates were analyzed per condition ( N = 2 , ±SEM ) . The number of peptides and peptide intensity in fractions 6–9 for all proteins identified are listed in Figure 3—source data 1 . ( C ) Correlation coefficient ( R ) of the sum of the normalized peptide intensity profile through fractions 6–9 for each protein identified in the analysis with respect to eIF2B4 was plotted . The Correlation coefficient ( R ) of the sum of the normalized peptide intensity profile with respect to eIF2B4 of each protein identified are listed in Figure 3—source data 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 00810 . 7554/eLife . 07314 . 009Figure 3—source data 1 . Number of peptides and peptide intensity in fractions 6–9 for all proteins identified . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 00910 . 7554/eLife . 07314 . 010Figure 3—source data 2 . Correlation coefficient ( R ) of the sum of the normalized peptide intensity profile through fractions 6–9 with respect to eIF2B4 for each protein identified . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 01010 . 7554/eLife . 07314 . 011Figure 3—figure supplement 1 . Structures of ISRIB ( ISRIB-A1 ) and ISRIBinact ( ISRIB-A18 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 01110 . 7554/eLife . 07314 . 012Figure 3—figure supplement 2 . Analysis of the gradients subjected to mass spectrometric analysis in Figure 3B . ( A ) Western blot analysis as in Figure 3A . The protein composition of fractions 6–9 was analyzed by mass spectrometry ( Figure 3B ) . ( B ) Total protein across the sucrose gradient visualized by Coomassie blue staining . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 012 To determine if eIF2B's ostensible increase in molecular mass was due to dimerization of a complete eIF2B complex , we used mass spectrometry to validate the shift of all of its five subunits . To this end , we treated cells with ISRIB or with an inactive analog ( ‘ISRIBinact’ [ISRIB-A18] , Figure 3—figure supplement 1 ) and subjected extracts to fractionation on sucrose gradients . We used ISRIBinact to control for non-specific hydrophobic interactions of ISRIB with proteins in the extract . We determined the complete protein composition in the fractions in which eIF2B peaked in the presence of ISRIB ( fractions 6–9 , Figure 3—figure supplement 2 ) by mass spectrometry . This analysis revealed a significant ISRIB-dependent enrichment of all five eIF2B subunits ( Figure 3B ) . Notably , eIF2B subunits in ISRIB samples exhibited a characteristic profile in which all subunits collectively peaked in fraction 7 . By contrast eIF2B subunits in ISRIBinact samples were most abundant in fraction 6 and trailed further into the gradient . As expected , two other large protein complexes , the proteasome ( Figure 3B; data shown for subunit PSMD1 ) and eIF3 ( Figure 3B; data shown for subunit eIF3A ) , showed no displacement upon ISRIB treatment . Because the mass spectrometric analysis of the gradient was performed with a non-targeted method , it allowed us to ask whether additional proteins would associate with eIF2B potentially contributing to the shift in size . To address this question , we correlated the intensity profiles of all other proteins identified through the analyzed fractions to the sedimentation profile exhibited by a representative subunit , eIF2B4 . We plotted the correlation coefficient ( R-value ) for each comparison . We were excited to find that all eIF2B subunits ( eIF2B1 , eIF2B2 , eIF2B3 , eIF2B5 ) stood out as most strongly correlated to eIF2B4 , all exhibiting correlation coefficients ( R-values ) > 0 . 98 ( Figure 3C ) , strongly indicating that the increase in molecular mass of eIF2B upon ISRIB addition indeed resulted from eIF2B dimerization . Moreover , these analyses strongly support the notion that eIF2B forms a complete complex upon ISRIB treatment . To identify the subunit of eIF2B targeted by ISRIB , we monitored drug-target engagement , utilizing a cellular extract thermal shift assay ( CETSA ) ( Martinez Molina et al . , 2013 ) . This method relies on the principle that ligand binding can stabilize protein folding and hence increase the protein's resistance to heat denaturation . To this end , we incubated a cell lysate with and without ISRIB and then heated aliquots to different temperatures , followed by centrifugation to separate soluble from precipitated denatured proteins . We then analyzed the soluble fractions by Western blotting with antibodies against eIF2B1 , eIF2B4 and eIF2B5 . When the lysate was pre-incubated with ISRIB , we observed an increase in thermal stability of eIF2B4 ( Figure 4 , lanes 4 and 5 , arrows ) . Although slight , the increase was highly reproducible and , as was the case for the analysis of the eIF2B shift in the sucrose gradients shown in Figure 3 , a background band that cross-reacts with the anti-eIF2B4 antibody ( red asterisk ) provided a convenient internal control for the exclusive stabilization of eIF2B4 . By contrast , no ISRIB-dependent increase in thermal stability was observed with the two other eIF2B subunits analyzed ( eIF2B1 and eIF2B5 ) , or with the translation initiation factors eIF2α or eIF3a ( Figure 4 ) . This analysis suggests that eIF2B subunits act autonomously in this assay , as eIF2B4 was stabilized while other subunits denatured and precipitated . We conclude that ISRIB binds eIF2B4 eliciting this stabilization . 10 . 7554/eLife . 07314 . 013Figure 4 . ISRIB enhances the thermo-stability of the regulatory subunit of eIF2B . Clarified HEK293 cell lysates were treated with DMSO ( -ISRIB ) or with 200 nM ISRIB ( +ISRIB ) for 20 min . Treated and untreated lysates were partitioned into smaller aliquots and heated to different temperatures for 3 min and then centrifuged to remove precipitated proteins . The supernatant fraction was loaded onto a SDS-PAGE gel and immunoblotted with the indicated antibodies . The red asterisk indicates a background band that cross-reacts with the eIF2B4 antibody . Shown is a representative blot ( N = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 013 To explore the functional consequences of ISRIB binding on eIF2B's GEF activity , we directly tested its effect on the rate of GDP release from eIF2 . To this end , we pre-loaded purified eIF2 with radioactive GDP ( [3H]-GDP ) and measured the fraction that remained bound as a function of time in the presence of an excess of unlabeled GDP . As expected , the intrinsic rate of nucleotide release was slow; after 20 min of incubation , only 20% of [3H]-GDP dissociated from the eIF2 complex ( Figure 5A , black dashed line ) . The intrinsic rate of GDP release was not affected by the addition of ISRIB ( Figure 5A , red dashed line ) . Upon addition of eIF2B , we observed a significant increase in the rate of GDP release ( t1/2 = 3 . 2 min ) , leading to an 80% release after 10 min ( Figure 5A , solid black line ) . Excitingly , GDP release was threefold faster upon addition of ISRIB ( t1/2 = 1 . 1 min ) ( Figure 5A , solid red line ) . 10 . 7554/eLife . 07314 . 014Figure 5 . ISRIB enhances the GEF activity of eIF2B in vitro . eIF2 was preloaded with [3H]-GDP and the fraction of binary complex remaining was measured by filter binding . Partially purified eIF2B or buffer was added at t = 0 min . An aliquot of the reaction was stopped at the indicated times , filtered through a nitrocellulose membrane and radioactivity was measured . ( A ) Purified eIF2 was incubated with buffer ( ± 100 nM ISRIB , dashed lines ) or partially purified eIF2B ( ± 100 nM ISRIB , solid lines ) for the indicated times and the remaining fraction of [3H]-GDP-eIF2 was measured ( N = 3 , ± SD ) . ( B ) Purified and phosphorylated eIF2 ( eIF2-P ) was preloaded with [3H]-GDP and incubated with buffer ( ± 100 nM ISRIB , dashed lines ) or partially purified eIF2B ( ± 100 nM ISRIB , solid lines ) for the indicated times and the remaining fraction of [3H]-GDP-eIF2 was measured ( N = 2 , ± SD ) . ( C ) eIF2 was preloaded with [3H]-GDP and mixed with eIF2-P at a ratio of 3:1 and then incubated with eIF2B with or without 100 nM ISRIB for the indicated times and the remaining fraction of [3H]-GDP-eIF2 was measured ( N = 2 , ± SD ) . ( D ) eIF2 was preloaded with [3H]-GDP and mixed with eIF2-P at a ratio of 1:1 and then incubated with eIF2B with or without 100 nM ISRIB for the indicated times and the remaining fraction of [3H]-GDP-eIF2 was measured ( N = 2 , ± SD ) . Purified human eIF2 and partially purified rabbit reticulocyte eIF2B are shown in Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 01410 . 7554/eLife . 07314 . 015Figure 5—figure supplement 1 . Purified human eIF2 ( panel A , lane 2 ) , recombinant GST-PERK ( panel A , lane 1 ) and partially purified rabbit reticulocyte eIF2B ( panel B ) were analyzed by SDS-PAGE and stained with Coomassie blue dye . Red asterisks indicate the migration of the five subunits of eIF2B . We utilized fractions 6 and 7 of the Mono-Q column for the guanine nucleotide exchange assays in Figure 5 . We estimate that the eIF2B complex represents ∼10% of the total protein in these fractions . DOI: http://dx . doi . org/10 . 7554/eLife . 07314 . 015 We next tested the behavior of phosphorylated eIF2 ( eIF2-P ) in these assays . To this end , we generated eIF2-P by incubating eIF2 with recombinantly expressed PERK kinase and ATP . We next loaded eIF2-P with [3H]-GDP and measured GDP release . As expected from the known inhibitory role of eIF2α phosphorylation on eIF2B , GDP release from eIF2-P remained virtually unchanged in the presence of eIF2B ( Figure 5B , black solid line ) . We next asked whether ISRIB allows eIF2-P to be a substrate for eIF2B . Our data show that ISRIB did not stimulate GDP release from eIF2-P ( Figure 5B , red solid line ) , indicating that this is not the case . We next explored whether ISRIB can overcome the inhibitory effects of eIF2-P on eIF2B . To this end , we tested if ISRIB can promote GDP release from unphosphorylated eIF2 in the presence eIF2-P by mixing [3H]-GDP-loaded eIF2 with eIF2-P in a 3:1 or 1:1 ratio . Although the exchange reaction was slower , ISRIB stimulated GDP release at the eIF2:eIF2-P ratio of 3:1 ( -ISRIB: t1/2 = 6 . 7 min , vs + ISRIB: t1/2 = 2 . 7 min ) ( Figure 5C ) , whereas we observed hardly any stimulation at the 1:1 ratio ( -ISRIB: t1/2 = 6 . 4 min , vs + ISRIB: t1/2 = 5 . 3 min ) ( Figure 5D ) . Thus , the relative ratio of substrate ( eIF2 ) to inhibitor ( eIF2-P ) emerges as an important parameter affecting ISRIB's ability to modulate eIF2B activity . Taken together , these functional data underscore the notion that ISRIB acts as an activator of eIF2B and that ISRIB alleviates inhibition by eIF2-P , as long as eIF2-P is present below threshold levels .
In this work , ISRIB emerged as an eIF2B activator . First , ISRIB promoted the formation of or stabilized eIF2B dimers ( ‘[eIF2B]2’ ) and enhanced GEF activity in biochemical assays . Second , knockdown of both eIF2B4 and eIF2B5 subunits rendered cells resistant to the action of ISRIB , presumably because under these conditions the total amount of eIF2B that can be activated in cells is reduced . Note that the three other subunits of eIF2B were not represented in our focused shRNA library and therefore could not have been identified in the screen . Functioning as an activator , ISRIB joins the still sparsely populated group of unnatural small molecule enzyme activators , while the vast majority of synthetic small molecules that modulate enzyme activity are inhibitors ( Wiseman et al . , 2010; Zorn and Wells , 2010; Wang et al . , 2014 ) . Conversely , knockdown of eIF4G1 sensitized cells to ISRIB . This can be explained because , under conditions of reduced eIF4G1 , overall cap-dependent translation initiation is reduced . A lower concentration of ISRIB could then suffice to generate sufficient amounts of GTP-loaded eIF2 to maintain normal rates of translation , even in the presence of eIF2α-P . Intriguingly , knockdown of other components of the cap-binding complex , such as eIF4A1 , or components of the eIF3 complex , such as eIF3f and eIF3b , not only reduced sensitivity to ISRIB but also affected induction of the reporter upon ER stress alone . In agreement with studies in yeast and plants ( Szamecz et al . , 2008; Roy et al . , 2010 ) , knockdown of the eIF3 subunits in the library ( eIF3a , eIF3b , and eIF3f ) reduced translational induction of the reporter , presumably due to eIF3's stimulatory effects on re-initiation after translation of short uORFs . Our data therefore provide the first evidence that the mechanism of re-initiation may be similar in mammalian cells . The differences observed between assorted initiation factors on reporter expression is likely to reflect the extent to which translation initiation was reduced under the different knockdown conditions . Importantly however , only knockdown of the eIF2B subunits targeted by shRNAs in the library conferred resistance to ISRIB . We previously proposed two models that could explain how ISRIB renders cells resistant to the inhibitory effects of eIF2α-P ( Sidrauski et al . , 2013 ) . First , ISRIB could weaken the effect of eIF2α-P on eIF2B by interfering with its tight and non-productive binding . In this way , more eIF2B would be available to reload eIF2 with GTP . Second , ISRIB could enhance the basal activity of eIF2B so that the fraction not engaged with eIF2α-P would produce sufficient levels of ternary complex to sustain translation in cells . Currently , our in vitro enzymatic data do not allow us to distinguish between these models . While we showed that the rate of GDP release from purified eIF2 by eIF2B was significantly enhanced upon addition of ISRIB ( and therefore can explain the effect of ISRIB in living cells ) , we do not know what fraction of our eIF2 preparation was isolated in a eIF2α ( Ser-51 ) -phosphorylated state . ISRIB could thus either increase the GEF activity of eIF2B on eIF2 or diminish the inhibitory effect of a small amount eIF2-P present in the assay , akin to the regime that we directly tested by adding increasing amounts of in vitro phosphorylated eIF2 to the assay . Our analyses confirmed however that eIF2α-P is not a substrate for eIF2B ( in agreement with previous reports [Kimball et al . , 1998] ) , and determined that ISRIB does not enable eIF2B to use eIF2-P as a substrate . While catalyzing guanine nucleotide exchange on other GTPases can be effected by relatively simple enzymes , eIF2B is a complex molecular machine composed of five different subunits . Much remains uncertain about the structural arrangement of the subunits and how eIF2B's activity is regulated ( Jennings and Pavitt , 2014 ) . Similarly , how ISRIB exerts its effects on eIF2B remains unknown . eIF2B subunits are organized into two modules , called the catalytic ( eIF2B3 and eIF2B5 ) and regulatory ( eIF2B1 , eIF2B2 and eIF2B4 ) sub-complexes , containing two and three homologous proteins , respectively . The subunits of the regulatory subcomplex are characterized by highly homologous Rossman folds that bind nucleotides and are adorned by N-terminal extensions of lesser homology between the subunits . Intriguingly , recombinantly expressed eIF2B1 purified and crystallized as a stable homodimer , with an extensive buried interface contributed by the nucleotide-binding domains ( Bogorad et al . , 2014 ) . The residues contributing to the interface are highly conserved among its homologs in the complex . Combined with the SAR analyses indicating ISRIB's obligate twofold symmetry , the discovery that ( eIF2B ) 2 exist in both yeast and mammalian cells was instrumental in suggesting to us that eIF2B is the target of ISRIB ( Gordiyenko et al . , 2014; Wortham et al . , 2014 ) . According to this model , ISRIB binds to two regulatory eIF2B subunits that form part of the interface linking two pentamers . Native mass spectrometry of mammalian eIF2B revealed the existence of stable subcomplexes that lack the eIF2B1 subunit , indicating that this subunit is more loosely associated , as we confirmed here by sedimentation of the non-ISRIB treated control extracts ( Wortham et al . , 2014 ) . We have shown by biochemical analysis that ISRIB binding stabilizes ( eIF2B ) 2 , rendering it resistant to dissociation of eIF2B1 in the high-salt buffers used in the sucrose gradient analysis . Importantly , we showed by mass spectrometric proteomic analysis that no other protein co-profiled with ( eIF2B ) 2 in the gradients , demonstrating that the observed ISRIB-dependent effects were confined exclusively to eIF2B subunits . Given the relative stability of the eIF2B1 homodimer ( Kd < 1 nM; [Bogorad et al . , 2014] ) and our observation that ISRIB stabilized complete ( eIF2B ) 2 , it is likely that two opposing eIF2B1 subunits form an essential part of the interface that links two eIF2B pentamers . ISRIB could favor this interaction by adding to the affinity provided by a ( eIF2B1 ) 2 tether via the stabilization of an additional interface formed between homologous regions of two eIF2B4 subunits . This view would be in agreement with our data that showed protection by ISRIB of eIF2B4 to thermal denaturation . For symmetry reasons , as elegantly discussed in ( Bogorad et al . , 2014 ) , this arrangement would leave the interfaces of the two identical eIF2B2 subunits in the complex unpaired . Alternatively , ISRIB may stabilize interfaces between eIF2B4 in one eIF2B pentamer and eIF2B2 in an opposing pentamer . If this were the case , ISRIB would bind at a pseudo-symmetric interface formed by two different , yet strongly homologous components . We note in this scenario , two ISRIB molecules binding to two identical interfaces of opposite polarity ( eIF2B2 → eIF2B4 and eIF2B4 → eIF2B2 ) may bind and stabilize one ( eIF2B ) 2 , which may contribute to its potency . This would open the possibility that design and synthesis of non-symmetric analogs could further improve ISRIB's efficacy . A definite assignment of ISRIB's binding site will have to await the structural determination of ISRIB-bound ( eIF2B ) 2 or genetic analyses in which loss-of-function mutations are suppressed by compensating changes in ISRIB analogs . Consistent with the notion that the regulatory sub-complex provides binding sites for eIF2 , mutations in eIF2B in yeast that render cells resistant to phosphorylation of eIF2α map to eIF2B1 and eIF2B4 ( Pavitt et al . , 1997 ) . Moreover , two different variants in mammalian eIF2B4 ( generated by alternative splicing ) contain different N-terminal extension domains and exclusive expression of the longer variant desensitizes cells to eIF2α phosphorylation ( Martin et al . , 2010 ) , phenocopying the effects elicited by ISRIB in mammalian cells . In the structure of ( eIF2B1 ) 2 the N-terminal domains reach across the interface and interact with the nucleotide binding domain of the partnering eIF2B1 molecule . We speculate that the extended N-terminal domain of eIF2B4 may stabilize ( eIF2B ) 2 , mimicking the effects of ISRIB . Phosphorylation of eIF2 is important in long-term depression ( LTD ) , and we have recently shown that this modulation of synaptic plasticity can explain cognitive enhancement elicited by ISRIB treatment of wild type rodents ( Di Prisco et al . , 2014 ) . Engagement of metabotropic glutamate receptors ( mGluR ) in post-synaptic hippocampal cells leads to eIF2 phosphorylation and preferential translation of neuronally expressed oligophrenin-1 ( encoded by OPHN1 ) , a protein that mediates the initial steps of downregulation of postsynaptic AMPA receptors by endocytosis ( Nadif Kasri et al . , 2011 ) . Like ATF4 , the 5′-UTR of OPHN1 mRNA contains two uORFs that repress expression of the downstream coding sequence unless eIF2 is phosphorylated . Importantly , both genetic ablation of eIF2 phosphorylation and treatment with ISRIB but not the inactive analog ISRIB-A18 abolished the reduction in surface AMPARs and blocked mGluR-LTD ( Di Prisco et al . , 2014 ) . These findings hold promise that targeting the effects of phosphorylation of eIF2 by pharmacologically modulating eIF2B with drugs such as ISRIB could result in therapies for cognitive disorders . Activation of the ISR with its characteristic increase in eIF2 phosphorylation has been reported in numerous neurodegenerative diseases , including Alzheimer's disease , Parkinson's disease , Frontotemporal Dementia , Amyotrophic Lateral Sclerosis , and prion neurodegenerative diseases , but its role in disease progression has just recently begun to be interrogated ( Kim et al . , 2013; Leitman et al . , 2014; Ma et al . , 2013; Moreno et al . , 2013 , 2012 ) . The importance of eIF2 and eIF2B in brain function is underscored by the existence of mutations in these factors that cause human disease . A familial intellectual disability syndrome was mapped to a mutation in the γ subunit of eIF2 ( encoded by EIF2S3 ) . When an analogous mutation was introduced into yeast cells , it impaired eIF2-mediated translation initiation ( Borck et al . , 2012 ) . Mutations in the different subunits of eIF2B cause childhood ataxia with central nervous system ( CNS ) hypomyelination ( CACH ) or vanishing white matter disease ( VWMD ) . All affected individuals have two altered copies of a single eIF2B gene ( autosomal recessive inheritance ) and the majority are missense mutations that cause a single amino acid change while the remainder is a mixture of premature nonsense mutations , some causing a frame-shift and others altered splicing . All subunits of eIF2B are essential and the biochemical analysis of 40 different VWMD mutations revealed that the majority are hypomorphs , that is , cause partial loss-of function of eIF2B GEF activity ( Leegwater et al . , 2001; Li et al . , 2004; Fogli and Boespflug-Tanguy , 2006 ) . Whether ISRIB can reverse the deleterious effects of mutations in eIF2B in VWMD patients is not known , but we speculate that it may protect from a further reduction in GEF activity by stress-induced eIF2α-P . Intriguingly , the onset of VWMD is varied but generally exacerbated by head trauma and febrile illnesses . Interestingly , two VWMD mutations have been characterized that affect the integrity and dimerization of the eIF2B complex . A mutation in eIF2B1 ( V183F ) maps to the dimerization interface and the mutant recombinant protein is predominantly in the monomeric form and a mutation in eIF2B4 ( A391D ) affects complex integrity in the absence of eIF2B1 and dimerization ( Wortham et al . , 2014 ) . ISRIB induces dimerization and complex stability and thus may rescue the effects of such mutations . Given the wide spectrum of potential applications for ISRIB in neurological diseases , the identification of its molecular target is an important step . Having established a proof-of-principle that eIF2B can be pharmacologically modulated , now enables directed screening efforts to identify new series of compounds and thereby enhance the probability of developing clinically useful pharmaceuticals that address currently unmet needs . While this work was under review , Sekine et al . reported the independent identification of eIF2B as the molecular target of ISRIB ( Sekine et al . , 2015 ) .
Thapsigargin ( Tg ) was obtained from Sigma–Aldrich ( St Louis , MO ) . Tunicamycin ( Tm ) was obtained from Calbiochem EMB Bioscience ( Billerica , CA ) . The GSK PERK inhibitor ( G797800 ) was obtained from Toronto Research Chemicals ( North York , ON , Canada ) . HEK293T and K562 cells were maintained at 37C , 5% CO2 in either DMEM ( HEK293T ) or RPMI ( K562 ) media supplemented with 10% FBS , L-glutamine and antibiotics ( penicillin and streptomycin ) . The lentiviral reporter vector , pMK1163 , contains a CMV promoter driving expression of a fusion transcript with the following elements: the 5′ end of the human ATF4 mRNA up to the start codon of the ATF4-encoding ORF , an ORF encoding Venus ( adapting a previously published strategy [Lu et al . , 2004; Vattem and Wek , 2004] ) , followed by an IRES driving translation of tagBFP . The elements of this vector were generated as follows: we PCR-amplified the ATF4 region from human cDNA prepared from K562 cells using primers oMK305 ( 5′-CGTACTCGAGTTTCTACTTTGCCCGCCCACAG-3′ ) and oMK306 ( 5′-GCTCCTCGCCCTTGCTCACCATGTTGCGGTGCTTTGCTGGAATCG-3′ ) . Venus was amplified from DAA307 ( gift from Diego Acosta-Alvear ) , using primers oMK272 ( 5′-ATGGTGAGCAAGGGCGAGGAGC-3′ ) and oMK308 ( 5′-GCTAGAATTCTTACTTGTACAGCTCGTCCATGCC-3′ ) . The ATF4-Venus fusion was generated by PCR reaction using the two PCR products described above as templates , and oMK305 and oMK308 as primers . The EMCV IRES was amplified from plasmid pPPCX-IRES-GFP ( gift from Diego Acosta-Alvear ) . tagBFP was amplified from a tagBFP plasmid ( Evrogen , Moscow , Russia ) . The plasmid pMK1163 is in the lentiviral vector pSicoR ( Ventura et al . , 2004 ) , and its sequence is provided in Figure 1—source data 1 . Human K562 cells were transduced with pMK1163 and monoclonal cell lines were generated using FACS . One clone was selected as our reporter cell line based on low base-line expression of Venus and high expression following thapsigargin treatment ( high dynamic range ) . The reporter cell line was transduced with a pooled next-generation shRNA library . We used a sub-library that targets 2933 human genes associated with proteostasis , each with on average 25 independent shRNAs , and contains >1000 negative control shRNAs . After transduction , transduced cells were selected with puromycin ( 0 . 65 μg/ml ) for 2 days , and then grown in the absence of puromycin for 2 days . Cells were then separated into two populations , which were treated for 7 hr with either 300 nM thapsigargin alone or 300 nM thapsigargin and 15 nM ISRIB . Cells were then sorted based on reporter fluorescence using a BD FACS Aria2 . Cells from the thirds of the population with the highest and lowest reporter levels were collected . Genomic DNA was isolated from FACS-sorted populations , and shRNA-encoding cassettes were PCR-amplified and subjected to deep sequencing as previously described ( Kampmann et al . , 2014 ) . Using our previously described analysis pipeline ( Kampmann et al . , 2013 , 2014 ) , we calculated a quantitative phenotype ε for each shRNA , which represents the log2 ratio of its frequency in the high-fluorescence population over its frequency in the low-fluorescence population , from which the median of the negative control phenotypes was subtracted ( Kampmann et al . , 2013 ) . For each gene , ε phenotypes for the ∼25 shRNAs targeting the gene were compared to ε phenotypes for the negative control shRNAs , and p values were calculated using the Mann–Whitney U test to detect genes whose knockdown significantly modulated activation of the uORFs-ATF4-venus reporter in response to thapsigargin in the absence or presence of ISRIB . p values for all 2933 genes targeted by the sublibrary we used are listed in Figure 1—source data 2 . HEK293T cells carrying an ATF4 luciferase reporter ( as previously described in [Sidrauski et al . , 2013] ) were plated on poly-lysine coated 96 well plates ( Greiner Bio-One , Monroe , NC ) at 30 , 000 cells per well . Cells were treated the next day with tunicamycin ( 1 μg/ml ) and different concentrations ( serial dilution ) of each compound for 7 hr . Luminescence was measured using One Glo ( Promega , Madison , WI ) as specified by the manufacturer . EC50 values were calculated by plotting log10 [μM] for each compound as a function of the relative luminescence intensity or response . The EC50 corresponds to the concentration that provokes a half-maximal response . HEK293T cells were plated on 150 mm plates , treated with or without 200 nM ISRIB for 20 min , washed twice with ice-cold PBS , collected and centrifuged for 3 min at 800 rcf at 4°C . The pellets were resuspended in ice-cold lysis buffer: 50 mM Tris pH = 7 . 5 , 400 mM KCl , 4 mM Mg ( OAc ) 2 , 0 . 5% Triton X-100 and protease inhibitors ( EDTA-free protease inhibitor tablets , Roche , South San Francisco , CA ) . The lysates were clarified at 20 , 000×g for 15 min at 4°C and the supernatant was then subjected to a high-speed spin at 100 , 000×g for 30 min at 4°C to pellet the ribosomes . The supernatants were then loaded on a 5–20% sucrose gradient and centrifuged in a SW55 rotor for 14 hr at 40 , 000 rpm 4°C . 13 fractions were collected , protein was chloroform-methanol precipitated , resuspended in SDS-PAGE loading buffer and loaded on SDS-PAGE 10% gels ( Bio-Rad , Hercules , CA ) . Proteins were transferred to nitrocellulose and probed with primary antibodies diluted in phosphate-buffered saline supplemented with 0 . 1% Tween 20 and 5% bovine serum albumin . The following antibodies were used: eIF2B1 ( 1:1000; Proteintech 18010-1-AP , Chicago , IL ) , eIF2B2 ( 1:500; Proteintech 11034-1-AP ) , eIF2B4 ( 1:1000; Proteintech 11332-1-AP ) , eIF2B5 ( 1:500; Santa Cruz Biotechnologies sc-5558 , Dallas , TX ) , eIF3a ( 1:1500; Cell Signaling Technology #3411 , Danvers , MA ) and eIF2α ( 1:1500; Cell Signaling Technology #5324 ) . Following primary antibody incubation , either HRP-conjugated secondary antibody ( Promega ) or IRdye conjugated secondary antibodies ( LI-COR Biosciences , Lincoln , NE ) was used . Immunoreactive bands were detected using either enhanced chemi-luminescence ( Bio-Rad ) or the LI-COR Odyssey imaging system . HEK293T cells were treated with ISRIB or ISRIBinact ( ISRIB-A18 , Figure 3—figure supplement 1 ) at 200 nM for 20 min . Cells were then subjected to three liquid nitrogen freeze–thaw cycles in a modified lysis buffer devoid of Triton X-100 and supplemented with ISRIB or ISRIBinact at 50 nM . Lysates were loaded onto a 5–20% sucrose gradient . Proteins in fractions 6–9 were chloroform-methanol precipitated and re-suspended in 0 . 1 M tetraethylammonium bromide ( TEAB ) , 150 mM NaCl and 8M Urea and digested with trypsin as previously described ( Ramage et al . , 2015 ) . Digested peptide mixtures were analyzed in technical duplicate by LC-MS/MS on a Thermo Scientific LTQ Orbitrap Elite mass spectrometry system equipped with a Proxeon Easy nLC 1000 ultra high-pressure liquid chromatography and autosampler system . Samples were injected onto a C18 column ( 25 cm × 75 μm I . D . ) packed with ReproSil Pur C18 AQ ( 1 . 9 μm particles ) in 0 . 1% formic acid and then separated with a 1-hr gradient from 5% to 30% ACN in 0 . 1% formic acid at a flow rate of 300 nl/min . The mass spectrometer collected data in a data-dependent fashion , collecting one full scan in the Orbitrap at 120 , 000 resolution followed by 20 collision-induced dissociation MS/MS scans in the dual linear ion trap for the 20 most intense peaks from the full scan . Dynamic exclusion was enabled for 30 s with a repeat count of one . Charge state screening was employed to reject analysis of singly charged species or species for which a charge could not be assigned . Raw mass spectrometry data were analyzed using the MaxQuant software package ( version 1 . 3 . 0 . 5 ) ( Cox and Mann , 2008 ) . Data were matched to the SwissProt human proteins ( downloaded from UniProt on 2/15/13 , 20 , 259 protein sequence entries ) . MaxQuant was configured to generate and search against a reverse sequence database for false discovery rate calculations . Variable modifications were allowed for methionine oxidation and protein N-terminus acetylation . A fixed modification was indicated for cysteine carbamidomethylation . Full trypsin specificity was required . The first search was performed with a mass accuracy of ± 20 parts per million and the main search was performed with a mass accuracy of ± 6 parts per million . A maximum of five modifications were allowed per peptide . A maximum of two missed cleavages were allowed . The maximum charge allowed was 7+ . Individual peptide mass tolerances were allowed . For MS/MS matching , a mass tolerance of 0 . 5 Da was allowed and the top six peaks per 100 Da were analyzed . MS/MS matching was allowed for higher charge states , water and ammonia loss events . The data were filtered to obtain a peptide , protein , and site-level false discovery rate of 0 . 01 . The minimum peptide length was 7 amino acids . Results were matched between runs with a time window of 2 min for technical duplicates . CETSA were adapted from a previously described protocol ( Martinez Molina et al . , 2013 ) . HEK293T cells were lysed in a buffer containing: 50 mM Tris pH = 7 . 5 , 400 mM KCl , 4 mM Mg ( OAc ) 2 , 0 . 5% Triton X-100 and protease inhibitors ( EDTA-free protease inhibitor tablets , Roche ) . The lysates were clarified at 20 , 000×g for 15 min at 4°C . The supernatant was then incubated with ISRIB ( 1 μM , 0 . 1% DMSO ) or DMSO ( 0 . 1% ) at 30°C for 20 min , and subsequently spun at 100 , 000×g for 30 min at 4°C to pellet ribosomes . Supernatants following the high-speed spin were divided into PCR tubes and subjected to a gradient of temperatures for 3 min using the thermal cycler's built-in gradient function , such that column one corresponded to 52°C and column 12 corresponded to 62°C ( Tetrad 2 Thermal Cycler , Bio-Rad ) . Samples were allowed to cool for 3 min at room temperature , transferred to microfuge tubes , and spun at 20 , 000×g for 20 min at 4°C to separate the soluble fraction from the insoluble precipitates . The soluble fraction was then loaded on a 10% SDS-PAGE gel ( Bio-Rad ) and analyzed by Western blotting as described above . Rabbit reticulocyte lysate was obtained from Greenhectares ( http://greenhectares . com ) . eIF2B was purified as previously described ( Oldfield and Proud , 1992 ) . In brief , the reticulocyte lysate was thawed and protease inhibitor added ( EDTA-free protease inhibitor tablets , Roche ) . Ribosomes were precipitated by centrifugation ( 45 , 000 rpm for 4 . 5 hr , Beckman 50 . 2 Ti at 4°C ) and the supernatant was used as a source of eIF2B . KCl was added slowly to 100 mM final concentration and filtered using a 0 . 2 μM conical tube filter unit . The filtrate was loaded on a SP-Sepharose fast flow column ( 20 ml ) pre-equilibrated with Buffer A ( 20 mM Hepes/NaOH pH = 7 . 6 , 10% glycerol , 100 mM KCl , 0 . 1 mM EDTA and 2 mM DTT ) . A step gradient was used ( 100 , 200 and 400 mM KCl ) . eIF2B eluted at 400 mM KCl . The eluate was diluted slowly by adding Buffer A ( with no KCl ) to 100 mM KCl and then loaded on a Q-Sepharose ( 20 ml ) pre-equilibrated with Buffer A . A step gradient was used ( 300 mM and 500 mM KCl ) with eIF2B eluting at 500 mM KCl . The eluate was dialyzed overnight with Buffer A and loaded to a Mono Q ( 5-50 GL , GE Healthcare , Wauwatosa , WI ) equilibrated with buffer A ( a continuous gradient 100–500 mM KCl was used ) and eIF2B eluted at 350 mM KCl . The eluate was buffer exchanged with Buffer A and aliquots were flash frozen in liquid N2 . Human eIF2 was purified from HeLa cells as described previously ( Fraser et al . , 2007 ) . In brief , from the 40–50% ammonium sulfate precipitate of post-nuclear HeLa cell lysate , eIF2 was purified through a series of chromatographic steps which included a Mono Q 10/10 column ( GE Healthcare ) , a Mono S 10/10 column ( GE Healthcare ) , a CHT5-1 ceramic hydroxyapatite column ( Bio-Rad ) , and a Superose 6 16/60 column ( GE Healthcare ) . The protein was stored at −80°C in buffer containing 20 mM Hepes-K pH 7 . 5 , 150 mM KCl , 1 mM DTT , and 10% glycerol . GDP dissociation assays were adapted from a previously described protocol ( Sokabe et al . , 2012 ) . For each reaction purified eIF2 ( 21 pmol ) was incubated with 0 . 6 μCi [3H]-GDP ( 40 Ci/mmol , PerkinElmer , Waltham , MA ) in a reaction buffer ( 20 mM HEPES pH 7 . 5 , 80 mM KCl , 1 mM DTT , 1 mg/ml creatine phosphokinase ( EMD Millipore , Billerica , MA ) , 5% glycerol ) without magnesium at 37°C for 10 min , and then further incubated with 1 mM Mg ( OAc ) 2 at 30°C for 3 min with or without ISRIB ( 100 nM ) in a total volume of 60 μl . The reaction was initiatied by the addition of 60 nmol unlabeled GDP with or without eIF2B ( 0 . 6 μl of partially purified rabbit reticulocyte eIF2B , which correspond to approximately 0 . 3 pmoles of the complex ) . At each time point , an aliquot was taken ( 10 μl ) and the reaction was stopped by addition to 300 μl ice-cold stop buffer ( reaction buffer with 5 mM Mg ( OAc ) 2 ) , immediately filtered through a HAWP nitrocellulose membrane filter ( EMD Millipore ) on a vacuum manifold , and washed twice with 1 ml ice-cold stop buffer . Filters were dried and remaining [3H]-GDP bound to eIF2 was counted by liquid scintillation in Ecoscint ( National Diagnostics , Atlanta , GA ) . Data collected were fitted to a first-order exponential decay . eIF2-P was synthesized by incubating eIF2 ( 1 . 76 μM ) with recombinant GST-PERK ( 500 nM ) at 37°C for 45 min in a reaction buffer containing: 0 . 5 mM ATP , 50 mM Tris–HCl pH 7 . 5 , 4 mM MgCl2 , 100 mM NaCl , 1 mM tris ( 2-carboxyethyl ) phosphine ( TCEP ) , 1% glycerol . The phosphorylation reaction was stopped by the addition of 1 μM GSK PERK inhibitor ( Toronto Research Chemicals ) and 4 mM EDTA to chelate magnesium ions . For eIF2-P•GDP dissociation reactions ( Figure 5B ) , eIF2-P ( 21 pmol ) was loaded with [3H]-GDP . For experiments where eIF2 was mixed with eIF2-P ( Figure 5C , D ) , unphosphorylated eIF2 was loaded with [3H]-GDP and mixed ( 3:1 or 1:1 ) with eIF2-P , which was not loaded with [3H]-GDP , such that the sum of eIF2 and eIF2-P equaled 21 pmol . GDP dissociation assays were conducted as described above in the presence of 50 nM GSK PERK inhibitor to ensure that the residual PERK kinase did not phosphorylate eIF2 during the course of the dissociation assay . Cytosolic human PERK was codon-optimized for Escherichia coli expression by Genewiz Inc . A construct was then cloned into a PGEX-6P-2 vector for expression using two rounds of In-Fusion cloning ( Clontech , Mountain View , CA ) ( 535–1093 Δ660–868 ) . The cytosolic portion of PERK , lacking the unstructured loop region ( amino acids 535–1093 Δ660–868 ) was then co-expressed with a tag-less lambda phosphatase to produce a fully dephosphorylated PERK protein in BL21 star ( DE3 ) ( Life Technologies , Carlsbad , CA ) . Cells were grown to an OD600 of 0 . 5 before induction with 0 . 1 mM IPTG at 15°C for 24 hr . Cells where harvested and lysed using AVESTIN Emulsiflex-C3 in a buffer containing 50 mM Tris-HCl , pH 8 . 0 , 500 mM NaCl , 5% glycerol , 5 mM TCEP ( buffer A ) and EDTA-free COMPlete protease inhibitor cocktail ( Roche ) . The lysate was cleared by centrifugation at 100 , 000×g before batch-binding to a GST-Sepharose resin . The resin was washed 5 times with buffer A . The protein was loaded onto a HiTrap Q HP column to remove remaining lambda phosphatase . The PERK ( 535–1093 Δ660–868 ) protein was then concentrated and fractionated on a Superdex 200 GL ( GE Healthcare ) to remove protein aggregates . To a solution of the carboxylic acid ( 1 equiv . ) in N , N-dimethylformamide , were sequentially added 1-hydroxybenzotriazole hydrate ( 1 . 2 equiv . ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 1 . 2 equiv . ) , 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid ( 1 . 0 equiv . , prepared as described in the synthesis of ISRIB-A8 , below ) and N , N-diisopropylethylamine ( 1 . 5 equiv ) . The reaction mixture was stirred at room temperature until judged complete by LC-MS and then diluted with water ( 2 ml ) . The mixture was vigorously vortexed , centrifuged and the water was decanted . This washing protocol was repeated with water ( 2 ml ) and then with diethyl ether ( 2 ml ) . The wet solid was dissolved in dichloromethane ( 10 ml ) and dried over anhydrous magnesium sulfate . The solids were removed by filtration and the filtrate was concentrated by rotary evaporation to obtain the product . To a solution of the carboxylic acid ( 2 equiv . ) in N , N-dimethylformamide were sequentially added 1-hydroxybenzotriazole hydrate ( 2 equiv . ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 2 equiv . ) , the diamine ( 1 . 0 equiv . ) and N , N-diisopropylethylamine ( 6 equiv ) . The reaction mixture was stirred at room temperature until judged complete by LC-MS and then diluted with water . The precipitate formed was washed with water and 10% diethyl ether in dichloromethane . The precipitate was dried in vacuo to obtain the product . To a solution of ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 1 equiv . ) in N , N-dimethylformamide were added the carboxylic acid ( 2 equiv . ) , 1-[bis ( dimethylamino ) methylene]-1H-1 , 2 , 3-triazolo[4 , 5-b]pyridinium 3-oxid hexafluorophosphate ( 2 . 1 equiv . ) and N , N-diisopropylethylamine ( 4 equiv . ) . The reaction mixture was vigorously stirred at room temperature until judged complete by LC-MS . Water ( 2 ml ) was added . The mixture was centrifuged and the water was decanted . This washing protocol was repeated thrice and the resulting wet solid was concentrated down with toluene ( 10 ml ) in a rotary evaporator . The residual product was washed with diethyl ether ( 10 ml ) and concentrated using rotary evaporation to obtain the product . To a solution of 1 , 4-diaminobutane ( 0 . 032 g , 0 . 2 mmol ) in tetrahydrofuran ( 1 . 0 ml ) , were added 4-chlorophenoxyacetyl chloride ( 0 . 062 ml , 0 . 4 mmol ) and N , N-diisopropylethylamine ( 0 . 173 ml , 1 . 0 mmol ) . The reaction mixture was stirred at room temperature for 20 hr and then partitioned between 1:1 mixture of water/dichloromethane ( 20 ml ) . The organic layer was washed with 10% aqueous potassium hydrogen sulfate , water and brine . The organic phase was then dried over magnesium sulfate , filtered , and concentrated to obtain a brownish orange solid . The brownish orange solid was triturated with diethyl ether and the resulting solids were separated by centrifugation and dried to obtain 26 mg ( 31% ) of the title compound as tan powder . 1H NMR ( 400 MHz , DMSO-d6 ) δ 8 . 06 ( t , J = 5 . 6 Hz , 2H ) , 7 . 30–7 . 32 m , 4H ) , 6 . 93–6 . 95 ( m , 4H ) , 4 . 43 ( s , 4H ) , 3 . 08 ( d , J = 5 . 7Hz , 4H ) , 1 . 37 ( br . s , 4H ) LC-MS: m/z = 425 [M + H , 35Cl ]+ , 427 [M + H , 37Cl]+ . To a cooled ( 0°C ) solution of tert-butyl N-[ ( 1r , 3r ) -3-aminocyclobutyl]carbamate ( 0 . 05 g , 0 . 277 mmol ) in 1 , 2-dichloroethane ( 1 . 38 ml ) , was added trifluoroacetic acid ( 1 . 38 ml ) . The reaction mixture was stirred at room temperature for 2 hr and then concentrated down to dryness to obtain 100 mg of ( 1r , 3r ) -cyclobutane-1 , 3-bis ( aminium ) ditrifluoroacetate which was used without further purification . To a solution 4-chlorophenoxyacetic acid ( 0 . 19 g , 0 . 63 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 12 g , 0 . 63 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 175 g , 0 . 63 mmol ) , ( 1r , 3r ) -cyclobutane-1 , 3-bis ( aminium ) ditrifluoroacetate ( 0 . 1 g , 0 . 31 mmol ) and N , N-diisopropylethylamine ( 0 . 34 ml , 1 . 91 mmol ) . The reaction mixture was stirred at room temperature for 2 hr and then subjected to conditions described in procedure B to afford 72 mg ( 54% ) of the title compound . 1H NMR ( 300 MHz , CDCl3 ) δ 7 . 29–7 . 35 ( m , 4H ) , 6 . 91 ( dd , J = 9 , 2 . 2 Hz , 4H ) , 6 . 80 ( d , J = 7 . 6 Hz , 2H ) , 4 . 60–4 . 62 ( m , 2H ) , 4 . 48 ( s , 4H ) , 2 . 46–2 . 51 ( m , 4H ) LC-MS: m/z = 423 [M + H]+ . To a cooled ( 0°C ) solution of tert-butyl N-[ ( 1 s , 3 s ) -3-aminocyclobutyl]carbamate ( 0 . 05 g , 0 . 277 mmol ) in 1 , 2-dichloroethane ( 1 . 38 ml ) , was added trifluoroacetic acid ( 1 . 38 ml ) . The reaction mixture was stirred at room temperature for 1 . 5 hr and then concentrated down to dryness to obtain 100 mg of ( 1 s , 3 s ) -cyclobutane-1 , 3-bis ( aminium ) ditrifluoroacetate which was used without further purification . To a solution 4-chlorophenoxyacetic acid ( 0 . 19 g , 0 . 63 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 12 g , 0 . 63 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 175 g , 0 . 63 mmol ) , ( 1 s , 3 s ) -cyclobutane-1 , 3-bis ( aminium ) ditrifluoroacetate ( 0 . 1 g , 0 . 31 mmol ) and N , N-diisopropylethylamine ( 0 . 34 ml , 1 . 91 mmol ) . The reaction mixture was stirred at room temperature for 2 hr . The reaction mixture was then diluted with 5% methanol in dichloromethane , washed with water and brine . The organic layer was dried over magnesium sulfate , filtered and concentrated . The crude mixture was purified by flash column chromatography ( 40% acetone/hexanes ) to obtain 34 mg ( 25% ) of the title compound . 1H NMR ( 300 MHz , CDCl3 ) δ 7 . 26–7 . 29 ( m , 4H ) , 6 . 84–6 . 87 ( m , 4H ) , 6 . 77 ( d , J = 6 . 5 Hz , 2H ) , 4 . 42 ( m , 4H ) , 4 . 17–4 . 25 ( s , 2H ) , 2 . 84–2 . 93 ( m , 2H ) , 2 . 02–2 . 12 ( m , 2H ) LC-MS: m/z = 423 [M + H]+ . To a solution of 1 , 3-diaminopropane ( 0 . 017 ml , 0 . 2 mmol ) in tetrahydrofuran ( 0 . 6 ml ) , was added 4-chlorophenoxyacetyl chloride ( 0 . 062 ml , 0 . 4 mmol ) and N , N-diisopropylethylamine ( 0 . 08 ml , 0 . 5 mmol ) . The reaction mixture was stirred at room temperature for an hour and then partitioned between 1:1 mixture of water/dichloromethane ( 20 ml ) . The organic layer was washed with 10% aqueous potassium hydrogen sulfate , water and brine . The organic phase was then dried over magnesium sulfate , filtered and concentrated to obtain a brownish orange oil . The brownish orange oil was purified by flash column chromatography ( 5–80% acetone/dichloromethane ) to obtain 41 mg ( 49% ) of the title compound . 1H NMR ( 400 MHz , CDCl3 ) δ 7 . 24–7 . 26 ( m , 4H ) , 7 . 15 ( br . s , 2H ) , 6 . 85–6 . 87 ( m , 4H ) , 4 . 45 ( s , 4H ) , 3 . 08 ( quint , J = 6 . 3 Hz , 4H ) , 1 . 37 ( quint , J = 6 . 2 Hz , 2H ) LC-MS: m/z = 411 [M + H , 35Cl ]+ , 413 [M + H , 37Cl]+ . Step 1: To a mixture of tert-butyl N-[ ( 1r , 4r ) -4-aminocyclohexyl]carbamate ( 0 . 750 g , 3 . 5 mmol ) in THF ( 20 ml ) were sequentially added N , N-diisopropylethylamine ( 0 . 914 ml , 5 . 25 mmol ) and 4-chlorophenoxyacetyl chloride ( 0 . 573 ml , 3 . 78 mmol ) . The reaction mixture was vigorously stirred at room temperature for 3 hr and then diluted with water ( 100 ml ) . The precipitate was filtered and the solid was washed with water . The resulting solid was then diluted with diethyl ether and vacuum filtered . The filter cake was washed with diethyl ether . The residual ether was removed under vacuum to afford 1 . 103 g ( 82% ) of tert-butyl N-[ ( 1r , 4r ) -4-[2- ( 4-chlorophenoxy ) acetamido]cyclohexyl]carbamate as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 88 ( d , J = 7 . 87 Hz , 1H ) , 7 . 25–7 . 37 ( m , 2H ) , 6 . 93 ( d , J = 8 . 97 Hz , 2H ) , 6 . 68 ( d , J = 7 . 69 Hz , 1H ) , 4 . 41 ( s , 2H ) , 3 . 51 ( m , 1H ) , 3 . 13 ( br . s . , 1H ) , 1 . 72 ( t , J = 13 . 19 Hz , 4H ) , 1 . 34 ( s , 9H ) , 1 . 09–1 . 30 ( m , 4H ) ; LC-MS: m/z = 405 [M + Na , 35Cl ]+ , 407 [M + Na , 37Cl ]+ , 765 [2M + H , 35Cl × 2]+ , 767 [2M + H , 35Cl , 37Cl]+ . Step 2: To a suspension of tert-butyl N-[ ( 1r , 4r ) -4-[2- ( 4-chlorophenoxy ) acetamido]cyclohexyl]carbamate ( 0 . 5 g , 1 . 31 mmol ) in dichloromethane ( 9 ml ) were sequentially added triethylsilane ( 0 . 3 ml , 1 . 88 mmol ) , water ( 0 . 2 ml , 11 . 1 mmol ) , and trifluoroacetic acid ( 3 . 0 ml , 39 . 2 mmol ) . The suspension quickly clarified and turned yellow upon addition of trifluoroacetic acid . The reaction mixture was vigorously stirred at room temperature for 30 min and then the solvent was removed by rotary evaporation . The resulting colorless oil was triturated with diethyl ether . After decanting the ether washes , residual solvent was removed under vacuum to afford 499 mg ( 96% ) of 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 95 ( d , J = 7 . 87 Hz , 1H ) , 7 . 77 ( br . s . , 3H ) , 7 . 31 ( d , J = 8 . 97 Hz , 2H ) , 6 . 93 ( d , J = 8 . 97 Hz , 2H ) , 4 . 43 ( s , 2H ) , 3 . 54 ( m , 1H ) , 2 . 93 ( br . s . , 1H ) , 1 . 90 ( d , J = 9 . 16 Hz , 2H ) , 1 . 77 ( d , J = 9 . 34 Hz , 2H ) , 1 . 31 ( sxt , J = 11 . 50 Hz , 4H ) ; LC-MS: m/z = 283 [M + H , 35Cl ]+ , 285 [M + H , 37Cl]+ . Step 3: To a solution of 4-fluorophenoxyacetic acid ( 0 . 009 g , 0 . 050 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 009 g , 0 . 055 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 012 g , 0 . 057 mmol ) , 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid ( 0 . 02 g , 0 . 050 mmol ) and N , N-diisopropylethylamine ( 0 . 013 ml , 0 . 12 mmol ) . The reaction mixture was subjected to conditions described in procedure A to obtain 14 mg ( 60% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 88–7 . 92 ( M , 2H ) , 7 . 31 ( d , J = 9 Hz , 2H ) , 7 . 10 ( t , J = 8 . 8 Hz , 2H ) , 6 . 92–6 . 95 ( m , 4H ) , 4 . 39–4 . 42 ( m , 4H ) , 3 . 57 ( br . s , 2H ) , 1 . 74 ( d , J = 5 . 9 Hz , 4H ) , 1 . 29–1 . 33 ( m , 4H ) LC-MS: m/z = 435 [M + H , 35Cl ]+ , 437 [M + H , 37Cl]+ . To a solution 4-fluorophenoxyacetic acid ( 0 . 12 g , 0 . 7 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 094 g , 0 . 7 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 140 g , 0 . 7 mmol ) , ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 040 g , 0 . 35 mmol ) and N , N-diisopropylethylamine ( 0 . 372 ml , 2 . 1 mmol ) . The reaction mixture was subjected to conditions described in procedure B to afford 73 mg ( 50% ) of the title compound . 1H NMR ( 300 MHz , CDCl3 ) δ 7 . 02 ( t , J = 8 . 3 Hz , 4H ) , 6 . 89–6 . 90 ( m , 4H ) , 6 . 38 ( d , J = 7 . 5 Hz , 2H ) , 4 . 43 ( s , 4H ) , 3 . 88 ( br . s , 2H ) , 2 . 07 ( d , J = 5 . 7 Hz , 4H ) , 1 . 36–1 . 39 ( m , 4H ) LC-MS: m/z = 419 [M + H]+ . To a solution 4-methyl-phenoxyacetic acid ( 0 . 016 g , 0 . 101 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 014 g , 0 . 101 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 02 g , 0 . 101 mmol ) , 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid ( 0 . 04 g , 0 . 101 mmol ) and N , N-diisopropylethylamine ( 0 . 06 ml , 0 . 303 mmol ) . The reaction mixture was subjected to conditions described in procedure A to obtain 7 mg ( 16% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 91 ( d , J = 8 Hz , 1H ) , 7 . 84 ( d , J = 7 . 8 Hz , 1H ) , 7 . 31 ( d , J = 8 . 8 Hz , 2H ) , 7 . 06 ( t , J = 8 . 3 Hz , 2H ) , 6 . 94 ( d , J = 8 . 8 Hz , 2H ) , 6 . 80 ( d , J = 8 . 4 Hz , 2H ) , 4 . 42 ( s , 2H ) , 4 . 35 ( s , 2H ) , 3 . 56 ( br . s , 2H ) , 2 . 20 ( s , 3H ) , 1 . 73 ( d , J = 6 . 6 Hz , 4H ) , 1 . 22–1 . 33 ( m , 4H ) LC-MS: m/z = 431 [M + H]+ . To a solution 4-methylphenoxyacetic acid ( 0 . 116 g , 0 . 7 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 094 g , 0 . 7 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 14 g , 0 . 7 mmol ) , ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 04 g , 0 . 35 mmol ) and N , N-diisopropylethylamine ( 0 . 372 ml , 2 . 1 mmol ) . The reaction mixture was stirred at 52°C for 24 hr and then subjected to conditions described in procedure B to afford 84 mg ( 58% ) of the title compound . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 84 ( d , J = 6 . 8 Hz , 2H ) , 7 . 05 ( d , J = 6 . 8 Hz , 4H ) , 6 . 80 ( d , J = 6 . 6 Hz , 4H ) , 4 . 35 ( s , 4H ) , 3 . 56 ( br . s , 2H ) , 2 . 19 ( s , 6H ) , 1 . 73 ( br . s , 4H ) , 1 . 31 ( br . s , 4H ) LC-MS: m/z = 411 [M + H]+ . To a solution 4-cyanophenoxyacetic acid ( 0 . 009 g , 0 . 050 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 009 g , 0 . 055 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 012 g , 0 . 057 mmol ) , 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid ( 0 . 02 g , 0 . 050 mmol ) and N , N-diisopropylethylamine ( 0 . 013 ml , 0 . 12 mmol ) . The reaction mixture was subjected to conditions described in procedure A to obtain 14 mg ( 65% ) of the title compound as a beige solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 99 ( d , J = 7 . 9 Hz , 1H ) , 7 . 91 ( d , J = 8 . 1 Hz , 1H ) , 7 . 76 ( d , J = 8 . 8 Hz , 1H ) , 7 . 31 ( d , J = 9 . 1 Hz , 1H ) , 7 . 07 ( d , J = 8 . 8 Hz , 2H ) , 6 . 94 ( d , J = 8 . 8 Hz , 2H ) , 4 . 55 ( s , 2H ) , 4 . 42 ( s , 2H ) , 3 . 56 ( br . s , 2H ) , 1 . 74 ( d , J = 7 . 7 Hz , 4H ) , 1 . 28–1 . 32 ( m , 4H ) LC-MS: m/z = 442 [M + H , 35Cl ]+ , 444 [M + H , 37Cl]+ . To a solution 4-cyanophenoxyacetic acid ( 0 . 124 g , 0 . 7 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 094 g , 0 . 7 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 14 g , 0 . 7 mmol ) , ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 04 g , 0 . 35 mmol ) and N , N-diisopropylethylamine ( 0 . 372 ml , 2 . 1 mmol ) . The reaction mixture was subjected to conditions described in procedure B to afford 54 mg ( 36% ) of the title compound . 1H NMR ( 300 MHz , DMSO-d6 ) δ 8 . 01 ( d , J = 5 . 8 Hz , 2H ) , 7 . 76 ( d , J = 6 . 8 Hz , 4H ) , 7 . 08 ( d , J = 6 . 8 Hz , 4H ) , 4 . 55 ( s , 4H ) , 3 . 56 ( br . s , 2H ) , 1 . 75 ( br . s , 4H ) , 1 . 31 ( br . s , 4H ) LC-MS: m/z = 433 [M + H]+ . To a solution 3 , 4-dichlorophenoxyacetic acid ( 0 . 011 g , 0 . 050 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 009 g , 0 . 055 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 012 g , 0 . 057 mmol ) , 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4-aminocyclohexyl]acetamide trifluoroacetic acid ( 0 . 020 g , 0 . 050 mmol ) and N , N-diisopropylethylamine ( 0 . 013 ml , 0 . 12 mmol ) . The reaction mixture was subjected to conditions described in procedure A to obtain 21 mg ( 86% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 94 ( d , J = 8 . 2 Hz , 1H ) , 7 . 91 ( d , J = 8 . 2 Hz , 1H ) , 7 . 51 ( d , J = 8 . 8 Hz , 1H ) , 7 . 31 ( d , J = 9 Hz , 2H ) , 7 . 22 ( d , J = 2 . 9 Hz , 1H ) , 6 . 92–6 . 95 ( m , 3H ) , 4 . 48 ( s , 2H ) , 4 . 42 ( s , 2H ) , 3 . 56 ( br . s , 2H ) , 1 . 74 ( d , J = 6 Hz , 4H ) , 1 . 26–1 . 31 ( m , 4H ) LC-MS: m/z = 485 [M + H , 35Cl ]+ , 487 [M + H , 37Cl]+ . To a solution of ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 025 g , 0 . 2 mmol ) in N , N-dimethylformamide ( 1 ml ) were added 3 , 4-dichlorophenoxyacetic acid ( 0 . 097 g , 0 . 4 mmol ) , 1-[bis ( dimethylamino ) methylene]-1H-1 , 2 , 3-triazolo[4 , 5-b]pyridinium 3-oxid hexafluorophosphate ( 0 . 175 g , 0 . 5 mmol ) and N , N-diisopropylethylamine ( 0 . 153 ml , 0 . 9 mmol ) . The reaction mixture was subjected to conditions described in procedure C to obtain 107 mg ( 94% ) of the title compound as a cream colored solid . 1H NMR ( 400 MHz , CDCl3 ) δ 7 . 37 ( d , J = 8 . 8 Hz , 2H ) , 7 . 04 ( s , 2H ) , 6 . 78 ( d , J = 8 . 8 Hz , 2H ) , 6 . 26 ( d , J = 8 . 1 Hz , 2H ) , 4 . 42 ( s , 4H ) , 3 . 85 ( br . s , 2H ) , 2 . 05 ( d , J = 6 Hz , 4H ) , 1 . 31–1 . 39 ( m , 4H ) ; LC-MS: m/z = 519 [M + H , 35Cl]+ , 521 [M + H , 37Cl]+ . Step 1: To a cooled solution ( 0°C ) of ( 1r , 4r ) -4-[2- ( 4-chlorophenoxy ) acetamido]cyclohexan-1-aminium trifluoroacetate ( 0 . 550 g , 1 . 4 mmol ) in THF and N , N-diisopropylethylamine ( 0 . 966 ml , 5 . 5 mmol ) slowly added chloroacetyl chloride ( 0 . 121 ml , 1 . 5 mmol ) . The mixture was stirred at ambient temperature for 20 min . The reaction mixture was diluted in dichloromethane , washed with 0 . 1 N hydrochloric acid , water and brine . The organic layer was dried over magnesium sulfate , filtered and concentrated in a rotary evaporator to obtain about 430 mg of crude 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4- ( 2-chloroacetamido ) cyclohexyl]acetamide that was used without further purification . Step 2: To a suspension of 2- ( 4-chlorophenoxy ) -N-[ ( 1r , 4r ) -4- ( 2-chloroacetamido ) cyclohexyl]acetamide ( 0 . 036 g , 0 . 1 mmol ) and 4-chloro-3-fluorophenol ( 0 . 015 g , 0 . 1 mmol ) in acetone ( 1 . 0 ml ) , added potassium carbonate ( 0 . 021 g , 0 . 2 mmol ) and stirred at 120°C in the microwave reactor for 20 min . The reaction mixture was concentrated down and suspended in water ( 10 ml ) . The mixture was vigorously vortexed then centrifuged , and the water was decanted . This washing protocol was repeated with water and then with diethyl ether ( 10 ml ) . The wet solid was dissolved in dichloromethane ( 10 ml ) and dried over anhydrous magnesium sulfate . The solids were removed by filtration , and the filtrate was concentrated by rotary evaporation to afford 28 mg ( 60% ) of the title compound as a tan solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 9 ( t , J = 8 . 9 Hz , 2H ) , 7 . 46 ( t , J = 8 . 9 Hz , 1H ) , 7 . 31 ( d , J = 9 Hz , 2H ) , 7 . 03 ( dd , J = 11 . 4 , 2 . 7 Hz , 1H ) , 6 . 94 ( d , J = 9 Hz , 2H ) , 6 . 81 ( dd , J = 8 . 5 , 2 . 3 Hz , 1H ) , 4 . 46 ( s , 2H ) , 4 . 42 ( s , 2H ) , 1 . 74 ( d , J = 6 . 2 Hz , 4H ) , 1 . 29–1 . 35 ( m , 4H ) LC-MS: m/z = 469 [M + H , 35Cl ]+ , 471 [M + H , 37Cl]+ . Step 1: To a solution 4-chloro-3-fluorophenol ( 0 . 100 g , 0 . 7 mmol ) in N , N-dimethylformamide ( 2 ml ) , were added potassium carbonate ( 0 . 189 g , 1 . 4 mmol ) and tert-butyl bromoacetate ( 0 . 111 ml , 0 . 8 mmol ) and stirred at 65°C for 2 hr . The reaction mixture was diluted with ethyl acetate ( 10 ml ) , washed with water ( 3 × 10 ml ) and brine ( 10 ml ) . The organic layer was dried over magnesium sulfate and concentrated in a rotary evaporator to obtain 177 mg of tert-butyl 2- ( 4-chloro-3-fluorophenoxy ) acetate as a colorless oil which was used without further purification . Step 2: To a solution of tert-butyl 2- ( 4-chloro-3-fluorophenoxy ) acetate ( 177 mg , 0 . 7 mmol ) in methanol/water ( 4 . 5 ml , 2:1 ) was added aqueous 5 N NaOH solution ( 0 . 7 ml , 3 . 5 mmol ) and stirred at ambient temperature for an hour . The reaction mixture was concentrated in a rotary evaporator to remove methanol , diluted with water ( 5 ml ) and extracted with ethyl acetate ( 5 ml ) . The aqueous layer was adjusted to about pH 2 with 1 N hydrochloric acid and extracted with ethyl acetate ( 3 × 5 ml ) . The organic extract was washed with brine ( 5 ml ) , dried over magnesium sulfate and concentrated to obtain 108 mg of 2- ( 4-chloro-3-fluorophenoxy ) acetic acid as a white solid which was used without further purification . Step 3: To a solution of ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 02 g , 0 . 2 mmol ) in N , N-dimethylformamide ( 1 ml ) were added 2- ( 4-chloro-3-fluorophenoxy ) acetic acid ( 0 . 072 g , 0 . 4 mmol ) , 1-[bis ( dimethylamino ) methylene]-1H-1 , 2 , 3-triazolo[4 , 5-b]pyridinium 3-oxid hexafluorophosphate ( 0 . 14 g , 0 . 4 mmol ) and N , N-diisopropylethylamine ( 0 . 122 ml , 0 . 7 mmol ) . The reaction mixture was subjected to conditions described in procedure C to obtain 85 mg ( >95% ) of the title compound as a white solid . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 23–7 . 28 ( m , 2H ) , 6 . 72 ( d , J = 8 Hz , 2H ) , 6 . 61–6 . 64 ( m , 4H ) , 4 . 36 ( s , 4H ) , 3 . 56 ( m , 2H ) , 1 . 95 ( d , J = 6 . 2 Hz , 4H ) , 1 . 28–1 . 33 ( m , 4H ) ; LC-MS: m/z = 487 [M + H , 35Cl ]+ , 489 [M + H , 37Cl ]+ . To a solution 3- ( 4-chlorophenyl ) propionic acid ( 0 . 129 g , 0 . 7 mmol ) in N , N-dimethylformamide ( 1 . 0 ml ) were sequentially added 1-hydroxybenzotriazole hydrate ( 0 . 094 g , 0 . 7 mmol ) , 1- ( 3-dimethylaminopropyl ) -3-ethylcarbodiimide hydrochloride ( 0 . 14 g , 0 . 7 mmol ) , ( 1r , 4r ) -cyclohexane-1 , 4-diamine ( 0 . 04 g , 0 . 35 mmol ) and N , N-diisopropylethylamine ( 0 . 372 ml , 2 . 1 mmol ) . The reaction mixture was stirred at 52°C for 18 hr and then subjected to conditions described in procedure B to afford 103 mg ( 66% ) of the title compound . 1H NMR ( 400 MHz , DMSO-d6 ) δ 7 . 65 ( d , J = 7 . 5 Hz , 2H ) , 7 . 28 ( d , J = 8 . 1 Hz , 4H ) , 7 . 17–7 . 19 ( m , 4H ) , 3 . 41 ( br . s , 2H ) , 2 . 73–2 . 76 ( m , 4H ) , 2 . 26–2 . 30 ( m , 4H ) , 1 . 66–1 . 68 ( m , 4H ) , 1 . 10–1 . 12 ( m , 4H ) LC-MS: m/z = 447 [M + H , 35Cl ]+ , 449 [M + H , 37Cl]+ . | Proteins are often described as life's ‘workhorse’ molecules , and cells must be able to build new proteins to stay alive . This ability is also vital for storing new memories . A protein called eIF2 carries out a critical step in the process that cells use to make proteins; and a decrease in the activity of eIF2 has been linked with memory loss in diseases such as Parkinson's and Alzheimer's disease . When a cell experiences stressful conditions—such as virus infection or starvation—it triggers the ‘integrated stress response’ . This response helps the cell conserve its resources and take corrective steps to restore its normal working conditions . As part of the integrated stress response , an enzyme adds a phosphate group onto eIF2 . The ‘phosphorylated’ eIF2 blocks protein production , which causes the cell to make fewer proteins . In 2013 , researchers revealed that a small drug-like molecule , called ISRIB , could prevent this decline in protein production following eIF2 phosphorylation; and when ISRIB was administered to mice and rats , it enhanced their long-term memories . However , this early work did not identify the molecule that is targeted by ISRIB . Now Sidrauski , Tsai et al . —including many of researchers involved in the 2013 work—have used a combination of techniques including genetics , chemistry and biochemistry to reveal the target of ISRIB . The experiments show that ISRIB's molecular target is a protein complex called eIF2B . Artificially reducing the production of eIF2B made cells resistant to the effects of ISRIB . The eIF2B protein normally works to activate eIF2; Sidrauski , Tsai et al . observed that ISRIB boosts the activity of eIF2B and renders it insensitive to blockage by phosphorylated eIF2 . This in turn increases protein production in the cell . But how does ISRIB activate eIF2B ? It was known that two copies of eIF2B can bind to each other; and Sidrauski , Tsai et al . found that ISRIB acts by stabilizing these larger protein complexes that are more active and less sensitive to inhibition by phosphorylated eIF2 . Finally , in further experiments , new versions of ISRIB were synthesized that are ten-times as active as the original molecule inside cells . Importantly , the discovery that eIF2B is the molecular target for ISRIB has recently been independently validated by other researchers , and it looks promising that this discovery will guide future efforts to develop clinically useful drugs to treat memory disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2015 | Pharmacological dimerization and activation of the exchange factor eIF2B antagonizes the integrated stress response |
The chloride-proton exchanger CLC-7 plays critical roles in lysosomal homeostasis and bone regeneration and its mutation can lead to osteopetrosis , lysosomal storage disease and neurological disorders . In lysosomes and the ruffled border of osteoclasts , CLC-7 requires a β-subunit , OSTM1 , for stability and activity . Here , we present electron cryomicroscopy structures of CLC-7 in occluded states by itself and in complex with OSTM1 , determined at resolutions up to 2 . 8 Å . In the complex , the luminal surface of CLC-7 is entirely covered by a dimer of the heavily glycosylated and disulfide-bonded OSTM1 , which serves to protect CLC-7 from the degradative environment of the lysosomal lumen . OSTM1 binding does not induce large-scale rearrangements of CLC-7 , but does have minor effects on the conformation of the ion-conduction pathway , potentially contributing to its regulatory role . These studies provide insights into the role of OSTM1 and serve as a foundation for understanding the mechanisms of CLC-7 regulation .
CLC-7 is a member of the CLC family of chloride ( Cl- ) channels and chloride ( Cl- ) /proton ( H+ ) transporters and is expressed in the lysosome and the resorption lacuna of osteoclasts ( Graves et al . , 2008; Ishida et al . , 2013; Kornak et al . , 2001; Weinert et al . , 2010 ) . In the membranes of these acidic compartments , CLC-7 uses the large pH gradient to catalyze the uptake of two Cl- ions for each H+ released ( Graves et al . , 2008; Leisle et al . , 2011; Ludwig et al . , 2013 ) . Dysfunction of CLC-7 is associated with dysregulation of ion and pH homoeostasis of the lysosome and the resorption lacuna ( Graves et al . , 2008; Ishida et al . , 2013; Kasper et al . , 2005; Kornak et al . , 2001; Lange et al . , 2006; Undiagnosed Diseases Network et al . , 2019; Steinberg et al . , 2010; Weinert et al . , 2010 ) . As both of these compartments rely on high proton concentrations to perform their physiological roles , disruption of CLC-7 function is associated with human diseases driven by impaired lysosomal and/or osteoclast function ( Kasper et al . , 2005; Kornak et al . , 2001; Lange et al . , 2006; Undiagnosed Diseases Network et al . , 2019; Pressey et al . , 2010 ) . In particular , osteopetrosis , a disease characterized by dense and brittle bones , is the most common disease associated with CLC-7 mutation , with more than 50 distinct pathogenic mutations identified to date ( Chalhoub et al . , 2003; Cleiren et al . , 2001; Kasper et al . , 2005; Kornak et al . , 2001; Lange et al . , 2006; Sartelet et al . , 2014; Schulz et al . , 2010; Weinert et al . , 2010 ) . Extensive structural and functional characterization of prokaryotic and eukaryotic CLC channels and transporters have established a framework for Cl-/H+ exchange and identified several key residues that participate in the transport cycle ( Accardi et al . , 2004; Accardi and Miller , 2004; Accardi et al . , 2005; Basilio et al . , 2014; Chavan et al . , 2020; Dutzler et al . , 2002; Dutzler et al . , 2003; Feng et al . , 2010; Feng et al . , 2012; Jayaram et al . , 2008; Park et al . , 2017; Park and MacKinnon , 2018; Picollo et al . , 2012; Wang et al . , 2019; Zdebik et al . , 2008 ) . Within the Cl--conduction pathway , the gating glutamate ( Glugate ) that is conserved in CLC transporters is proposed to oscillate between at least four different conformations ( Chavan et al . , 2020; Dutzler et al . , 2002; Dutzler et al . , 2003; Feng et al . , 2010 ) . The movement and changes in the protonation state of Glugate are coupled to the binding and release of Cl- ions in the highly conserved external and central binding sites ( Picollo et al . , 2012 ) . Near the center of the transporter , the anion and H+-conduction pathways diverge with the anion pathway passing through the internal binding site before reaching the cytosol , while the H+-conduction pathway passes through a hydrophobic gap before reaching a conserved internal glutamate ( Gluin ) ( Accardi and Miller , 2004; Accardi et al . , 2005; Chavan et al . , 2020; Leisle et al . , 2020; Lim and Miller , 2009; Zdebik et al . , 2008 ) . This conserved Gluin is dispensable for coupled transport and water molecules has been proposed to mediate H+ transport through the hydrophobic gap ( Feng et al . , 2010; Han et al . , 2014; Wang and Voth , 2009 ) . Despite these extensive efforts , the precise mechanisms by which Cl- and H+ transport are coupled remains poorly understood as are the mechanisms that underlie the gating of CLC transporters . Unique among mammalian CLC transporters , CLC-7 requires a β-subunit , osteopetrosis-associated transmembrane protein 1 ( OSTM1 ) , for transport activity ( Lange et al . , 2006; Leisle et al . , 2011 ) . CLC-7 and OSTM1 co-localize in lysosomes and the ruffled border of osteoclasts ( Lange et al . , 2006; Leisle et al . , 2011; Schulz et al . , 2010 ) . There , CLC-7 and OSTM1 stabilize the expression of one another and are both required for Cl-/H+ exchange ( Lange et al . , 2006; Leisle et al . , 2011 ) . OSTM1 is predicted to be a glycosylated , single-pass transmembrane protein and mutations in OSTM1 , like mutations in CLC-7 , can lead to osteopetrosis and neurodegeneration in humans and mice ( Chalhoub et al . , 2003; Kasper et al . , 2005; Kornak et al . , 2001; Lange et al . , 2006; Majumdar et al . , 2011; Pressey et al . , 2010 ) . However , the mechanisms by which OSTM1 and CLC-7 cooperate to enable proper ion transport remains an open question . To begin to understand the mechanisms of CLC-7 function and its unique requirement for OSTM1 , we have determined electron cryomicroscopy ( cryo-EM ) structures of CLC-7 and of a CLC-7/OSTM1 complex .
Following an evaluation of multiple CLC-7 orthologues , we decided to focus our structural studies on the chicken and human CLC-7 proteins based on their expression levels and their biochemical stabilities . Full-length chicken CLC-7 ( ggCLC-7 ) and human CLC-7 , which are 86 . 4% identical , were expressed in HEK293S GnTI- cells as mEGFP-fusions , purified to homogeneity in the detergent lauryl maltose neopentyl glycol ( LMNG ) , cholesterol hemisuccinate ( CHS ) , 150 mM KCl and 50 mM Tris-HCl pH 8 . 0 , and analyzed by cryo-EM . Vitrified human CLC-7 transporters displayed a strongly preferred orientation that was confirmed by two-dimensional classification ( Figure 1—figure supplement 1 ) . Because of the very limited views of the transporter , we were not able to reconstruct a three-dimensional density map of human CLC-7 . In contrast , two-dimensional classification of ggCLC-7 revealed a wide range of views and was suitable for three-dimensional structure determination ( Figure 1—figure supplement 2 ) . Three-dimensional classification of the imaged ggCLC-7 transporters identified a single class that displayed both well-ordered transmembrane and cytosolic domains . Reconstruction of these particle images with twofold symmetry imposed yielded a structure of dimeric ggCLC-7 at a resolution of 2 . 9 Å that enabled model building ( Figure 1A , Figure 1—figure supplement 2 , Figure 1—figure supplement 3 and Table 1 ) . The final refined model , which lacks the disordered N- and C-termini , fits well into the density with good geometry ( Figure 1B and Figure 1—figure supplement 2 , Figure 1—figure supplement 3 and Table 1 ) . Each protomer of dimeric ggCLC-7 contains a transmembrane domain composed of 18-transmembrane helices and a cytoplasmic domain composed of an N-terminal domain and two C-terminal cystathionine β-synthase ( CBS ) domains ( Figure 1C ) . Both the transmembrane and cytosolic domains contribute to the large ( ~3700 Å2 ) ggCLC-7 dimer interface ( Figure 1D ) . The transmembrane domain of ggCLC-7 adopts the canonical CLC architecture with each protomer possessing discrete ion permeation pathways that extend from the cytosol to the lysosomal lumen ( Figure 2A ) . Structural and functional analysis of CLC transporters and channels have defined the Cl--conduction pathway and its three conserved Cl--binding sites ( Accardi et al . , 2004; Accardi and Miller , 2004; Accardi et al . , 2005; Basilio et al . , 2014; Chavan et al . , 2020; Dutzler et al . , 2002; Dutzler et al . , 2003; Feng et al . , 2010; Feng et al . , 2012; Park et al . , 2017; Park and MacKinnon , 2018; Picollo et al . , 2012; Wang et al . , 2019; Zdebik et al . , 2008 ) , and this architecture is well-preserved in ggCLC-7 . In the present conformation of ggCLC-7 , constrictions too narrow to accommodate Cl- ions exist on both ends of the Cl--conduction pathway ( Figure 2B , C ) . On the cytosolic side of the pathway between the central and internal Cl--binding sites , the side chains of Ser200 , Tyr501 and Tyr598 form a constriction with a minimum radius of 0 . 6 Å ( Figure 2B ) . The luminal side of the Cl--conduction pathway contains two additional constrictions ( Figure 2C ) . The first constriction , which has a minimum radius of 0 . 7 Å is immediately adjacent to the external Cl--binding site and is formed by the side chains of Glu243 , Ile511 and the backbone of Lys242 and Glu243 . The second constriction ( 1 . 1 Å minimum radius ) is at the luminal entrance and is formed by the side chains of Lys242 , Glu467 and the backbone of Gly241 . Together the three constrictions yield an occluded state for the transporter , sealing off the external and central Cl--binding sites from the cytosol and the lysosomal lumen . The most dynamic residue in the ion transport pathways of CLC transporters is Glugate , whose conformation changes during the transport cycle . In previous structures of CLC transporters , the side chain carboxylic moiety of Glugate has occupied four different positions: ‘middle’ where it occupies the central Cl--binding site , ‘up’ where it moves toward the extracellular vestibule , ‘down’ where it occupies the central Cl- binding site and most recently ‘out’ where it reaches away from the Cl--conduction pathway toward the H+-conduction pathway ( Chavan et al . , 2020; Dutzler et al . , 2002; Dutzler et al . , 2003; Feng et al . , 2010; Last et al . , 2018; Figure 2—figure supplement 1 ) . In ggCLC-7 , the Glugate ( Glu243 ) adopts the ‘up’ conformation , where it participates in establishing one of the luminal constrictions ( Figure 2D and Figure 2—figure supplement 1 ) . A non-protein density was resolved between Glugate and Glu467 that we assigned as a water molecule . This water may help to stabilize the conformations of Glugate and Glu467 , which may both be protonated at pH 8 . 0 . Within the Cl--conduction pathway , non-protein densities that we attributed to Cl- ions were resolved at the external , central and internal Cl--binding sites ( Figure 2D ) . The external Cl- site is formed by the backbone nitrogens of Glu243 and Gly244 on helix αF and Phe510 and Ile511 on helix αO . The intensity of the external Cl--binding site ( ~14 σ ) is the strongest of the three Cl--binding sites and is nearly equivalent to that of backbone atoms of nearby residues , suggesting a high Cl- occupancy . The density for the central Cl- site has a slightly lower intensity ( ~12 σ ) and is formed by the side chains of the highly conserved Ser200 from helix αD and Tyr598 on helix αS and the backbone nitrogens of Val509 and Phe510 of helix αN . The internal Cl--binding site , which is located in a solvent-exposed vestibule on the cytoplasmic side of the transporter , has the lowest intermediate intensity ( ~8 σ ) and is formed by the backbone nitrogens of Ser200 and Gly201 and the side chain of Gln204 , all on helix αD . The relative intensities of the three Cl--binding sites are consistent with structural and biochemical studies performed with E . coli CLC1 ( ecCLC ) that showed that the central and external binding sites have much higher affinity for Cl- ions than the internal site ( Lobet and Dutzler , 2006; Picollo et al . , 2009 ) . Near the center of the transporter , the H+-conduction pathway of CLC transporters diverges from the Cl--conduction pathway as they approach the cytosolic side of the transporter ( Accardi et al . , 2005; Chavan et al . , 2020; Han et al . , 2014; Leisle et al . , 2020; Park and MacKinnon , 2018; Wang and Voth , 2009; Zdebik et al . , 2008 ) . This bifurcation occurs near the central Cl--binding site and is proposed to extend through a hydrophobic gap to the conserved Gluin ( Chavan et al . , 2020 ) . In the ggCLC-7 structure , Gluin ( Glu310 ) on helix αG is located more than 15 Å away from Glugate , where it extends into a solvent filled cavity between the transmembrane and cytosolic domains that is continuous with the cytosol ( Figure 2E , F ) . Within the loosely packed hydrophobic gap between Glugate and Gluin , several non-protein densities were resolved that we have tentatively modeled as water molecules ( Figure 2F ) . Water molecules have previously been detected within the hydrophobic gap in structures and in molecular dynamics simulations of CLC transporters and have been proposed to serve as a proton-conducting water-wire ( Chavan et al . , 2020; Han et al . , 2014; Leisle et al . , 2020; Wang and Voth , 2009 ) . In ggCLC-7 , the water molecules in the hydrophobic gap can access the Cl--conduction pathway through an opening with a minimum radius of ~1 . 4 Å between Gly244 , Phe297 and Phe510 ( Figure 2F ) . However , the pathway is not continuous with the cytosol as the hydrophobic gap is sealed near Gluin by a 1 . 0 Å constriction formed Ile202 , Phe306 and Met558 . In a recent structure of a mutant of ecCLC , the constrictions between Glugate and Gluin were both expanded , creating a continuous pathway that would facilitate H+ conduction ( Chavan et al . , 2020 ) . It is possible that a similar conformational change may occur during the transport cycle of ggCLC-7 to open the constrictions and allow protons to pass through the hydrophobic gap . The cytoplasmic domain of CLC-7 is composed of the N-terminal domain and the two C-terminal CBS domains ( Figure 3A ) . The N-terminal domain , which has not been resolved in previous CLC structures , is comprised of a 14-amino acid extended segment and helix αA that are well-defined in the ggCLC-7 density map ( Figure 3B and Figure 1—figure supplement 3 ) . The extended segment is positioned at the center of the three-way interface between the transmembrane domain , CBS1 and CBS2 . Because of its central position , the N-terminal domain is a major contributor to the tertiary and quaternary structure of ggCLC-7 . Indeed , the N-terminal domain forms a larger interface with the transmembrane domain than either of the CBS domains . Immediately adjacent to the N-terminal domain in a groove between the two CBS domains is a large density that cannot be attributed to the protein ( Figure 3C ) . We modeled this non-protein density as a Mg2+-bound ATP based on its shape and a comparison with the ATP-bound structure of the isolated CBS domains of CLC-5 ( Meyer et al . , 2007; Figure 3—figure supplement 1 ) . Notably , no nucleotides were added during the 30-hour purification of ggCLC-7 so any ATP present must have been co-purified with the transporter . In the ggCLC-7 structure , the transporter forms multiple interactions with all three components of the ATP molecule ( Figure 3C ) . The adenine base of the ATP is sandwiched between the side chains of His654 on CBS1 , with which it forms π-stacking interactions , and Met778 on CBS2 . The adenine base also forms polar interactions with the side-chain oxygen and backbone nitrogen of Thr632 and the backbone nitrogen of Gly656 that contribute to the specificity for adenine nucleotides ( Meyer et al . , 2007 ) . The ribose sugar forms polar interactions with side chains of Ser628 and Asp783 . The triphosphate group is coordinated by residues from both CBS domains as well as the N-terminal domain . The α-phosphate interacts with the side chain of His654 and the backbone oxygen of Asn653 of CBS1 , the side chain of Lys782 of CBS2 . Coordinating the α- and β-phosphate is a Mg2+ that is partially coordinated by Glu91 of the N-terminal domain . Additionally , the β-phosphate also interacts with the side chain of Asn655 and the backbone oxygen of His654 of CBS1 and the side chain of His765 of CBS2 . The γ-phosphate interacts with the side chain and backbone nitrogen of Ser92 of the N-terminal domain and the side chain of Arg764 of CBS2 . Together , the numerous interactions between ATP and ggCLC-7 and the slow-off rate of ATP during the purification are consistent with ATP binding to ggCLC-7 with high affinity . The ATP-binding site in ggCLC-7 shares many features with the ATP-binding site resolved in the structure of the isolated CBS domains of CLC-5 , including the coordination of the adenine base and the ribose sugar ( Figure 3—figure supplement 1; Meyer et al . , 2007 ) . The major differences between the ATP interactions of human CLC-5 and ggCLC-7 are the additional interactions with the triphosphate group of ATP in ggCLC-7 . In particular , both ATP and Mg2+ directly interact with the N-terminal domain of CLC-7 ( Figure 3C and Figure 3—figure supplement 1 ) . As the N-terminal domain was not present in the CLC-5 CBS domain crystallization construct ( Meyer et al . , 2007 ) , such interactions were not previously identified . Binding studies performed with the CLC-5 CBS domains found that ATP binds with an affinity of ~100 μM , similar to affinities measured for ADP and AMP ( Meyer et al . , 2007 ) . In contrast , the densities corresponding to the α-phosphate and γ-phosphate are nearly equivalent in the ggCLC-7 density map , indicating that ATP was predominant species co-purified with transporter ( Figure 3C ) . Sequence alignment of ggCLC-7 with the nine human CLC proteins reveals the N-terminal domain resolved in the ggCLC-7 structure ( residues 87–114 ) is conserved in human CLC-7 as well as the closely-related human CLC-6 transporter ( 45% identity ) . The conservation is less clear in the more distantly related CLC-3 , CLC-4 and CLC-5 transporters and no conservation is apparent in the CLC-1 , CLC-2 , CLC-Ka and CLC-Kb channels ( Figure 3B ) . Furthermore , Glu91 and Ser92 , the residues that interact with the triphosphate group of ATP in ggCLC-7 , are only conserved in CLC-6 and CLC-7 , indicating that ATP binding may vary among CLC proteins . We therefore speculate that ATP is the preferred ligand for CLC-7 and that further studies resolving the N-terminal domain of other CLC transporters will reveal to what extent CLC transporters bind specific adenine nucleotides . Non-protein densities that likely correspond to either ordered lipids or detergents were resolved around the periphery of the transmembrane domain of ggCLC-7 ( Figures 1A and 4A , B ) . Because it is difficult to distinguish lipids from detergents based on cryo-EM density maps alone , we were able to assign only one of the densities . The well-resolved head group allowed us to model the density as a phosphoinositol-3-phosphate ( PI3P ) , which is a low-abundance constituent of lysosomal membranes ( Figure 4A ) . Similar to ATP , PI3P was co-purified with the transporter from the HEK293S GnTI- cell membranes . The PI3P molecule is located at the interface between the transmembrane domain and the cytosolic domain and interacts with residues from both domains . An amphipathic β-hairpin between helices αF and αG containing multiple positively charged residues surrounds the inositol phosphate head group and glycerol backbone , separating them from the rest of the membrane ( Figure 4A–C ) . The head group also interacts with Lys213 , Arg219 and Arg714 , which coordinate the phosphate group at the 3 position of inositol ring , and with Val218 , Leu220 , Val252 , Ala255 , Ser262 , Thr263 and Lys281 . Two 8-carbon acyl chains were modeled into PI3P density in a groove on the surface of the transporter formed by helices αB , αE and αF . In ggCLC-7 , PI3P is largely coordinated by helices αF and αG and the intervening β-hairpin ( Figure 4 ) . Among human CLCs , the elaborated loop between helices αF and αG present in ggCLC-7 is conserved in CLC-7 as well as CLC-6 . The loop is also present in the vacuolar nitrate/H+ antiporter CLC-a from Arabidopsis thaliana ( atCLC-a ) ( De Angeli et al . , 2006; Figure 4C ) . Based on the structure of ggCLC-7 and alignment of the sequences , three positively charged residues were identified within the β-hairpin of the transporters that may facilitate access of the negatively-charged lipid into the binding pocket ( Figure 4C ) . In addition , several other residues that participate in the coordination of the PI3P head group , including Lys281 and Arg714 from ggCLC-7 , are also conserved among hsCLC-6 , hsCLC-7 and atCLC-a . Together , these data suggest that the PI-binding site may be a conserved feature among a subset of CLC transporters . Unlike the other mammalian CLC transporters expressed in endosomes and lysosomes , CLC-7 is not active by itself . CLC-7 activity is dependent on the presence of its β-subunit , OSTM1 ( Lange et al . , 2006; Leisle et al . , 2011 ) . To better understand the role of OSTM1 in CLC-7-mediated Cl-/H+ exchange , we next co-expressed human CLC-7 and human OSTM1 in HEK293S GnTI- cells , purified the complex to homogeneity and analyzed its structure by cryo-EM . Similar to vitrified human CLC-7 by itself , human CLC-7/OSTM1 particles also adopted a preferred orientation in the ice . However , the effect was less severe for the CLC-7/OSTM1 particles and by collecting a large data set , we were able to resolve additional views . Two-dimensional and three-dimensional classification revealed the presence of intact CLC-7/OSTM1 complexes in the data set as well as a minor population of free CLC-7 dimers ( Figure 5—figure supplement 1 ) . Due to a low abundance and a preferred orientation of free CLC-7 particles , structural reconstitution of the CLC-7 homodimer was not possible . By employing a hierarchical classification approach , we were able to identify a population of intact CLC-7/OSTM1 complexes in which the cytoplasmic , transmembrane and luminal domains were all clearly resolved . Reconstruction of these particle images with two-fold symmetry imposed yielded a structure of CLC-7 in complex with OSTM1 at a resolution of 2 . 8 Å ( Figure 5A , Figure 5—figure supplement 1 and Table 1 ) . 3D variability analysis of the selected particles revealed that the luminal domain of OSTM1 is flexibly attached to the transmembrane domain and adopts a range of different orientations . We observed up to a 6 Å displacement of the peripheral regions of the luminal domain of OSTM1 when the different states were aligned by their transmembrane domains ( Figure 5—figure supplement 2 ) . We therefore applied masks and performed local refinements , which yielded separate density maps at resolutions between 2 . 8 and 3 . 1 Å with improved interpretability for the transmembrane and cytosolic domains of CLC-7 and OSTM1 and the luminal domain of OSTM1 ( Figure 5—figure supplement 1 and Table 1 ) . Despite the preferred orientation of the raw data set , the focus refined reconstructions determined with the selected particles display only minimal anisotropy and , following merging into a single composite map , were suitable for model building and coordinate refinement ( Figure 5B , Figure 5—figure supplement 1 , Figure 5—figure supplement 3 and Table 1 ) . The final refined structure contains two symmetrical copies of both CLC-7 and OSTM1 that fit well into the density with good geometry . When viewed from the side , the two copies of OSTM1 wrap around three sides of the CLC-7 dimer ( Figure 5 ) . The luminal domains of OSTM1 form a dimeric cap-like structure that covers the luminal surface of CLC-7 while the transmembrane helices pack against the periphery of the CLC-7 transmembrane domain . The C-terminal cytoplasmic domain of OSTM1 is disordered and no cytoplasmic interactions with CLC-7 were resolved . The luminal domain of OSTM1 is composed of two three-helix bundles ( Figure 6A ) . Connecting both within and between the helical bundles are five disulfide bonds that constrain the organization of the luminal domain . In the first bundle , a disulfide bond connects helix 1 to the short helix 3 . In the second bundle , disulfide bonds connect helix 6 to helix 5 and to the linker between helix 7 and transmembrane helix 8 . Between the two bundles , disulfide bonds connect the linker between helices 4 and 5 to helix 2 of the first bundle and to helix 7 of the second bundle ( Figure 6A , B ) . The two helical bundles create a large dimer interface that buries ~4600 Å2 of shared surface area ( Figure 6C ) . The core of the dimer interface is formed by an antiparallel packing of helices 1 and 4 with helices 3 and 7 and several of the inter-helical linkers also making substantial contributions . While most of the interactions that stabilize the OSTM1 dimer interface are hydrophobic including the entirety of the helix 1-helix 1 and helix 3-helix 3 interactions , several polar interactions are also present including an ionic interaction between Arg107 and Asp150 ( Figure 6D ) . At the periphery of the luminal domain , non-protein densities were resolved extending from seven exposed asparagine residues ( 93 , 128 , 163 , 184 , 194 , 263 and 274 ) on OSTM1 ( Figure 6A ) . As previous computational analysis had identified these residues as consensus sites for N-linked glycosylation ( Lange et al . , 2006 ) , we modeled these non-protein densities as carbohydrate moieties . The quality and interpretability of the carbohydrate densities varied between the seven sites , allowing us to model chains of different length . For example , density for a single N-linked N-acetyl-glucosamine group was resolved for Asn93 and Asn163 , while a branched five-sugar carbohydrate moiety was resolved for Asn263 ( Figure 5—figure supplement 3 ) . While only minimal carbohydrate moieties can be added to N-linked glycosylation sites in the HEK293S GnTI- cell line used for protein expression due to a mutation in N-acetyl-glucosaminyltransferase I , in non-glycoslyation-defective mammalian cells these glycosylation sites would be elaborately decorated and likely encase the entire surface of the luminal domain . Because CLC-7 lacks any N-linked glycosylation sites , the glycosylation shell surrounding the rigid , disulfide-linked core of OSTM1 likely protects the luminal domain of CLC-7 from the harsh degradative environment of the lysosomal lumen . We next compared the structure of ggCLC-7 with the structure of the human CLC-7/OSTM1 complex to determine how OSTM1 binding influences the conformation of CLC-7 . Overall , the CLC-7 dimers show good alignment ( RMSD 0 . 4 Å ) in the presence and absence of OSTM1 ( Figure 7A ) . The cytosolic domains and most of the transmembrane domains are essentially identical . Moreover , densities corresponding to ATP and PI3P molecules were resolved in their respective binding sites and the ligands interact with CLC-7 in a similar fashion regardless of the presence or absence of OSTM1 ( Figure 5A , Figure 5—figure supplement 3 and Figure 5—figure supplement 4 ) . The only detectable rearrangements in CLC-7 occur near interfaces where CLC-7 directly contacts OSTM1 ( Figure 7A ) . The largest CLC-7/OSTM1 interface is formed between helix 8 of OSTM1 and the transmembrane domain of CLC-7 ( Figure 7B ) . Binding of OSTM1 is accompanied by a bend in helix αB of CLC-7 at Gly149 that results in 9 Å shift of the luminal end of helix αB ( measured at Cα of Glu168 ) toward helix αK ( Figure 7—figure supplement 1 ) . Small ( <2 Å ) movements are resolved in the luminal ends of the nearby helices αC and αK toward OSTM1 helix 8 . The rearrangements in helices αB , αC and αK of CLC-7 allow helix 8 of OSTM1 , which is slightly kinked near Pro296 , to pack against the surface of CLC-7 ( Figure 7B ) . The interaction between transmembrane domains is largely mediated by the packing of hydrophobic residues , but polar interactions are formed between Tyr300 of OSTM1 and Glu416 and Trp503 of CLC-7 that may contribute to the specificity of the CLC-7/OSTM1 interaction ( Figure 7C ) . OSTM1 also contacts and stabilizes the linker between helices αK and αL , which was too poorly ordered to be modeled in absence of OSTM1 in the ggCLC-7 density map ( Figure 7A , D ) . In the CLC-7/OSTM1 complex , the 25-residue linker between helices αK and αL forms two interactions with the luminal domain of OSTM1 . Asp456 of CLC-7 forms a polar interaction with Tyr228 of helix 6 . Gly443 and Gly444 of CLC-7 form a small interface with a portion of the loop between helices 6 and 7 that includes Pro250 , Gly251 and His253 . In addition to the direct protein-protein interactions , a non-protein density that may correspond to a cholesterol based on its size and shape was resolved at the interface between the transmembrane domains of CLC-7 and OSTM1 ( Figure 5—figure supplement 3 ) . Notably , no density was present at this site in the ggCLC-7 map , suggesting that CLC-7 and OSTM1 together may form an additional lipid-binding site . To examine the influence of OSTM1 on Cl- and H+ transport , we compared the ion conduction pathways in the CLC-7/OSTM1 and ggCLC-7 structures ( Figure 2 , Figure 7E , Figure 7—figure supplement 2 and Figure 7—figure supplement 4 ) . In both structures , the Cl--conduction pathways adopt similar occluded states with narrow constrictions present at either end and ions occupying the three binding sites . Superpositioning reveals that most of the pore-lining residues are positioned similarly in the presence and absence of OSTM1 . The only residues that adopt differing conformations are Phe301 ( Phe297 in ggCLC-7 ) and Phe514 ( Phe510 in ggCLC-7 ) , both of which are highly conserved among CLC family members and whose mutation in human CLC-7 leads to defects in Cl-/H+ exchange coupling , voltage-dependence and activation ( Leisle et al . , 2020 ) . Inspection of the CLC-7/OSTM1 density map reveals that the side-chain density for Phe514 is distorted in a manner consistent with the side chain adopting both the modeled conformation ( conformation 1 ) as well as a conformation similar to that resolved for Phe510 in ggCLC-7 ( conformation 2 ) ( Figure 7—figure supplement 3 ) . Inspection of the ggCLC-7 density map revealed no density consistent with the alternative rotamer , indicating that conformation two is the predominant state for Phe510 . While the existence of the two conformations for Phe514 in CLC-7/OSTM1 is clearer in a map sharpened to emphasize the high-resolution features , due to the limited resolution and the anisotropy present in the data , it is difficult to estimate occupancy of the different rotamers . Thus , while we modeled Phe514 as the predominant conformation 1 , the data also supports the existence of conformation 2 . Because Phe514 is located at the interface between the hydrophobic gap and the central Cl- binding site , changes in its conformation can modify the H+-conduction pathway . In CLC-7/OSTM1 , the side chain of Phe514 ( conformation 1 ) narrows the constriction of this pathway to a minimum radius of 0 . 8 Å , which is too narrow to allow water molecules to enter the hydrophobic gap ( Figure 7—figure supplement 4 ) . In contrast , the alternative conformation adopted by Phe510 in the ggCLC-7 structure widens the pathway sufficiently to allow water molecules to cross ( 1 . 4 Å minimum radius ) ( Figure 2F ) . We next compared the relative intensities of the ion binding sites to assess the effect on OSTM1 binding on Cl- binding in the permeation pathway . In ggCLC-7 , the external and central sites exhibit strong densities that are only slightly weaker than nearby protein atoms , indicating a high occupancy for Cl- ions at these sites , while the density at the internal Cl-binding site is significantly weaker ( Figure 2D ) . The relative order of intensities differs in the CLC-7/OSTM1 structure ( Figure 7—figure supplement 2 ) . In CLC-7/OSTM1 , the density at the external site is the strongest and has a similar intensity as nearby protein atoms ( ~14 σ ) . However , unlike in ggCLC-7 , the central site in CLC-7/OSTM1 is the weakest and is only slightly above the background ( ~4 σ ) . While we must be cautious in interpreting the densities occupying the Cl--binding sites of CLC-7/OSTM1 because of its anisotropic nature , the differences in relative intensities of the Cl--binding site peaks between ggCLC-7 and CLC-7/OSTM1 suggest that there may be a change in Cl- occupancy of the central site when CLC-7 is bound to OSTM1 . A change in occupancy of the central Cl- site may be associated with the different conformation of Phe510/Phe514 , which is located ~4 Å from the central Cl- site in both structures . Such as association would be consistent with molecular dynamics simulations performed using ecCLC that identified a coupling between Cl- occupancy at the central site and the conformation of Phe297 , which is equivalent to Phe514 in human CLC-7 ( Leisle et al . , 2020 ) . Together , these data indicate that OSTM1 binding does not greatly perturb the conformation of the ion conduction pathways in CLC-7 and that its influence on CLC-7 transport activity does not occur through large-scale rearrangements . Rather , these data suggest that OSTM1 binding can potentially induce subtle conformational changes in key residues and provide critical structural support for CLC-7 . Moreover , by virtue of its heavy glycosylation , OSTM1 can protect the un-glycosylated CLC-7 from degradation in the acidic lysosomal lumen .
In this study , we present structures of the lysosomal Cl-/H+ exchanger CLC-7 alone and in complex with its obligatory β-subunit OSTM1 . The structure of the CLC-7/OSTM1 complex reveals that OSTM1 forms a heavily-glycosylated cap that covers the luminal surface of CLC-7 ( Figures 5 and 6 ) . OSTM1 associates with CLC-7 largely through interactions mediated by the transmembrane domains , consistent with analyses that demonstrated that deletion of the transmembrane domain of OSTM1 phenocopies the Ostm1 null in mice ( Pata and Vacher , 2018 ) . When complexed with CLC-7 , OSTM1 does not adopt the structure of a RING finger domain as had previously been suggested ( Fischer et al . , 2003 ) . Instead , the luminal domain of OSTM1 forms a tightly packed core composed of helical bundles linked together by numerous disulfide bonds ( Figure 6 ) . This stable core , together with the glycosylated periphery make the luminal domain of OSTM1 well-suited to survive the harsh degradative environment of the lysosomal lumen . In contrast , CLC-7 alone among the human CLC transporters lacks any N-glycosylation sites of its own and is consequently unstable in the lysosome when expressed in the absence of OSTM1 ( Lange et al . , 2006 ) . The structure of CLC-7/OSTM1 is thus consistent with OSTM1 serving a protective role to shield CLC-7 from proteolysis and degradation . Comparison of the CLC-7 structures in the presence and absence of OSTM1 reveals that OSTM1 binding induces subtle changes to the conformations of the ion permeation pathways . Among the changes are the conformations of residues essential for proper transport activity and the occupancies of the Cl--binding sites . Notably , these conformational changes appear to occur only in a subset of the CLC-7/OSTM1 complexes . It is therefore possible that OSTM1 binding alters the equilibrium between different CLC-7 conformations . However , the current data do not enable accurate modeling of alternative rotamers , and thus it is not possible to compare the fraction of CLC-7 transporters adopting each possible state in the presence and absence of OSTM1 . Moreover , we do not know precisely to which functional state these conformations correspond . In the CLC-7/OSTM1 and ggCLC-7 structures , the Cl--conduction pathways resemble those of the ecCLC E148Q mutant where the three Cl--binding sites are occupied and the Glugate adopts the ‘up’ conformation where it can potentially exchange protons with the lumen ( Dutzler et al . , 2003 ) . Because the ‘up’ conformation is a coordinate along the proposed transport cycle ( Feng et al . , 2012 ) , and because CLC-7 functionality has been previously detected in the absence of OSTM1 using solid-supported membranes and in plant vacuoles ( Costa et al . , 2012; Schulz et al . , 2010 ) , it is possible that CLC-7 structures both in the presence and absence of OSTM1 represent states that are competent for Cl-/H+ transport activity . Notably , even in conditions where CLC-7 activity could be detected without OSTM1 , current levels were significantly increased by its co-expression , consistent with OSTM1 potentiating transport activity ( Schulz et al . , 2010 ) . Future investigations will thus be necessary to precisely determine how OSTM1 stimulates CLC-7 activity - whether by inducing conformational changes in the ion conduction pathways or merely stabilizing lysosomal expression of CLC-7 . Initial characterizations of CLC-7/OSTM1 demonstrated that its activity is dependent on membrane potential and luminal pH ( Leisle et al . , 2011; Ludwig et al . , 2013 ) . Our structures reveal the presence of ATP and phosphatidylinositol-binding sites , suggesting that additional signals may also regulate CLC-7 activity . Indeed , ATP has been demonstrated to influence the activity of multiple CLCs , but the precise effect of ATP on transporter activity has been controversial with evidence supporting both stimulatory and inhibitory roles . For example , addition of ATP increased transporter activity of CLC-4 by two-fold but inhibited activity of CLC-2 channels ( Stölting et al . , 2013; Vanoye and George , 2002 ) . Moreover , the particular adenine nucleotide species that can influence CLC activity has been unclear . Binding studies conducted with the CBS domains of CLC-5 detected affinities of ~100 µM for ATP , ADP and AMP ( Meyer et al . , 2007 ) . Recent studies revealed that CLC-3 , CLC-4 and CLC-5 are able to distinguish between different nucleotide moieties , and showed that Mg2+ ions modify the effect of ADP binding ( Grieschat et al . , 2020 ) . In the structures of ggCLC-7 and CLC-7/OSTM1 , we observe direct coordination of all three phosphates not only through interactions with the CBS domains but also with the N-terminal domain ( Figure 3 and Figure 3—figure supplement 1 ) . Based on our data , we suggest that the actual binding affinity of CLC-7 to ATP is much higher than that detected for the CBS domains alone . As ATP levels are in excess of 1 mM under physiological conditions , it is likely that ATP is constitutively bound to CLC-7 and may therefore serve a structural role rather than a regulatory role . While ATP is a regulatory factor for numerous proteins , a structural role for nucleotides has been previously described for AMP-activated protein kinase ( AMPK ) , inositol 1 , 4 , 5-trisphosphate ( IP3 ) receptors and some prokaryotic regulator of potassium conductance ( RCK ) -gated channels ( Bezprozvanny and Ehrlich , 1993; Cao et al . , 2013; Hardie and Hawley , 2001; Kong et al . , 2012; Kröning et al . , 2007; Teixeira-Duarte et al . , 2019 ) . Among the CLC family , the residues in the N-terminal domain of CLC-7 that interact with ATP and Mg2+ are not broadly conserved , suggesting that the structural divergence of the ATP-binding site may contribute to varied effects that been reported among CLC family members . Dysregulated ATP binding to CLC-7 may play a role in human disease . While CLC-7 remains fully functional in the absence of ATP ( Leisle et al . , 2011 ) , mapping disease mutations onto CLC-7/OSTM1 reveals a hotspot of mutations on CBS2 near the ATP-binding site ( Figure 8 ) . Several distinct mutations of Arg767 , which directly participates in the ATP γ-phosphate coordination , as well as mutations of neighboring Gly765 and Leu766 have been identified as leading to osteopetrosis ( Cleiren et al . , 2001; Leisle et al . , 2011; Sartelet et al . , 2014 ) . Previous work characterizing the function of the Arg767 mutants revealed distinct phenotypes , with the R767P and R767W mutants displaying no activity while the R767Q mutants displayed increased activation kinetics ( Leisle et al . , 2011 ) . Together , these results indicate that while ATP may be constitutively bound and serve a structural role , disruption of the binding site has functional consequences for CLC-7 . The second ligand co-purified with ggCLC-7 and CLC-7/OSTM1 is PI3P , a phosphatidylinositol ( PI ) lipid species enriched in endolysosomal membranes that constitutes between 0 . 1% and 0 . 5% of the total PI content in cells ( Falasca and Maffucci , 2006 ) . While PI3P has not been previously characterized as a regulator of ion transport protein activity , the related phosphatidylinositol 3 , 5-bisphosphate ( PI ( 3 , 5 ) P2 ) is potent modulator of ion transport proteins whose abundance is tightly regulated ( Fine et al . , 2018; Hasegawa et al . , 2017; She et al . , 2018; She et al . , 2019 ) . Under basal conditions , PI ( 3 , 5 ) P2 concentrations are very low ( <0 . 1% of cellular PI content ) , but in yeast can rise more than 20-fold upon hyperosmotic shock ( Duex et al . , 2006 ) . In endosomes and lysosomes , PI ( 3 , 5 ) P2 binding activates the Ca2+ channel TRPML1 ( Dong et al . , 2010 ) and the Na+ channel TPC1 ( Wang et al . , 2012 ) . In plant vacuoles , which share many features with lysosomes , PI ( 3 , 5 ) P2 potently regulates atCLC-a , inhibiting its activity with an IC50 of ~10 nM ( Carpaneto et al . , 2017 ) . Because the PI3P-binding site appears to be conserved between atCLC-a and CLC-7 , we modeled in a PI ( 3 , 5 ) P2 lipid into the binding site in the CLC-7/OSTM1 structure to gain insights into its effect ( Figure 4—figure supplement 1 ) . In the model , a phosphate at 5-position of inositol ring could not be accommodated due to steric clashes with Lys281 on helix αG . We therefore speculate that binding of the regulatory PI ( 3 , 5 ) P2 may induce conformational changes to CLC-7 that may alter its activity , analogous to the inhibitory effect of PI ( 3 , 5 ) P2 on atCLC-a . Consistent with the PI lipids influencing transporter activity , a mutation of Tyr715 , which is located near the PI3P-binding site , to cysteine was recently identified in the gene encoding human CLC-7 that causes a novel lipid storage disease without osteopetrosis ( Undiagnosed Diseases Network et al . , 2019 ) . Functional analysis of this mutant revealed that it displays increased current levels when expressed in oocytes compared to wild-type CLC-7 and leads to a hyper-acidification phenotype in lysosomes ( Figure 8; Undiagnosed Diseases Network et al . , 2019 ) . For decades CLC-7 was perhaps the most enigmatic CLC family member ( Brandt and Jentsch , 1995 ) . Functional and structural characterization was limited by its lysosomal expression and its absolute requirement for a β-subunit , OSTM1 . OSTM1 has been shown to have dual functions , both stabilizing CLC-7 in the lysosome and serving as an essential activator of the transporter ( Lange et al . , 2006; Leisle et al . , 2011; Stauber and Jentsch , 2010 ) . Our studies reveal how OSTM1 interacts with CLC-7 protecting the transporter from the acidic environment of the lysosomal lumen and lay the groundwork for future studies to elucidate how it , as well as ATP and the lipids identified in the structures , regulate CLC-7 transport activity and contribute to pH homeostasis in the lysosome and osteoclasts .
The gene encoding CLCN7 from Gallus gallus was synthesized ( SynBio ) and subcloned into a BacMam expression vector with a C-terminal mEGFP-tag fused via a short linker containing a PreScission protease site ( Goehring et al . , 2014 ) . The plasmid was mixed with PEI 25K ( Polysciences , Inc ) at a 1 : 3 ratio for 30 min and then used to transfect HEK293S GnTi– cells ( ATCC: CRL-3022 ) . For a 1 L cell culture 1 mg plasmid and 3 mg PEI 25K were used . After 24 hr incubation at 37 °C , sodium butyrate ( Sigma ) was added to a final concentration of 10 mM , and cells were allowed to grow at 37 °C for an additional 72 hr before harvesting . Cell pellets were washed in phosphate-buffered saline solution and flash frozen in liquid nitrogen . Membrane proteins were solubilized in 2% lauryl maltose neopentyl glycol ( LMNG , Anatrace ) , 0 . 2% cholesteryl hemisuccinate tris salt ( CHS , Anatrace ) , 20 mM HEPES pH 7 . 5 , 150 mM KCl supplemented with protease-inhibitor cocktail ( 1 mM PMSF , 2 . 5 μg/mL aprotinin , 2 . 5 μg/mL leupeptin , 1 μg/mL pepstatin A ) and spatula of DNaseI for 1 hr . Solubilized proteins were separated by centrifugation 75 , 000 g for 40 mins , followed by binding to 2 . 5 ml anti-GFP nanobody resin for 1 hr , which was equilibrated with washing buffer containing 0 . 1% LMNG , 50 mM Tris-HCl pH 8 , 150 mM KCl , 2 mM DTT ( BufferA ) . Anti-GFP nanobody affinity chromatography was performed by 20 column volumes of washing with BufferA , followed by overnight PreScission digestion , and elution with wash buffer . Protein sample was concentrated to a volume of 500 µl using CORNING SPIN-X concentrators ( 100 kDa cutoff ) , followed by centrifugation 10 , 000 g for 10 mins . Concentrated proteins were further purified by size exclusion chromatography on a Superose 6 Increase 10/300 GL ( GE healthcare ) in BufferA . Peak fractions were pooled and concentrated to ~2 mg/mL using CORNING SPIN-X concentrators ( 100 kDa cutoff ) . Genes encoding human CLCN7 and OSTM1 were synthesized ( SynBio ) and subcloned into BacMam expression vectors with C-terminal mCerulean- and mVenus- tags , respectively , fused via a short linker containing a PreScission protease site ( Goehring et al . , 2014 ) . Transient transfection was carried out as described above for chicken CLCN7 , with a single modification; for gene expression , valproic acid ( VPA , Sigma ) was added to induce expression at a final concentration of 2 . 2 mM . Equal amounts of plasmids encoding CLCN7 and OSTM1 were added to the reaction mix . Cell pellets were washed in phosphate-buffered saline solution and flash frozen in liquid nitrogen . Membrane proteins were solubilized in 2% lauryl maltose neopentyl glycol ( LMNG , Anatrace ) , 0 . 2% cholesteryl hemisuccinate tris salt ( CHS , Anatrace ) , 20 mM HEPES pH 7 . 5 , 150 mM KCl supplemented with protease-inhibitor cocktail ( 1 mM PMSF , 2 . 5 μg/mL aprotinin , 2 . 5 μg/mL leupeptin , 1 μg/mL pepstatin A ) and spatula of DNaseI for 1 hr . Solubilized proteins were separated by centrifugation 75 , 000 g for 40 min , followed by binding to 2 . 5 ml anti-GFP nanobody resin for 1 hr , which was equilibrated with washing buffer containing 0 . 01% LMNG , 50 mM Tris-HCl pH 8 , 150 mM KCl , 2 mM DTT ( BufferB ) . Anti-GFP nanobody affinity chromatography was performed by 20 column volumes of washing with BufferB , followed by overnight PreScission digestion , and elution with wash buffer . Protein sample was concentrated to a volume of 500 µl using CORNING SPIN-X concentrators ( 100 kDa cutoff ) , followed by centrifugation 10 , 000 g for 10 min . Concentrated proteins were further purified by size exclusion chromatography on a Superose 6 Increase 10/300 GL ( GE healthcare ) in BufferB . Peak fractions were pooled and concentrated to ~2 . 5 mg/mL using CORNING SPIN-X concentrators ( 100 kDa cutoff ) . For CLC-7 from Gallus gallus ( ggCLC-7 ) , 3 μl of purified protein at a concentration of 2 mg/ml was applied to glow-discharged Au 400 mesh QUANTIFOIL R1 . 2/1 . 3 holey carbon grids ( Quantifoil ) , and then plunged into liquid nitrogen-cooled liquid ethane with a FEI Vitrobot Mark IV ( FEI Thermo Fisher ) . Grids were transferred to a 300 keV FEI Titan Krios microscopy equipped with a K2 summit direct electron detector ( Gatan ) . Images were recorded with Leginon ( Suloway et al . , 2005 ) in super-resolution mode at 22 , 5000x , corresponding to pixel size of 0 . 536 Å . Dose rate was eight electrons/pixel/s , and defocus range was −1 . 2 to −2 . 5 µm . Images were recorded for 8 s with 0 . 2 s subframes ( total 40 subframes ) , corresponding to a total dose of 61 electrons/Å2 . For the CLC-7/OSTM1 complex from Homo sapiens ( hsCLC-7/OSTM1 ) , 3 μl of purified protein at a concentration of 2 mg/ml was supplemented with 1 mM ATP and 0 . 1% LMNG and was applied to glow-discharged Au 400 mesh QUANTIFOIL R1 . 2/1 . 3 holey carbon grids ( Quantifoil ) , and then plunged into liquid nitrogen-cooled liquid ethane with a FEI Vitrobot Mark IV ( FEI Thermo Fisher ) . Grids were transferred to a 300 keV FEI Titan Krios microscopy equipped with a K3 summit direct electron detector ( Gatan ) . Images were recorded with SerialEM ( Mastronarde , 2005 ) in super-resolution mode at 22 , 5000x , corresponding to pixel size of 0 . 532 Å . Dose rate was 13 electrons/pixel/s , and defocus range was −1 . 2 to −2 . 7 µm . Images were recorded for 4 s with 100 ms subframes ( total 40 subframes ) , corresponding to a total dose of 44 electrons/Å2 . 40-frame super-resolution movies ( 0 . 536 Å/pixel ) of ggCLC-7 were gain corrected , Fourier cropped by two and aligned using whole-frame and local motion correction algorithms by MotionCor2 ( Zheng et al . , 2017 ) ( 1 . 0723 Å/pixel ) . Whole-frame CTF parameters were determined using CTFfind 4 . 1 . 10 ( Rohou and Grigorieff , 2015 ) . Approximately 500 particles were manually selected to generate initial templates for autopicking that were improved by several rounds of two-dimensional classification in Relion 3 . 0 ( Scheres , 2016 ) , resulting in 6 , 542 , 536 particles . False-positive selections and contaminants were excluded from the data using multiple rounds of heterogeneous classification in cryoSPARC v2 ( Punjani et al . , 2017 ) using models generated from the ab initio algorithm in cryoSPARC v2 , resulting in a stack of 343 , 094 particles . Heterogeneous classification in cryoSPARC v2 was then used to identify 137 , 234 particles displaying both the transmembrane and cytosolic domains . After particle polishing in Relion and local CTF estimation and higher order aberration correction in cryoSPARC v2 , a reconstruction was determined at resolution of 2 . 9 Å by non-uniform refinement in cryoSPARC v2 ( Punjani et al . , 2019 ) . The final reconstruction was further improved by employing density modification on the two unfiltered half-maps with a soft mask in Phenix ( Terwilliger et al . , 2019 ) . 40-frame super-resolution movies ( 0 . 532 Å/pixel ) of hsCLC-7/OSTM1 complex were gain corrected , Fourier cropped by two and aligned using whole-frame and local motion correction algorithms by MotionCor2 ( 1 . 064 Å/pixel ) . Approximately 500 particles were manually selected to generate initial templates for autopicking that were improved by several rounds of two-dimensional classification in Relion and autopicking using Relion , resulting in 15 , 288 , 379 particles . False-positive selections and contaminants were excluded through iterative rounds of heterogeneous classification in cryoSPARC v2 using models generated from the ab initio algorithm in cryoSPARC v2 , resulting in a stack of 932 , 232 particles . Heterogeneous classification in cryoSPARC v2 was then used to identify 327 , 619 particles displaying the luminal , transmembrane and cytosolic domains . After particle polishing in Relion and local CTF estimation and higher order aberration correction in cryoSPARC v2 , a reconstruction was determined to 2 . 8 Å . 3D variability analysis in cryoSPARC v2 was then employed to characterize conformational heterogeneity ( Punjani and Fleet , 2020 ) . To interpret the results of the 3D variability analysis , CLC-7 and the luminal domain of OSTM1 were rigid-body docked into the two extreme states and the midpoint . Masks were generated for the luminal domain and the transmembrane and cytosolic domains that were used for local refinement . Local refinements yielded a reconstruction for the luminal domain at an estimated resolution of 3 . 0 Å , a reconstruction for the transmembrane domain at an estimated resolution of 2 . 9 Å and the cytosolic domain at an estimated resolution of 2 . 8 Å . The final reconstructions were then further improved by employing density modification on the two unfiltered half-maps with a soft mask in Phenix ( Terwilliger et al . , 2019 ) . A composite map was generated from the local refinement maps in Phenix that was used for model building and refinement ( Terwilliger et al . , 2019 ) . The structure of CLC from Cyanidioschyzon merolae ( cmCLC ) ( Feng et al . , 2010 ) was manually docked into the ggCLC-7 density map using chimera ( Pettersen et al . , 2004 ) . The model was then manually rebuilt according to the density using coot ( Emsley et al . , 2010 ) . Atomic coordinates were refined against the density modified map using phenix . real_space_refinement with geometric and Ramachandran restraints maintained throughout ( Adams et al . , 2010 ) . The refined ggCLC-7 structure was manually docked into the CLC-7/OSTM1 density map using Chimera ( Pettersen et al . , 2004 ) . The human CLC-7 model was manually rebuilt using COOT to fit the density . OSTM1 was manually built into the density by first placing poly-alanine helices and then using large side chains and glycosylation sites to register the helices . N-linked glycosylation trees were built and refined using the ‘carbohydrate’ module in COOT ( Emsley and Crispin , 2018 ) . Notably , the density for the carbohydrate residues was of poorer quality than the nearby protein and due to the difficulty in modeling carbohydrates in cryo-EM density maps ( Emsley and Crispin , 2018 ) they are less precisely modeled than the protein . Atomic coordinates were refined against the density modified map using phenix . real_space_refinement with geometric and Ramachandran restraints maintained throughout ( Adams et al . , 2010 ) . Figures were prepared with UCSF Chimera ( Pettersen et al . , 2004 ) , UCSF ChimeraX ( Goddard et al . , 2018 ) , MOLE ( Pravda et al . , 2018 ) , PyMol ( Schrödinger LLC , 2020 ) and Jalview ( Waterhouse et al . , 2009 ) . | Inside the cells of mammals , acidic compartments called lysosomes are responsible for breaking down large molecules and worn-out cells parts so their components can be used again . Similar to lysosomes , specialized cells called osteoclasts require an acidic environment to degrade tissues in the bone . Both osteoclasts and lysosomes rely on a two-component protein complex to help them digest molecules . Mutations in the genes for both proteins are directly linked to human diseases including neurodegeneration and osteopetrosis – a disease characterized by dense and brittle bones . For the main protein in this complex , called CLC-7 , to remain stable and perform its roles , it requires an accessory subunit known as OSTM1 . CLC-7 is a transporter that funnels electrically charged particles into and out of the lysosome , which helps to maintain the environment inside the lysosome compartment . However , due to the tight partnership between CLC-7 and OTSM1 , how they influence each other is poorly understood . To determine the roles of CLC-7 and OSTM1 , Schrecker et al . looked at the structure of the complex using a technique called single particle electron microscopy , which allows proteins to be visualized almost down to the individual atom . The analysis revealed that OSTM1 covers almost the entire inside surface of CLC-7 , protecting it from the acidic environment inside the lysosome and contributing to its stability . When the two subunits are bound together , OSTM1 also slightly changes the structure of the pore formed by CLC-7 , suggesting that OSTM1 may regulate CLC-7 activity . Schrecker et al . have laid the foundation for understanding more about the activity and regulation of CLC-7 and OSTM1 in lysosomes and osteoclasts . The structures described also help explain previous findings , including why OSTM1 is important for the stability of CLC-7 . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2020 | Cryo-EM structure of the lysosomal chloride-proton exchanger CLC-7 in complex with OSTM1 |
The genome-scale transcriptional programs that specify the mammalian trachea and esophagus are unknown . Though NKX2-1 and SOX2 are hypothesized to be co-repressive master regulators of tracheoesophageal fates , this is untested at a whole transcriptomic scale and their downstream networks remain unidentified . By combining single-cell RNA-sequencing with bulk RNA-sequencing of Nkx2-1 mutants and NKX2-1 ChIP-sequencing in mouse embryos , we delineate the NKX2-1 transcriptional program in tracheoesophageal specification , and discover that the majority of the tracheal and esophageal transcriptome is NKX2-1 independent . To decouple the NKX2-1 transcriptional program from regulation by SOX2 , we interrogate the expression of newly-identified tracheal and esophageal markers in Sox2/Nkx2-1 compound mutants . Finally , we discover that NKX2-1 binds directly to Shh and Wnt7b and regulates their expression to control mesenchymal specification to cartilage and smooth muscle , coupling epithelial identity with mesenchymal specification . These findings create a new framework for understanding early tracheoesophageal fate specification at the genome-wide level .
Proper specification of the trachea and esophagus is critical for the function of the respiratory and digestive systems . Formation of the trachea and esophagus requires signals from the splanchnic mesenchyme which specify ventral and dorsal domains of the foregut endoderm tube , resulting in lung bud outgrowth and the initiation of physical separation of the trachea and esophagus ( Billmyre et al . , 2015; Cardoso and Lü , 2006; Domyan et al . , 2011; Goss et al . , 2009; Harris-Johnson et al . , 2009; Morrisey et al . , 2013; Rankin et al . , 2018; Rankin et al . , 2016; Shannon et al . , 1998; Stevens et al . , 2017 ) . Ultimately , the tracheal epithelium is composed of multi-ciliated , secretory , and basal cells within a pseudostratified columnar monolayer which is surrounded by mesenchyme-derived ventral cartilaginous rings , and dorsal smooth muscle of the trachealis . The esophagus consists ultimately of a stratified squamous keratinized epithelium surrounded by smooth muscle . Early failure of tracheoesophageal ( TE ) fate specification can result in an unseparated common foregut endoderm tube , producing common congenital pathologies known as tracheoesophageal fistula ( TEF ) or tracheal agenesis ( TA ) ( Billmyre et al . , 2015; Morrisey and Hogan , 2010; Sher and Liu , 2016 ) . Other anomalies of foregut development , such as tracheomalacia and congenital tracheal stenosis , involve the improper formation of the adjacent tracheal mesenchyme-derived cartilage and smooth muscle ( Morrisey and Hogan , 2010; Sher and Liu , 2016 ) . While foregut malformations in humans are common , we know very little of the genetic causes of these defects , in part due to a paucity of information of normal tracheoesophageal transcriptional patterning . The earliest marker of tracheal and lung fate is the transcription factor NKX2-1 ( TTF1 ) which is expressed in the ventral foregut in a pattern complementary to the dorsally enriched expression of the transcription factor SOX2 ( Guazzi et al . , 1990; Minoo et al . , 1999; Mizuno et al . , 1991; Que et al . , 2007 ) . Disruption of Nkx2-1 in mice resulted in upregulation of SOX2 in the ventral endoderm and differentiation of the adjacent mesenchyme into smooth muscle rather than tracheal cartilage ( Minoo et al . , 1999; Que et al . , 2007 ) . Conversely , hypomorphic disruption of Sox2 in mice resulted in upregulation of dorsal NKX2-1 and a conversion of the stratified esophageal epithelium to a simple columnar epithelium surrounded by smooth muscle that histologically resembles that of the trachea ( Que et al . , 2007; Teramoto et al . , 2019 ) . Similarly , knockdown of SOX2 in human induced pluripotent stem cell ( hiPSC ) -derived dorsal foregut cells resulted in upregulation of NKX2-1 , and forced expression of SOX2 in hiPSC-derived ventral foregut cells repressed NKX2-1 ( Trisno et al . , 2018 ) . Together these data have given rise to a model in which NKX2-1 and SOX2 form a co-repressive master regulatory switch to define tracheal and esophageal cell fates ( Billmyre et al . , 2015; Domyan et al . , 2011; Que et al . , 2007; Teramoto et al . , 2019; Trisno et al . , 2018 ) . The regulatory programs downstream of NKX2-1 and SOX2 are not known and , therefore , the extent to which each promotes or represses tracheal and esophageal cell fates is not clear . Moreover , beyond these two transcription factors , we currently know very little about the transcriptional identity of the early dorsoventral endodermal populations that ultimately give rise to the trachea and esophagus . The mechanisms coupling epithelial and mesenchymal fate specification in the trachea and esophagus are not well understood , but involve epithelial to mesenchymal signaling . For example , loss of WNT signaling from the endoderm to the tracheal mesenchyme results in a loss of tracheal cartilage and a corresponding expansion of smooth muscle ( Hou et al . , 2019; Kishimoto et al . , 2019; Snowball et al . , 2015 ) . SHH signaling regulates smooth muscle specification in multiple contexts ( Huycke et al . , 2019; Mao et al . , 2010 ) and loss of SHH signaling from the airway and intestinal epithelium results in loss of smooth muscle formation ( Kim et al . , 2015; Litingtung et al . , 1998; Pepicelli et al . , 1998 ) and mispatterning of tracheal cartilage ( Miller et al . , 2004; Sala et al . , 2011 ) . Thus , while WNT and SHH signaling are critical for foregut mesenchymal differentiation , how these signals are transcriptionally regulated in the tracheal and esophageal epithelium is currently unknown . In this study , we dissect the transcriptional regulation of tracheal and esophageal fate specification by combining multiple genomic approaches . By single cell RNA-sequencing ( scRNA-seq ) we define the transcriptional identity of the trachea , esophagus , and lung at their initial stages of development , and identify new and robust markers of tracheoesophageal specification . We then dissect the NKX2-1 regulatory program that specifies TE identity using our scRNA-seq datasets , in combination with bulk RNA-sequencing of Nkx2-1-/- mutant tracheas , and NKX2-1 chromatin immunoprecipitation and sequencing ( ChIP-seq ) of wild type tracheas . We discover a previously unknown NKX2-1-independent transcriptional program that encompasses the majority of the newly-identified tracheal and esophageal transcriptomes . We assay the NKX2-1 transcriptional program in functional compound mouse mutant experiments to test whether NKX2-1 regulates these TE genes through repression of SOX2 or independently of SOX2 . These data uncover a role for NKX2-1 in regulating epithelial-to-mesenchymal signaling , thereby coupling TE epithelial identity with cartilage and smooth muscle fate specification . This study therefore establishes a new framework for understanding key regulators of early cell fate specification in the trachea and esophagus .
To understand how the trachea and esophagus are specified on a transcriptome-wide scale , we performed droplet-based single-cell RNA sequencing ( scRNA-seq ) on E10 . 5 mid-separation and E11 . 5 post-separation dissected mouse foregut epithelial cells . We generated 6407 single-cell transcriptomes at E10 . 5 , comprising 5 cell clusters , and 10 , 493 single-cell transcriptomes at E11 . 5 , comprising 7 cell clusters , and visualized these clusters using Uniform Manifold Approximation and Projection ( UMAP ) dimensional reduction ( Becht et al . , 2019; Stuart et al . , 2018; Figure 1a , b ) . We delineated the dorsoventral axis in our scRNA-seq data at E10 . 5 and E11 . 5 by projecting the expression levels of Nkx2-1 and Sox2 on the UMAP ( Figure 1—figure supplement 1c , d ) . Similarly , Sox9 , a marker of developing lung epithelium ( Herriges et al . , 2012; Perl et al . , 2005; Rockich et al . , 2013 ) , marked a subset of Nkx2-1-positive respiratory cells , enabling us to distinguish distal lung from trachea ( Figure 1—figure supplement 1c , d ) . Using differential expression analysis for each cell cluster and RNAscope fluorescent in-situ hybridization , we defined cell clusters as lung , trachea , esophagus , and pharynx ( Figure 1a–l’’ , Figure 1—figure supplement 2a–m , Figure 1—figure supplement 3a–c , Figure 1—figure supplement 4a ) as well as a cluster corresponding to the ultimobranchial body , a derivative of the pharyngeal endoderm that gives rise to the follicular cells of the thyroid ( Nilsson and Fagman , 2017; Figure 1—figure supplement 4b ) . Within the E11 . 5 lung , we identified a cluster with unique expression of known distal lung markers such as Bmp4 ( Weaver et al . , 1999; Figure 1—figure supplement 3c , Figure 1—source data 1 ) , demonstrating that the proximodistal lung axis can be identified by markers in our dataset . Signatures of proliferation also divided both the trachea and lung clusters at E11 . 5 ( Figure 1b , Figure 1—source data 1 ) . Our scRNA-seq data revealed a wealth of genes previously unknown to mark cell types of the ventral and dorsal foregut prior to TE separation ( Figure 1c , Figure 1—source data 1 ) , as well as new genes that distinguish the trachea , lung , and esophagus after TE separation ( Figure 1—figure supplement 3c , Figure 1—source data 1 ) . To visualize their spatial expression , we performed RNAscope for several genes marking each of these cell types in the undivided E10 . 5 foregut and lung ( Figure 1d–l; Figure 1—figure supplement 2a–m ) , and the E11 . 5 trachea , esophagus , and lung ( Figure 1—figure supplement 3 ) . In all cases , RNAscope analysis confirmed our scRNA-seq finding and provided additional information about the spatial patterns of marker gene expression . We found many genes , including Klf5 , Krt19 , Dcn , Pitx1 , and Lrig1 that exhibited expression specifically within the dorsal foregut/esophageal cells ( Figure 1c , j–l’’; Figure 1—figure supplement 2k–m” , Figure 1—figure supplement 3 , Figure 3—figure supplement 1b ) . Interestingly , within the foregut endoderm , all of these genes were more dorsally restricted than SOX2 and were more precisely complementary to NKX2-1 , indicating that they may serve as better markers of the esophagus during its early development ( Figure 1—figure supplement 4 ) . Notably , we also discovered genes such as Tppp3 , Pcdh10 , Ly6h , and Cldn18 that exhibited enrichment in the ventral foregut at E10 . 5 ( Figure 1c , g–i” , Figure 1—figure supplement 2e-j' ) and specifically marked the trachea and proximal airway at E11 . 5 ( Figure 1—figure supplement 3 ) . Our discovery of this gene class indicates that the trachea and proximal airway are actively transcriptionally specified and at least somewhat distinct from the lung during early respiratory development . All tracheal and esophageal markers we identified and validated were dorsoventrally restricted in the common foregut of E10 . 5 embryos prior to physical separation of the trachea and esophagus , consistent with extensive fate specification before TE separation ( Billmyre et al . , 2015 ) . Together , these data uncover a multitude of previously unknown genes that define initial dorsoventral patterning of the foregut and the earliest stages of the trachea , lung , and esophagus . Given our discovery of a broad network of previously unknown genes expressed during early TE specification , we sought to examine how tracheoesophageal fates are dysregulated upon Nkx2-1 loss at the transcriptome-wide scale . We performed RNA-sequencing ( RNA-seq ) of dissected and FACS-purified foregut epithelium from E11 . 5 Nkx2-1-/- and wild-type ( WT ) embryos . Differential expression analysis between Nkx2-1-/- and WT foregut epithelium identified 257 NKX2-1-dependent genes , with 109 genes upregulated and 148 genes downregulated in Nkx2-1-/- foreguts ( Figure 2a , Figure 2—source data 1 ) . We examined the expression of these NKX2-1-dependent genes in our scRNA-seq dataset at E11 . 5 to determine , on a global-scale , where they were expressed in the developing foregut . We found that genes that were upregulated in Nkx2-1-/- mutants were enriched in cells of the esophagus and pharynx ( Figure 2b , Figure 2—figure supplement 1a ) , and genes that were downregulated in Nkx2-1-/- mutants were enriched in tracheal and lung cells ( Figure 2c , Figure 2—figure supplement 1b ) . Together , these data define the NKX2-1 transcriptional program in early TE development and support the hypothesis that , within this program , NKX2-1 positively regulates tracheal genes and negatively regulates esophageal genes . Surprisingly , our scRNA-seq dataset identified many genes that mark tracheal and esophageal cells that did not appear to exhibit a change in expression in our Nkx2-1-/- mutant RNA-seq analysis ( Figure 1—source data 1 , Figure 2—source data 1 ) . We examined the spatial expression of several of these NKX2-1-independent genes using an Nkx2 . 5-cre strain which mediates recombination in the ventral foregut ( Figure 3—figure supplement 1a; Stanley et al . , 2004 ) to generate Nkx2-1lox/lox; Nkx2 . 5cre/+ ( Nkx2-1-TrKO ) embryos lacking NKX2-1 in the trachea . All Nkx2-1-TrKO embryos we examined exhibited a complete failure of foregut separation ( n = 8/8 embryos ) . Using RNAscope , we found Irx2 , Ly6h , and Nrp2 to be tracheal-specific and maintained in the ventral epithelium of the unseparated foregut tube in E11 . 5 Nkx2-1-TrKO embryos ( Figure 3a–f’; Figure 3—figure supplement 1b ) . Likewise , we found Dcn , Ackr3 , and Meis2 to be esophageal-specific and maintained in the dorsal region of the common foregut tube in Nkx2-1-TrKO foreguts ( Figure 3g–l’; Figure 3—figure supplement 1b ) . Immunofluorescent staining also showed that the esophageal genes LRIG1 and PITX1 were maintained in the dorsal region of Nkx2-1-/- foreguts ( Figure 3—figure supplement 1b ) . These data indicate the presence of an NKX2-1-independent transcriptional program within the developing trachea and esophagus , and suggest that Nkx2-1-/- mutant foreguts do not undergo a complete tracheal-to-esophageal fate conversion . We therefore examined the extent to which tracheoesophageal patterning is independent of NKX2-1 by generating bulk transcriptional profiles of WT tracheal and esophageal epithelium at E11 . 5 to compare with our Nkx2-1-/- mutant bulk RNA-seq dataset ( Figure 3m ) . Differential expression analysis identified 1126 genes enriched in the trachea and 809 genes enriched in the esophagus of WT embryos ( Figure 3n , Figure 3—source data 1 ) . Notably , only 11% of tracheal-enriched genes and 6% of esophageal-enriched genes in WT foreguts were affected by loss of NKX2-1 , and the majority of tracheal- or esophageal-enriched genes were unchanged in Nkx2-1-/- foreguts ( Figure 3n ) . Together , these findings define relevant NKX2-1 transcriptional targets during early tracheoesophageal development and uncover a significant NKX2-1-independent gene regulatory program . To identify direct targets of NKX2-1 during TE specification , we performed NKX2-1 chromatin immunoprecipitation followed by sequencing ( ChIP-seq ) on dissected E11 . 5 WT trachea . We identified 15 , 861 genomic regions ( peaks ) shared between two biological replicates that showed NKX2-1 binding ( Figure 4a , FDR < 0 . 00001 ) and were centrally enriched for the known NKX2-1 motif ( Figure 4b , p=3 . 4e-37 ) . We next looked closely at NKX2-1 binding at or near the loci of select NKX2-1-dependent tracheal and esophageal genes . We observed NKX2-1 binding at the Nkx2-1 promoter and five binding sites within 10 kb of the Nkx2-1 gene , consistent with previous data suggesting that NKX2-1 autoregulates its own expression ( Nakazato et al . , 1997; Oguchi and Kimura , 1998; Tagne et al . , 2012 ) , and also observed binding of NKX2-1 at the Sox2 promoter and four binding sites within the Sox2 locus , suggesting direct repression of Sox2 by NKX2-1 ( Figure 4c ) . In addition , NKX2-1 binding was observed at the promoter of genes such as Pcdh10 , Tppp3 , and Klf5 that we identified as specific markers of the tracheal and esophageal lineages by scRNA-seq ( Figure 4c ) , suggesting that these genes may also be direct targets of NKX2-1 regulation . Genome-wide comparison of NKX2-1-bound genes with the Nkx2-1-/- de-regulated transcriptome revealed that NKX2-1-dependent genes were associated with NKX2-1 ChIP-seq peaks at a higher frequency than observed at random ( Figure 4d , Fisher’s exact test , p<0 . 0001 ) . Furthermore , when we divided NKX2-1-dependent genes into those that were upregulated or downregulated in Nkx2-1-/- mutants , we found that genes that are downregulated in Nkx2-1-/- mutants are more frequently associated with NKX2-1 ChIP-seq peaks ( Figure 4d , Fisher’s exact test , p<0 . 0001 ) . These data suggest that whereas NKX2-1 is a direct positive regulator of tracheal-specific genes , repression of esophageal-specific genes may more often be indirect . Based on previous studies , NKX2-1 indirect regulation may be mediated through its repression of Sox2 ( Billmyre et al . , 2015 ) . Indeed , given the well-established genetically co-repressive relationship of SOX2 and NKX2-1 in the foregut ( Domyan et al . , 2011; Que et al . , 2007; Teramoto et al . , 2019; Trisno et al . , 2018 ) , it is challenging to determine whether the transcriptional changes we observed in Nkx2-1-/- mutants are solely due to the loss of NKX2-1 or also due to the subsequent gain of SOX2 . Thus , we devised a genetic strategy to uncouple NKX2-1 and SOX2 regulation of tracheal and esophageal genes by generating Nkx2-1; Sox2 compound mutant embryos . To achieve this , we again utilized Nkx2 . 5-cre to generate Nkx2-1lox/lox; Nkx2 . 5cre/+ ( Nkx2-1-TrKO ) embryos lacking NKX2-1 , and Nkx2-1lox/lox; Sox2lox/lox; Nkx2 . 5cre/+ ( Nkx2-1-TrKO; Sox2-TrKO ) embryos lacking both NKX2-1 and SOX2 in the ventral foregut/trachea cells ( Figure 5a–c ) . Similar to the Nkx2-1-TrKO embryos , all Nkx2-1-TrKO; Sox2-TrKO embryos we examined exhibited a complete failure of foregut separation ( n = 6/6 embryos ) . We then examined NKX2-1-regulated genes that were determined by our scRNA-seq analysis to be markers of E11 . 5 trachea or esophagus . The expression of the tracheal-specific genes Pcdh10 and Tppp3 was decreased in Nkx2-1-TrKO and Nkx2-1-TrKO; Sox2-TrKO ventral foreguts indicating that their expression requires NKX2-1 but not upregulation of SOX2 ( Figure 5d–i’ ) , Conversely , expression of the esophageal-specific genes Klf5 and Has2 was upregulated in the ventral foreguts of Nkx2-1-TrKO , as predicted from our RNA-seq analysis ( Figure 5j–o’ ) . Interestingly , whereas Klf5 expression was also increased in the ventral foregut of Nkx2-1tr/tr; Sox2tr/tr mutants ( arrowhead Figure 5l , l’ ) , Has2 expression was not , with the exception of a few cells that retain SOX2 expression ventrally ( arrowhead in Figure 5o , o’ ) . Thus , while Klf5 and Has2 were both repressed by NKX2-1 , upregulation of Has2 in the ventral foregut of Nkx2-1-TrKO embryos appears to also depend on the upregulation of SOX2 in this region . Together , these results revealed that NKX2-1 regulation results from both direct activation of tracheal genes and repression of esophageal genes , as well as the indirect suppression of target genes through the repression of Sox2 , illustrating the complex relationship between these two transcription factors with opposing expression patterns . The apparent fate transformation of Nkx2-1-/- mutant trachea to esophagus is supported by the conversion of ventral mesenchymal cell fates from tracheal cartilage to smooth muscle ( Minoo et al . , 1999; Que et al . , 2007; Yuan et al . , 2000 ) . We confirmed previous reports of a dramatic reduction and disorganization of tracheal cartilage with expansion of the smooth muscle in Nkx2-1-/- mutants , and additionally observed malformation of the thyroid and cricoid cartilage ( n = 3/3 embryos , Figure 6a , o , p ) . Our finding that the endoderm in Nkx2-1-/- mutants did not exhibit a complete fate transformation led us to ask whether we could uncover the specific NKX2-1 targets that impact epithelial to mesenchymal signaling amongst this more limited set of candidates . Notably , Wnt7b expression decreased and Shh increased in Nkx2-1-/- foregut epithelium compared to WT ( Figure 2a ) , and both signaling genes have known functions in tracheal cartilage and smooth muscle formation . In WT E11 . 5 and E13 . 5 embryos , Wnt7b is expressed in the tracheal epithelium and Shh is expressed more strongly in the esophageal epithelium ( Gerhardt et al . , 2018; Litingtung et al . , 1998; Rajagopal et al . , 2008; Snowball et al . , 2015; Figure 6c , f , i , m ) . We examined our NKX2-1 ChIP-seq data to determine whether NKX2-1 binds directly to Wnt7b and Shh in the trachea and observed binding at three sites 20 kb and 60 kb upstream and 20 kb downstream of the Wnt7b gene . Though a previous study detected binding of NKX2-1 to the Wnt7b promoter in a lung epithelial cell line ( Weidenfeld et al . , 2002 ) , we instead observe binding at three sites 20 kb and 60 kb upstream and 20 kb downstream of the Wnt7b gene . We also observed two NKX2-1 ChIP-seq peaks within the Shh gene , and two additional peaks 50 kb and 150 kb upstream of Shh , consistent with the possibility of direct regulation by NKX2-1 ( Figure 6b ) . Together , these data indicate that Wnt7b and Shh are targets of NKX2-1 regulation in the developing foregut . To determine whether NKX2-1 regulates Wnt7b and Shh through a change in SOX2 expression , we examined Nkx2-1; Sox2 compound mutants . Wnt7b expression was lost in both Nkx2-1-TrKO and Nkx2-1-TrKO; Sox2-TrKO foreguts , suggesting that NKX2-1 regulates Wnt7b independently of regulation by SOX2 ( Figure 6c–e’ ) . However , the increase in Shh expression in the ventral foregut observed in Nkx2-1-TrKO embryos was not observed in Nkx2-1-TrKO; Sox2-TrKO foreguts ( Figure 6f–h’ ) , indicating that SOX2 is required for upregulation of Shh in the ventral foregut . Thus , Wnt7b expression is positively regulated by NKX2-1 but not subject to regulation by SOX2 , whereas the dorsally restricted expression of Shh is mediated by the combined action of NKX2-1 and SOX2 . To determine whether NKX2-1 regulation of Wnt7b and Shh impacts epithelial-mesenchymal signaling , we first examined the expression of Wnt7b and downstream canonical WNT signaling transcriptional target Axin2 in E11 . 5 and E13 . 5 Nkx2-1-/- and WT foreguts . Whereas in WT embryos Axin2 expression was localized to the ventral tracheal mesenchyme where SOX9+ cartilage progenitor cells are found , in Nkx2-1-/- mutants Axin2 expression was decreased and correlated with disorganized SOX9 expression ( Figure 6i–l , Figure 6—figure supplement 1a ) . This decrease in Axin2 expression was more dramatic in the caudal foregut , consistent with the increased severity of cartilage specification defects in this region seen in skeletal preparations ( Figure 6a ) and as visualized by SOX9 localization in Nkx2-1-/- foreguts ( Figure 6—figure supplement 1c–e ) . We next examined the expression of Shh and downstream transcriptional target Ptch1 in Nkx2-1-/- mutant and WT foreguts . In E11 . 5 and E13 . 5 Nkx2-1-/- foreguts , increased expression of Shh in the ventral foregut epithelium compared to WT was mirrored by an increase in Ptch1 expression in the surrounding mesenchyme ( Figure 6m , n; Figure 6—figure supplement 1b ) . This increase in Ptch1 expression also corresponded with the increase in smooth muscle as visualized by ACTA2 ( i . e . α-SMA ) surrounding Nkx2-1-/- foreguts compared to WT ( Figure 6m–p , S8c , f , g ) . Thus , our data show that in addition to regulating a subset of genes that define tracheoesophageal epithelial identity at this stage , NKX2-1 and SOX2 regulate epithelial-mesenchymal crosstalk via WNT and SHH signaling to instruct mesenchymal differentiation , explaining this aspect of the observed TE fate transformation in Nkx2-1-/- mutant foreguts ( Figure 7 ) .
NKX2-1 is a critical regulator of respiratory development across vertebrates ( Minoo et al . , 1999; Rankin et al . , 2018; Small et al . , 2000; Yuan et al . , 2000 ) , but how it directs TE specification was unknown . It has been previously demonstrated in adenocarcinoma and in the adult distal lung , that loss of NKX2-1 leads to adoption of a gastric transcriptional phenotype ( Little et al . , 2019; Snyder et al . , 2013 ) . Further , mesenchymal cues that induce lung and trachea development converge to establish ventral NKX2-1 expression , and Nkx2-1-/- mutant embryos exhibit dramatic epithelial and mesenchymal phenotypes . Together , these findings have led to the proposal that NKX2-1 is a master regulator of initial tracheal fate specification ( Billmyre et al . , 2015; Domyan et al . , 2011; Goss et al . , 2009; Harris-Johnson et al . , 2009; Morrisey and Hogan , 2010; Que et al . , 2007 ) . By performing scRNA-seq profiling on the endoderm at the onset of tracheal and esophageal development , and comparing this profile to Nkx2-1-/- mutant embryos , we find that NKX2-1 loss does not cause a transcriptome-wide tracheal-to-esophageal conversion , but rather results in changes in expression of a key subset of dorsoventrally restricted genes that can explain apparent fate-conversion phenotypes in the mesenchyme ( Figure 7 ) . The functional significance of the NKX2-1-independent tracheal program , and how it is established , remains to be explored . It is possible that mesenchymal signals such as WNT and BMP induce other tracheal transcriptional regulators in parallel . Interestingly , the ISL1 transcription factor has been recently identified to exhibit ventrally-restricted expression in the trachea and is required for normal NKX2-1 expression ( Kim et al . , 2019 ) . Further , our data reveal that Isl1 is independent of NKX2-1 ( Figure 2—source data 1 ) , indicating that it resides hierarchically upstream of NKX2-1 . It remains to be determined , however , the extent to which ISL1 regulates tracheal transcriptional identity . Our profiling of tracheoesophageal specification , combined with genomic analyses of additional mouse mutants , provide an exciting opportunity for further exploration of additional regulators of tracheoesophageal specification within the foregut epithelium . Our scRNA-seq profiling reveals many new markers of dorsoventral foregut identity beyond Nkx2-1 and Sox2 . This of particular importance for human iPSC/ESC-derived models where Nkx2-1 and Sox2 expression are often used to indicate respiratory or esophageal/gut fate decisions ( Dye et al . , 2015; Hawkins et al . , 2017; Trisno et al . , 2018; Zhang et al . , 2018 ) . Our data suggest that in addition to using Nkx2-1 as a marker of respiratory fate , differentiation protocols might be more thoroughly characterized using a panel of NKX2-1-dependent and -independent respiratory markers . Further , though we have not explored it in depth , we also provide here the first scRNA-seq dataset characterizing the lung at early stages of branching morphogenesis , which may additionally aid in more precise characterization of early lung development . Our study therefore provides a rich resource for expanding our understanding of early cell fate decisions in stem cell models of human foregut and lung development . By defining the NKX2-1-dependent transcriptomic profile , we elucidate how NKX2-1 regulates initial tracheal and esophageal development at the genome-wide level . A dual repressive genetic relationship with SOX2 has been previously demonstrated , and our ChIP-seq data support direct repression of Sox2 by NKX2-1 . We find , however , that direct binding of NKX2–1 occurs more often at genes that are normally expressed in the trachea , and loss of tracheal gene expression in Nkx2-1-/- mutants is not indirectly mediated by upregulation of SOX2 . Together , these findings broadly support a model in which NKX2-1 more often positively regulates the tracheal gene expression program directly . While we demonstrate that SOX2 is a mediator of NKX2-1-indirect regulation of some genes , SOX2 targets in the foregut have not been identified at the transciptomic level , and it is therefore unclear the extent to which SOX2 controls TE fates . Additionally , our identification of other NKX2-1-dependent transcription factors such as KLF5 provides candidates potentially mediating NKX2-1 indirect repression of esophageal genes . Tracheal cartilage and smooth muscle phenotypes in Nkx2-1-/- embryos support a TE fate conversion phenotype ( Minoo et al . , 1999; Que et al . , 2007 ) ; here we provide a likely explanation of these phenotypes and reveal how NKX2-1 may regulate development of the tracheal cartilage and restrict smooth muscle formation . Normal tracheal cartilage specification requires WNT signaling from the tracheal epithelium to the mesenchyme ( Hou et al . , 2019; Kishimoto et al . , 2019; Snowball et al . , 2015 ) . We find that NKX2-1 positively regulates Wnt7b resulting in activation of a canonical WNT-signaling response in the adjacent ventral tracheal mesenchyme . A recent report suggested that epithelial-mesenchymal WNT signaling during cartilage specification occurs through NKX2-1-independent mechanisms ( Kishimoto et al . , 2019 ) . Our results differ from this finding but do support the possibility that an NKX2-1 independent mechanism also contributes to tracheal cartilage specification , as some disorganized cartilage still forms in Nkx2-1-/- mutants . In addition to regulating WNT signaling , we found that NKX2-1 loss resulted in upregulation of Shh and increased SHH signal reception in the ventral mesenchyme . As HH signaling is a key regulator of smooth muscle formation in multiple contexts , and its disruption also results in perturbation of tracheal cartilage formation ( Huycke et al . , 2019; Mao et al . , 2010; Miller et al . , 2004; Sala et al . , 2011 ) , it is likely that this upregulation also contributes to disruption of tracheal cartilage and expansion of smooth muscle formation . Notably , similar to the other trachea and esophagus markers we examined , NKX2-1 regulation of Wnt7b does not depend on SOX2 , whereas the upregulation of Shh in Nkx2-1-/- mutants does . This fits with the broader model of NKX2-1 directly activating a defined tracheal program , but more often repressing an esophageal program through ventral repression of SOX2 . While few genetic causes of foregut anomalies in humans have been identified , human stem cell models have indicated that key players in respiratory and gut specification are conserved between mice and humans ( Ostrin et al . , 2018; Trisno et al . , 2018 ) . Therefore , future work to understand genetic causes of foregut malformations in humans including tracheoesophageal fistula ( TEF ) , tracheal agenesis , esophageal atresia , tracheomalacia , and tracheal stenosis will benefit greatly from the improved understanding of normal tracheal and esophageal specification presented in this study . Together , these findings dramatically advance our understanding of the earliest stages of tracheoesophageal development , while revising our perspective on the role of NKX2-1 in this process .
All animal procedures were performed at the University of California San Francisco ( UCSF ) under approval from the UCSF Institutional Animal Care and Use Committee ( mouse protocol # AN164190 ) . Mouse embryos were collected from pregnant females via cesarean section at the described timepoint following observation of a vaginal plug . Noon the day of the plug was considered embryonic day 0 . 5 . For single cell sequencing experiments , timed-pregnant CD1 female mice were obtained from Harlan/Envigo ( Cat: 030 ) and embryos were staged using somite counts . For mutant analysis , the following alleles were used: Nkx2-1lox/lox ( MGI: 3653645 ) , Sox2lox/lox ( MGI: 4366453 ) , Nkx2 . 5-Cre ( MGI: 2448972 ) , Actin-Cre ( MGI: 2176050 ) . Mouse embryos were dissected at E10 . 5 or E11 . 5 in cold PBS and fixed overnight in 4% paraformaldehyde at 4C . For cryopreservation , embryos were subjected to a sucrose gradient of 12 . 5% sucrose in PBS for 8 hr , followed by 25% sucrose in PBS overnight at 4°C , and embedded in OCT . Tissue sections of 12 um were cut using a cryostat and used for immunofluorescence , or in-situ hybridization followed by immunofluorescence with standard protocols . Primary antibodies used for immunofluorescence were: NKX2-1 ( Millipore 07601 , 1:200 ) , SOX2 ( Neuromics GT15098 , 1:250 ) , SOX9 ( Santa Cruz , sc-20095 , 1:250 ) , LRIG1 ( R&D , AF3688 , 1:200 ) , PITX1 ( NovusBio , NBP188644 , 1:250 ) . For in-situ hybridization , 12 µm cryosections were stained using the RNAscope Multiplex Fluorescent Reagent Kit v2 ( Advanced Cell Diagnostics , cat# 323100 ) with the following adjustments to the manufacturer’s protocol: antigen retrieval step was bypassed , protease step used ProteasePlus for 10 min . Following in-situ hybridization , slides were washed 2x in PBS and subjected to immunofluorescent staining as described in ‘Immunofluorescence’ above . Probes used for in-situ hybridization against mouse RNA were obtained from Advanced Cell Diagnostics as follows: Irx2 ( 519901 ) , Wnt7b ( 401131 ) , Pcdh10 ( 477781-C3 ) , Tppp3 ( 586631 ) , Ly6h ( 587811 ) , Krt19 ( 402941 ) , Klf5 ( 444081 ) , Foxe1 ( 509641 ) , Crabp1 ( 474711-C3 ) , Bmp4 ( 401301-C2 ) , Nrp2 ( 500661 ) , Dcn ( 413281-C3 ) , Has2 ( 465171-C2 ) , Meis2 ( 436371-C3 ) , Ackr3 ( 482561-C2 ) , Axin2 ( 400331-C3 ) , Shh ( 314361 ) , Ptch1 ( 402811-C2 ) . Skeletal preps were performed as previously described ( Martin et al . , 1995 ) . Foregut tissue was dissected in cold PBS and dissociated to single cells using TrypLE Express ( phenol-red free , Thermo cat# 12604013 ) at 37°C for 5 min , followed by trituration for 1–3 min at 37°C . Cells were washed twice with FACS buffer ( 2 mM EDTA and 5% fetal bovine serum in phenol-red free HBSS ) . To identify epithelial cells , cells were stained with PerCP/Cy5 . 5 anti-mouse CD326/EpCAM ( BioLegend , cat# 118219 , used at 1:100 ) at 4°C for 30 min followed by two washes with FACS buffer . Cells were resuspended in FACS buffer with Sytox Blue nucleic acid stain ( Thermo , S11348 , used at 1 µM ) to stain dead cells , and passed through a 35 µm cell strainer . Cells were sorted using a BD FACS Aria II . Single live epithelial cells were collected after size selection and gating for Sytox-negative , EpCAM-positive cells . For scRNA-seq , cells were sorted into EDTA-free FACS buffer and processed as described below . For bulk-RNA-seq , cells were sorted directly into RNA lysis buffer ( Qiagen RNeasy Micro kit , cat# 74004 ) with 1% beta-mercaptoethanol and processed as described below . Foregut tissue was dissected from 20 embryos at E10 . 5 ( 6-9ts ) and 28 embryos at E11 . 5 ( 16-20ts ) . To ensure for representation of tracheal and esophageal cells at E11 . 5 , lung tissue was separated from E11 . 5 foreguts and processed in parallel . Tissue from each timepoint was pooled and single-cell suspension and epithelial purification was performed as described above . 25 , 000 live epithelial cells from each sample were loaded into individual wells for single-cell capture using the Chromium Single Cell 3’ Reagent Kit V2 ( 10X Genomics ) . Library preparation for each sample was also performed using the Chromium Single Cell 3’ Reagent Kit V2 , and each sample was given a unique i7 index . Libraries were pooled and subjected to sequencing in a single lane of an Illumina NovaSeq6000 . Sequencing data were processed , and downstream analysis performed as described below . For bulk RNA-sequencing experiments , whole foregut tissue was dissected from E11 . 5 Nkx2-1-/- or WT embryos , and lungs were removed at the time of dissection . For RNA-sequencing of wild type ( WT ) trachea and esophagus , trachea and esophagus were manually separated at the time of dissection . Foreguts of individual embryos were dissociated , stained , and sorted as described above . Each biological replicate consisted of RNA pooled from two Nkx2-1-/- or WT embryos , with a total of 3 biological replicates from different litters . RNA was purified using the RNeasy Micro kit ( Qiagen , cat# 74004 ) and quantification was performed using the RNA 6000 Pico kit ( Agilent , cat# 5067–1513 ) on an Agilent 2100 Bioanalyzer . RNA-sequencing libraries were prepared from 4 ng of input RNA using the SMARTer Stranded Total RNAseq kit V2 ( Takara , cat# 634411 ) with 13 amplification cycles . Library size and quality was checked using an Agilent 2100 Bioanalyzer with the High Sensitivity DNA kit ( Agilent , cat# 5067–4626 ) , and library concentration was determined with the QuBit dsDNA HS Assay kit ( Invitrogen , cat# Q32854 ) . Libraries were normalized to 7 nM , pooled , and sequenced across two lanes of an Illumina HiSeq 4000 to generate 50 base pair single-end reads . Data processing and downstream analysis was performed as described below . For ChIP-seq experiments , tracheas were dissected from WT E11 . 5 mouse embryos and cross-linked with 1% cold PFA for 9 min . Crosslinking was stopped with glycine for a final concentration of 0 . 125M . The crosslinked tissue was washed 2x in PBS and stored at −80°C . For each replicate 175 trachea were pooled and the tissue was thawed and dissociated in cold PBS by passing through a 25G needle until fully dissociated . The cells were lysed in 500 µl lysis buffer ( 50 mM Tris-HCl pH8 , 2 mM EDTA pH8 , 0 . 1% NP-40 , 10% glycerol in DNase/RNase-free water ) with protease inhibitors ( Aprotinin , Pepstatin A , Leupeptin , 1 mM PMSF ) for 5 min on ice . Nuclei were pelleted by spinning cells at 845xg for 5 min at 8°C , then lysed with 500 µl SDS lysis buffer ( 50 mM Tris-HCl + 10 mM EDTA + 1% SDS in sterile water ) for 5 min on ice . The chromatin from lysed nuclei in SDS lysis buffer was sheared to obtain 200–500 bp DNA fragments using a Diagenode Bioruptor with 35 cycles ( 30 s on/off ) submerged in cold water . Fragment sizes were determined by running a 20 µl aliquot of reverse-crosslinked chromatin on a 1 . 5% agarose gel . The sheared chromatin was diluted 1:10 with ChIP dilution buffer ( 50 mM Tris-HCl , 2 mM EDTA , 0 . 5M NaCl , 0 . 1% SDS in sterile water ) then pre-cleared with washed Dynabeads Protein G for 1 hr at 4°C . The Dynabeads were magnetically isolated from chromatin and 1% of chromatin was separated and reverse crosslinked to be used as input . The remaining sample was incubated with Nkx2-1 antibody ( Millipore 07601 ) overnight at 4°C . To isolate antibody-bound DNA , washed Dynabeads Protein G were added to each sample ( 50 ul beads/sample ) and incubated for 30 min at 4°C . The Dynabeads with antibody-bound chromatin were isolated magnetically and subjected to 3x washes each with wash buffer ( 10 mM Tris-HCl , 2 mM EDTA , 0 . 5M NaCl , 0 . 1% SDS , 1% NP-40 in sterile water ) , LiCl buffer ( 10 mM Tris-HCl , 2 mM EDTA , 0 . 5M LiCl , 0 . 1% SDS , 1% NP-40 ) , and TE buffer ( 1 mM Tris-HCl , 1 mM EDTA ) for 5 min on ice . The chromatin was eluted in 100 µl of 2% SDS in TE on a 65°C heatblock with vigorous shaking ( 1400 rpm ) for 15 min . Input DNA and immunoprecipitated DNA were reverse crosslinked by adding 5 µl 5M NaCl to 100 µl eluate and incubating overnight at 65°C , followed by a 30 min treatment with RNase . The reverse crosslinked DNA was purified using MicroChIP Diapure columns ( Diagenode , cat# C03040001 ) and eluted in 10 µl of elution buffer . The entire eluate of Nkx2-1 immunoprecipitated DNA and 0 . 5 ng of input DNA were used to prepare ChIP libraries . The libraries were prepared using the Microplex Library Preparation Kit v2 ( Diagenode , cat# C05010012 ) according to manufacturer's instructions with 14 amplification cycles . Library quality and size were calculated using an Agilent 2100 Bioanalyzer with the High Sensitivity DNA kit ( Agilent , cat# 5067–4626 ) , and library concentration was quantified with the QuBit dsDNA HS Assay kit ( Invitrogen , cat# Q32854 ) . The libraries were pooled to 5 nM and sequenced in one lane of an Illumina HiSeq 4000 . The sequencing data were processed and downstream analysis was performed as described below . We used the Cell Ranger v2 . 1 . 1 pipeline from 10X Genomics for initial processing of raw sequencing reads . Briefly , raw sequencing reads were demultiplexed , aligned to the mouse genome ( mm10 ) , filtered for quality using default parameters , and UMI counts were calculated for each gene per cell . Filtered gene-barcode matrices were then analyzed using the Seurat v3 . 0 R package ( Stuart et al . , 2018 ) . Seurat objects were generated with CreateSuratObject ( min . cells = 10 , min . features = 200 ) for E10 . 5 foregut and lung cells , E11 . 5 foregut cells , and E11 . 5 lung cells . E11 . 5 foregut and lung cells were merged to create a single gene-barcode matrix . Cells were further filtered based on the distribution of number of genes ( nFeature ) and percent mitochondrial genes ( percent . mito ) per cell across the dataset as follows . nFeature_RNA ( E10 . 5 ) :>2000 , <7000 , nFeature_RNA ( E11 . 5 ) :>2500 , <8500 , percent . mito:>0 . 5 , <7 . 5 . Data were normalized for sequencing depth , log-transformed , and multiplied by a scale factor of 10000 using the default parameters of NormalizeData . Linear regression was performed to eliminate variability across cell cycle stage ( CellCycleScoring ) and mitochondrial content using ScaleData . For E11 . 5 merged foregut and lung , nCount_RNA was also regressed out as these datasets retained slight variability in sequence depth that was not eliminated with ScaleData . The top 2000 variable genes within each dataset were selected based on a variance stabilizing transformation ( FindVariableGenes , selection . method = ‘vst’ ) and used in downstream principal component analysis ( PCA ) . The principal components ( PCs ) were identified with RunPCA and PCs to include in downstream analysis were empirically determined with visualization of PCs in an ElbowPlot . Cell clusters were identified by construction of a shared nearest neighbor graph ( FindNeighbors ) and a modularity optimization-based clustering algorithm ( FindClusters ) using the PCs determined by PCA ( E10 . 5 dims = 1:20 , E11 . 5 dims = 1:12 ) . Clustering was performed at multiple resolutions between 0 . 2 and 2 , and optimal resolution was determined empirically based on the expression of known population markers and the FindMarkers function ( E10 . 5 resolution = 0 . 55 , E11 . 5 resolution = 0 . 45 ) . Several outlying clusters of mesenchymal contamination were removed , and cells were re-clustered for visualization purposes . Cells and clustering were visualized using Uniform Manifold Approximation and Projection ( UMAP ) dimensional reduction ( RunUMAP ) . Markers for each cluster were identified with FindAllMarkers using default parameters , and cluster identity was determined based on the presence of known markers , as well as experimental evidence of RNA localization in specific cell types . Analysis of RNA-seq reads was performed as described previously ( Percharde et al . , 2018 ) . Differential expression analysis ( Nkx2-1-/- vs WT , WT trachea vs WT esophagus ) was performed using DESeq2 ( Love et al . , 2014 ) ( test = c ( ‘Wald’ ) , betaPrior = T ) and genes with a log2 fold change > 0 . 7 or<−0 . 7 and an adjusted p-value<0 . 05 were determined to be differentially expressed . Differential expression results were visualized using the ggplot2 package . FASTQ files of raw sequencing reads for NKX2-1 ChIP and input libraries were processed using a custom script ( github . com/akelakuwahara/foregut/run ChIPseq ) . Quality and length trimming and generation of fastqc files to examine sequence quality were performed using Trim Galore ( Krueger , 2014 ) . Trimmed reads were aligned to the mouse genome ( mm9 ) using bowtie2 and sorted deduplicated bam files were generated using samtools . Peak calling was performed with MACS2 ( Zhang et al . , 2008 ) using a false discovery rate less than 1e-5 ( macs2 callpeak -t chip . sorted . bam -c input . sorted . bam -f BAM -q 0 . 00001 g mm -n nkx_peaks --outdir macs/ ) . Peaks shared between both replicates were identified by finding the intersection of both replicates using the Intersect tool in Galaxy ( usegalaxy . org ) . Motif analysis to test for the enrichment of the NKX2-1 motif was performed with MEME ChIP using a 500 bp region flanking the peak summit for all peaks shared between both replicates . NKX2-1 binding at specific loci was visualized in the Integrative Genomics Viewer . Peak-gene associations were generated with GREAT using the basal-plus-extension rule ( McLean et al . , 2010 ) . All code used for data analysis is available at https://github . com/akelakuwahara/foregut ( Kuwahara , 2020; copy archived at https://github . com/elifesciences-publications/foregut ) . | The trachea or windpipe is a tube that connects the throat to the lungs , while the esophagus connects the throat to the stomach . The trachea has cartilage rings that help to ensure clear airflow to the lungs , while the esophagus walls are lined with muscles that help to move food to the stomach . Although there are many differences between them , both the trachea and esophagus form from the same group of cells during development . Proteins called transcription factors help to control the formation of different body parts by switching different groups of genes on and off in different subsets of cells . Existing research has suggested that a transcription factor called NKX2 . 1 drives trachea formation , while another , called SOX2 , is important in esophagus formation . An absence of either of these two proteins is thought to be associated with serious birth defects including loss of the trachea or esophagus , or failure of the two to separate fully . How these two transcription factors interact and drive the development of the trachea and esophagus , however , is currently unclear . Kuwahara et al . used mice to study the role of NKX2 . 1 and SOX2 in the formation of the trachea and esophagus . The findings identify many new genes that are active in the trachea and esophagus and reveal that NKX2 . 1 is not a master regulator that controls all of the genes involved in trachea formation . However , NKX2 . 1 does control several key genes , particularly those involved in forming cartilage in the trachea instead of muscle in the esophagus . The investigation also revealed many genes that are not controlled by NKX2 . 1 suggesting that other , currently unknown , systems play a major role in trachea formation . More work is required to understand the wider genetic regulators involved in differentiating the trachea from the esophagus . The findings in this study will help researchers to understand birth defects in the trachea and esophagus that result from genetic errors . They also reveal information about gene regulation processes that are relevant to the formation of other body parts and in the context of other diseases . In the long term , they could support regenerative medicine to regrow or replace lost or damaged body parts using lab-grown stem cells . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"genetics",
"and",
"genomics"
] | 2020 | Delineating the early transcriptional specification of the mammalian trachea and esophagus |
Porokeratosis ( PK ) is a heterogeneous group of keratinization disorders . No causal genes except MVK have been identified , even though the disease was linked to several genomic loci . Here , we performed massively parallel sequencing and exonic CNV screening of 12 isoprenoid genes in 134 index PK patients ( 61 familial and 73 sporadic ) and identified causal mutations in three novel genes ( PMVK , MVD , and FDPS ) in addition to MVK in the mevalonate pathway . Allelic expression imbalance ( AEI ) assays were performed in 13 lesional tissues . At least one mutation in one of the four genes in the mevalonate pathway was found in 60 ( 98% ) familial and 53 ( 73% ) sporadic patients , which suggests that isoprenoid biosynthesis via the mevalonate pathway may play a role in the pathogenesis of PK . Significantly reduced expression of the wild allele was common in lesional tissues due to gene conversion or some other unknown mechanism . A G-to-A RNA editing was observed in one lesional tissue without AEI . In addition , we observed correlations between the mutations in the four mevalonate pathway genes and clinical manifestations in the PK patients , which might support a new and simplified classification of PK under the guidance of genetic testing .
Porokeratosis ( PK , MIM 175800 ) is a heterogeneous group of keratinization disorders that exhibit an autosomal dominant mode of inheritance . PK is also a skin-specific autoinflammatory disease which was often inherited and linked to ultraviolet light exposure and immunosuppression ( Schamroth et al . , 1997; Abramovits and Oquendo , 2013 ) . For example , eruptive pruritic papular porokeratosis exemplifies the inflammatory manifestation , and complications to inflammatory conditions such as localized cutaneous amyloidosis are seen in PK patients ( Biswas , 2015 ) . Indeed , PK and psoriasis share some features at both clinical and molecular levels and sometimes coexist in the same patients ( Zhang et al . , 2008 ) . As a histological hallmark that unifies all variants of PK , cornoid lamella ( CL ) is a vertical ‘column’ of parakeratosis . The pattern of CL can be slender , broad , or confluent , which is related to epidermal hyperplasia and dermal inflammation . However , CL is not a unique feature of PK because it can be seen in some inflammatory and inherited cutaneous disorders and also as an incidental finding ( Biswas , 2015 ) . PK is currently classified according to the clinical manifestations , such as number , size , morphology , and distribution of the histological lesions . A better system of classification is expected because some variants of PK are fraught with confusing terminology ( Schamroth et al . , 1997; Sertznig et al . , 2012; Biswas , 2015 ) . For example , it is sometimes hard to completely differentiate disseminated superficial actinic porokeratosis ( DSAP ) from disseminated superficial porokeratosis ( DSP ) by age of onset and sun-exposed areas . In addition to the heterogeneity in clinical manifestations , genetic heterogeneity is also observed in PK . At least five linkage loci ( i . e . , 12q23 . 2-24 . 1 , 15q25 . 1-26 . 1 , 18p11 . 3 , 1p31 . 3-p31 . 1 , 16q24 . 1-24 . 3 ) have been reported for the disseminated forms of PK which include DSAP , DSP , porokeratosis palmaris et plantaris disseminata ( PPPD ) , and immunosuppression-induced porokeratosis ( ISIP ) ( Schamroth et al . , 1997; Luan et al . , 2011 ) . However , only one causal gene , the mevalonate kinase gene ( MVK ) at 12q24 , has been identified in DSAP and porokeratosis of Mibille ( PM ) ( Zhang et al . , 2012; Li and Zhang , 2014; Zeng et al . , 2014 ) . Here we performed genetic analysis in 134 index patients with PK to identify additional causal genes and to establish new parameters for better differential diagnosis of PK . Allelic expression imbalance ( AEI ) and cDNA mutation were analyzed in 13 pairwise lesional tissues ( LTs ) and neighboring normal-appearing skin ( NNS ) to investigate the underlying pathogenicity of PK .
We previously identified a linkage locus for DSAP on chromosome 16q24 . 1-24 . 3 in a four-generation Chinese DSAP family ( Luan et al . , 2011 ) . We applied the candidate gene approach and hypothesized that the MVD gene in the 16q24 . 1-24 . 3 region was the causal gene of DSAP in the PK family . We performed Sanger sequencing of all MVD exons and identified a c . 746T>C ( p . Phe249Ser ) mutation in MVD . The c . 746T>C mutation displayed 100% co-segregation with the PK phenotype in this family ( Figure 1 ) . 10 . 7554/eLife . 06322 . 003Figure 1 . Identification of a MVD mutation in a porokeratosis ( PK ) family . ( A ) c . 746T>C ( p . Phe249Ser ) in MVD displayed 100% co-separation with PK phenotype in this family ( Luan et al . , 2011 ) . ( B ) Sanger sequencing chromatograms of proband ( II-7 , affected ) and normal control ( II-8 , unaffected ) at the c . 746T>C mutation site indicated by arrow . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00310 . 7554/eLife . 06322 . 004Figure 1—figure supplement 1 . Examples of six pedigree charts showing that each mutation displayed 100% co-segregation with the porokeratosis ( PK ) phenotype in the family . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00410 . 7554/eLife . 06322 . 005Figure 1—figure supplement 2 . Two MVD mutations , c . 302C>G ( p . Pro101Arg ) and c . 683 G>A ( p . Arg228Gln ) , for S-62 were located in the trans position because his daughter ( S-62-D ) carried only one of them . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 005 Since both MVD and MVK are involved in isoprenoid biosynthesis via the mevalonate pathway ( Thurnher et al . , 2013 ) , we hypothesized that mutations in other members of the mevalonate pathway exist in PK and screened for mutations in 12 genes in the mevalonate pathway including MVD and MVK in 134 PK patients ( 61 familial and 73 sporadic ) . The 12 genes are: acetyl-CoA acetyltransferase ( ACAT ) 1 and 2 , 3-hydroxy-3-methylglutaryl-CoA synthase ( HMGCS ) 1 and 2 , 3-hydroxy-3-methylglutaryl-CoA reductase ( HMGCR ) , MVK , phosphomevalonate kinase ( PMVK ) , MVD , isopentenyl-diphosphate delta isomerase ( IDI ) 1 and 2 , farnesyl diphosphate synthase ( FDPS ) , and geranylgeranyl diphosphate synthase 1 ( GGPS1 ) ( Kanehisa et al . , 2012 ) ( Figure 2 ) . A cycled primer extension and ligation-dependent amplification ( CPELA ) reaction for enrichment of target genomic fragments and next generation sequencing ( NGS ) were performed in each sample . The NGS data were analyzed for rare single nucleotide variants ( SNVs ) with no record in dbSNP or a minor allele frequency of <1% in Chinese from 1000 genome database . After Sanger sequencing validation , 60 out of 61 SNVs called from NGS data were demonstrated to be true SNVs , which were predicted to impact protein function . Of these , 12 non-pathogenic SNVs were excluded from subsequent analysis ( Supplementary file 1 ) . In total , 48 mutation sites were identified in MVK , PMVK , MVD , and FDPS ( Figure 3 , Figure 3—source data 1 ) . 10 . 7554/eLife . 06322 . 006Figure 2 . Isoprenoid biosynthesis via the mevalonate pathway . 12 member genes ( ACAT1 , ACAT2 , HMGCS1 , HMGCS2 , HMGCR , MVK , PMVK , MVD , IDI1 , IDI2 , FDPS , GGPS1 ) were subject to mutation screening . The genomic loci of the 12 member genes are provided in parentheses . The illustration is adapted from the 00900 interactive map of the Kyoto Encyclopedia of genes and genomes ( KEGG ) ( Kanehisa et al . , 2012 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00610 . 7554/eLife . 06322 . 007Figure 3 . Mutational spectrum of MVK , PMVK , MVD and FDPS in 113 of the 134 porokeratosis ( PK ) patients . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00710 . 7554/eLife . 06322 . 008Figure 3—source data 1 . Sanger sequencing chromatograms of normal control and porokeratosis ( PK ) patients at 48 mutation sites in MVK , PMVK , MVD and FDPS . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00810 . 7554/eLife . 06322 . 009Figure 3—figure supplement 1 . Breakpoint analysis for three large deletion mutations in MVK and FDPS genes . ( A ) Agarose electrophoresis analysis of long PCR products from patients with deletion mutations and a normal control . ( B ) Sequencing chromatograms of long PCR products from deletion mutations . ( C ) Alignment of the sequences from three deletion mutations with human genome reference GRch38 primary assembly indicates a deletion of 10 , 076 bp , 3362 bp and 3304 bp , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 00910 . 7554/eLife . 06322 . 010Figure 3—figure supplement 2 . Illustration of amplification of multiple target DNA fragments mediated by cycled primer extension and ligation . In brief , for each target region , an extension primer and a block probe are designed . The extension primer has a 5′ exonuclease-resistance modification and the block probe has a 5′ phosphorate and 3′exonuclease-resistance modification . These two oligos are mixed with genomic DNA , heat-denatured , and then annealed to the same strand of the target DNA fragment . The primer extends and stops until it meets the block probe , and the extension product is then ligated with the block probe by a thermal stable ligase . The above procedure can be repeated n times on the PCR machine using a two-step PCR cycling program . The extension and ligation product is then purified by an exonuclase mixture digestion to remove any DNA fragment with no exonuclease-resistance modification at both ends such as remained primers or probes , primer dimers and genomic DNA , and then amplified using universal NGS PCR primer pairs . A CPELA reaction can include hundreds of extension primer-blocking probe sets and simultaneously amplify hundreds of target DNA fragments for subsequent massively parallel sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 01010 . 7554/eLife . 06322 . 011Figure 3—figure supplement 3 . The principle of CNVplex technology . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 011 For 27 samples with no confirmed point mutation in MVK , PMVK , MVD , and FDPS genes , large deletion or duplication mutations were interrogated by analyzing the copy number of promoters and all exons in these genes . As a result , three deletion mutations were identified by the CNVplex assay . These three deletions were identified in four ( F-26 , F-40 , S-69 and S-21 ) , one ( S-56 ) and one ( F-6 ) cases , respectively . The breakpoints of three large deletions were also determined ( Figure 3 , Figure 3—figure supplement 1 ) . We determined the genomic variations by Sanger sequencing in other family members apart from one with an unknown mutation . In 20 pedigrees with more than two blood samples collected , all patients carried the same mutation with the proband . On the other hand , no mutation was detected in any of the healthy family members . Except for one common MVD mutation of c . 746T>C , all deleterious candidate SNVs and large deletions were undetected in healthy family members and 270 unrelated ethnically matched controls . In total , we identified 13 mutations in MVD , six mutations in PMVK , four mutations in FDPS , and 28 mutations in MVK in 113 of the 134 PK patients ( Supplementary file 2 ) . Notably , 21 of the 28 MVK mutations are novel ( Li and Zhang , 2014 ) . These mutations include 30 missense ( 58% ) , two start codon ( 4% ) , five nonsense ( 10% ) , seven indels ( 14% ) , four splicing defects ( 8% ) , and three large deletions ( 6% ) , and predictably affect functions of the respective enzymes . The two MVD mutations ( c . 746T>C and c . 875A>G ) were identified in 50 unrelated patients , accounting for 81% of all patients with MVD mutations . Among the 60 familial patients in whom at least one mutation was found , 30 had MVD mutations , three had PMVK mutations , two had FDPS mutations , 24 had MVK mutations , and one had both MVK and MVD mutations . Among the 53 sporadic patients in whom at least one mutation was found , 31 had MVD mutations , six had PMVK mutations , two had FDPS mutations , and 14 had MVK mutations . No mutation was found in the remaining 21 PK patients . For the 21 PK patients in whom no pathogenic mutation in MVK , PMVK , MVD , or FDPS was found , we sequenced all exons of the SLC17A9 gene by Sanger sequencing . Mutations in the SLC17A9 gene were recently reported to be another genetic cause for DSAP ( Cui et al . , 2014 ) . However , no mutation in the SLC17A9 gene was found in the 21 PK patients studied . To explore the genotype–phenotype correlations , we studied the clinical records of 134 index patients ( 92 males and 42 females ) with PK . In all patients , the clinical diagnosis was supported by histopathological examinations . The onset of the disease ranged from birth to 78 years ( mean age 31 years ) . The distribution of lesions could be categorized as localized and disseminated forms ( Schamroth et al . , 1997 ) . Several classical variants were found to coexist in the same patient . However , no malignant degeneration was found in the lesions of these patients . Although all affected members in the same family carried the same mutation , each affected member showed different clinical manifestations and severity . We observed some interesting correlations between gene mutations and clinical phenotypes of PK ( Table 1 and Figure 4 ) . First , giant plaque-type porokeratosis ptychotropica ( PPt ) with lesion diameters at least 5 cm appeared to be a unique phenotype associated with MVK mutations . This feature was observed in 50% ( 19/38 ) of index patients with MVK mutations , but not in any patients with MVD , PMVK , or FDPS mutations . In addition , patients with MVK mutations generally showed the widest range of phenotypes in terms of both the number and size of lesions . Second , localized genital PK ( Chen et al . , 2006 ) and porokeratoma ( Walsh et al . , 2007 ) appeared to be unique phenotypes associated with PMVK mutations . Third , in PK patients with MVD mutations , the age of onset spanned from 5 to 70 years , and the diameter of the lesions was generally less than 2 cm . Five females and one male with MVD mutations manifested mild solar facial PK ( Sharquie and Al-Baghdady , 2003 ) , which was not found in any patients with mutations in other genes . Fourth , in patients with FDPS mutations , the number of lesions was generally more than 500 and the diameter of the lesions was less than 1 cm . 10 . 7554/eLife . 06322 . 012Table 1 . Clinical characteristics and genetic causes of 134 index patients with PKDOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 012Genetic causes of 134 index patients with PKMVK ( 39* ) PMVK ( 9 ) MVD ( 62* ) FDPS ( 4 ) Unknown ( 21 ) Sex Male27938117 Female1202434Number of lesions 0–1044006 10–1002051305 100–500503905 >500†1001045Diameter of lesions Minimum2 mm5 mm1 mm1 mm1 mm Maximum20 cm5 cm2 cm1 cm2 cmAge of onset At birth00003 0–202342010 20–401232819 40–60311225 >6011204Variants of PK DSAP/DSP26056410 SFP00600 PM239006 HPM133100 Giant plaque of PPt190000 Genital PK ( localized ) 04000 Porokeratoma05000 LP11305Comorbidity Psoriasis vulgaris40200*One PK patient ( proband of family-28 , female ) , who has both the mutation c . 235G>A ( MVK ) and the mutation c . 746T>C ( MVD ) , was included in both MVK and MVD groups . †The number of lesions is more than 500 . DSAP , disseminated superficial actinic porokeratosis; DSP , disseminated superficial porokeratosis; HPM , hyperkeratotic porokeratosis , LP , linear porokeratosis , PK , porokeratosis; PM , porokeratosis of Mibelli , PPt , porokeratosis ptychotropica; SFP , solar facial porokeratosis . 10 . 7554/eLife . 06322 . 013Figure 4 . Representative clinical phenotypes and histopathology associated with the four genotypes . From left to right , pedigree charts , clinical phenotypes and the corresponding histopathology photos are shown correspondingly . ( A ) Family ( F ) -42 proband with MVK mutation showed giant hyperkeratotic plaque-type porokeratosis ptychotropica . ( B ) F-60 proband with PMVK mutation showed tumor-like porokeratoma in the genitogluteal region . ( C ) F-36 proband with MVD mutation showed discrete , red-brown annular keratotic papules or maculopapules on the chest . ( D ) F-47 proband with FDPS mutation showed multiple , small , superficial , annular papules with thread-like ridges on the legs . All histopathology showed cornoid lamella , a histological hallmark of porokeratosis with vertical columns of parakeratosis overlying an area of hypogranulosis with dyskeratotic cells . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 013 As a whole , the lesions harboring mutations in MVD and FDPS genes tended to be more homogeneous and superficial than those carrying mutations in MVK and PMVK genes . The phenotypes of MVD and FDPS mutations on the natal cleft were mild and unobvious . However , some patients harboring MVK and PMVK mutations firstly presented with untypical lesions of genitogluteal porokeratosis ( GGP ) . The expression of the wild allele in 10 out of 13 LTs was significantly reduced at the 1% significance level ( Figures 5–7 , Supplementary file 3 ) . In contrast , the expression of the mutant allele was significantly reduced in almost all NNS ( except one from F-42 ) carrying a mutation resulting in nonsense , frameshift , or splicing defect . As for LT from F-31 , the mutant to wild allelic ratio in genomic DNA was >3 compared with the ratio of about 1 in the corresponding NNS genomic DNA . However , no deletion or duplication of all MVK exons was detected in F-31 , which might indicate that gene conversion of the wild to the mutant allele occurred in the LT ( Figure 5 ) . Gene conversion might also occur in the thigh LT of F-43 ( Figure 7 ) . No copy number change or CpG island methylation was detected at the corresponding mutated genes in all LTs . A mutation of c . 1003G>A ( p . Gly335Ser ) on the wild allele ( T ) of MVK: c . 1093T>A was identified in the cDNA of the LT without AEI but not the cDNA of NNS collected from the left forearm of F-38 ( Figure 6 ) . Surprisingly , this mutation of c . 1003G>A was not identified in the genomic DNA of this LT either , which indicated that G-to-A RNA editing occurred on the wild allele of MVK: c . 1093T>A . No mutation was observed in cDNA samples from other tissues . 10 . 7554/eLife . 06322 . 014Figure 5 . Gene conversion of the wild to mutant allele was identified in a buttock lesion from F-31 carrying a MVK mutation of c . 395delT . ( A ) The mutant/wild allelic ratios in genomic DNA ( gDNA ) and complementary DNA ( cDNA ) of lesional tissue ( LT ) and neighboring normal-appearing skin ( NNS ) . The quantity of the mutant allele was about threefold and 10-fold more than the quantity of the wild allele in gDNA and cDNA , respectively . ( B ) The chromatograms of single nucleotide extension targeting the c . 395delT mutation using the SNaPshot kit for five DNA samples ( blood gDNA , NNS gDNA , LT gDNA , NNS cDNA and LT cDNA ) . The mutant peak was overpresented in both LT gDNA and LT cDNA . ( C ) No copy number change in genomic DNA of NNS and LT . PA and PB are probes in the promoter region , E01 to E11 designates exon 1 to exon 11 , and A or B indicate two different probes in the same exon . ( D ) The bisulfite sequencing of NNS gDNA and LT gDNA . No methylation was observed for the targeted CpG sites in the promoter region of MVK . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 01410 . 7554/eLife . 06322 . 015Figure 6 . G-to-A RNA editing at position 1003 of the wild allele ( T ) of c . 1093T>A in MVK was detected in a left forearm lesion from F-38 . ( A ) No allelic expression imbalance was observed in lesional tissue ( LT ) from F-38 . ( B ) A mutation of c . 1003G>A was identified in LT cDNA , but not in neighboring normal-appearing skin ( NNS ) cDNA or LT gDNA . ( C ) Sequencing the c . 1093A and c . 1093T allele-specific PCR products indicated the mutant allele ( A ) of c . 1003G>A was in the cis position with the wild allele ( T ) of c . 1093T>A . A-SPCR , c . 1093A–specific PCR; T-SPCR , c . 1093T-specific PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 01510 . 7554/eLife . 06322 . 016Figure 7 . Significantly reduced expression of the wild allele in other nine lesion tissues . The Student t-test was performed to measure the difference in mutant/wild allelic ratios in genomic DNA ( gDNA ) or complementary DNA ( cDNA ) from lesion tissue ( LT ) and neighboring normal-appearing skin ( NNS ) for each mutation . The test score ( p value ) is presented above the LT bars . The asterisk ( * ) designates a significance level of 1% . F-43a and F-43b indicates the tissue sets of the left forearm and left thigh , respectively , from the same F-43 patient . DOI: http://dx . doi . org/10 . 7554/eLife . 06322 . 016
In this present study we were able to identify mutations in four genes ( MVK , PMVK , MVD , and FDPS ) underlying PK by massively parallel sequencing and copy number variation analysis in 134 index patients with PK . In the 12 PK families where the affected members span three to six generations , we found mutations only in MVK or MVD but not in PMVK or FDPS ( Figure 1 , Figure 1—figure supplement 1 ) . We also observed the correlation of genotypes and several clinical manifestations in the PK patients ( Table 1 ) , which might be helpful to simplify the classification of PK under the guidance of genetic testing . Interestingly , two patients were identified to carry two mutations: c . 302C>G ( p . Pro101Arg ) and c . 683G>A ( p . Arg228Gln ) in MVD for S-62 , and c . 746T>C ( p . Phe249Ser ) in MVD and c . 235G>A ( p . Asp79Asn ) in MVK for F-28 . According to the prediction programs of bioinformatics , the c . 235G>A in MVK and c . 302C>G in MVD were less conservative than the other mutation carried in F-28 and S-62 , respectively . The two mutations in MVD for S-62 were located in the trans position , for his daughter carried only one of them ( Figure 1—figure supplement 2 ) . S-62 had late-onset PK at the age of 50 and his 30-year-old daughter had no lesions to date . According to our observation , index patients F-28 and S-62 did not present with more severe phenotypes or earlier age of onset or any other unique clinical features . In this study we did not detect pathogenic mutation in 21 patients: five cases with linear PK ( LP ) , six cases with solitary plaque-type PK , and 10 cases with disseminated PK ( only one had family history ) . There were three possible explanations: ( 1 ) these patients might have other non-classical or mosaic/somatic mutations that could not be detected by current mutation screening methods; ( 2 ) these patients might have mutations in unknown genes; ( 3 ) these patients might be acquired ISIP without mutation . Recently , mutations in the SLC17A9 gene were reported to be another genetic cause for DSAP ( Cui et al . , 2014 ) . We sequenced all exons of SLC17A9 in these 21 patients by Sanger sequencing but found no mutations in SLC17A9 . RNA editing is a process of some discrete changes to specific nucleotide sequences of a RNA molecule which may be critical during normal development and diseases ( Witkin , et al . , 2015 ) . G-to-A RNA editing in PK was first reported in our study . Interestingly , the RNA editing seemed to occur selectively on the one copy of MVK and this selective editing pattern seemed to be capable of being transmitted during mitosis . The mechanism underlying this observation should be investigated further . Using AEI assays , we observed that the expression of wild allele was significantly reduced to various extents in most LTs ( >77% ) . The variation in the expression reduction level might be due to the LT containing various amounts of normal-appearing keratinocytes and other different types of cells . Somatic loss of function of the wild allele in the epidermal stem cells might play a key role in the formation of lesions , and it might occur spontaneously and be induced by environmental factors such as ultraviolet light . There was only one RNA editing identified in the LT , therefore the somatic loss of function of the wild allele in LTs of PK could be mainly caused by suppression of the wild allele through gene conversion in a minor proportion of cases or an unknown DNA methylation-independent epigenetic mechanism in most cases . Allelic imbalance of histone methylation modification unlinked to DNA methylation , whose pattern is able to be transmitted during cell mitosis , could be the DNA methylation-independent epigenetic mechanism ( Xu et al . , 2011 ) . Further investigation would be necessary to test this hypothesis . We did not find reduced expression or somatic mutation of the wild allele in only two LTs . The reason for this could be that the LT might contain many normal cells affecting the sensitivity of AEI and mutation detection . As a starter button of PK , gene conversion of the wild allele to the mutant allele was first identified in this study . This observation is generally consistent with the phenomenon known as ‘revertant mosaicism’ , in which the mutant allele is corrected in some stem cells . Revertant mosaicism is mainly identified in blood diseases such as Wiskott–Aldrich syndrome and Fanconi’s anemia , and skin diseases such as ichthyosis and epidermolysis bullosa ( May , 2011 ) . Revertant mosaicism shows the bright side of gene conversion , but our observation revealed the opposite side . Therefore , gene conversion could be a natural ‘rescue angel’ in some recessive diseases or a ‘damaging devil’ in some dominant diseases , which could mostly affect tissues with extensive ability to self-renew for life . Our findings support the notion that Mendelian diseases can be caused by multiple genes involved in the same metabolic or signal transduction pathway , or producing proteins on interaction . The four genes MVK , PMVK , MVD , and FDPS are members of the mevalonate pathway , suggesting that this pathway is involved in the pathogenesis of PK . The mevalonate pathway of isoprenoid biosynthesis provides precursors of isoprenoids , which are ubiquitous in living species and diverse in biological function ( Smit and Mushegian , 2000 ) and serve as precursors of cholesterol , heme A , ubiquinones , dolichol , and isoprenylated proteins ( e . g . , RAS and Lamin B ) ( Figure 2 ) . Moreover , it is also the subject of many pharmacological regulating drugs such as statins . It is well known that cholesterol is an important constituent of the cell membrane of most eukaryotic cells , and several inherited disorders have been linked to defects in cholesterol biosynthesis . The isoprenylated proteins play a role in the regulation of cell growth , division , and differentiation ( Moir and Spann , 2001; Zwerger et al . , 2011; Davidson and Lammerding , 2014 ) and are probably associated with the retained nuclei in the stratum corneum ( i . e . , parakeratosis ) . We propose that the accumulation of abnormal metabolites or shortage of isoprenoids might predispose patients to idiopathic inflammation of the skin , and that correction of the abnormal isoprenoid biosynthesis might be a novel therapeutic direction for PK . Further functional studies will be necessary to confirm this hypothesis .
All procedures followed the guidelines of the Helsinki Declaration and were approved by the ethics committee and by the Scientific Ethical Committee of Fudan University . Participating centers provided local Institutional Review Board approval for genetic analysis . Study participants provided informed consent for genetic testing . From 2001 to 2013 , peripheral blood samples were collected from 134 index patients with PK ( 92 males and 42 females ) and 180 family members ( 103 males and 77 females ) . Thirteen pairwise LTs and NNS were dissected from 12 index patients with a MVK , MVD , or FDPS mutation ( Supplementary file 3 ) . In addition , blood samples were collected from 270 healthy adult individuals undergoing routine medical examination in the hospital . The 134 index patients were diagnosed by at least two experienced dermatologists , based on both clinical features and histological examinations . In view of the family history , these index patients were divided into family ( F ) and sporadic ( S ) cases . Among the 61 familial cases , 40 index patients were probands from each family . We examined members of 21 families from different provinces of China , whose generations ranged from 2 to 6 . Genomic DNA was extracted from the peripheral blood using the QIAamp DNA blood mini kit ( Qiagen , Germany ) . Total RNA was isolated from the skin using the RNeasy Protect Mini Kit ( Qiagen , Germany ) according to the manufacturer's instructions . PCR products were sequenced by BigDye3 . 1 ( Applied Biosystems , Foster City , CA , USA ) . Raw data were collected on an ABI 3130XL sequencer and mutations were called by the PolyPhred program ( Nickerson et al . , 1997 ) . The 60 SNVs in 270 control samples were genotyped using the ABI PRISM SNaPshot Multiplex Kit ( Applied Biosystems , Foster City , CA , USA ) and ABI3730xl sequencer . A total of 162 fragments covering the promoter region , 5′UTR , coding sequence and splicing site of 12 genes in the mevalonate pathway ( ACAT1 , ACAT2 , HMGCS1 , HMGCS2 , HMGCR , MVK , PMVK , MVD , IDI1 , IDI2 , FDPS , and GGPS1 ) were amplified using the EasyTarget® amplification kit ( Genesky Biotechnologies Inc , Shanghai , China ) which was developed according to the CPELA method . The CPELA was a new method developed by Genesky Biotechnologies for fast and simple enrichment of multiple gene regions for massively parallel sequencing . The principle of CPELA is described in Figure 3—figure supplement 2 . The amplification products by CEPLA were mixed and size-separated by 2% agarose gel electrophoresis . Products of length 200–400bp were recovered . The final concentration of the library mixture was determined by real-time quantitative PCR and the average concentration for each library was estimated . The libraries tagged with different index sequences from several projects were well mixed in an appropriate concentration for each library corresponding to its aimed sequencing depth . The final library mixture was sequenced on MiSeq sequencer ( Illumina ) using Miseq reagent kit v2 . The sequencing reads were separated for each sample by running CASAVA ( Illumina Inc , San Diego , CA , USA ) and target reads were determined by comparing them with fragment reference sequences ( hg19 ) using the Blat program ( Kent , 2002 ) . In order to reduce the reading error from the sequencing reaction step , we first compared the two reads from the same cluster and integrated them into one correction read . The correction reads were then aligned to hg19 using the Burrows–Wheeler Aligner ( BWA ) ( Li and Durbin , 2010 ) . SNV calling was performed using both GATK ( McKenna et al . , 2010 ) and Varscan programs ( Koboldt et al . , 2012 ) , and the called SNV data were then combined . The Annovar program was used for SNV annotation ( Wang et al . , 2010 ) . The functional effect of non-synonymous SNVs was assessed by the PolyPhen-2 , SIFT , and MutationTaster ( Ng and Henikoff , 2003; Adzhubei et al . , 2010; Schwarz et al . , 2010 ) . Non-synonymous SNVs with SIFT score of <0 . 05 , Polyphen-2 score of >0 . 85 or MutationTaster score of >0 . 85 were considered as significant of not being benign . To sort potentially deleterious variants from benign polymorphisms , perl scripts were used to filter the SNVs against those of dbSNP135 . Any SNV recorded in dbSNP135 and with a minor allele frequency of ≥1% in Chinese from 1000 genome database was considered as benign polymorphisms and therefore removed for subsequent analysis . The copy number of the target regions was measured by a CNVplex assay , a high-throughput multiplex CNV analysis method recently developed by Genesky Biotechnologies . The principle of CNVplex technology is described in Figure 3—figure supplement 3 . We utilized this technology for quantitative analysis of copy numbers of all 37 exons and upstream promoter regions in the MVK , MVD , PMVK , and FDPS genes for the blood DNA samples with no point mutation identified and the tissue DNA samples . Based on the copy number measurements for all target sequences , the breakpoints were estimated to be located between two neighboring probe target sites showing different copy numbers . Several primer sets flanking the two probe target sites were tested to amplify the target region from case and control DNA samples using a long PCR protocol . Specific PCR products from case samples were sequenced using the ABI BigDye3 . 1 and the breakpoints were determined by blasting the sequences with human reference genome assembly . Both DNA and RNA were extracted from pairwise LT and NNS . Each RNA sample was reversely transcripted into cDNA twice using polyA and N9 primer mix and Reverse Transcriptase M-MLV ( RNase H- ) ( New England Biolabs , England ) . For each pairwise tissue set , single nucleotide extension was used to quantitate the ratio of the mutant to the wild allele both in the tissue cDNA and DNA using the ABI PRISM SNaPshot Multiplex Kit ( Applied Biosystems ) , followed by normalization to the ratio value in the corresponding patient’s blood DNA as a reference of 1:1 . As for cDNA from F-42 and S-36 patients with splicing defect mutations of MVK: c . 371+2T>A and FDPS: c . 684+1G>A respectively , fluorescent PCR followed by capillary electrophoresis was used to quantitate the ratios of mutant to wild transcripts . Total RNA was reversely transcribed into cDNA . Two or three sets of primers were designed to amplify the cDNA of MVK , MVD , or FDPS , covering all coding sequences for each gene . The sequences of PCR products were determined by Sanger sequencing . Two allele-specific PCRs for c . 1093T>A followed by Sanger sequencing were performed in order to determine whether the mutant allele ( A ) of c . 1003G>A was in the cis position with the wild allele ( T ) of c . 1093T>A . DNA from each pairwise LT and NNS were subject to bisulfite conversion using the EZ DNA Methylation Kit ( Zymo Research Corporation , Irvine , CA , USA ) . PCR products amplifying the bisulfite converted DNA at CpG islands of MVK , MVD , or FDPS were sequenced on MiSeq sequencer ( Illumina ) using Miseq reagent kit v2 or on the ABI 3130xl genetic analyzer using the ABI BigDye3 . 1 . | Porokeratosis refers to a group of around twenty skin conditions that involve a build-up of a protein called keratin in skin cells . Keratin forms the tough fibres that give strength to hair and nails , and people suffering from porokeratosis develop hardened skin lesions . Porokeratosis is an uncommon condition; most cases are inherited and have been linked to exposure to ultraviolet light and having a weakened immune system . Mutations in one gene called MVK are known to cause two forms of the disorder , but it is suspected that other genetic causes of porokeratosis will also be identified . The MVK gene encodes an enzyme that is involved in making chemicals called isoprenoids . This large and diverse class of chemicals provides the building blocks for making many other important molecules in all living species . Zhang , Li et al . have now analysed genetic material from 134 different porokeratosis patients to search for mutations in other genes involved in the production of isoprenoids . The patients examined include 61 people with a family history of the disorder , and 73 cases in which the condition seems to be a one-off occurrence . This search identified mutations in three additional genes ( called PMVK , MVD and FDPS ) that are all linked to porokeratosis . Further analysis of these three genes and MVK revealed that about half of the patients with mutations in the MVK gene developed large lesions ( that were over 5 centimetres in diameter ) . However , those with mutations in the other three genes did not develop such large lesions . Mutations in some of the newly identified genes were instead linked to porokeratosis affecting specific areas of the body; for example , PMVK and MVD mutations are linked to porokeratosis localized to the genitals and around the eyes , respectively . This means that , in the future , doctors might be able to simplify the diagnosis of the different varieties of porokeratosis based on information gained via genetic tests . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"genetics",
"and",
"genomics"
] | 2015 | Genomic variations of the mevalonate pathway in porokeratosis |
The experience of rewarding or aversive stimuli is encoded by distinct afferents to dopamine ( DA ) neurons of the ventral tegmental area ( VTA ) . Several neuromodulatory systems including oxytocin regulate DA neuron excitability and synaptic transmission that process socially meaningful stimuli . We and others have recently characterized oxytocinergic modulation of activity in mouse VTA DA neurons , but the mechanisms underlying oxytocinergic modulation of synaptic transmission in DA neurons remain poorly understood . Here , we find that oxytocin application or optogenetic release decrease excitatory synaptic transmission , via long lasting , presynaptic , endocannabinoid-dependent mechanisms . Oxytocin modulation of excitatory transmission alters the magnitude of short and long-term depression . We find that only some glutamatergic projections to DA neurons express CB1 receptors . Optogenetic stimulation of three major VTA inputs demonstrates that oxytocin modulation is limited to projections that show evidence of CB1R transcripts . Thus , oxytocin gates information flow into reward circuits in a temporally selective and pathway-specific manner .
Dopamine ( DA ) neurons in the ventral tegmental area ( VTA ) play a pivotal role in signaling reward-related stimuli ( Wise , 2004 ) . Alterations in the function of VTA DA neurons have been linked to drug abuse ( Kauer and Malenka , 2007; Niehaus et al . , 2009 ) , as well as neurodevelopmental and psychiatric disorders including autism spectrum disorders ( ASDs ) ( Bariselli et al . , 2016 ) , schizophrenia , and depression ( Grace , 2016 ) . DA neurons exhibit two activity patterns in vivo – low frequency tonic firing ( 1–5 Hz ) and burst , or phasic , firing patterns ( Floresco et al . , 2003; Grace and Bunney , 1984 ) . Phasic activity of DA neurons is observed when an animal is in an environment with strong reward salience ( Schultz , 1998; Schultz et al . , 2015 ) , resulting in large transient increases in DA concentration in the ventral striatum . These DA increases have been directly assayed using fast-scan cyclic voltammetry ( Tsai et al . , 2009 ) and are well-matched by recent optical imaging studies that demonstrate unpredicted reward-associated calcium transients in the axons of VTA DA neurons expressing genetically encoded calcium sensors ( Howe and Dombeck , 2016 ) . Glutamatergic afferents to DA neurons are thought to control transitions between tonic and phasic activity ( Canavier and Landry , 2006; Floresco et al . , 2003; Overton and Clark , 1997 ) . For the VTA , these inputs arise from multiple brain regions , including the medial prefrontal cortex ( mPFC ) , lateral habenular nucleus ( LHb ) and pedunculopontine tegmental nucleus ( PPN ) ( Geisler et al . , 2007; Good and Lupica , 2009; Morales and Margolis , 2017 ) . Neuronal activity and glutamatergic synaptic transmission in the VTA are tightly regulated by a diverse group of neuromodulators acting on G protein-coupled receptors ( GPCRs ) . VTA DA neurons express numerous GPCRs , including dopamine Drd2 receptor ( Vallone et al . , 2000 ) , as well as receptors for serotonin ( Doherty and Pickel , 2000 ) , corticotropin-releasing factor ( Ungless et al . , 2003 ) , orexin ( Korotkova et al . , 2003 ) , vasopressin , and oxytocin ( Skuse and Gallagher , 2009; Xiao et al . , 2017 ) , among others . Several GPCR families are known to modulate neuronal activity and synaptic transmission in VTA DA neurons ( Bonci and Malenka , 1999; Borgland et al . , 2006; Ungless et al . , 2003 ) , but our understanding of how neurohypophyseal peptides control synaptic transmission in the midbrain dopamine system is limited . Oxytocin is one of the two major neurohypophyseal peptides that function centrally as a neuromodulator and has been linked to the processing of socially rewarding stimuli via actions in the striatum and in the neocortex ( Choe et al . , 2015; Dölen et al . , 2013; Marlin et al . , 2015 ) . In the nucleus accumbens , oxytocin weakens glutamatergic synaptic transmission through presynaptic mechanisms involving serotonergic inputs from the dorsal raphe nucleus ( Dölen et al . , 2013 ) , and in the mPFC , oxytocin application dampens glutamatergic synaptic transmission through presynaptic endocannabinoid signaling ( Ninan , 2011 ) . Yet , in the auditory and piriform cortex , oxytocin application or evoked release dampen inhibitory transmission without altering excitatory input processing ( Mitre et al . , 2016 ) . Oxytocin neurons residing in the paraventricular nucleus of the hypothalamus ( PVN ) directly project to the VTA and activate oxytocin receptors ( OxtRs ) to regulate the tonic activity of DA neurons and social behavior ( Beier et al . , 2015; Hung et al . , 2017; Tang et al . , 2014; Xiao et al . , 2017 ) . Consistent with oxytocin-driven enhancement in the activity of VTA DA neurons , oxytocin infusions into the VTA potentiate social reward ( Song et al . , 2016 ) and modulate reward-based behavior ( Mullis et al . , 2013 ) . Given these observations , we reasoned that oxytocin is poised to regulate synaptic transmission in the VTA . Recently , OxtR agonist was found to regulate synaptic inputs to a specific subpopulation of VTA DA neurons that project to medial nucleus accumbens , altering the balance of excitation and inhibition through cellular mechanisms that remain incompletely understood ( Hung et al . , 2017 ) . Here , we delve deeper into the function of oxytocin in regulating evoked excitatory postsynaptic currents ( EPSCs ) in genetically targeted VTA DA neurons by relying on pharmacological methods , optogenetics , retrograde tracing , quantitative triple-channel in situ hybridization , and 2-photon imaging with single synapse 2-photon uncaging , or photorelease , of glutamate . We find that oxytocin inhibits excitatory synaptic transmission through OxtRs and retrograde endocannabinoid signaling in a cell-autonomous manner , with timing and effect magnitudes sufficient to modulate both short- and long-term plasticity in VTA DA neurons .
To determine whether oxytocin regulates synaptic transmission in VTA DA neurons we carried out voltage-clamp recordings of tdTomato+ or eYFP+ neurons in horizontal slices prepared from P25-40 Slc6a3i-Cre; Ai14 or Ai32 reporter mice ( Figure 1A ) . Many DA neurons in ventromedial VTA express oxytocin receptors ( Figure 1—figure supplement 1A–C ) , and the majority of DA neurons in this region increase tonic activity in response to oxytocin application ( Tang et al . , 2014; Xiao et al . , 2017 ) , so we targeted these cells for recordings in this study . Excitatory postsynaptic currents ( EPSCs ) were evoked by electrical stimulation at a holding potential of −70 mV in the presence of 10 μM GABA ( A ) receptor antagonist ( SR95531 ) at room temperature . Evoked EPSCs were largely blocked by the AMPA receptor antagonist NBQX and fully abolished following subsequent application of the NMDA receptor antagonist CPP ( Figure 1—figure supplement 1D ) . Bath application of 1 μM oxytocin significantly reduced the amplitude of evoked EPSCs ( 77 . 43 ± 1 . 55% of baseline , p<0 . 001 , paired t-test , n = 15 neurons from 10 mice ) ( Figure 1A , B and F ) , without changing the input resistance of VTA DA neurons or decay time constant of EPSCs ( Figure 1—figure supplement 1E–F ) . This oxytocin-induced EPSC depression in VTA DA neurons was similar in males and females ( ♂ , 79 . 12 ± 5 . 58% of baseline , n = 5 neurons; ♀ , 77 . 36 ± 3 . 39% of baseline , n = 10 neurons . p=0 . 440 , Mann-Whitney test ) , consistent with a prior report on sex invariance of oxytocinergic regulation of midbrain DA systems in structure and function ( Xiao et al . , 2017 ) . To build a dose-response curve we also evaluated the effect of oxytocin on excitatory synaptic transmission at concentrations below 1 µM ( 0 nM , 10 nM , 100 nM , and 500 nM ) , and observed that oxytocin application in concentrations at or above 100 nM is sufficient to decrease EPSC amplitude in a dose-dependent manner ( 0 nM , 103 . 50 ± 3 . 05% of baseline; 10 nM , 94 . 74 ± 1 . 38% of baseline; 100 nM , 87 . 73 ± 0 . 88% of baseline; 500 nM , 83 . 31 ± 3 . 18% of baseline , p<0 . 01 , Kruskal-Wallis test with Dunn’s Multiple Comparison post hoc test , n = 5–8 neurons from 2 to 5 mice/group ) ( Figure 1F , Figure 1—source data 1 ) . These results suggest that although oxytocin release or application enhance spontaneous tonic activity of VTA DA neurons ( Tang et al . , 2014; Xiao et al . , 2017 ) , the same peptide dampens excitatory synaptic transmission in VTA DA neurons . In the central nervous system , oxytocin primarily binds to a single variant of the oxytocin receptor ( OxtR ) to regulate neuronal activity ( Gimpl and Fahrenholz , 2001; Stoop , 2012 ) . OxtR is expressed in some VTA DA neurons ( Hung et al . , 2017; Vaccari et al . , 1998; Xiao et al . , 2017 ) . We next tested whether oxytocin regulates the excitatory synaptic transmission to VTA DA neurons by activating OxtR . In the presence of OxtR antagonist ( 10 μM L368 , 899 ) we observed that bath application of 1 μM oxytocin did not have a significant effect on the amplitude of evoked EPSCs ( 98 . 89 ± 3 . 39% of baseline , p=0 . 844 , Wilcoxon matched-pairs signed rank test , n = 8 neurons from six mice ) ( Figure 1C and G , Figure 1—source data 1 ) . OxtR antagonist application alone had no effect on the amplitude of evoked EPSCs ( Figure 1—figure supplement 1G ) , which suggests that no oxytocin tone is present in the acute brain slice of the VTA . OxtRs primarily couple to Gαq protein effectors , which activate the phospholipase C ( PLC ) pathway to mobilize intracellular Ca2+ store release ( Gimpl and Fahrenholz , 2001; Stoop , 2012 ) . We evaluated the contribution of the PLC pathway to the observed dampening of synaptic transmission using a selective PLC inhibitor U73122 . In the presence of 10 μM U73122 , oxytocin failed to decrease evoked EPSC amplitude ( 98 . 92 ± 5 . 33% of baseline , p=0 . 949 , Wilcoxon matched-pairs signed rank test , n = 5 neurons from four mice ) ( Figure 1D and G , Figure 1—source data 1 ) . Thus , oxytocin binds to Gαq protein-coupled OxtR to regulate synaptic transmission in VTA DA neurons through the canonical PLC pathway . In addition to OxtR , oxytocin has been shown to bind structurally related vasopressin receptors ( Gimpl and Fahrenholz , 2001; Tribollet et al . , 1988 ) . Indeed , vasopressin 1a receptor ( V1aR ) transcripts are present in the rodent VTA , including in DA neurons ( Dubois-Dauphin et al . , 1996; Xiao et al . , 2017 ) . To determine whether oxytocin modulates evoked EPSCs by binding to V1aR we used a potent and selective V1aR antagonist ( 10 μM SR49059 ) in the bath solution . Under these conditions , oxytocin application decreased the amplitude of the evoked EPSCs to the same degree as in the baseline condition ( 80 . 94 ± 2 . 33% of baseline , p<0 . 001 , Wilcoxon matched-pairs signed rank test , n = 9 neurons from six mice ) ( Figure 1E and G , Figure 1—source data 1 ) . To determine whether endogenous oxytocin release is sufficient to regulate synaptic transmission in VTA DA neurons , we relied on the Oxti-Cre mice injected with a Cre-dependent rAAV expressing a fusion of ChR2 and the red fluorescent protein mCherry ( ChR2-mCherry ) into the PVN ( Figure 1H ) ( Xiao et al . , 2017 ) . PVN oxytocin neuron projections that co-release glutamate have been reported in the parabrachial nucleus ( Ryan et al . , 2017 ) and in brainstem vagal neurons ( Piñol et al . , 2014 ) . To evaluate potential co-release of glutamate in the VTA , we used single 10 ms-long light pulses to optically stimulate PVN oxytocin fibers while recording VTA DA neurons . All recorded neurons were validated using post hoc immunolabeling against tyrosine hydroxylase ( TH ) , the rate-limiting enzyme for canonical dopamine synthesis ( Figure 1H ) ( Xiao et al . , 2017 ) . In contrast to parabrachial and vagal neurons , no fast responses in VTA DA cells were observed ( light-evoked response change −0 . 556 ± 0 . 285 pA , p=0 . 303 , Wilcoxon matched-pairs signed rank test; n = 16 neurons from five mice ) . We then used trains of 10 ms-long light pulses at 20 Hz delivered in 30 second-long bursts to evoke axonal release of oxytocin directly in the acute slices from VTA ( Knobloch et al . , 2012; Xiao et al . , 2017 ) . Electrically evoked EPSCs were recorded before and after optical stimulation of oxytocin-positive axons ( Figure 1I , Figure 1—source data 1 , Figure 1—figure supplement 2 ) . Optical stimulation of oxytocin fibers depolarized VTA DA neurons in voltage-clamp recordings , evidenced by a small but consistent change in holding current ( 2 . 955 ± 0 . 871 pA , p<0 . 01 , Wilcoxon signed rank test , n = 15 neurons from eight mice ) , consistent with prior observations of tonic activity changes in VTA DA neurons in response to oxytocin . As for bath application experiments , we observed that the amplitude of evoked EPSC in VTA DA neurons was decreased , evident 30 s after optical stimulation ( 61 . 71 ± 4 . 39% of baseline , p<0 . 01 , Wilcoxon matched-pairs signed rank test , n = 9 neurons from six mice ) ( Figure 1I , Figure 1—figure supplement 2 ) . This repeatable effect was abolished in the presence of OxtR antagonist ( L368 , 899 ) ( 114 . 30 ± 8 . 04% of baseline , p=0 . 156 , Wilcoxon matched-pairs signed rank test , n = 6 neurons from four mice ) ( Figure 1I , Figure 1—figure supplement 2 ) . We also observed that 30 s long light activation of oxytocinergic fibers in VTA is sufficient to lead to a relatively long-lasting decrease in the amplitude of EPSCs ( 5–15 min after light stimulation: 69 . 22 ± 6 . 72% of baseline; 15–25 min after light stimulation: 57 . 77 ± 7 . 01% of baseline , p<0 . 01 , Kruskal-Wallis test with Dunn’s Multiple Comparison post hoc test , n = 6 neurons from four mice ) ( Figure 1J–K , Figure 1—source data 1 ) . Both presynaptic neurotransmitter release and postsynaptic receptor properties contribute to the strength of synaptic transmission . We first assessed whether oxytocin regulates excitatory synaptic transmission by decreasing glutamate release . To accomplish this we started by measuring the paired-pulse ratio ( PPR ) following two electrical stimuli ( 50 ms interstimulus interval , ISI ) . Oxytocin application significantly increased PPR at a 50 ms ISI but not 80 or 100 ms ISI ( Baseline , 1 . 120 ± 0 . 057; Oxt , 1 . 721 ± 0 . 116; p<0 . 01 , Wilcoxon matched-pairs signed rank test , n = 8 neurons from six mice ) ( Figure 2A–B , Figure 2—source data 1 , Figure 2—figure supplement 1 ) , indicating that oxytocin decreases the probability of glutamate release . Blocking either OxtR or the PLC pathway abolished the effect on PPR . However , in the presence of V1aR antagonist , oxytocin application still significantly increased PPR ( Baseline , 0 . 948 ± 0 . 076; Oxt , 1 . 326 ± 0 . 149; p<0 . 01 , Wilcoxon matched-pairs signed rank test , n = 8 neurons from five mice ) ( Figure 2A–B ) . Optical activation of oxytocin-positive axons in VTA was sufficient to increase PPR 30 s after light stimulation , consistent with the effect of light stimulation on EPSC amplitude changes ( Figure 2C , Figure 2—figure supplement 2 ) . These light-evoked changes in PPR , as for the amplitude attenuation , were likewise blocked by the OxtR antagonist ( Control: Baseline , 1 . 220 ± 0 . 127; Light , 1 . 634 ± 0 . 195; p<0 . 01 , Wilcoxon matched-pairs signed rank test , n = 9 neurons from six mice; L368 , 899: Baseline , 1 . 454 ± 0 . 232; Light , 1 . 147 ± 0 . 113; p=0 . 156 , Wilcoxon matched-pairs signed rank test , n = 6 neurons from four mice ) ( Figure 2D , Figure 2—source data 1 , Figure 2—figure supplement 2 ) . To further probe potential effects of oxytocin on presynaptic glutamate release , we evaluated spontaneous EPSCs ( sEPSC ) of VTA DA neurons . Bath application of oxytocin decreased sEPSC frequency ( Baseline , 2 . 328 ± 0 . 308 Hz; Oxt , 1 . 679 ± 0 . 237 Hz; p<0 . 05 , Wilcoxon matched-pairs signed rank test , n = 6 neurons from five mice ) , but had no significant effect on sEPSC amplitude ( Baseline , 7 . 741 ± 1 . 060 pA; Oxt , 8 . 830 ± 1 . 389 pA; p=0 . 094 , Wilcoxon matched-pairs signed rank test ) ( Figure 2E–F , Figure 2—source data 1 ) . Together , these results are consistent with the interpretation that oxytocin reduces the probability of glutamate release in excitatory inputs to VTA DA neurons . While oxytocin appears to alter presynaptic glutamate release , the experiments above do not exclude the possibility of oxytocin also regulating postsynaptic receptor properties . In order to explicitly test this possibility , we carried out two-photon glutamate uncaging experiments to directly stimulate individual spiny protrusions on VTA DA dendrites using bath-applied MNI-L-Glutamate ( Figure 2G ) . VTA DA neurons from Slc6a3i-Cre; Ai14 mice were filled with Alexa Fluor 488 through the recording pipette . Dendritic spines were visualized using laser-scanning 2-photon imaging at 910 nm , and glutamate was uncaged via two brief pulses ( 0 . 5 ms duration , 50 ms ISI ) of 720 nm light with a second mode-locked Ti-Sapphire laser . Uncaging-evoked EPSCs ( uEPSCs ) were recorded at the soma at a holding potential of −70 mV in the presence of 10 μM GABA ( A ) receptor antagonist . The amplitude of uEPSCs evoked by the first light pulse recorded in the baseline condition was similar to that recorded in the presence of 1 μM oxytocin ( Figure 2H–I , Figure 2—source data 1 ) . Moreover , oxytocin did not alter the PPR of uEPSCs , evaluated using both paired measurements on single spines following oxytocin flow-in ( five dendritic spines from four neurons in three mice ) or in group data ( 17 dendritic spines for each condition from 16 neurons in four mice ) ( Figure 2H and J , Figure 2—source data 1 ) . These results suggest that oxytocin does not modulate glutamatergic receptor properties of VTA DA neurons , confirming its selective presynaptic effects on synaptic transmission . Previous studies demonstrate that increasing the activity of VTA DA neurons triggers these neurons to release endocannabinoids , which can activate presynaptic cannabinoid CB1 receptors to regulate neurotransmitter release ( Melis et al . , 2004; Oleson and Cheer , 2012; Riegel and Lupica , 2004 ) . Since oxytocin weakens evoked EPSCs in VTA DA neurons by reducing presynaptic glutamate release , retrograde endocannabinoid signaling could provide a possible mechanism for the observed effects . We also observed that the activation of CB receptors with a selective agonist ( WIN55212-2 ) decreases evoked EPSC amplitude ( 79 . 77 ± 1 . 75% of baseline , n = 7 neurons from five mice , p<0 . 01 , Wilcoxon signed rank test ) and increases PPR ( Baseline , 0 . 780 ± 0 . 047; WIN , 0 . 947 ± 0 . 061; n = 7 neurons from five mice , p<0 . 01 , Wilcoxon matched-pairs signed rank test ) ( Figure 3A , G , Figure 3—source data 1 , Figure 3—figure supplement 1 ) . In the presence of WIN55212-2 , occluded by CB1 receptor activation , oxytocin application failed to modulate the amplitude and PPR of evoked EPSCs in VTA DA neurons ( Figure 3B , G , Figure 3—source data 1 , and Figure 3—figure supplement 1 ) . Conversely , blocking cannabinoid CB1 receptors with a selective , high affinity CB1 receptor inverse agonist ( AM251 ) or antagonist ( CP945598 ) abolished oxytocinergic modulation of EPSC amplitude ( AM251: 97 . 12 ± 2 . 97% of baseline , p=0 . 438 , Wilcoxon signed rank test , n = 6 neurons from three mice; CP945598: 98 . 76 ± 2 . 70% of baseline , p=0 . 688 , Wilcoxon signed rank test , n = 6 neurons from four mice ) and of the PPR of evoked EPSCs ( Figure 3C , Figure 3G , Figure 3—source data 1 , and Figure 3—figure supplement 1 ) . The two major mammalian endocannabinoids are the bioactive lipid 2-arachidonoylglycerol ( 2-AG ) and anandamide , and previous evidence suggests that 2-AG is a primary endocannabinoid synthesized and released by VTA DA neurons ( Gantz and Bean , 2017; Mátyás et al . , 2008; Merrill et al . , 2015 ) . PLC cleaves phosphatidylinositol 1 , 4 , 5-bisphosphate into diacylglycerol ( DAG ) and inositol 1 , 4 , 5-trisphosphate; DAG is subsequently converted into 2-AG by DAG lipase ( Di Marzo et al . , 1998; Piomelli , 2003 ) . To determine the identify and function of endocannabinoids in oxytocinergic modulation of synaptic transmission in VTA DA neurons , we blocked the 2-AG synthesizing enzyme DAG lipase α ( DGLα ) with orlistat ( THL ) ( Gantz and Bean , 2017; Tanimura et al . , 2010 ) . Brain slices were treated with 1 μM THL for ~1 hr before and during whole-cell recording . Blocking 2-AG synthesis with THL abolished oxytocinergic effects on the amplitude of evoked EPSCs ( 99 . 40 ± 5 . 54% of baseline , p=0 . 938 , Wilcoxon signed rank test , n = 6 neurons from four mice ) ( Figure 3D , G , Figure 3—source data 1 ) and on the PPR ( Figure 3—figure supplement 1 ) . Since bath application of orlistat cannot distinguish cell-autonomous from circuit-level effects of inhibiting 2-AG synthesis , we next carried out experiments using intracellular inhibitors of PLC or 2-AG synthesis ( U73122 or THL ) in internal recording solution . Intracellular blockade of either the PLC pathway or 2-AG synthesis was sufficient to abolish oxytocinergic effects on the amplitude of evoked EPSCs and PPR ( Figure 3E–G , Figure 3—source data 1 , Figure 3—figure supplement 1 ) . Therefore , in the VTA , oxytocin activates the PLC/DAG lipase pathway to increase 2-AG release , weakening glutamatergic synaptic transmission in a cell-autonomous manner ( Figure 3H ) . Synaptic plasticity of glutamatergic transmission in VTA DA neurons provides a critical substrate relevant to learning and addiction ( Chen et al . , 2010; Kauer and Malenka , 2007; Niehaus et al . , 2009 ) , and endocannabinoids play a pivotal role in synaptic plasticity by depressing synaptic transmission on short- and long-term scales ( Gerdeman et al . , 2002; Di Marzo et al . , 1998 ) . Given the observed effects of oxytocin on endocannabinoid-dependent regulation of excitatory synaptic transmission and PPR , we further probed the potential for oxytocinergic modulation of plasticity-inducing stimuli in VTA DA neurons . Ten electrical stimuli at 50 ms ISI were used to induce short-term synaptic plasticity . The amplitude of EPSCs in most recorded VTA DA neurons ( 6/8 neurons ) decreased to a steady state when stimulated using this protocol in the control condition . During oxytocin perfusion , the magnitude of EPSC attenuation decreased , as measured by the ratio of last to first evoked EPSC amplitude . This decrement is accounted for by the drop in the amplitude of EPSCs evoked by the first two electrical stimulations ( Figure 4A–B , Figure 4—source data 1 ) . Oxytocin perfusion significantly increased the ratio between the amplitudes of the last and first EPSCs in a stimuli train ( Control: 0 . 655 ± 0 . 131; Oxt: 0 . 974 ± 0 . 228; p<0 . 05 , Wilcoxon matched-pairs signed rank test , n = 8 neurons from five mice ) ( Figure 4C , Figure 4—source data 1 ) . Given the relatively long-lasting , presynaptic effects of oxytocin on glutamatergic synaptic transmission , we questioned whether this modulatory mechanism can contribute to long-term depression ( LTD ) of excitatory inputs to VTA DA neurons , an important plasticity form for these cells . We used 10 Hz repetitive electrical stimulation ( 5 min ) or 100 Hz stimulation ( 1 second-long period , repeated 4 times with 10 second-long inter-train intervals ) in separate experiments to induce LTD ( Good and Lupica , 2010; Kreitzer and Malenka , 2005; Pan et al . , 2008 ) . In the control condition , EPSC amplitude decreased to 64 . 22 ± 4 . 70% of baseline following 10 Hz stimulation ( seven neurons from five mice ) ( Figure 4D–E , Figure 4—source data 1 ) and to 64 . 00 ± 8 . 19% of baseline following 100 Hz stimulation ( five neurons from five mice ) ( Figure 4F–G , Figure 4—source data 1 ) . In the presence of oxytocin , LTD magnitude induced by either stimulation protocol was substantially increased ( 10 Hz: 39 . 42 ± 6 . 19% of baseline , eight neurons from six mice , p<0 . 05 , Mann-Whitney test vs . control; 100 Hz: 34 . 77 ± 4 . 18% of baseline , six neurons from six mice , p<0 . 05 , Mann-Whitney test vs . control ) ( Figure 4D–H , Figure 4—source data 1 ) . Input resistance of VTA DA neurons was similar before and after LTD induction and was not altered by oxytocin ( Figure 4—figure supplement 1A and C ) . The PPR of EPSCs remained unchanged through LTD induction in the control condition ( Figure 4—figure supplement 1B and D ) , but increased following LTD induction in the presence of oxytocin for both stimulation protocols ( Figure 4—figure supplement 1B and D ) . To confirm the involvement of OxtRs and CB1Rs in the additive presynaptic component to LTD in the presence of oxytocin , brain slices were treated with 10 μM L368 , 899 or 5 μM AM251 for ~30 min before and during whole-cell recording . In the presence of oxytocin and either OxtR antagonist L368 , 899 or the CB1 receptor inverse agonist AM251 , LTD magnitude induced by 100 Hz stimulation was similar to that in the control condition and smaller than the LTD magnitude in the presence of oxytocin ( L368 , 899: 63 . 46 ± 8 . 48% of baseline , six neurons from four mice; AM251: 55 . 77 ± 4 . 69% of baseline , seven neurons from four mice , p<0 . 05 , Kruskal-Wallis test with Dunn’s Multiple Comparison post hoc test ) ( Figure 4F–H , Figure 4—source data 1 , and Figure 4—figure supplement 2 ) . Together , oxytocin modulates both short- and long-term synaptic plasticity in VTA DA neurons by presynaptic mechanisms mediated by endocannabinoids . While brain regions that send glutamatergic projections to the VTA – but not the VTA DA neurons themselves – are known to express the CB1 receptors ( Mátyás et al . , 2008 ) , it remains unknown whether specific groups of VTA-projecting neurons selectively express this receptor . To address this we carried out combined retrograde labeling using green retrobeads ( GRB ) injected into the VTA and quantitative multi-channel fluorescence in situ hybridization ( FISH ) assays for mRNA of CB1 receptor gene ( Cnr1 ) in several brain regions . Although there are not many published reports using the combination of retrobead injections with FISH assays , we found this combination robust and useful for evaluating transcripts within pathway-specific populations of neurons . Confirming prior reports ( Mátyás et al . , 2008 ) , we did not find substantial CB1 receptor expression in VTA DA neurons ( Figure 5A–B , Figure 5—source data 1 ) , even though DA neurons in the arcuate nucleus do express the CB1 receptor ( Figure 5—figure supplement 1A ) . DA neurons were identified by labeling for tyrosine hydroxylase ( Th ) , and few Th+ neurons in the VTA co-localized with Cnr1+ puncta ( <7% , n = 1258 Th+ neurons from three mice ) ( Figure 5B ) . Next , we quantified co-localization of GRB and Cnr1 signals with glutamatergic neurons , identified by the expression of Slc17a7 ( Vglut1 ) or Slc17a6 ( Vglut2 ) ( Figure 5C ) . We focused on three well-studied regions that project to the VTA: medial prefrontal cortex ( mPFC ) , lateral habenular nucleus ( LHb ) , and pedunculopontine nucleus ( PPN ) ( Beier et al . , 2015; Good and Lupica , 2010 ) . In the mPFC , Slc17a7 probe was used to label glutamatergic neurons; there , the vast majority of glutamatergic neurons projecting to VTA ( GRB+/Slc17a7+ ) co-localized with strong Cnr1 signal ( 143 GRB+/Slc17a7+/Cnr1+neurons of 188 GRB+/Slc17a7+ neurons from three mice ) ( Figure 5D–E , Figure 5—source data 1 , Figure 5—figure supplement 1B ) . We labeled glutamatergic neurons in LHb and PPN using a Slc17a6 probe , observing extensive overlap between GRB signal and Slc17a6 puncta in both of these regions . However , LHb had almost no Cnr1 expression , so triple-labeled neurons were rare ( 4 GRB+/Slc17a6+/Cnr1+neurons of 178 GRB+/Slc17a6+ neurons from two mice ) ( Figure 5D–E , Figure 5—source data 1 , Figure 5—figure supplement 1B ) . In contrast , over 70% of GRB+/Slc17a6+ neurons in the PPN co-localized with Cnr1 signal ( 306 GRB+/Slc17a6+/Cnr1+neurons of 401 GRB+/Slc17a6+ neurons from three mice ) ( Figure 5D–E , Figure 5—source data 1 , Figure 5—figure supplement 1B ) . The observation that CB1 receptor transcripts express in a subset of VTA glutamatergic afferents leads to a strong prediction that oxytocin modulates only those inputs , allowing CB1R-negative inputs to pass unchanged . To test this , we turned to optogenetic stimulation of specific glutamatergic afferents , using Slc17a6i-Cre mice , with a Cre-dependent rAAV expressing ChR2-mCherry delivered into the mPFC , LHb , or PPN , respectively ( Figure 6A ) . After 4–5 weeks , 2 ms 470 nm light was used to activate glutamatergic axons in the VTA while we recorded light-evoked EPSCs in VTA DA neurons ( validated with post hoc TH immunolabeling , Figure 6A ) . For neurons that responded to light stimulation with time-locked EPSCs , 1 μM oxytocin was applied following >10 min long baseline acquisition . Following mPFC targeting , 9/30 VTA DA neurons from four mice responded to light stimulation , and oxytocin application attenuated the amplitude of light-evoked EPSCs ( 73 . 35 ± 3 . 83% of baseline , p<0 . 05 , Wilcoxon signed rank test , n = 6 neurons from four mice ) ( Figure 6B–C , Figure 6—source data 1 , Figure 6—figure supplement 1 ) . After PPN transduction , 8/10 VTA neurons from two mice responded to light stimulation , and oxytocin also significantly dampened light-evoked EPSCs ( 82 . 55 ± 0 . 76% of baseline , p<0 . 05 , Wilcoxon signed rank test , n = 7 neurons from two mice ) ( Figure 6B–C , Figure 6—source data 1 , Figure 6—figure supplement 1 ) . In contrast , with ChR2 virus targeted to the LHb , while 8/21 VTA DA neurons responded to light stimulation , oxytocin had no effect on the amplitude of evoked EPSCs ( 94 . 84 ± 2 . 27% of baseline , p=0 . 219 , Wilcoxon signed rank test , n = 7 neurons from four mice ) ( Figure 6B–C , Figure 6—source data 1 , Figure 6—figure supplement 1 ) . These data likely indicate that distinct sources of glutamate converging on individual VTA DA neurons are differentially regulated by oxytocin based on CB1 receptor expression , but an alternative interpretation of these results is that VTA DA neurons receiving inputs from lateral habenula selectively lack OxtRs .
Oxytocin has been previously reported to modulate the excitability of VTA DA neurons ( Hung et al . , 2017; Tang et al . , 2014; Xiao et al . , 2017 ) , but how this peptide regulates synaptic transmission to these same DA neurons has not been investigated at the mechanistic level , precluding an integrative conceptual framework of peptidergic actions in midbrain DA systems . Here , we find that oxytocin inhibits excitatory synaptic transmission via OxtR and retrograde endocannabinoid signaling , modulating both short- and long-term plasticity in a subset of VTA DA neurons through presynaptic mechanisms . Furthermore , we observe a heterogeneity of endocannabinoid CB1 receptor expression among the glutamatergic regions projecting to the VTA . Specifically , VTA-targeting glutamatergic neurons from the mPFC and the PPN , but not the LHb , express the CB1 receptor . Our observations of pathway-specific modulation by oxytocin demonstrate a mechanism for selective gating of information flowing into VTA reward systems . For OxtR-expressing VTA DA neurons , one possibility is that CB1 receptor positive and negative glutamatergic signals converge onto single DA neurons , leading to selective filtering of inputs in the presence of oxytocin . However , an alternative hypothesis consistent with our observations is that some VTA DA neurons receive exclusively CB1 receptor positive or negative glutamatergic inputs . In addition , not all VTA DA neurons express oxytocin receptors . The details of topographic mapping between glutamatergic inputs and VTA DA neurons , with regard to their oxytocin responsiveness , remain to be characterized in future studies . Regardless of where the pathway-selectivity arises , evidence in support of pathway-specific oxytocin modulation of synaptic transmission to VTA DA neurons is likely to have functional consequences for information processing and behavior . Distinct glutamatergic inputs to VTA DA systems are associated with varied behavioral output . For example , the lateral habenula acts as a source of negative reward signals for DA neurons ( Lammel et al . , 2012; Matsumoto and Hikosaka , 2007 ) . Meanwhile , mPFC glutamatergic afferents increase VTA activity and may promote the development of addiction ( Gariano and Groves , 1988; Wu et al . , 2013 ) , and pedunculopontine glutamatergic afferents to VTA DA neurons drive motivated behavior ( Yoo et al . , 2017 ) . One possibility that can be explored in future studies is that oxytocin selectively gates inputs of specific valence . Oxytocin in the VTA has recently been shown to promote socially rewarding behavior ( Hung et al . , 2017 ) . This study also reported a direct facilitation of tonic activity in VTA DA neurons , in response to oxytocin receptor agonist application , consistent with our prior observations ( Xiao et al . , 2017 ) . Hung et al . propose that an increase in excitatory drive to VTA DA neurons , with a change in excitation/inhibition balance , accounts for the enhancement of tonic activity in response to oxytocin . These data appear to contradict our observations of oxytocin-evoked endocannabinoid-mediated dampening of EPSC amplitudes , but differences in neuronal identity and recording conditions likely account for the observed discrepancies . First , Hung et al . targeted VTA DA neurons retrograde-labeled from medial nucleus accumbens ( NAc ) , while we recorded non-selectively targeted ventromedial DA neurons . This difference in selection is important because DA neurons projecting to distinct regions of NAc receive different glutamatergic inputs ( Beier et al . , 2015 ) . Medial NAc-projecting VTA DA neurons are innervated by glutamatergic axons from lateral hypothalamus , lateral habenula and dorsal raphe nuclei ( Beier et al . , 2015 ) , where we do not observe extensive presence of CB1Rs in glutamatergic neurons . Additionally , Hung et al . include spermine in their recording pipette solution , presumably in order to intracellularly block NMDARs ( Araneda et al . , 1999 ) for tight pharmacological control important when recording EPSCs and IPSCs from the same neurons . However , spermine can modulate and inhibit the PLC cascade ( Sechi et al . , 1978; Wojcikiewicz and Fain , 1988 ) which is downstream of the OxtR . The inclusion of spermine in the internal solution is expected to emphasize circuit-level effects of oxytocin application or release over cell-autonomous effects that rely on retrograde signaling . Altogether , both studies using distinct targeting and pharmacological recording conditions produce important insights into oxytocin signaling in the VTA , highlighting the complexity of this multifaceted modulation and raising questions for further study . In the VTA , oxytocin is poised to function as a selective high-pass filter for synaptic transmission in several distinct ways . First , the heterogeneity of CB1 receptor expression in glutamatergic projections means that some inputs are allowed to pass through with no dampening in the presence of oxytocin . This mechanism also operates in the striatum , where cortical glutamatergic inputs , but not thalamostriatal ones , display CB1 receptor-dependent LTD ( Wu et al . , 2015 ) . Second , in our experiments only the initial stimuli in pairs or trains appear to be decreased by oxytocin . Therefore , on a short time-scale repetitively active inputs , presumably carrying behaviorally salient information , are allowed to pass . Moreover , transient exposure to oxytocin is sufficient to lead to relatively long-lasting changes in synaptic transmission . On a longer time-scale , oxytocin signaling does not potentiate the canonical LTD mechanism itself , but it can provide a separate endocannabinoid-dependent , presynaptic component to forms of long-term plasticity that are important for VTA DA neurons . VTA DA LTD has been implicated in addiction , fear learning , satiety , and aversion behaviors ( Labouèbe et al . , 2013; Liu et al . , 2010; Pignatelli et al . , 2017; Thomas and Malenka , 2003 ) . Therefore , the additive effects of oxytocin to classical LTD increase the range of physiological contexts where neurohypophyseal peptide signaling impacts the VTA . Altogether , our findings suggest that oxytocin gates inputs into reward systems both spatially and temporally . How would oxytocin-mediated dampening of excitatory synaptic transmission operate in the context of oxytocin-evoked enhancement of tonic activity in VTA DA neurons that we and others have reported ( Hung et al . , 2017; Tang et al . , 2014; Xiao et al . , 2017 ) ? In midbrain DA neurons , the activation of Gαq-coupled cascades enhances endocannabinoid release ( Gantz and Bean , 2017 ) . OxtR signaling canonically occurs through Gαq cascades; mechanistically , the corresponding increases in calcium concentration may provide the intracellular calcium required for endocannabinoid synthesis and release . Conceptually , an increase in the excitability of VTA DA neurons , together with a decoupling from a subset of inputs , may leave these neurons primed for action in response to the inputs that show persistent activity or are not dampened by endocannabinoid retrograde actions . This new evidence for oxytocinergic control of synaptic transmission in the VTA has several clinical implications , given the involvement of midbrain DA regions in reward-based behavior ( Howe and Dombeck , 2016; Russo and Nestler , 2013; Schultz , 1998 ) , drug abuse and addiction ( Jones and Bonci , 2005; Kauer and Malenka , 2007; Niehaus et al . , 2009; Stelly et al . , 2016 ) , as well as neurodevelopmental and neurodegenerative disorders ( Dichter et al . , 2012 ) . The existence of this modulatory system opens new possibilities for indirectly controlling endocannabinoid or dopaminergic signaling by leveraging interacting neuromodulators like oxytocin . Endocannabinoid signaling is suggested to be neuroprotective across multiple brain regions , including the hippocampus and midbrain dopamine regions ( Melis and Pistis , 2007; Xu and Chen , 2015 ) . Pharmacologically enhancing endocannabinoid-mediated striatal plasticity suffices to rescue motor deficits observed in rodent models of Parkinson’s disease ( Kreitzer and Malenka , 2007 ) . Given the vast literature on the neuroprotective properties of oxytocin ( Ceanga et al . , 2010; Kaneko et al . , 2016; Tyzio et al . , 2006 ) and the broad expression of oxytocin receptors throughout the vertebrate brain ( Mitre et al . , 2016 ) , the therapeutic potential of developing oxytocin receptor-targeting adjunctive pharmacological agents could be considerable . Because endocannabinoids in midbrain dopamine systems are linked to the processing of socially and non-socially rewarding stimuli , including drugs of abuse , oxytocinergic control over endocannabinoid signaling establishes new research questions as well as possibilities for therapeutic interventions .
Animals were handled according to protocols approved by the Northwestern University Animal Care and Use Committee . Weanling and young adult mice ( postnatal days 25–40 ) of both sexes were used in this study . C57BL/6 mice used for breeding were acquired from Charles River ( Wilmington , MA ) ; other mouse lines were acquired from the Jackson Laboratory ( Bell Harbor , ME ) . B6 . SJL-Slc6a3 tm1 . 1 ( cre ) Bkmn/J mice ( Slc6a3i-Cre ) , which express Cre recombinase under the control of the dopamine transporter promoter ( Bäckman et al . , 2006 ) , were used to identify dopaminergic neurons via reporter crosses . B6 . 129S-Oxt tm1 . 1 ( cre ) Dolsn/J mice ( Oxti-Cre , #024234 ) , which express the enzyme Cre recombinase under control of the oxytocin promoter ( Shah et al . , 2014 ) , were used to target oxytocinergic neurons in PVN . Slc17a6tm2 ( cre ) Lowl/J mice ( Slc17a6i-Cre , # 016963 ) , which express Cre recombinase under the control of vesicular glutamate transporter 2 promoter ( Vong et al . , 2011 ) , were used to target glutamatergic neurons . Slc6a3i-Cre mice were crossed to a floxed tdTomato reporter strain ( Ai14 , Jackson Lab , #007914 ) , or , for a small subset of experiments , a floxed ChR2-eYFP strain ( Ai32 , Jackson lab , # 012569 ) ( Madisen et al . , 2010 ) . Mice heterozygous for Cre were used for experiments; genotyping followed standard procedures available on the Jackson Lab website . P23-25 mice were anesthetized with 1 . 5–2% isofluorane , received ketoprofen for analgesia , and were placed on a small animal stereotaxic frame ( David Kopf Instruments , Tujunga , CA ) . Green retrobeads ( Lumafluor , Naples , FL ) were delivered unilaterally into the VTA ( 2 . 7 mm posterior to bregma , 0 . 5 mm lateral , and 4 . 5 mm below the pia ) through a pulled glass pipette at a rate of 50 nl/minute for a total of 100 nl using an UltraMicroPump ( World Precision Instruments , Sarasota , FL ) . AAV9-EF1a-DIO-hChR2 ( H134R ) -mCherry rAAV ( 1 . 24 × 1013 GC/ml ) was unilaterally injected into mPFC ( 2 . 4 mm anterior to bregma , 0 . 4 mm lateral , and 2 . 4 mm below the pia ) , LHb ( 1 . 0 mm posterior to bregma , 0 . 3 mm lateral , and 2 . 6 mm below the pia ) , or PPN ( 4 . 7 mm posterior to bregma , 1 . 3 mm lateral , and 3 . 7 mm below the pia ) at a rate of 100 nl/minute for a total of 150 nl . For targeting oxytocinergic neurons , the same virus was injected into the PVN ( 1 . 0 mm posterior to bregma , 0 . 3 mm lateral , 4 . 5 mm and 4 . 7 mm below the pia ) at a rate of 100 nl/minute for a total of 1000 nl . Injection coordinates for green retrobeads in the VTA: 2 . 7 mm posterior to bregma , 0 . 5 mm lateral , and 4 . 5 mm below the pia . The pipette was held at the injection location for 15–20 min after retrobead injection and 10 min after virus injection . Coordinates were slightly adjusted for mouse age and size . Mice recovered for 7–9 days following retrograde tracer injections , and 4–5 weeks after virus injection . To confirm neuronal identity as dopaminergic cells in all electrophysiology experiments combined with optogenetics , slices were fixed in 4% paraformaldehyde ( PFA ) overnight after recording , washed in 0 . 1 M phosphate buffed saline ( PBS ) , and processed for immunostaining against tyrosine hydroxylase . Sections were pretreated in 0 . 2% Triton-X100 for an hour at RT , and then incubated for 24 hr at 4°C with primary antibody solution in PBS with 0 . 2% Triton-X100 ( rabbit anti-tyrosine hydroxylase , 1:1000; AB152 , Abcam , Cambridge , UK ) . Tissue was rinsed in PBS , reacted with secondary antibody for 2 hr at RT ( goat anti-rabbit Alexa 647 , 1:500 , Life Technologies , Carlsbad , CA ) , rinsed again , then mounted onto Superfrost Plus slides ( ThermoFisher Scientific , Waltham , MA ) , dried and coverslipped under glycerol:TBS ( 3:1 ) with Hoechst 33342 ( 1:1000; ThermoFisher Scientific ) . OxtR-2 antibody staining ( a generous gift of R . Froemke ) was conducted following previously published procedures ( Mitre et al . , 2016 ) . Mice were deeply anaesthetized with isoflurane and transcardially perfused with 4% PFA . Brains were post-fixed for 2 hr , and transferred to 30% sucrose solution in PBS and stored at 4°C overnight . Then , brains were embedded in Tissue-Tek O . C . T . compound ( VWR ) , stored overnight at −80°C , and sectioned on a cryostat at 18 µm thickness . Sections were rinsed in PBS , blocked for 2 hr in PBS with 0 . 2% Triton X-100% and 5% donkey serum , and incubated with OxtR-2 antibody serum at 1:250 dilution . Following a 2 day-long incubation at 4°C in a humidified chamber , sections were rinsed three times in PBS and incubated for 2 hr in Alexa Fluor 647-conjugated goat anti-rabbit antibody ( Thermo Fisher Scientific , 1:500 ) . Whole sections were imaged with an Olympus VS120 slide scanning microscope ( Olympus Scientific Solutions Americas , Waltham , MA ) . Confocal images were acquired with a Leica SP5 confocal microscope ( Leica Microsystems ) . Depth-matched z-stacks of 1 µm-thick optical sections were analyzed in ImageJ ( FIJI ) ( Schindelin et al . , 2012 ) . FISH labeling and analyses were conducted according to previously published procedures ( Xiao et al . , 2017 ) . Briefly , brains were quickly removed from deeply anesthetized mice and frozen in tissue-freezing medium prior to storage at −80˚C , sectioned at 20 μm ( Leica CM1850 ) , adhered to Superfrost Plus slides , and frozen . Samples were fixed with 4% PFA in 0 . 1 M PBS at 4˚C , processed according to RNAscope Fluorescent Multiplex Assay manual for fresh frozen tissue ( Advanced Cell Diagnostics , Newark , CA ) , and coverslipped using ProLong Gold antifade reagent with DAPI ( Molecular Probes ) . The following probes were used: Cnr1 and Oxtr in channel 1 , with tyrosine hydroxylase ( Th ) , Slc17a6 ( Vglut2 ) , or Slc17a7 ( Vglut1 ) in channel 2 . Sections were subsequently imaged on a Leica SP5 confocal microscope in three channels with a 40x or 100x objective lens , with 1 μm between adjacent z-sections . Probe omission negative controls were carried out for every reaction . FISH images were analyzed as previously ( Xiao et al . , 2017 ) with a MATLAB script utilizing imreadBF for file loading and a modified version of Fast 2D peak finder . Three adjacent z-stack slices were combined , for a total of ~3 μm of tissue . In general , combining between 2 and 3 μm was optimal to ensure that differences in subcellular localization of RNA transcripts do not lead to missed co-localization , while minimizing false positive co-localization driven by signal from other cells . All channels were thresholded for intensity to remove background signal . Watershed segmentation of the image was performed using Th , Slc17a6 , or Slc17a7 channel information to localize somata . Puncta of FISH molecules were counted within established cell boundaries . Whether a cell was considered positive for a given marker was determined by setting transcript-dependent thresholds for the number of puncta , but data for all imaged cells are shown throughout the study . This threshold was set by comparing manual counts of cells to histograms of puncta per cell for several images . The established puncta number threshold was used for all remaining images of a given channel/probe combination . Brain slice preparation was adapted from previously published procedures ( Xiao et al . , 2017 ) . Briefly , animals were deeply anesthetized by isoflurane , followed by a transcardial perfusion with ice-cold , oxygenated artificial cerebrospinal fluid ( ACSF ) containing ( in mM ) 127 NaCl , 2 . 5 KCl , 25 NaHCO3 , 1 . 25 NaH2PO4 , 2 . 0 CaCl2 , 1 . 0 MgCl2 , and 25 Glucose ( osmolarity ~310 mOsm/L ) . After perfusion , the brain was removed and immersed in ice-cold ACSF . Tissue was blocked and transferred to a slicing chamber containing ice-cold ACSF , supported by a block of 4% agar . Horizontal slices of 250 μm thickness were cut on a Leica VT1000s in ventral-dorsal direction and transferred into a holding chamber with ACSF , equilibrated with 95%O2/5%CO2 . Slices were incubated at 34 ˚C for ~30 min prior to electrophysiological recording . Slices were transferred to a recording chamber perfused with oxygenated ACSF at a flow rate of 2–4 ml/min at room temperature . Whole-cell recordings were obtained from neurons visualized under infrared DODT contrast video microscopy using patch pipettes of ~2–5 MΩ resistance . For electrical stimulation experiments , VTA dopamine neurons were identified based on tdTomato or eYFP signal in Slc6a3 i-Cre; Ai14 or Slc6a3i-Cre; Ai32 mice . For optogenetic experiments , dopamine neurons were identified on the basis of the combination of their electrophysiological and morphological properties , and further validated with post hoc TH immunolabeling ( Xiao et al . , 2017 ) . Recording electrodes contained the following ( in mM ) : 120 CsMeSO4 , 15 CsCl , 10 HEPES , 10 Na-phosphocreatine , 2 MgATP , 0 . 3 NaGTP , 10 QX314 , and 1 EGTA ( pH 7 . 2-7 . 3 , ~295 mOsm/L ) . For optogenetic and two-photon uncaging experiments , Alexa Fluor 488 dye ( 20 µM ) was added to the internal solution . SR95531 ( 10 μM ) was added to ACSF for recording electrical stimulation-evoked EPSCs , light stimulation-evoked EPSCs , and spontaneous EPSCs , all acquired at a holding potential of −70 mV . Recordings were made using 700B amplifiers ( Axon Instruments , Union City , CA ) ; data were sampled at 10 kHz and filtered at 4 kHz with a MATLAB-based acquisition script ( MathWorks , Natick , MA ) . Series resistance and input resistance were monitored using a 100 ms , 5 mV hyperpolarizing pulse at every sweep , and experiments were started after series resistance had stabilized ( ~20 MΩ , uncompensated ) . For electrical stimulation experiments , a concentric bipolar micro electrode ( CBAPB75 , FHC , Inc ) was placed approximately 100 μm away from the recording electrode and 80 μs electrical pulses applied at intervals of 30 s were used to evoke EPSCs . The amplitudes of EPSCs were calculated by taking a 1 ms window around the peak of the EPSC and comparing this to a 1 s window immediately prior to the onset of the electrical stimulation artefact . Paired stimuli were delivered using inter-stimulus intervals of 50 , 80 and 100 ms , and the paired-pulse ratio ( PPR ) was defined as the ratio between the amplitudes of the second and the first EPSCs . EPSC decay time was calculated based on a single exponential fit and reported as the time constant , averaged for all recorded EPSCs within each neuron in baseline condition and following oxytocin application . Spontaneous EPSCs ( sEPSCs ) were pharmacologically isolated in the presence of GABA ( A ) R antagonist SR 95531 ( 10 µM ) . A 5 min , 10 Hz repetitive electrical stimulation was used to induce LTD of EPSCs in VTA DA neurons ( Pan et al . , 2008 ) . In a subset of experiments LTD was induced using 1 s long periods of 100 Hz stimulation , repeated 4 times with 10 second-long inter-train intervals ( Kreitzer and Malenka , 2005 ) . Paired stimuli delivered at 50 ms inter-stimulus interval and 30 second-long inter-train intervals were used to evoke EPSCs before and after inducing LTD . The amplitude of first EPSC in every train was used to quantify EPSC amplitude change , and PPR was defined as the ratio between the amplitudes of the second and the first EPSCs . For quantifying the magnitude of LTD , EPSCs were normalized based on 5 min baseline recordings . LTD magnitude reflects the average normalized EPSC amplitude during a 10 min-long period starting 10 min after the end of induction period . To activate ChR2-expressing fibers of glutamatergic neurons in the VTA , 2 ms-long light pulses ( 470 nm , ~5 mW ) at intervals of 30 s were delivered at the recording site using whole-field illumination through a 40X water-immersion objective ( Olympus , Tokyo , Japan ) with a PE300 CoolLED illumination system ( CoolLED Ltd . , Andover , UK ) . The amplitudes of optical evoked EPSCs were calculated by taking a 1 ms window around the peak of the EPSC and comparing this to a 1 s window prior to the onset of the light stimulation . To activate ChR2-expressing fibers of oxytocinergic neurons in the VTA , 10 ms-long light pulses ( 470 nm , 20 Hz for 30 s , ~5 mW ) were delivered at the recording site using whole-field illumination . Electrically evoked EPSCs were recorded 60 s , 30 s , and 0 s before light stimulation , and 0 s and 30 s after light stimulation . Three to four consecutive responses at 6 min-long intervals were acquired for each neuron . Slc6a3i-Cre; tdTomato mice were used for all two-photon uncaging and imaging experiments . We recorded dopamine neurons in the VTA using cesium-based internal solution containing 20 μM Alexa Fluor 488 . After a 10–15 min long wash-in period , cell morphology was visualized using Alexa Fluor signal . Two mode-locked Ti:Sapphire lasers ( Mai Tai eHP DS , Newport ) were used for imaging and uncaging at the wavelengths of 910 nm and 720 nm , respectively . The beam of the laser was directed by a two-dimensional galvanometer scanning mirror system ( HSA Galvo 8315K , Cambridge Technology ) . Fluorescence emission was collected by two PMTs above and below the sample ( H10770P , Hamamatsu ) after passing through a dichroic beamsplitter ( FF670-SDi01−26 × 38 , Semrock ) and a bandpass filter ( FF02- 520/28 , Semrock ) . 2 . 5 mM MNI-L-glutamate ( Tocris ) was perfused in the recirculating bath and two 0 . 5 ms long laser pulses at 50 ms ISI were delivered to a target spot near a dendritic spine to photoactivate glutamate . We used a version of Scanimage to control scanning parameters and image acquisition ( Kozorovitskiy et al . , 2015; Pologruto et al . , 2003 ) . Laser intensity was controlled by a Pockels cell and laser power at the sample plane was 10–15 mW . After selecting a dendritic spine for uncaging , several spots around the spine were sampled for maximal uncaging-evoked EPSC ( uEPSC ) response . Location of the maximal response for a given dendritic spine was selected for the experiment . At least ten consecutive AMPAR-uEPSCs were acquired at 1 second-long intervals from each dendritic spine . Pharmacological agents were acquired from Tocris ( Bristol , UK ) or Sigma-Aldrich ( St . Louis , MO ) . Drugs were applied by bath perfusion: oxytocin ( 10 nM - 1 μM ) , L368 , 899 hydrochloride ( 10 μM ) , U73122 ( 10 μM ) , SR49059 ( 10 μM ) , CP945598 hydrochloride ( 10 μM ) , WIN55212-2 ( 5 μM ) , AM251 ( 5 μM ) , orlistat ( THL , 5 μM ) ) , MNI-L-glutamate ( 2 . 5 mM ) , SR95531 hydrobromide ( 10 μM ) , NBQX ( 10 μM ) , and CPP ( 10 μM ) . Offline analyses of electrophysiology were performed using MATLAB ( Mathworks , Natick , MA ) and IgorPro ( Wavemetrics , Portland , OR ) . Whenever possible , data were analyzed blind to condition . For sample sizes , both the number of neurons analyzed and the number of animals are provided . Sex and age were balanced across groups . Statistical analyses were carried out using GraphPad Prizm 5 software ( GraphPad , LaJolla , CA ) and SPSS ( IBM , New York , NY ) . Group data are expressed as group means ±SEM . Only non-parametric statistical tests are used throughout the study . For two-group comparisons , statistical significance was determined by two-tailed Wilcoxon matched-pairs signed rank test or Mann-Whitney test . For multiple group comparisons , Friedman’s 2-Way ANOVA by ranks test and Kruskal-Wallis test with Dunn’s Multiple Comparison post hoc test were used . p<0 . 05 was considered statistically significant . | The mammalian brain contains millions of nerve cells or neurons that communicate with each other via a process called neurotransmission . To send a message to its neighbor , a neuron releases a chemical called a neurotransmitter into the space between the cells . The neurotransmitter then binds to receiver proteins on the target cell . Another group of chemicals , known as neuromodulators , regulate this process , adjusting the way that neurons respond to neurotransmitters . In doing so , they help regulate many types of behavior in mammals . The neuromodulator oxytocin , for example , has earned the nickname ‘the love hormone’ because it promotes social behavior and bonding . It does this in part by altering the activity of neurons in a brain region called the ventral tegmental area ( VTA ) . These neurons produce the brain’s main reward signal , dopamine , which is itself a neuromodulator . But exactly how oxytocin affects the activity of dopamine-producing neurons is unclear . By recording from individual neurons in slices of mouse brain tissue , Xiao et al . show that oxytocin filters inputs to dopamine neurons in the VTA . It does this by making the dopamine neurons release another group of reward signals , known as endocannabinoids . These are the brain’s own version of the chemicals found inside cannabis plants . The endocannabinoids bind to neurons that provide input to the VTA dopamine neurons . Some of these input neurons normally activate the VTA by releasing a neurotransmitter called glutamate . However , the binding of endocannabinoids decreases their ability to do this , and thereby lowers the activation of the VTA dopamine neurons . But not all glutamate neurons are sensitive to endocannabinoids . Moreover , oxytocin affects glutamate neurons that fire repeatedly less than it affects those that fire only occasionally . Oxytocin thus acts as a filter . It allows certain inputs – those that are repeatedly active and those that are insensitive to endocannabinoids – to continue activating VTA dopamine neurons . At the same time , it weakens the influence of other inputs . Dopamine release in the VTA drives drug abuse and addiction . Understanding how oxytocin affects VTA neurons may thus open up new avenues for the treatment of addiction disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Oxytocin functions as a spatiotemporal filter for excitatory synaptic inputs to VTA dopamine neurons |
Cerebral small vessel disease ( SVD ) is a leading cause of stroke and dementia . CADASIL , an inherited SVD , alters cerebral artery function , compromising blood flow to the working brain . TIMP3 ( tissue inhibitor of metalloproteinase 3 ) accumulation in the vascular extracellular matrix in CADASIL is a key contributor to cerebrovascular dysfunction . However , the linkage between elevated TIMP3 and compromised cerebral blood flow ( CBF ) remains unknown . Here , we show that TIMP3 acts through inhibition of the metalloprotease ADAM17 and HB-EGF to regulate cerebral arterial tone and blood flow responses . In a clinically relevant CADASIL mouse model , we show that exogenous ADAM17 or HB-EGF restores cerebral arterial tone and blood flow responses , and identify upregulated voltage-dependent potassium channel ( KV ) number in cerebral arterial myocytes as a heretofore-unrecognized downstream effector of TIMP3-induced deficits . These results support the concept that the balance of TIMP3 and ADAM17 activity modulates CBF through regulation of myocyte KV channel number .
Small vessel disease ( SVD ) of the brain is a leading cause of stroke and age-related cognitive decline and disability for which there are currently no treatments ( Pantoni , 2010 ) . Our limited understanding of the pathogenesis of cerebral SVD is a major obstacle to the development of treatments . Monogenic forms of these diseases , such as CADASIL ( Cerebral Autosomal Dominant Arteriopathy with Subcortical Infarcts and Leukoencephalopathy ) , provide a window into the mechanism underlying much more common , but largely indistinguishable , sporadic forms of SVD ( Joutel and Faraci , 2014 ) . CADASIL , the most common hereditary cerebral SVD , is caused by dominant mutations in the NOTCH3 receptor that stereotypically lead to the extracellular deposition of NOTCH3 ectodomain ( Notch3ECD ) and aggregates of other proteins on vessels ( Joutel et al . , 2000; Chabriat et al . , 2009; Monet-Leprêtre et al . , 2013 ) . A deficit in cerebral blood flow ( CBF ) hemodynamics is an early feature of the disease , suggesting that cerebrovascular dysfunction may have a key role in disease pathogenesis ( Chabriat et al . , 2000; Pfefferkorn et al . , 2001; Liem et al . , 2009 ) . Small vessels of the brain have unique functional properties that ensure that the brain , which has a limited capacity to store energy , maintains an adequate supply of blood-borne nutrients in the face of variations in blood pressure and changing neuronal energy demands . Cerebral arteries exist in a partially constricted state called 'myogenic tone' , which reflects an intrinsic contractile response of arterial myocytes to increases in intravascular pressure . Thus , these arteries are positioned to dilate , and thereby increase local CBF , in response to elevated neuronal activity . This phenomenon , known as functional hyperemia , serves to satisfy enhanced glucose and oxygen demands of active neurons ( Iadecola and Nedergaard , 2007 ) . Impaired functional hyperemia and CBF autoregulation , attenuated CBF responses to topical application of vasodilators , and diminished myogenic responses of cerebral arteries and arterioles are early and prominent features of the well-established TgNotch3R169C CADASIL mouse model ( Joutel et al . , 2010; Dabertrand et al . , 2015; Capone et al . , 2016 ) . The mechanisms underlying this cerebrovascular dysfunction are poorly understood . Recently , we found that TIMP3 ( tissue inhibitor of metalloproteinases-3 ) forms complexes with Notch3ECD and abnormally accumulates in the extracellular matrix of brain vessels of patients and mice with CADASIL ( Monet-Leprêtre et al . , 2013 ) . Remarkably , genetic overexpression of TIMP3 recapitulates both CBF and myogenic-response deficits of the CADASIL model; conversely , genetic reduction of TIMP3 in CADASIL model mice restores normal function ( Capone et al . , 2016 ) . TIMP family members are key regulators of the metalloproteinases that degrade the extracellular matrix . Within the TIMP family , TIMP3 has the broadest spectrum of substrates , which extends to members of the ADAM ( a disintegrin and metalloproteinases ) family . These metalloproteinases proteolytically release the extracellular domains of membrane-bound cytokines , cell adhesion molecules and growth factors , such as tumor necrosis factor-α and several ligands of the epidermal growth factor receptor ( EGFR ) family , including HB-EGF ( heparin-binding EGF-like growth factor ) ( Brew and Nagase , 2010; Khokha et al . , 2013; Arpino et al . , 2015 ) . As such , in addition to being a powerful regulator of extracellular matrix remodeling in various organs ( Arpino et al . , 2015 ) , TIMP3 is a key player in inflammatory pathologies and autoimmune diseases through regulation of cell surface proteins ( Brew and Nagase , 2010; Khokha et al . , 2013 ) . However , how metalloproteinase inhibition might dynamically regulate arterial tone and CBF hemodynamics is unclear . In another recent study , we established that upregulation of voltage-gated potassium ( KV ) channels at the plasma membrane of arterial myocytes is responsible for the diminished myogenic responses of cerebral arteries and penetrating arterioles in the TgNotch3R169C CADASIL model . Notably , an influence of the endothelium in myogenic tone deficit was ruled out ( Dabertrand et al . , 2015 ) . KV channels play an important and dynamic role in opposing pressure-induced depolarization and vasoconstriction ( Longden et al . , 2015 ) . Furthermore , we ( Dabertrand et al . , 2015 ) and our collaborators ( Koide et al . , 2007 ) have found that down-regulation of plasma membrane KV channels through activation of the EGFR pathway restores normal responses to pressure . Collectively , our results suggest a fundamental linkage between the activity of TIMP3 in the extracellular matrix of cerebral arterial myocytes and cerebral arterial tone . Here , we find that the ADAM17/HB-EGF/EGFR ( ErbB1/ErbB4 ) signaling axis is a key TIMP3-sensitive signaling module that regulates CBF responses and the myogenic tone of cerebral arteries . We further provide evidence that disruption of this TIMP3-sensitive pathway mediates cerebrovascular dysfunction in the TgNotch3R169C CADASIL model and identify upregulated KV channel current density in cerebral arterial myocytes as a heretofore-unrecognized effector of this pathway . These insights into the relationship between TIMP3 activity and cerebral arterial tone may ultimately lead to therapeutic options in cerebral SVD .
To explore the role of TIMP3 in the regulation of CBF , we monitored CBF responses in wild-type mice equipped with an open cranial window over the somatosensory cortex , before and after the application of recombinant TIMP3 as well as TIMP1 or TIMP2 ( Figure 1A; Figure 1—source data 1 ) . We initially ensured that a recombinant protein applied over the cranial window could enter the brain . In the absence of robust anti-TIMP3 antibody suitable for immunohistochemistry and of an in situ assay to detect TIMP3 activity , we assessed brain penetration of Fluorescein isothiocyanate-labeled albumin ( FITC-BSA , 66 kDa ) . After 30 min of continuous superfusion , fluorescence imaging of fixed vibratome slices showed that FITC-BSA entered the cortex along penetrating arteries beneath the cranial window ( Figure 1—figure supplement 1 ) , consistent with transport via the glymphatic system ( Iliff et al . , 2012 ) . We found that TIMP3 ( 40 nM ) did not affect resting CBF ( Figure 1B ) , but did strongly reduce the increase in CBF evoked by whisker stimulation ( Figure 1C , D; Figure 1—source data 2 , 3 ) . Superfusion of 8 nM TIMP3 similarly attenuated functional hyperemia ( Figure 1—figure supplement 2; Figure 1—source data 2 , 3 ) . To rule out a possible effect of TIMP3 on neural activity , we recorded evoked neural activity during TIMP3 application ( Figure 1E ) . We found that the amplitude of the somatosensory fields potentials produced by electrical stimulation of the whisker pad was unaltered by TIMP3 superfusion ( Figure 1F ) . 10 . 7554/eLife . 17536 . 003Figure 1 . Exogenous TIMP3 specifically impairs cerebrovascular reactivity . ( A ) Schematic representation of the experimental protocol used to test the effects of recombinant TIMP1 ( 50 nM ) , TIMP2 ( 50 nM ) or TIMP3 ( 40 nM ) superfusion on the somatosensory cortex of 2-month-old wild-type mice . ( B–D ) Resting CBF ( B ) and CBF responses to whisker stimulation ( C , D ) were evaluated upon superfusion of vehicle or TIMP proteins . ( C ) Representative trace of CBF responses to whisker stimulation upon superfusion of vehicle or TIMP proteins ( C ) . ( E ) Representative trace of the field potentials evoked by whisker stimulation upon vehicle or TIMP3 superfusion , showing typical sharp positive ( P1 ) -negative ( N1 ) waves followed by a slower positive-negative waveform occurring within 80 ms post stimulus ( Di and Barth , 1991 ) . ( F ) The amplitude of the negative wave ( N1 , asterisk in E ) of the field potential was not affected by TIMP3 superfusion ( p=0 . 79 ) . ( G , H ) CBF responses to topical application of adenosine ( G ) or acetylcholine ( H ) upon superfusion of vehicle or TIMP proteins . Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test ( B , D , G , H ) or unpaired Student’s t-test ( F ) . ( *p<0 . 05 , ***p<0 . 001 compared to vehicle; n = 5 mice/groups ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00310 . 7554/eLife . 17536 . 004Figure 1—source data 1 . Reagents used for Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00410 . 7554/eLife . 17536 . 005Figure 1—source data 2 . Main physiological variables of mice studied in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00510 . 7554/eLife . 17536 . 006Figure 1—source data 3 . Numerical data that were used to generate the bar charts in Figure 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00610 . 7554/eLife . 17536 . 007Figure 1—figure supplement 1 . Assessment of brain penetration of fluorescein isothiocyanate labelled serum albumin ( FITC-BSA ) superfused over the cranial window . ( A ) FITC-BSA was topically superfused over the somatosensory cortex for 30 min , the brain was removed at the time of killing , post-fixed , sectioned in 50-µm-thick coronal slices through the perfusion site using a vibratome and immunostained with anti-smooth muscle alpha actin conjugated to Alexa 594 ( a-SMA ) . Shown is a representative vibratome coronal section counterstained with DAPI and examined by epifluorescence microscopy ( merge of DAPI and FITC images ) . The pia matter as well the penetrating vessels under the window ( left side ) display spontaneous FITC fluorescence . ( B–D ) Higher magnification of selected regions ipsilateral to the window ( 1-B , 2-C ) or contralateral to the window ( 3-D ) , delineated in A , showing that FITC-BSA entered the cortex along the penetrating arteries ( white arrowheads ) beneath the cranial window ( B , C ) . Scale bar represents 500 µm ( A ) and 100 µm ( B–D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00710 . 7554/eLife . 17536 . 008Figure 1—figure supplement 2 . Exogenous TIMP3 ( 8 nM ) impairs cerebrovascular reactivity . ( A–D ) Resting CBF ( A ) and CBF responses to whisker stimulation ( B ) or topical application of acetylcholine ( C ) or adenosine ( D ) were evaluated upon superfusion of TIMP3 ( 8 nM ) or vehicle . Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test . ( ***p<0 . 001 compared with vehicle; n = 5 mice/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 008 The known TIMPs share 38–49% amino acid identity and inhibit most matrix metalloproteinases ( MMPs ) . However , differences in substrate selectivity and inhibitory properties between different TIMPs have been described ( Khokha et al . , 2013; Stetler-Stevenson , 2008 ) ( Figure 1—source data 1 ) . This prompted us to assess the effects of other TIMPs on functional hyperemia . In sharp contrast to TIMP3 , neither exogenous TIMP1 ( 50 nM ) nor TIMP2 ( 50 nM ) altered functional hyperemia ( Figure 1C , D; Figure 1—source data 2 , 3 ) . We further assessed CBF responses to topical application of the endothelium-dependent and smooth muscle-dependent vasodilators , acetylcholine and adenosine , respectively , upon superfusion of TIMP3 ( 8 and 40 nM ) , TIMP1 ( 50 nM ) or TIMP2 ( 50 nM ) . Again , the increases in CBF induced by acetylcholine or adenosine were profoundly attenuated by TIMP3 but were unaffected by TIMP2 or TIMP1 , with the exception of a modest attenuation of the adenosine-induced increase in CBF by TIMP1 ( Figure 1G , H; Figure 1—figure supplement 2; Figure 1—source data 2 , 3 ) . Thus , these findings establish that elevation of TIMP3 is sufficient to induce CBF deficits in vivo and suggest that a TIMP3-specific target accounts for these deficits . Our efforts to identify the target of TIMP3 focused on ADAM17 , which is uniquely inhibited by TIMP3 ( Xu et al . , 2012 ) and is expressed in brain arteries , as demonstrated by our immunoblot analyses ( Figure 2A , B ) . If TIMP3 does indeed act through inhibition of ADAM17 , its effects on CBF responses should be mimicked by pharmacological inhibition of ADAM17 . Here , we used the hydroxamate-based GW413333X inhibitor , which specifically blocks both ADAM10 and ADAM17; the ADAM10 inhibitor GI254023X was used as a control ( Hundhausen et al . , 2003 ) ( Figure 2—source data 1 ) . We found that GW413333X ( 5 µM ) , but not GI254023X ( 5 and 20 µM ) , strongly attenuated the increase in CBF produced by whisker stimulation or topical application of acetylcholine or adenosine ( Figure 2C–E; Figure 2—figure supplement 1A; Figure 2—source data 2 , 3 ) . To further support a specific role for ADAM17 in these defects , we assessed CBF responses following reduction of ADAM17 levels using a genetic approach . Complete ablation of ADAM17 is lethal ( Peschon et al . , 1998 ) . Therefore , we used hypomorphic mice with dramatically reduced expression of ADAM17 ( Adam17ex/ex ) using the exon-induced translational stop strategy . These mice are viable , but develop eye , skin and heart defects as a consequence of impaired EGFR signaling ( Chalaris et al . , 2010 ) . We found that genetic depletion of ADAM17 strongly attenuated CBF responses in a dose-dependent manner ( Figure 2F , G; Figure 2—figure supplement 1B; Figure 2—source data 2 , 3 ) . To confirm that the reduction in evoked CBF responses in these mice is caused by reduced ADAM17 expression , we examined whether an enzymatically active extracellular domain of ADAM17 ( sADAM17 ) applied exogenously could prevent these CBF deficits . Topical neocortical application of sADAM17 ( 16 nM ) ( Figure 2—source data 1 ) did not affect cerebrovascular responses in wild-type mice , but did fully restore CBF responses in Adam17ex/+ mice with half-reduced ADAM17 levels ( Figure 2H–J; Figure 2—figures supplements 1C , 2A; Figure 2—source data 2 , 3 ) . Together , these results indicate that decreasing ADAM17 activity compromises CBF regulation . 10 . 7554/eLife . 17536 . 009Figure 2 . Cerebrovascular responses are impaired by pharmacological or genetic inhibition of ADAM17 , and rescued by exogenous sADAM17 . ( A ) Immunoblot of cerebral arteries dissected from Adam17+/+ and Adam17ex/ex mice ( n = 3 biological samples/genotype ) incubated with anti-ADAM17 or anti-smooth muscle α-actin ( α-SMA ) antibody . ( B ) Quantification of relative protein level of ADAM17 in ( A ) . ( C–E ) Resting CBF ( C ) and CBF responses to whisker stimulation ( D ) or topical application of adenosine ( E ) were evaluated upon superfusion of the dual ADAM10/ADAM17 inhibitor GW413333X ( GW; 5 µM ) or the ADAM10 inhibitor GI254023X ( GI; 5 and 20 µM ) . ***p<0 . 05 compared with vehicle . ( F , G ) CBF responses to whisker stimulation ( F ) or topical application of adenosine ( G ) were strongly reduced in Adam17ex/+ mice and further reduced in Adam17ex/ex mice compared with wild-type littermate controls . ( H–J ) Exogenous sADAM17 ( 16 nM ) significantly ameliorated CBF responses to whisker stimulation ( I ) or topical application of adenosine ( J ) in Adam17ex/+ mice , whereas ADAM17 had no effect on wild-type littermates . ( K–M ) Resting CBF and CBF responses were evaluated in TgBAC-TIMP3 mice and non-transgenic littermates ( WT ) before and after superfusion of ADAM17 . CBF responses to whisker stimulation ( L ) or topical application of adenosine ( M ) were strongly reduced in TgBAC-TIMP3 mice compared with those in WT mice , as previously reported ( Capone et al . , 2016 ) , and were normalized by sADAM17 superfusion . Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test ( B–G ) and two-way repeated measure ANOVA followed by Bonferroni post-hoc test ( H–M ) ( n = 5 mice/group ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 00910 . 7554/eLife . 17536 . 010Figure 2—source data 1 . Reagents used for Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01010 . 7554/eLife . 17536 . 011Figure 2—source data 2 . Main physiological variables of mice studied in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01110 . 7554/eLife . 17536 . 012Figure 2—source data 3 . Numerical data that were used to generate the bar charts in Figure 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01210 . 7554/eLife . 17536 . 013Figure 2—figure supplement 1 . CBF responses to acetylcholine are attenuated by pharmacological or genetic inhibition of ADAM17 but rescued upon superfusion of exogenous sADAM17 . ( A ) CBF responses to topical application of acetylcholine were evaluated in 2-month-old wild-type mice upon superfusion of the dual ADAM10/ADAM17 inhibitor GW413333X ( GW; 5 µM ) or the ADAM10 inhibitor GI254023X ( GI; 5 and 20 µM ) . ***p<0 . 001 compared with vehicle . ( B ) CBF responses to topical application of acetylcholine were strongly reduced in Adam17ex/+ mice and further reduced in Adam17ex/ex mice compared to littermate wildtype ( WT ) mice . ( C ) Exogenous soluble active ectodomain of ADAM17 ( sADAM17; 16 nM ) significantly ameliorated CBF responses to topical application of acetylcholine in Adam17ex/+ mice , whereas it had no effect on wild-type littermates . ( D ) CBF responses to topical application of acetylcholine were strongly reduced in TgBAC-TIMP3 mice compared with those in WT mice , as previously reported ( Capone et al . , 2016 ) , and were normalized by sADAM17 superfusion . Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test ( A , B ) and two-way repeated measure ANOVA followed by Bonferroni post-hoc test ( C , D ) ( n = 5 mice/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01310 . 7554/eLife . 17536 . 014Figure 2—figure supplement 2 . Absolute measurements of resting CBF in Adam17ex/+ and TgBAC-TIMP3 mice in the presence and absence of sADAM17 . Resting CBF , expressed as Laser Doppler flow arbitrary units ( LDFU ) , was evaluated in Adam17ex/+ mice ( A ) , TgBAC-TIMP3 mice ( B ) and appropriate wild-type littermates , before and after superfusion of soluble ADAM17 ( 16 nM ) . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 5 mice/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 014 To further confirm the direct connection between increased TIMP3 expression and reduced ADAM17 activity and CBF deficits , we tested whether exogenous sADAM17 is capable of preventing the CBF deficits produced by genetic overexpression of TIMP3 . Superfusion with the enzymatically active extracellular domain of ADAM17 ( 16 nM ) increased resting CBF in TgBAC-TIMP3 mice towards the same absolute values as wild-type mice and improved all evoked cerebrovascular responses ( Figure 2K–M; Figure 2—figures supplements 1D , 2B; Figure 2—source data 2 , 3 ) suggesting that TIMP3 induces CBF deficits by decreasing ADAM17 activity . To elucidate the molecular factors that operate downstream of ADAM17 in the context of cerebrovascular regulation , we examined the role of the EGFR signaling pathway . This pathway consists of four related receptor tyrosine kinases of the ErbB family—ErbB1/EGFR ( Her1 ) , ErbB2/Neu ( Her2 ) , ErbB3 ( Her3 ) and ErbB4 ( Her4 ) —which are regulated by 11 different ligands , all of which are produced as membrane-bound precursor proteins and cleaved by cell surface proteases to yield the active soluble species; ADAM17 is the critical sheddase of at least six of these ligands ( Sahin et al . , 2004; Roskoski , 2014 ) ( Figure 3A , B ) . A critical role for ADAM17 in EGFR signaling is supported by the observation that mice deficient for ADAM17 ( Peschon et al . , 1998; Chalaris et al . , 2010 ) resemble mice lacking EGFR , exhibiting perinatal lethality , generalized epithelial defects , and defective cardiac valves ( Miettinen et al . , 1995; Sibilia and Wagner , 1995; Threadgill et al . , 1995 ) . To investigate the role of the EGFR pathway , we recorded CBF responses evoked by whisker stimulation or vasodilators before and after topical application of blockers of this pathway ( Figure 3—source data 1 ) . 10 . 7554/eLife . 17536 . 015Figure 3 . Full CBF responses require ErbB1/ErbB4 and HB-EGF . ( A , B ) Schematic representation of the ErbB signaling pathway . Ligands are all produced as membrane-bound precursor proteins that are cleaved by cell-surface sheddases to yield the active growth factor species . Binding of the soluble form of the ligand induces ErbB receptor homodimerization or heterodimerization , converting the receptor to an active dimeric conformation ( A ) . Ligands are grouped in four rows according to their receptor specificity ( top; arrows ) ; the six ligands for which ectodomain shedding is primarily mediated by ADAM17 appear in black characters , and the remaining five are in grey characters ( B ) . ( C–K ) Resting CBF ( C , F , I ) and CBF responses to whisker stimulation ( D , G , J ) or topical application of adenosine ( E , H , K ) were evaluated before and after superfusion of various inhibitors of the ErbB signaling pathway , including the ErbB1/ErbB4 inhibitor AG1478 ( 10 and 20 µM ) ; the ErbB2 inhibitor AG825 ( 50 and 200 µM ) ( C–E ) , the soluble ErbB receptor traps ( ErbB1-Fc , 66 . 7 nM; ErbB3-Fc , 71 . 4 nM; ErbB4-Fc , 71 . 4 nM ) and the respective control IgG1-Fc and IgG2-Fc fragments ( 286 nM ) ( F–H ) , heparin and the synthetic peptide p21 ( 12 µM ) and the control inactive peptide p21-mut ( 12 µM ) ( I–K ) . None of these compounds affected resting CBF , except ErbB4-Fc , which produced a slight increase . ( C–K ) Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test ( *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 compared to vehicle; n = 5/group ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01510 . 7554/eLife . 17536 . 016Figure 3—source data 1 . Reagents used for Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01610 . 7554/eLife . 17536 . 017Figure 3—source data 2 . Main physiological variables of mice studied in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01710 . 7554/eLife . 17536 . 018Figure 3—source data 3 . Numerical data that were used to generate the bar charts in Figure 3 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 01810 . 7554/eLife . 17536 . 019Figure 3—figure supplement 1 . Blockade of ErbB1/ErbB4 or HB-EGF impairs CBF responses to acetylcholine . ( A , B ) CBF responses to application of acetycholine were evaluated before and after superfusion of various inhibitors of the ErbB signaling pathway , including the soluble ErbB receptor traps ( ErbB1-Fc , 66 . 7 nM; ErbB3-Fc , 71 . 4 nM; ErbB4-Fc , 71 . 4 nM ) and the respective control IgG1-Fc and IgG2-Fc fragments ( 286 nM ) ( A ) , heparin and the synthetic peptide p21 ( 12 µM ) and the control inactive peptide p21-mut ( 12 µM ) ( B ) . Significance was determined by one-way ANOVA followed by Tukey’s post-hoc test . ( ***p<0 . 001 compared with vehicle; n = 5 /group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 019 Based on the decrease in evoked CBF responses to elevation of TIMP3 or reduction of ADAM17 , we predicted that inhibition of ErbB pathway would have a similar effect . Indeed , we found that neocortical application of the selective ErbB1/EGFR and ErbB4 inhibitor , tyrphostin AG1478 ( 10 and 20 µM ) , strongly attenuated the evoked CBF responses , but did not impair resting CBF . In contrast , CBF responses were unaffected by the selective ErbB2 inhibitor , tyrphostin AG825 , at both 50 and 200 µM ( Figure 3C–E; Figure 3—source data 2 , 3 ) . Of note , ErbB3 lacks kinase activity ( Roskoski 2014 ) . We then tested the effect of soluble recombinant decoy ErbB receptors , known as chimeric ErbB receptor traps , which comprise the truncated extracellular domain of the ErbB receptor fused with the constant region ( Fc ) of human immunoglobulin ( Stratman et al . , 2010 ) . Evoked CBF responses were attenuated by superfusion of either ErbB1/EGFR ( 67 nM ) or ErbB4 ( 71 nM ) receptor traps , which block the function of ErbB1 and ErbB4 ligands , respectively , but not by superfusion of the ErbB3 ( 71 nM ) receptor trap or by control IgG1 Fc ( 286 nM ) and IgG2 Fc ( 286 nM ) fragments ( Figure 3F–H; Figure 3—figure supplement 1A; Figure 3—source data 2 , 3 ) . Notably , the effects of ErbB1 and ErbB4 receptor traps on evoked CBF responses were not additive ( Figure 3G–H; Figure 3—source data 2 , 3 ) , even though neither ErbB1 nor ErbB4 receptor traps achieved maximum inhibition . ErbB2 has no known ligand ( Roskoski , 2014 ) ; thus , these data are consistent with a role for ErbB1/EGFR or ErbB4 activation in CBF responses , and suggest the involvement of bispecific ligands with dual-specificity toward ErbB1 and ErbB4 . Next , we sought to pinpoint which ErbB ligand that requires ADAM17 cleavage for activation is involved in CBF regulation . Heparin-binding EGF-like growth factor ( HB-EGF ) is one of three ligands that can bind to both ErbB1 and ErbB4 and is expressed in the vasculature ( Zhang et al . , 2014 ) . ADAM17 is the major sheddase of HB-EGF ( Sahin et al . , 2004 ) , and ADAM17-mediated shedding of proHB-EGF largely regulates soluble , mature HB-EGF binding to and activating ErbB receptors ( Yamazaki et al . , 2003 ) . Moreover , mice lacking HB-EGF have reduced postnatal viability with defective cardiac valvulogenesis , similar to mice lacking ADAM17 ( Jackson et al . , 2003 ) , prompting us to study the role of HB-EGF in cerebrovascular regulation . To do this , we examined the impact of HB-EGF inhibition on CBF responses . Unlike all other EGF ligands apart from amphiregulin , HB-EGF has a heparin-binding domain , and interactions through this domain with cell surface-associated heparan sulfate proteoglycans ( HSPGs ) are necessary for binding and activation of ErbB receptors ( Higashiyama et al . , 1993 ) . We found that superfusion of heparin ( 40 ui/mL ) , which competitively inhibits binding of HB-EGF to cell surface HSPGs ( Higashiyama et al . , 1993 ) , impaired evoked CBF responses without affecting resting CBF ( Figure 3I–K; Figure 3—figure supplement 1B; Figure 3—source data 2 , 3 ) . To further support a role for HB-EGF in evoked CBF responses , we examined the effects of the synthetic peptide p21 , which corresponds to the heparin-binding domain of murine HB-EGF and similarly inhibits binding of HB-EGF to cell surface HSPGs ( Higashiyama et al . , 1993 ) . We found that superfusion of p21 ( 12 µM ) similarly impaired evoked CBF responses without affecting resting CBF ( Figure 3I–K; Figure 3—figure supplement 1B; Figure 3—source data 2 , 3 ) ; in contrast , a mutated inactive p21 peptide ( p21-mut; 12 µM ) had no effect on evoked or resting CBF . To assess the connection between HB-EGF and ADAM17 in the context of cerebrovascular regulation , we tested the ability of a soluble form of HB-EGF ( sHB-EGF ) to counteract cerebrovascular dysfunction induced by ADAM17 inhibition or depletion ( Figure 4—source data 1 ) . TIMP3 or the ADAM10/17 inhibitor , GW413333X , was topically applied over the neocortex and CBF responses were measured before and after superfusion with sHB-EGF . We found that sHB-EGF ( 20 nM ) prevented TIMP3 and GW-induced cerebrovascular deficits ( Figure 4D–I; Figure 4—figure supplement 1A , B; Figure 4—source data 2 , 3 ) . Also , sHB-EGF significantly improved evoked CBF responses in Adam17ex/ex mice ( Figure 4—figure supplement 2 ) . Notably , sHB-EGF could not prevent CBF deficits induced by pharmacological blockage of ErbB1/EGFR and ErbB4 ( Figure 4A–C; Figure 4—source data 1–3 ) . These findings , collectively , suggest that ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) is a key TIMP3-sensitive signaling pathway for cerebrovascular regulation . 10 . 7554/eLife . 17536 . 020Figure 4 . sHB-EGF overcomes CBF deficits induced by ADAM17 inhibition . Effects of sHB-EGF ( 20 nM ) on resting CBF ( A , D , G ) and whisker stimulation ( B , E , H ) - and adenosine ( C , F , I ) -induced CBF responses were assessed in the presence and absence of the ErbB1/ErbB4 inhibitor AG1478 ( 10 and 20 µM ) ( A–C ) , TIMP3 ( 40 nM ) ( D–F ) or the ADAM10/ADAM17 inhibitor GW413333X ( GW; 5 µM ) ( G–I ) using a cranial window model . Significance was determined by repeated measure ANOVA followed by Tukey’s post-hoc test ( *p<0 . 05 , ***p<0 . 001 compared to vehicle; ###p<0 . 001 TIMP3+sHB-EGF versus TIMP3 and GW+sHB-EGF versus GW; n = 5/group ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02010 . 7554/eLife . 17536 . 021Figure 4—source data 1 . Reagents used for Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02110 . 7554/eLife . 17536 . 022Figure 4—source data 2 . Main physiological variables of mice studied in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02210 . 7554/eLife . 17536 . 023Figure 4—source data 3 . Numerical data that were used to generate the bar charts in Figure 4 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02310 . 7554/eLife . 17536 . 024Figure 4—figure supplement 1 . Acetylcholine-induced CBF responses impaired by ADAM17 inhibition are ameliorated by exogenous sHB-EGF . ( A , B ) Effects of sHB-EGF ( 20 nM ) on acetylcholine-induced CBF responses were assessed in the presence and absence of TIMP3 ( 40 nM ) ( A ) or the ADAM10/ADAM17 inhibitor GW413333X ( GW; 5 µM ) ( B ) . ***p<0 . 001 compared to vehicle; ###pDOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02410 . 7554/eLife . 17536 . 025Figure 4—figure supplement 2 . CBF deficits induced by ADAM17 deficiency are improved by sHB-EGF . Effects of sHB-EGF ( 20 nM ) on resting CBF ( A ) and whisker stimulation ( B ) and adenosine ( C ) -induced CBF responses were assessed in Adam17ex/ex and wildtype littermate ( Adam17+/+ ) mice . Significance was determined by repeated measure ANOVA followed by Bonferroni post-hoc test ( **p< 0 . 01 , ***p<0 . 001; n = 5/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 025 Our data support the concept that the ADAM17/HB-EGF/ ( ErbB1/ErB4 ) axis regulates CBF responses , and any genetic or pharmacological maneuver that inhibits this pathway impairs CBF responses to diverse stimuli , including topical application of vasodilators and neural activity ( Figures 1–3 ) . Notably , responses cannot be enhanced by stimulation of this pathway , implying that this pathway is maximally activated in a physiological in vivo setting . Given that myogenic tone sets the resting arterial diameter from which other stimuli can induce vasoconstriction or vasodilation , we hypothesized that a reduction in the myogenic tone of cerebral arteries could represent a common mechanism underlying these CBF deficits . To test this hypothesis , we assessed the effects of inhibitors of this pathway on pressure-induced constriction of brain arteries ( Figure 5—source data 1 ) . We found that pre-incubation of arterial segments with TIMP3 ( 8 nM ) strongly attenuated myogenic tone at pressures of 40 mmHg and above compared to arteries incubated with vehicle , whereas recombinant TIMP2 ( 10 nM ) had no effect ( Figure 5A–C; Figure 5—source data 2 ) . Notably , attenuation of myogenic responses by TIMP3 was even more pronounced in intracerebral penetrating arterioles ( Figure 5—figure supplement 1 ) . Likewise , myogenic constriction to pressure was strongly attenuated by the dual ADAM10/ADAM17 inhibitor GW413333X ( 1 µM ) , but not by the ADAM10 inhibitor GI254023X ( 1 µM ) . Also , myogenic tone was reduced in heterozygous ADAM17 hypomorphic mice ( Adam17ex/+ ) compared to wild-type littermates ( Adam17+/+ ) but restored by pre-incubating arterial segments of Adam17ex/+ mice with sADAM17 ( 3 . 2 nM ) ( Figure 5D , I; Figure 5—figure supplement 2; Figure 5—source data 2 ) . Moreover , pre-incubation of arteries with the ErbB1/ErbB4 inhibitor AG1478 ( 2 µM ) or the HB-EGF inhibitor p21 peptide ( 2 . 4 µM ) , but not with the p21-mut peptide ( 2 . 4 µM ) , strongly attenuated myogenic responses ( Figure 5E , F; Figure 5—source data 2 ) . Thus , these data indicate that myogenic tone is increased by tonic activity of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway . 10 . 7554/eLife . 17536 . 026Figure 5 . The ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) signaling module is involved in regulating the myogenic tone of cerebral arteries . ( A–C ) Effects of TIMP proteins on the myogenic responses of posterior cerebral arteries to increasing intraluminal pressure . ( A , B ) Representative internal diameter recordings in the presence of TIMP2 ( 10 nM ) ( A ) or TIMP3 ( 8 nM ) ( B ) . ( C ) Summary data of results in ( A ) and ( B ) . ( D–F ) Myogenic tone of posterior cerebral arteries , tested in the presence and absence of the dual ADAM10/ADAM17 inhibitor GW413333X ( GW; 1 µM ) , the ADAM10 inhibitor GI254023X ( GI; 1 µM ) ( D ) , the ErbB1/ErbB4 inhibitor AG1478 ( 2 µM ) ( E ) , the p21 peptide ( 2 . 4 µM ) , and the mutated inactive peptide p21-mut ( 2 . 4 µM ) ( F ) . ( C–F ) **p<0 . 01 , ***p<0 . 001 versus vehicle . ( G , H ) Effects of TIMP3 ( 8 nM ) ( g ) or GW ( 1 µM ) ( H ) on the myogenic tone of posterior cerebral arteries were tested in the presence of soluble HB-EGF ( 3 nM ) or vehicle . ##p<0 . 01 , ###p<0 . 001 , TIMP3+HB-EGF versus TIMP3 and GW+HB-EGF versus GW . ( I ) Myogenic tone of posterior cerebral arteries was tested in heterozygous Adam17ex/+ ( ex/wt ) and Adam17+/+ ( wt/wt ) mice in the presence and absence of soluble HB-EGF ( 3 nM ) . **p<0 . 01 , ***p<0 . 001 Adam17ex/+ versus Adam17+/+ ; ##p<0 . 01 , ###p<0 . 001 , Adam17ex/+/HB-EGF versus Adam17ex/+ ) . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 6–8 arteries/group ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02610 . 7554/eLife . 17536 . 027Figure 5—source data 1 . Reagents used for Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02710 . 7554/eLife . 17536 . 028Figure 5—source data 2 . Numerical data that were used to generate the graphs in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02810 . 7554/eLife . 17536 . 029Figure 5—figure supplement 1 . TIMP3 strongly impairs myogenic tone of parenchymal arterioles . ( A–C ) Effects of TIMP proteins on the myogenic responses of parenchymal arterioles to increasing intraluminal pressure . ( A , B ) Representative internal diameter recordings in the presence of TIMP2 ( 10 nM ) ( A ) or TIMP3 ( 8 nM ) ( B ) . ( C ) Summary data of results in ( A ) and ( B ) . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test . ( **p<0 . 01 , ****p<0 . 0001 n = 8 arterioles/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 02910 . 7554/eLife . 17536 . 030Figure 5—figure supplement 2 . sADAM17 ameliorates arterial tone in Adam17ex/+ mice . Myogenic tone of posterior cerebral arteries was tested in heterozygous Adam17ex/+ in the presence and absence of soluble ADAM17 ( 3 . 2 nM ) . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( **p< 0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . n = 5–6 arteries/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 030 To provide further support for the concept that ADAM17 and HB-EGF function as part of a signaling module to enhance the myogenic tone of cerebral arteries , we tested the ability of exogenous sHB-EGF to counteract the effects of ADAM17 inhibition ( Figure 5—source data 1 ) . Pressurized arteries were pre-incubated with recombinant TIMP3 ( 4 nM ) or the ADAM10/ADAM17 inhibitor GW413333 in the presence of sHB-EGF ( 3 nM ) or vehicle . We found that co-incubation of arterial segments with sHB-EGF significantly ameliorated the TIMP3-induced reduction in arterial tone ( Figure 5G; Figure 5—source data 2 ) . Likewise , co-incubation of arterial segments with sHB-EGF overcame the reduction in arterial tone caused by GW413333-mediated inhibition of ADAM17 ( Figure 5H; Figure 5—source data 2 ) . Moreover , sHB-EGF significantly increased myogenic tone in arteries from heterozygous ADAM17-hypomorphic mice ( Adam17ex/+ ) , restoring a near-normal myogenic phenotype ( Figure 5I; Figure 5—source data 2 ) . Collectively , these data support the concept that the TIMP3-sensitive pathway , ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) , increases myogenic constriction in brain arteries . Our findings above predict that excess TIMP3 impairs arterial tone and CBF responses in CADASIL by suppressing the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway . To test this , we examined whether recombinant sADAM17 and sHB-EGF could restore normal pressure-induced myogenic constriction of brain arteries and normal cerebrovascular responses in TgNotch3R169C CADASIL mice . We found that preincubation of arterial segments with the enzymatically active extracellular domain of ADAM17 ( 3 . 2 nM ) increased myogenic tone in arteries from TgNotch3R169C CADASIL mice , whereas sADAM17 had no detectable effect on arterial segments from wild-type mice at this concentration ( Figure 6A ) . We further found that sADAM17 ( 16 nM ) , locally applied on the necortex of TgNotch3R169C CADASIL mice , significantly improved resting CBF and rescued the impaired reactivity of brain vessels to whisker stimulation and vasodilators ( Figure 6B–D; Figure 6—figure supplement 1; Figure 6—source data 1 , 2 ) . We previously reported that sHB-EGF restores myogenic responses in parenchymal arteries from TgNotch3R169C CADASIL mice ( Dabertrand et al . , 2015 ) . Here , we extend these observations , showing that exogenous sHB-EGF ( 20 nM ) restored evoked CBF responses in TgNotch3R169C mice ( Figure 6E–G; Figure 6—source data 1 , 2 ) . Collectively , these findings support the concept that the diminished myogenic tone and CBF deficits in CADASIL are caused by TIMP3-mediated suppression of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway . 10 . 7554/eLife . 17536 . 031Figure 6 . Exogenous sADAM17 and sHB-EGF ameliorate CBF deficits and arterial tone in TgNotch3R169C mice . ( A ) Myogenic tone of posterior cerebral arteries from TgNotch3R169C mice ( TgN3R169C ) and non-transgenic littermates ( WT ) was tested in the presence of soluble ADAM17 or vehicle . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 versus WT+vehicle; #p<0 . 05 , ##p<0 . 01 , ###p<0 . 001 TgN3R169C+vehicle versus TgN3R169C+sADAM17 ( n = 5–7 arteries/ group; 1 artery/mouse ) . ( B–D ) Resting CBF ( B ) and CBF responses to whisker stimulation ( C ) or adenosine ( D ) were tested in TgN3R169C and WT mice , before and after superfusion of soluble ADAM17 . ( E–G ) Effects of soluble HB-EGF tested in a second batch of TgN3R169C and WT mice . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 5–6 mice/group ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03110 . 7554/eLife . 17536 . 032Figure 6—source data 1 . Main physiological variables of mice studied in Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03210 . 7554/eLife . 17536 . 033Figure 6—source data 2 . Numerical data that were used to generate the graphs and bar charts in Figure 6 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03310 . 7554/eLife . 17536 . 034Figure 6—figure supplement 1 . Resting CBF and acetylcholine-induced CBF responses impaired by the R169C Notch3 mutation are ameliorated by exogenous sADAM17 . Resting CBF , expressed as Laser Doppler flow arbitrary units ( LDFU ) ( A ) and CBF responses to acetylcholine ( B ) were tested in TgNotch3R169C mice ( TgN3R169C ) and non-transgenic littermates ( WT ) before and after superfusion of sADAM17 ( 16 nM ) . ***p<0 . 001 . Significance was determined by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 5–6 mice/group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 034 Our prior work established that upregulation of KV channels in the plasma membrane of cerebral arterial myocytes is responsible for the diminished myogenic response of cerebral arteries in the TgNotch3R169C CADASIL model . Importantly , application of sHB-EGF was found to normalize KV current density and restore myogenic responses in cerebral arteries from TgNotch3R169Cmice ( Dabertrand et al . , 2015 ) . In light of this and the above , we investigated the involvement of TIMP3 and ADAM17 in this upregulation of KV current density . We first asked whether reducing TIMP3 expression in the TgNotch3R169C mice would decrease the number of functional KV channels . To this end , we measured KV currents in freshly isolated myocytes from cerebral arteries of TgNotch3R169C mice with normal expression of TIMP3 ( TgNotch3R169C;Timp3+/+ ) , which have reduced myogenic tone , and in freshly isolated myocytes from cerebral arteries of TgNotch3R169C mice with reduced expression of TIMP3 ( TgNotch3R169C;Timp3+/- ) , in which myogenic responses are restored ( Dabertrand et al . , 2015 ) . Currents were recorded in response to 10-mV voltage steps from −70 mV to +60 mV . We found that KV current density was significantly lower in myocytes from TgNotch3R169C;Timp3+/- mice than in myocytes from TgNotch3R169C;Timp3+/+ at all voltage steps above +10 mV ( Figure 7A–C; Figure 7—source data 2 ) . Conversely , incubation of wild-type arterial myocytes with recombinant TIMP3 ( 8 nM ) resulted in a significant increase in KV current density compared with myocytes incubated with vehicle ( Figure 7—figure supplement 1A , B; Figure 7—source data 2 ) . Remarkably , half-maximal activation voltage ( V0 . 5 ) and slope ( k ) , determined by fitting normalized peak tail currents to the Boltzmann equation , were statistically indistinguishable among arterial myocytes from the different groups analyzed . Likewise , activation ( τact ) and deactivation ( τdeact ) time constants determined from exponential fits of individual voltage-evoked current traces and current decay , respectively , were comparable among the different groups . These current kinetics , attributable to KV1 . 5 channels , are consistent with our previous report ( Dabertrand et al . , 2015 ) . These results indicate that the TIMP3 pathway regulates the number of channels , and not channel properties ( Figure 7—figure supplement 2; Figure 7—source data 1 , 2 ) . Using the Goldman–Hodgkin–Katz constant field equation and a single-channel conductance of 15 pS ( Aiello et al . , 1998 ) , we estimated the average number of functional KV channels per myocyte . This analysis showed that exogenously applied TIMP3 increased the number of KV channels in arteries from wild-type mice by ~25% ( from 3120 to 3920 per myocyte ) . A similar increase in KV channel number was observed in the TIMP3-overexpressing TgNotch3R169C;Timp3+/+ genetic model , where channel density ( 4840/myocyte ) was ~38% greater than that in TgNotch3R169C;Timp3+/- mice ( 3510/myocyte ) . 10 . 7554/eLife . 17536 . 035Figure 7 . TIMP3 haploinsufficiency and exogenous sADAM17 decrease KV channel current density in cerebral smooth muscle cells from TgNotch3R169C mice . ( A , B ) Typical family of KV currents recorded in isolated cerebral smooth muscle cells from double-mutant TgNotch3R169C;Timp3+/- mice , with Timp3 haploinsufficiency in the context of Notch3R169C overexpression ( B ) , and TgNotch3R169C;Timp3+/+ mice , with wild-type Timp3 in the context of Notch3R169C overexpression ( A ) elicited by voltage pulses from −70 mV to +60 mV in the presence of 1 μM paxilline ( included to block BK channel currents ) . ( C ) Summary of current density results , showing that current density is decreased in myocytes of TgNotch3R169C;Timp3+/- mice compared with those of TgNotch3R169C;Timp3+/+ mice . ( D ) Typical family of KV currents recorded in isolated cerebral smooth muscle cells from TgNotch3R169C mice incubated with soluble ADAM17 ( 3 . 2 nM ) . ( E ) Summary of current density results , showing that the current density of TgNotch3R169C mice is decreased in the presence of sADAM17 . Significance was analyzed by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 7–8 cells/group; 1 cell/mouse ) . Error bars indicate SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03510 . 7554/eLife . 17536 . 036Figure 7—source data 1 . Comparison of cerebral KV current properties . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03610 . 7554/eLife . 17536 . 037Figure 7—source data 2 . Numerical data that were used to generate the graphs in Figure 7 . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03710 . 7554/eLife . 17536 . 038Figure 7—figure supplement 1 . Exogenous TIMP3 increases voltage-gated potassium ( KV ) channel current density in cerebral smooth muscle cells . ( A ) Typical family of KV currents recorded in isolated cerebral smooth muscle cells from non-Tg ( WT ) mice incubated with TIMP3 ( 8 nM ) or vehicle and elicited by voltage pulses from −70 mV to +60 mV in the presence of 1 μM paxilline ( included to block BK channel currents ) . ( B ) Summary of current density results , showing that exogenous TIMP3 increases current density . Significance was analyzed by two-way repeated measures ANOVA followed by Bonferroni post-hoc test ( n = 5 cells /group; 1 cell/mouse ) . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 03810 . 7554/eLife . 17536 . 039Figure 7—figure supplement 2 . Analyses of cerebral KV current properties . ( A ) Activation time constants ( τactivation ) were determined from an exponential fit of individual voltage-evoked current traces . ( B ) De-activation time constants ( τdeactivation ) , obtained from an exponential fit of tail currents at −40 mV . ( C ) Steady-state activation properties of KV currents measured from normalized tail currents . The voltage for half-maximal activation ( V1/2 ) and the factor k were obtained from a fit of the data to the Boltzman equation . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 039 Our model predicts that exogenous sADAM17 should counteract the increase in TIMP3 in cerebral arteries of TgNotch3R169Cmice by decreasing KV channel density . Consistent with the ability of exogenous sADAM17 to restore normal myogenic responses in TgNotch3R169C mice , we found that application of enzymatically active , sADAM17 ( 3 . 2 nM ) significantly reduced KV current density in TgNotch3R169C cerebral myocytes , decreasing the density of Kv channels by ~22% ( from 4840 to 3760 channels per myocyte ) ( Figure 7D , E; Figure 7—source data 2 ) . Thus , these new findings , taken together with our previous studies , indicate that excess TIMP3 in the TgNotch3R169CCADASILmodel drives increased KV channel density and diminished myogenic responses by reducing ADAM17 activity and subsequently reducing the release of sHB-EGF .
Although SVD of the brain is a heterogeneous group of disorders with different ultimate causes acting through specific pathways , the recently emerging view is that perturbations of proteins constituting or associated with the extracellular matrix of cerebral vessels could be a convergent pathway driving the functional and structural alterations of small brain vessels ( Joutel et al . , 2016 ) . Previously , we demonstrated that elevated TIMP3 , a protein tightly bound to the extracellular matrix of brain arteries , contributes to cerebrovascular dysfunction in CADASIL , a genetic paradigm of small vessel disease of the brain ( Monet-Leprêtre et al . , 2013; Capone et al . , 2016 ) . In the present study , we establish the novel concept that a TIMP3-sensitive pathway is constitutively engaged in the regulation of cerebral hemodynamics , and we unravel the mechanism by which excess TIMP3 in brain vessels compromises cerebrovascular regulation in a clinically relevant model of CADASIL . By combining genetic and pharmacological approaches with in vivo analyses of CBF regulation and ex vivo measurements of myogenic responses of brain arteries in physiological settings , we found that ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) is a key TIMP3-sensitive signaling module essential for maintaining robust CBF responses to evoked neural activity or topically applied vasodilators as well as for myogenic responses of brain arteries . Next , using the TgNotch3R169C model , we provided pharmacological evidence that , in the setting of CADASIL , attenuated ADAM17 and HB-EGF-dependent activation of ErbB1/ErbB4 underlies deficits in evoked CBF responses and cerebral arterial tone . Further , by using patch clamp electrophysiology in combination with genetic and pharmacological approaches , we identified upregulated KV channel current density in cerebral arterial myocytes as the heretofore-unrecognized downstream effector of this TIMP3-sensitive pathway by which excess TIMP3 reduces arterial tone in the TgNotch3R169C CADASIL model . Collectively , these data suggest that elevated TIMP3 blunts the activity of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway in cerebral arterial myocytes , thereby attenuating myogenic responses in brain arteries and compromising CBF regulation in CADASIL ( Figure 8 ) . 10 . 7554/eLife . 17536 . 040Figure 8 . Proposed model of TIMP3 regulation of cerebral arterial tone and CBF responses . ( A ) Under physiological conditions ( upper panel ) , TIMP3 is present in a low abundance in the extracellular matrix of brain arteries . ADAM17 at the cell surface of cerebral arterial myocytes is therefore active and able to cleave and release sHB-EGF , resulting in ErbB1/ErbB4 activation and KV1 channel endocytosis . The internalization of KV1 channels relieves the tonic hyperpolarizing influence of these channels on the membrane potential of arterial myocytes , thereby allowing full development of pressure-induced vasoconstriction ( myogenic tone ) of brain arteries and enabling full CBF responses to whisker stimulation and vasodilators . ( B ) In CADASIL ( lower panel ) , Notch3ECD accumulates at the surface of smooth muscle cells , leading to an increase in the amount of TIMP3 , which binds to and inhibits ADAM17 , blunting sHB-EGF release and ErbB1/ErbB4 activity , and thereby decreasing KV1 endocytosis . The resulting increase in KV1 current density hyperpolarizes arterial myocytes , acting as a brake to limit the development of myogenic tone and evoked CBF responses . DOI: http://dx . doi . org/10 . 7554/eLife . 17536 . 040 Our data provide the first evidence for a mechanistic link between a change in a component of the extracellular matrix of cerebral arteries—TIMP3—and a pathogenic alteration in the density of an ion channel—KV—in cerebral arterial myocytes . KV channels are powerful negative regulators of arterial tone , which act by exerting a tonic hyperpolarizing influence on the membrane potential of arterial smooth muscle cells that serves to limit pressure-induced depolarization and vasoconstriction ( Longden et al . , 2015 ) . Our results introduce the novel concept that the concentration of TIMP3 in brain vessels regulates arterial tone and blood flow by playing a critical role in adjusting KV channel density . We surmise that such an extracellular matrix-dependent paradigm may be at play in more common forms of cerebral small vessel disease where remodeling of the vascular extracellular matrix is a key feature ( Joutel et al . , 2016 ) . Our results indicate that , under physiological conditions , tonic activity of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway prevents excess accumulation of KV channels at the plasma membrane and thereby maintains myogenic tone and robust CBF responses to neural activity and vasodilators . Interestingly , we found that only factors that inhibit this pathway had a functional effect; activating this pathway in wild-type mice by providing sADAM17 or sHB-EGF did not enhance evoked CBF responses or myogenic tone . This suggests that the set point of this pathway in a physiological in vivo setting is already at maximum . In support of this interpretation is a recent study showing that genetic overexpression of ADAM17 protein does not result in enhanced shedding activity in vivo ( Yoda et al . , 2013 ) . On the other hand , decreasing KV current density in cerebral artery myocytes , which is at least 50% lower than that in peripheral artery myocytes ( Dabertrand et al . , 2015 ) , could be an in vivo rate-limiting step following physiological activation of this pathway . Notably however , studies in experimental models of aneurysmal subarachnoid hemorrhage indicate that this pathway can be further activated in a pathological context . Indeed , Wellman and colleagues have shown that the blood component , oxyhemoglobin , causes suppression of KV currents in cerebral arterial myocytes through HB-EGF–mediated activation of ErbB1/ErbB4 , resulting in membrane depolarization and enhanced tone of brain arteries ( Nystoriak et al . , 2011 ) . In the present study , we found that any genetic or pharmacological maneuver that blocked the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway attenuated both the myogenic tone of brain arteries and the increase in CBF responses evoked by diverse stimuli; conversely , exogenous sADAM17 and sHB-EGF could overcome the reduction in both myogenic tone and evoked CBF responses elicited by the R169C Notch3 mutation or elevated TIMP3 . Moreover , our previous ( Dabertrand et al . , 2015 ) and current results collectively indicate that upregulation of KV channels is sufficient to explain the decrease in myogenic tone , without an involvement of the endothelium or large conductance , voltage and Ca2+ activated K+ ( BK ) channels . Nonetheless , we cannot exclude an effect on other channels engaged by pressure . Given the key role of KV channels in the regulation of arterial tone , these findings are consistent with the interpretation that the smooth muscle ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) /KV pathway regulates evoked CBF responses by elevating the physiological tone of brain arteries , and that the reduction in myogenic tone caused by inhibition of this pathway by excess TIMP3 in the extracellular matrix surrounding smooth muscle cells likely accounts for the attenuation of evoked CBF responses in CADASIL ( Figure 8 ) . A previous study in acute brain slices provides additional support for this interpretation , showing that the initial degree of arteriolar tone determines the diameter changes elicited by functional hyperemia ( Blanco et al . , 2008 ) . On the other hand , a transient loss of myogenic tone is expected to increase resting CBF , and vice versa . Unexpectedly , we found that acute pharmacological blockade of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway did not affect resting CBF ( or inconsistently increased it ) , despite its ability to profoundly reduce myogenic responses of brain arteries ex vivo . Also , neocortical application of exogenous sADAM17 unexpectedly increased resting CBF in TgNotch3R169Cmice , despite its ability to increase and normalize myogenic tone in these mice . However , although myogenic tone and myogenic responses are known to contribute to the regulation of resting CBF , their relative importance are hard to quantify and poorly understood; their contribution may also change depending on conditions or disease states and other mechanisms —metabolic , neural , endothelial—also influence or contribute to resting CBF ( Cipolla , 2009 ) . It is also possible that overall resting CBF does not change despite seeing a change in myogenic tone in one portion of the vasculature because of compensatory adjustments in vessels downstream . Simultaneous in vivo recordings of blood flow and vessel diameters may be of interest to address this possibility . Moreover , it should be stressed that in our experiments cell populations targeted by pharmacological compounds likely differ depending on whether the compound is topically applied in vivo over the somatosensory cortex or incubated ex vivo with isolated brain arteries . In particular , proteins or peptides topically applied in vivo are thought to target only the abluminal surface of the vessel ( smooth muscle cells ) ( Park et al . , 2013 ) and may target other brain cells ( e . g . , astrocytes ) , whereas ex vivo incubation targets only vascular cells , including both abluminal and luminal ( endothelial cell ) surfaces . In this regard , involvement of the ADAM17/HB-EGF/ ( ErbB1/ErbB4 ) pathway in cells other than arterial myocytes cannot be ruled out . Finally , our results may point toward the involvement of other downstream effectors in addition to KV channels in the regulation of CBF by this pathway . The fact that increasing TIMP3 or decreasing ADAM17 caused a concentration-dependent impairment of cerebrovascular function taken together with the observation that exogenous sADAM17 is capable of overcoming elevated TIMP3-induced cerebrovascular dysfunction indicates that ADAM17 activity depends on the relative activity of ADAM17 and TIMP3 in brain arteries . Several lines of evidence from cell systems indicate that the bulk of ADAM17 is intracellular , whereas the majority of ADAM17 shedding activity occurs at the cell surface , where ADAM17 can associate with its natural inhibitor TIMP3 ( Xu et al . , 2012; Chapnick et al . , 2015 ) . Thus , the ratio of TIMP3 and ADAM17 at the cell surface is likely a key determinant of ADAM17 activity . Biochemical confirmation of this in brain arteries will require further investigation , although the lack of ADAM17 and TIMP3 antibodies suitable for immunohistochemistry , the tiny amount of material provided by segments of cerebral arteries for biochemical studies , and the lack of in situ or specific assay to assess ADAM17 activity in tissues remain major technical obstacles . Although many of the molecular details of the mechanism responsible for EGFR-mediated suppression of KV channels in cerebral arterial myocytes remain unsettled , previous studies have shown that activation of EGFR tyrosine kinase activity can suppress KV channel activity through enhanced endocytosis ( Koide et al . , 2007; Ishiguro et al . , 2006 ) . Functional homo- or heteromeric KV channels are formed from four α-subunits , plus additional β-subunits . KV1 . 5 , and to a lesser extent KV1 . 2 , are the predominant α-subunits in rodent brain arteries ( Thorneloe et al . , 2001; Straub et al . , 2009 ) . Whereas direct tyrosine phosphorylation of the channel has been identified as the mechanism regulating KV 1 . 2 endocytosis in HEK or neuronal cells ( Nesti et al . , 2004 ) , a role for this mechanism in KV1 . 5 endocytosis has not yet been demonstrated ( Ishiguro et al . , 2006 ) . Whether other subunits within the KV 1 . 5 channel complex or a closely associated protein is the target of phosphorylation remains to be tested . On the other hand , KV channel suppression could be mediated by enhanced lysosomal or proteasomal degradation , as recently shown for KV 1 . 5 in mesenteric arteries ( Kidd et al . , 2015 ) . Future experiments are needed to elucidate mechanisms responsible for the regulation and trafficking of KV1 channels in cerebral arterial myocytes . In summary , our study has uncovered a novel and central role for the ADAM17/HB-EGF/ErbB/KV signaling pathway in the physiological and pathological control of CBF and arterial tone . Our results highlight a heretofore-unrecognized mechanistic link between pathological alterations of the vascular extracellular matrix and KV channel density that underlies cerebrovascular dysfunction in CADASIL . We believe that this novel extracellular matrix-dependent mechanism establishes an important paradigm for cerebrovascular regulation . Importantly , illumination of its dysfunction in cerebral small vessel disease offers multiple points of potential therapeutic intervention that may prove to be more easily druggable than pathological changes in vascular extracellular matrix .
Acetylcholine , adenosine , the selective ErbB1/ErbB4 inhibitor tyrphostin AG1478 , and the selective ErbB2 inhibitor tyrphostin AG825 ( Levitzki and Gazit , 1995 ) were purchased from Sigma Aldrich ( St . Louis , MO ) . Heparin was purchased from Merck Millipore ( Molsheim , France ) . The ADAM inhibitors GI254023X ( ADAM 10 ) and GW413333X ( ADAM10/ADAM17 ) ( Hundhausen et al . , 2003 ) were synthesized by Iris Biotech ( Marktredwitz , Germany ) . Murine recombinant TIMP1 , murine soluble ErbB1 , ErbB3 and ErbB4 receptor traps ( ErbB1-IgG1 Fc , ErbB3-IgG2 Fc and ErbB4-IgG2 Fc ) , and control IgG1 Fc and IgG2 Fc fragments , as well as human bioactive ADAM17 were purchased from R&D Systems ( Lille , France ) . Murine recombinant TIMP2 and TIMP3 were purchased from Uscn Life Science ( Houston , TX , and murine sHB-EGF was purchased from BioVision ( Milpitias , CA ) . The p21 peptide ( H-KKK KKG KGL GKK RDP CLR KYK-OH ) , which competitively inhibits HB-EGF binding to heparan sulfate proteoglycan ( Higashiyama et al . , 1993 ) , and a mutated , inactive peptide in which all lysine residues that are important for inhibitory activity are replaced with alanines ( p21-mut; H-AAA AAG AGL GAA RDP CLR AYA-OH ) were purchased from Eurogenetec ( Seraing , Belgium ) and resuspended at a concentration of 25 mM in DMSO , following the manufacturer’s directions . Paxilline was purchased from A . G . Scientific ( San Diego , CA ) . Apamin and charybdotoxin were purchased from Enzo Life Sciences ( Farmingdale , NY ) . Papain and collagenase type 4 were purchased from Worthington Biochemical Corporation ( Lakewood , NJ ) . All other chemicals were obtained from Sigma Aldrich . Experiments were conducted in FVB/N mice ( Charles River Laboratories , France ) ; transgenic mice overexpressing the R169C mutation of Notch3 ( TgNotch3R169C , line 88 ) , bred on an FVB/N background ( Joutel et al . , 2010 ) ; transgenic mice overexpressing human TIMP3 ( TgBAC-TIMP3 ) , bred on a hybrid background ( 88% FVB/N/12% C57Bl/6 ) ( Capone et al . , 2016 ) ; homozygous Adam17ex/ex mice , which express profoundly reduced ADAM17 protein levels in all tissues; and heterozygous Adam17ex/+ mice and wild-type littermates ( Chalaris et al . , 2010 ) , maintained on a C57BL/6-SV129 hybrid background . Genotyping analyses were performed by polymerase chain reaction ( PCR ) using the following primer pairs: TgNotch3 , 5’-TCA ACG CCT TCT CGT TCT TC-3’ ( forward ) and 5’-AAT ACC GTC GTG CTT TCG AG-3’ ( reverse ) ; TgBAC-TIMP3 , 5’-CCA GGA GAC AGC AAG TAG CC-3’ ( forward ) and 5’-GCT GCT GTT TAG GGA TCT GC-3’ ( reverse ) ; Adam17 mutant and wild-type allele , 5’- TAT GTG ATA GGT GTA ATG -3’ ( forward ) and 5’ CTT ATT ATT CTC GTG GTC ACC -3’ ( reverse ) . Mice were bred and housed in pathogen-free animal facilities and fed a standard diet ad libitum with free access to water . All experiments described in this study were conducted in full accordance with the guidelines of our local Institutional Animal Care and Use Committee ( Lariboisière-Villemin ) , with every effort being made to minimize the number of animals used . All mice were male , aged 2 months , except for TgNotch3R169C , TgBAC-TIMP3 and non-transgenic littermate mice , which were 6 months old . We report this study in compliance with the ARRIVE guidelines . Protein extracts were prepared from cerebral pial arteries and immunoblotted using rabbit polyclonal anti-ADAM17 ( 18 . 2 ) ( 1:2000 ) ( Chalaris et al . , 2010 ) and anti-smooth muscle α-actin ( Clone 1A4 , Dako; Les Ulis , France ) antibodies , as previously described ( Monet-Leprêtre et al . , 2013 ) . Densitometric quantification of band intensity was performed using ImageJ ( version 10 . 2 , NIH ) . Mice were anesthetized and surgically prepared as described above . Field potentials were recorded using a stainless steel bipolar electrode placed in the somatosensory cortex contralateral to the activated whiskers ( 3 mm lateral and 1 . 5 mm caudal to bregma; depth , 0 . 5 mm ) . The somatosensory cortex was activated by two needle electrodes ( 21 gauge ) subdermally inserted in the whisker pad . Each stimulation trial lasted for 1 min ( 0 . 65 mA; 0 . 5 Hz; pulse duration , 1 ms ) and the interval between two trials was 10 min . Eight consecutive stimulation trials were performed on each mouse . The first four cycles were carried out in presence of vehicle , and the subsequent four trials were performed in the presence of recombinant TIMP3 ( 40 nM ) ; analyses were performed on the average of four trials . Data were obtained and recorded using the MP36R System ( Biopac System , CA ) and analyzed off-line using AcqKnowledge Software ( Biopac System , CA ) . After overdosing with CO2 , mice were decapitated and their brains were harvested . Arterial segments of the posterior cerebral artery and precapillary segments of parenchymal arterioles that arise from the middle cerebral artery M1 region and perfuse the neocortex were dissected , cannulated on two glass micropipettes in an organ chamber containing physiological salt solution ( PSS ) maintained at 37°C ( pH 7 . 4 ) , and pressurized using an arteriograph system ( Living Systems Instrumentation , Inc . , St . Albans , VT ) as previously described ( Joutel et al . , 2010 ) . Once prepared , arteries were allowed to stabilize for at least 60 min at a pressure of 60 mmHg until the development of basal tone . Pressure was then switched to 20 mmHg and compounds were added to the chamber for 20 to 60 min before increasing the intraluminal pressure to 40 , 60 , 80 and 100 mm Hg using a pressure-servo control pump . Vessel internal diameter was continuously recorded using a CCD camera and edge-detection software ( Biopac MP150; Biopac Systems Inc . , CA or AcqKnowledge Software; IonOptix , Milton , MA ) . Diameters measured in PSS were considered active diameters . At the conclusion of each experiment , maximal dilation was obtained in nominally Ca2+-free PSS containing EGTA ( 2–5 mM; Sigma ) . Artery diameters are given in micrometers . Myogenic tone was expressed as the percentage of passive diameter ( [passive diameter – active diameter]/passive diameter × 100 ) . Compound concentrations were based on initial experiments of cerebrovascular reactivity and used at approximately one fifth of the concentration used in vivo . Anterior , middle , and posterior cerebral arteries and arterioles were cleaned of connective tissue and placed in cell-isolation solution . Single smooth muscle cells were isolated from cerebral arteries by enzymatic digestion in papain ( 0 . 5 mg/mL ) and dithioerythritol ( 1 mg/mL ) for 12 min , followed by a second digestion in collagenase type 4 ( 1 mg/mL ) without Ca2+ for 10 min . Digested tissue was washed out and gently triturated with a fire-polished glass pipette . The single-cell suspension of myocytes was refrigerated until use ( typically 4–6 hr ) . Outward K+ currents were recorded from single cells in the presence of 1 µM paxilline ( to block BK currents ) at room temperature using the perforated-cell configuration of the patch-clamp technique . Recording electrodes with resistances of 2–4 MΩ were pulled from borosilicate glass and backfilled with a pipette solution of appropriate composition . Currents were recorded from cells on an Axopatch 200B amplifier , filtered at 2 kHz using a low-pass Bessel filter , and digitized at 10 kHz ( Digidata 1322A; Molecular Devices ) . pCLAMP-9 software ( Molecular Devices ) was used for data recording and analysis . The composition of cell isolation solution was 60 mM NaCl , 85 mM Na-glutamate , 3 mM KCl , 2 mM MgCl2 , 10 mM HEPES , 10 mM glucose , 7 mM mannitol , pH 7 . 4 . For patch-clamp experiments , the bath solution composition was 137 mM NaCl , 3 mM KCl , 0 . 1 mM CaCl2 , 4 mM glucose , 10 mM HEPES ( pH 7 . 3 ) , and contained paxilline ( 1 µM ) ; the pipette solution was 10 mM NaCl , 30 mM KCl , 110 mM K-aspartate , 1 mM MgCl2 , 10 mM HEPES ( pH 7 . 2 ) , and contained 250 µg/mL amphotericin B . Families of outward KV currents were elicited by series of 10-mV depolarizing steps from −70 mV to +60 mV , from a holding potential of −80 mV ( Figure 7a ) . Current density was calculated by dividing membrane current amplitude at the end of the pulse by cell capacitance . The relationship between myocyte membrane voltage and the amplitudes of tail currents ( I ) was fit to the Boltzmann equation , I=Imax1+e ( V0 . 5+V ) /k where Imax is the measured peak tail current , which allows determination of the half-maximal activation potential ( V0 . 5 ) and slope ( k ) . Data are expressed as means ± SEM . Sample size needed for CBF and myogenic tone analysis as well as for electrophysiology experiments was determined based on our prior works ( Dabertrand et al . , 2015 ) , ( Capone et al . , 2016 ) ; n values indicate the number of biological replicates . CBF responses were analyzed by one-way analysis of variance ( ANOVA ) or repeated-measure ANOVA followed by Bonferroni or Tukey post-hoc tests . Evoked potential fields were analyzed using unpaired Student’s t-test . Myogenic tone and current densities were analyzed by two-way repeated-measure ANOVA followed by Bonferroni post-hoc tests . All statistics were performed using Graph Pad Prism . Differences with p-values < 0 . 05 were considered statistically significant . The significance level was set at p<0 . 05 . | There are currently no effective treatments or cures for small blood vessel diseases of the brain , which lead to strokes and subsequent decreases in mental abilities . Normally , smooth muscle cells that surround the vessels relax or contract to regulate blood flow and ensure the right amount of oxygen and nutrients reaches the different regions of the brain . In a syndrome called CADASIL , which is the most common form of inherited small vessel disease , a genetic mutation causes the smooth muscle cells to weaken over time . The accumulation of several proteins – including one called TIMP3 – around the smooth muscle cells plays a key role in the smooth muscle cell weakening seen in CADASIL . Capone et al . have now studied mice that display the symptoms of CADASIL to investigate how TIMP3 decreases blood flow through blood vessels in the brain . This revealed that TIMP3 inactivates another protein called ADAM17 . The latter protein is normally responsible for starting a signaling pathway that helps smooth muscle cells to regulate blood flow according to the needs of the brain cells . Artificially adding more ADAM17 to the brains of the CADASIL mice reduced their symptoms of small vessel disease . Using smooth muscle cells freshly isolated from the brains of CADASIL mice , Capone et al . also demonstrated that abnormal TIMP3-ADAM17 signaling increases the number of voltage-dependent potassium channels in the membrane of the muscle cells . Having too many of these channels impairs the flow of blood through vessels in the brain . Further experiments are needed to investigate whether correcting TIMP3-ADAM17 signaling could prevent strokes in people with inherited CADASIL . It also remains to be seen whether similar signaling mechanisms are at play in other small vessel diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Mechanistic insights into a TIMP3-sensitive pathway constitutively engaged in the regulation of cerebral hemodynamics |
Lytic transglycosylases ( LT ) are enzymes involved in peptidoglycan ( PG ) remodeling . However , their contribution to cell-wall-modifying complexes and their potential as antimicrobial drug targets remains unclear . Here , we determined a high-resolution structure of the LT , an outer membrane lipoprotein from Neisseria species with a disordered active site helix ( alpha helix 30 ) . We show that deletion of the conserved alpha-helix 30 interferes with the integrity of the cell wall , disrupts cell division , cell separation , and impairs the fitness of the human pathogen Neisseria meningitidis during infection . Additionally , deletion of alpha-helix 30 results in hyperacetylated PG , suggesting this LtgA variant affects the function of the PG de-O-acetylase ( Ape 1 ) . Our study revealed that Ape 1 requires LtgA for optimal function , demonstrating that LTs can modulate the activity of their protein-binding partner . We show that targeting specific domains in LTs can be lethal , which opens the possibility that LTs are useful drug-targets .
Lytic transglycosylases ( LTs ) degrade peptidoglycan ( PG ) to produce N-acetylglucosamine ( GlcNAc ) −1 , 6-anhydro-N-acetylmuramic acid ( MurNAc ) -peptide ( G-anhM-peptide ) , a key cytotoxic elicitor of harmful innate immune responses ( Viala et al . , 2004 ) . LTs have been classified into four distinct families based on sequence similarities and consensus sequences . LTs belonging to family 1 of the glycoside hydrolase ( GH ) family 23 share sequence similarity with the goose-type lysozyme ( Blackburn and Clarke , 2001 ) . Family 1 can be further subdivided into five subfamilies , 1A through E , which are all structurally distinct ( Blackburn and Clarke , 2001 ) . Despite the overall structural differences among LTs , their active sites , enzymatic activities and substrate specificities are fairly well conserved . The crystal structure of the outer membrane lipoprotein LtgA , a homolog of Slt70 that belongs to family 1A of GH family 23 from the pathogenic Neisseria species , was previously determined at a resolution of 1 . 4 Å ( Figure 1a ) . ( Williams et al . , 2017; Williams et al . , 2018 ) . Briefly , LtgA is a highly alpha-superhelical structure consisting of 37 alpha helices ( Figure 1a ) . Although LTs have very diverse overall secondary structures , they exhibit similar substrate specificities and a preference for PG ( Vollmer et al . , 2008 ) . LtgA shares an overall weak sequence similarity with Slt70 ( 25% ) . However , the structural and sequence alignments of the catalytic domains of Slt70 and LtgA revealed absolute active site conservation ( Williams et al . , 2018 ) . The active site of LtgA is formed by ten alpha helices ( α 28 , 29 , 30 , 31 , 32 , 33 , 34 , 35 , 36 , 37 ) , with a six-alphahelix bundle ( α 29 , 30 , 31 , 32 , 33 , 34 ) constituting the core of the active site that firmly secures the glycan chain ( Figure 1a ) . LTs utilize a single catalytic residue , either a glutamate or aspartate , which plays the role of an acid and then that of a base ( Thunnissen et al . , 1994; van Asselt et al . , 1999; Scheurwater et al . , 2008; Reid et al . , 2004; van Asselt and Dijkstra , 1999 ) . In our recent study , active LtgA was monitored for the first time in the crystalline state , and the residues involved in the substrate and product formation steps were identified . Globally , conformational changes occurred in three domains , the U , C and L domains , between native LtgA and LtgA bound to the product ( Williams et al . , 2018 ) . Substantial conformational changes were observed in the active site , for example , during the product formation step , the active site adopted a more open conformation ( Williams et al . , 2018 ) . Many Gram-negative bacteria have multiple and redundant LTs; for example , Escherichia coli has eight ( MltA , MltB , MltC , MltD , MltE , MltF , MltG and Slt70 ) , and Neisseria species encode 5 ( LtgA , LtgB , LtgC , LtgD , and LtgE ) . Because the activity of LTs is redundant , the loss of one or more LTs in E . coli leads to no observable growth defects . When genes for six LTs were deleted from E . coli , a mild chaining phenotype was observed ( Heidrich et al . , 2002 ) . However , despite lack of strong observable phenotypic changes , it has been suggested that LTs may have well-defined roles in the cell . For example , the deletion of ltgA and ltgD in Neisseria gonorrhoeae eliminates the release of cytotoxic PG monomers suggesting the activities of LtgA and LtgD are redundant . Moreover , LtgA primarily localizes at the septum , indicating a role in the divisome machinery , whereas LtgD is distributed along the entire cell surface ( Schaub et al . , 2016 ) . The activities of LTs are known to be inhibited by β-hexosaminidase inhibitors ( for example , NAG-thiazoline ) ; by bulgecins A , B and C; and by PG-O-acetylation ( Williams et al . , 2017; Reid et al . , 2004; Templin et al . , 1992; Tomoshige et al . , 2018 ) . PG-O-acetylation ( Weadge et al . , 2005 ) is a process that allows pathogenic bacteria to subvert the host innate immune response ( Diacovich and Gorvel , 2010; Aubry et al . , 2011 ) . It should be noted that many Gram-positive and Gram-negative bacteria O-acetylate their PG , with a few notable exceptions such as E . coli and Pseudomonas aeruginosa ( Clarke et al . , 2010 ) . Peptidoglycan O-acetylation prevents the normal metabolism and maturation of PG by LTs ( Bera et al . , 2005 ) . Ape1 , a PG de O-acetylase , is present in Neisseria species and generally in Gram-negative bacteria that O-acetylate their PG . Ape1 catalyzes the hydrolysis of the O-acetyl modification specifically at the sixth carbon position of the muramoyl residue , thus assuring the normal metabolism of PG by LTs ( Weadge et al . , 2005; Weadge and Clarke , 2006; Pfeffer and Clarke , 2012 ) . LTs form protein complexes with other members of the PG biosynthetic apparatus , such as PBPs ( Vollmer et al . , 2008; Dijkstra and Thunnissen , 1994; Romeis and Höltje , 1994; van Heijenoort , 2011; Legaree and Clarke , 2008; Vollmer and Bertsche , 2008 ) . Most notable are the interactions between Slt70 and PBPs 1b , 1c , 2 and 3 ( von Rechenberg et al . , 1996 ) . PBPs are essential for bacterial cell wall synthesis and are required for proliferation , cell division and the maintenance of the bacterial cell structure . Previously , PBPs were thought to be primarily responsible for the polymerization of PG . Recently , RodA , a key member of the elongasome , and a shape , elongation , division and sporulation ( SEDS ) protein family member was shown to be a PG polymerase . RodA functions together with PBP2 to replicate the transglycosylase and transpeptidase activities found in bifunctional PBPs ( Cho et al . , 2016; Meeske et al . , 2016; Sjodt et al . , 2018 ) . SEDS proteins are widely distributed in bacteria and are important in both the cell elongation and division machinery . Neisseria species such as N . gonorrhoeae and N . meningitidis are coccoid in shape and lack an elongation machinery . Therefore , these species incorporate new PG through complex interactions in the divisome . Both N . gonorrhoeae and N . meningitidis have five PBPs , namely , PBP1 , PBP2 , PBP3 , PBP4 and PBP5 . PBP1 and PBP2 are homologous to E . coli PBP1a and PBP3 , while the Neisseria PBP3 and PBP4 are homologous to E . coli PBP4 and PBP7 ( Sauvage et al . , 2008 ) . PBP5 in both E . coli and Neisseria species are both predicted carboxypeptidases ( Zarantonelli et al . , 2013 ) . FtsW , a RodA homolog and a key component of the divisome machinery , forms a complex with FtsI ( PBP3 ) . The FtsW-PBP3 complex shares similar interacting regions with the RodA-PBP2 complex , and is the confirmed PG polymerase of the divisome ( Taguchi et al . , 2019 ) . Previous work by our group and others have demonstrated that PBPs and LTs can be targeted in a combined antibiotic regimen that could counter antibiotic resistance ( Bonis et al . , 2012 ) , highlighting the possibility of simultaneously inhibiting LTs and their binding partners , such as PBPs , to achieve a synergistic antibiotic effect . Here , we reveal the near-atomic-resolution crystal structure of a native version of LtgA with a disordered active site alpha helix . When LtgA , missing the alpha helix 30 motif , was expressed from an ectopic locus in N . meningitidis ( at an elevated level compared to wild type ) , bacterial growth , cell division and daughter cell separation were disrupted , compromising the integrity of the cell wall and PG composition , and diminishing bacterial fitness or virulence in a mouse infection model . It is known that LTs exist in multi-protein complexes . Here , we demonstrate that LTs can enhance the activity of one of their protein-binding partners thus ascribing a new role to LTs in the PG degrading machinery . This study demonstrates that despite the redundancy of LTs , they can be useful potential targets for future antibiotic development .
In the course of monitoring the LtgA reaction in the crystalline state , we captured a native version of LtgA with a distinctly disordered alpha helix 30 ( Figure 1a–b , Video 1 ) . This represents a newly identified conformational state of LtgA whereby alpha helix 30 transitions from an ordered to a disordered state ( Figure 1a–b ) . Interestingly , this disorder of alpha helix 30 did not affect the overall structural integrity of the active site ( Figure 1b , Video 1 ) because all the other helices making up the catalytic domain remained intact . Moreover , LtgA was already shown to be active in the crystalline state in our previous studies , although the molecular details of alpha helix 30 interactions with the ligand was not addressed ( Figure 1a–b , Video 1; Williams et al . , 2018 ) . Alpha helix 30 , with the sequence 501 ( MPATAREIAGKIGMD ) 516 ( Figure 1a–b , colored in light red ) , is structurally conserved among the closest homologs of LtgA , mainly , Slt’s , and other LTs such as MltE and MltC ( van Asselt et al . , 1999; Artola-Recolons et al . , 2011; Fibriansah et al . , 2012; Artola-Recolons et al . , 2014; Höltje , 1996; Figure 1—figure supplement 1 ) . Alpha helix 30 clamps the glycan strand during catalysis ( Figure 1c , Video 1 ) and undergoes conformational changes to a more open conformation after product formation ( Figure 1d ) . Met 501 and Glu 507 of alpha helix 30 lose hydrogen-bonding contact with the ligand after product formation ( Figure 1d , Video 1 ) . Consistent with the structural data showing the role of alpha helix 30 in substrate/product binding , a heterologously expressed and purified LtgAΔ30 showed severely diminished PG-binding capabilities when compared to wild-type LtgA or mutants of residues involved in the catalytic mechanism or substrate binding ( E481A , E580 ) of LtgA ( Figure 1—figure supplement 2 ) . This further emphasizes the potential critical structural role of alpha helix 30 in the function of LtgA and consequently in the proper metabolism of the PG . Given the important structural role of LtgA alpha helix 30 , we investigated its functional role in vivo by engineering the following constructs in N . meningitidis: i ) an LtgA knockout strain ( ΔltgA ) , ii ) a knockout strain complemented at an ectopic locus on the meningococcal chromosome with the wild-type gene ( ΔltgAltgA ) , or iii ) complemented with alpha helix 30 deletion ( ΔltgAltgAΔ30 ) . Similar to other LTs , a complete deletion of the ltgA gene from the chromosome did not affect the growth of the bacteria ( Figure 2a; Chan et al . , 2012 ) . Interestingly , the strain with ltgA lacking the alpha helix 30 coding sequence exhibited severely reduced growth ( Figure 2a ) , with an exponential phase growth rate ( 0 . 059 h−1 ±0 . 012 ) significantly different from that of the wild-type or ΔltgAltgA strain ( 0 . 72 h−1 ±0 . 15 or 0 . 21 h−1 ±0 . 043 , respectively ) based on the calculated slopes of the growth curves ( p<0 . 0001 ) ( Figure 2a ) . To exclude concerns about LtgA stability and to confirm that LtgAΔ30 continued to be expressed , the degradation of LtgA across all four strains was examined by western blot of lysates of bacteria harvested at various time points after incubation with chloramphenicol ( Figure 2b–c ) . As expected , LtgA was not detected in the ΔltgA knockout mutant ( Figure 2b–c ) . The levels of LtgA or LtgA Δ30 in the ΔltgAltgA and ΔltgAltgAΔ30 strains was 4 . 2 and 3 . 7 times higher , respectively , than observed in the wild-type strain at t0 , possibly because the transcription of ltgA was controlled by a stronger promoter in these strains when compared to the parental strain . After the addition of chloramphenicol , LtgA appeared to be maintained at comparable levels in the wild-type and complemented strains , and the levels decreased slowly during the sampling period , as revealed by quantitative measurement of relative protein abundance using densitometry ( t1/2 > 9 h ) ( Figure 2b–c ) . The promoter for ltgA has not yet been identified; therefore , ltgA was introduced in the chromosome of meningococcus and expressed under the control of a non-native promoter . Since ΔltgAltgAΔ30 exhibited reduced growth and this could be attributed to bacterial lysis or defects in cell division or cell separation , we examined all four strains using fluorescent microscopy ( labeling the membrane and intracellular DNA ) , and scanning electron microscopy ( SEM ) . Despite the reduced growth of strain ΔltgAltgAΔ30 , there was no physical evidence suggesting bacterial lysis . However , intriguingly in the ΔltgAltgAΔ30 strain , we observedstrong defects in cell separation and cell division , and the appearance of membrane stained extracellular material that was notably absent in the other three strains ( Figure 3 ( right panel ) , Figure 3—figure supplement 1 ) . Additionally , SEM revealed large blebs on the surface of some of the unseparated/undivided bacteria in the ΔltgAltgAΔ30 strain that were not observed in the other strains . A rather striking phenomenon is that the bacteria with blebs all had smooth surfaces that deviated from the normal rough surface appearance of N . meningitidis in the other strains ( Figure 3 ( left panel ) , Figure 3—figure supplement 1 ) . We also observed ghost cells; however , this phenomenon was not as pervasive as other abnormalities ( Figure 3 ( left panel ) ) . Interestingly , although the levels of LtgA or LtgA Δ30 expressed from an ectopic locus in N . meningitidis were higher in comparison to the natively expressed protein , severe morphological defects were only observed in the ΔltgAltgAΔ30 strain . We examined the PG profiles of wild-type N . meningitidis , ΔltgA , ΔltgAltgA and ΔltgAltgAΔ30 , to explore whether the integrity of the PG composition of ΔltgAltgAΔ30 strain was maintained . No notable differences were observed among the wild-type , ΔltgA and ΔltgAltgA strains ( Figure 4 , Figure 4—figure supplement 1 , Supplementary 1 ) . However , the PG of the ΔltgAltgAΔ30 strain was found to be markedly hyperacetylated when compared to that of the other strains , with a 102% increase in the amount of acetylated GlcNAc-anhMurNAc-tetrapeptide ( GM*4 ) , a 39% increase in acetylated GlcNAc-anhMurNAc-tetrapeptide crosslinked with GlcNAc-MurNAc-tetrapeptide ( GM*4-GM4 ) , and a 46% increase in doubly acetylated di-GlcNAc-anhMurNAc-tetrapeptide ( GM*4 GM*4 ) ( Figure 4 , Figure 4b ) . A 22% increase in the amount of GlcNAc-MurNAc tetrapeptide ( GM4 ) was also observed , while the amounts of GlcNAc-MurNAc tripeptide ( GM3 ) and GlcNAc-MurNAc pentapeptide ( GM5 ) decreased by 33% ( Figure 4 , Figure 4b ) . Overall , there was a marked increase in the amounts of acetylated PG monomers and dimers . The PG de-O-acetylase ( Ape1 ) is the enzyme responsible for removing the O-acetyl group from the C6-hydroxyl position of the glycan strand of the O-acetylated PG and ensures the continued metabolism of the PG by LTs ( LtgA , LtgD or LtgE and others ) ( Weadge et al . , 2005; Weadge and Clarke , 2006; Pfeffer and Clarke , 2012; Veyrier et al . , 2013 ) . Since the PG of the ΔltgAltgAΔ30 strain was hyperacetylated , the expression of Ape1 was assessed in all four strains ( N . meningitidis , ΔltgA , ΔltgAltgA , ΔltgAltgAΔ30 ) ( Figure 4—figure supplement 2; Figure 4—figure supplement 1 ) . Ape1 was comparably expressed in all four strains ( Figure 4—figure supplement 2; Figure 4—figure supplement 1 ) . Hyperacetylation of the PG was the most striking phenotype of the ΔltgAltgAΔ30 strain . We therefore explored whether: 1 ) LtgA and Ape1 form a PG degrading complex , or 2 ) the normal function of Ape1 depends on LtgA , or 3 ) Ape1 and LtgA work in concert enzymatically to de-O-acetylate the PG . To accomplish this , we purified Ape1 and LtgA , following their heterologous expression in E . coli ( Figure 5a ) . Each enzyme was purified individually and then combined prior to their application to size-exclusion columns ( Figure 5a ) . LtgA formed an approximately 100 kDa complex with Ape1 ( Figure 5a ) . We next examined the activity of LtgA and Ape1 against acetylated PG from N . meningitidis , or the activity of Ape1 alone , or Ape1 combined with LtgA toward 4-nitrophenyl acetate , a previously characterized substrate of Ape1 from N . gonorrhoeae that is not a substrate for LtgA ( Pfeffer et al . , 2013; Weadge and Clarke , 2007 ) . In the presence of equimolar ( 1 . 2 μM ) amounts of Ape1 , LtgA degrades the PG more efficiently ( Figure 5b ) . This result is consistent with previous studies that suggest O-acetylation blocks the function of LTs and lysozyme ( Weadge et al . , 2005; Weadge and Clarke , 2006; Pfeffer and Clarke , 2012; Veyrier et al . , 2013 ) . Surprisingly , in the absence of a common substrate and utilizing equimolar amounts ( 12 nM ) of LtgA and Ape1 , LtgA enhances the activity of Ape1 ( Figure 5c ) . The reaction remained well within the linear range for 60 min when both enzymes were present , which was in stark contrast to Ape1 , that showed less activity over the time course of 60 min . These data demonstrate that the enzymatic activities of LtgA and Ape1 are enhanced reciprocally when functioning together in a complex ( Figure 5c ) . It also appears that LtgA stabilizes and enhance the activity of Ape1 . Synergistic interaction between Ape1 and LtgA could reflect their coordinated function in PG structural regulation in vivo . Interestingly , LtgA , along with other enzymes such as PBP1a , and LtgE are co-conserved in all the proteobacteria that were surveyed ( Figure 5—figure supplement 1 ) . Meanwhile , Ape1 is exclusively co-conserved in Neisseria , Kingella , Snodgrassella , Morococcus , Azovibrio , and one isolate of Burkholderia ubortensis , suggesting Ape1 in contrast to LtgA and others was potentially acquired later by lateral gene transfer ( Figure 5—figure supplement 1 ) . The activity of Ape one and LtgA appears to be synergistic . Since a ΔltgA gave no noticeable phenotype , Neisseria meningitidis strains harboring a catalytically defective mutation of LtgA ( E481A ) , or a Δape1 strain of N . meningitidis were examined for morphological aberrations . The ltgA ( E481A ) strains showed no morphological abnormalities when compared to their parental strains ( Figure 3—figure supplement 2 ) . However , while the Δape1 strain showed no significant defects in cell division or cell separation , cell shape abnormalities and lysed bacteria were clearly evident ( Figure 3—figure supplement 2 ) . Additionally , in our previous study , we noted that diploid cells of the Δape1 strain were larger compared to the wild-type strain ( Veyrier et al . , 2013 ) . Altogether these data suggest that Ape 1 is an important cell shape determinant . In N . meningitidis and N . gonorrhoeae , the activity of LtgA and other LTs leads to a release of cytotoxic PG fragments , which are detected by the host and induce a Nod1-dependent inflammatory response ( Cloud and Dillard , 2002; Cloud and Dillard , 2004; Girardin et al . , 2003; Schaub et al . , 2016 ) . Since the alpha helix 30-deleted strain of LtgA compromised the PG composition the functional role of the alpha helix 30 was explored in vivo in a mouse infection model . For this purpose , we used transgenic mice expressing human transferrin as an experimental model that allows meningococcal growth by providing a human iron source during infection . The four N . meningitidis strains ( wild-type , ΔltgAltgA , ΔltgAltgAΔ30 and ΔltgA ) were used to infect the mice by intraperitoneal injection . Two hours after infection , the four strains induced similar levels of bacteremia ( Figure 6a ) , suggesting that the strains were not defective in their ability to reach the bloodstream . The ΔltgA strain appears to be cleared more slowly . However , the ΔltgAltgAΔ30 strain was cleared from the blood at a significantly faster rate than the other strains , exhibiting a 2-log difference in colony-forming units ( CFUs ) at the 6 hr time point compared to the wild-type strain ( Figure 6a ) . These results were also consistent with those at the cytokine production level , as the ΔltgAltgAΔ30 strain exhibited significantly decreased levels of IL-6 and KC ( the functional murine homolog of human IL-8 ) 6 hr after infection , while all the isolates exhibited similar levels 2 hr after infection ( Figure 6b ) . Overall , the ΔltgAltgAΔ30 strain displays impaired fitness in the host , suggesting LtgA alpha helix 30 plays a key role in bacterial virulence .
We devised a multidisciplinary approach using structural biology to show that it is possible to target a ‘hot spot’ on an LT in order to affect bacterial growth , cell division , and cell membrane integrity ( Figure 7a-c ) , which resulted in lethal consequences for the bacteria during host infection . Additionally , as we discovered with Ape1 , LTs can regulate the function and activity of their protein binding partners , revealing an additional role for LTs in the bacterium . This study shows the ripple effects of disrupting LtgA PG binding capabilities and sets the stage for future development of a class of antibiotics that may act by a dual action in vivo . A small molecule binding to alpha helix 30 could interfere with growth and simultaneously promote bacterial clearance , mimicking the enhanced clearance of the ltgAltgAΔ30 mutant in a murine infection model .
All constructs were created using standard molecular biological techniques . All constructs used for protein expression and purification in this study were GST fusions expressed from pGEX-4T1 ( GE Life Sciences ) . The native proteins without signal peptides were expressed in BL21 ( DE3 ) Gold competent cells ( Novagen ) . The gene encoding the LtgA deletion mutant lacking the alpha helix 503 ( ATAREIAGKIGMD ) 513 was chemically synthesized by ProteoGenix . The synthesized ltgA alpha helix deletion gene was cloned into a GST-fusion pGEX-4T1 ( GE Life Sciences ) plasmid as described above . The expression of all constructs was induced with 0 . 6 mM IPTG at an optical density at 600 nm ( OD600 ) of 0 . 7–0 . 8 and harvested after 4 hr of induction at 18°C . After glutathione-affinity chromatography and thrombin cleavage , proteins were purified to homogeneity by size-exclusion chromatography ( Superdex-200 , GE ) in 50 mM HEPES ( pH 7 . 4 ) , 150 mM NaCl , and 1 mM BME . After gel filtration , the proteins were immediately used for crystallization or flash frozen in liquid nitrogen and stored at −80°C . Crystallization screening was carried out by the sitting-drop vapor-diffusion method with a Mosquito ( TTP Labtech ) automated crystallization system . All crystals were grown at 18°C using the hanging-drop vapor-diffusion method . Crystals of 15–20 mg/ml LtgA were grown at 18°C and appeared within 2–3 days . LtgA was crystallized in a 1:1 ( v/v ) ratio against a well solution of 33% ( w/v ) PEG 6000 and 100 mM HEPES , pH 7 . 5 . Crystals were rectangular in shape and grew to approximately 200–300 μm in length . The data set was collected at the Soleil Synchrotron ( Beamline Proxima-1 ) ( Supplementary file 2 ) . Phasing by molecular replacement was performed using Phenix ( Adams et al . , 2010 ) . Building was performed using Coot ( Emsley and Cowtan , 2004 ) , and restrained refinement was carried out using a combination of Phenix and the ccp4 software suite ( Adams et al . , 2010; Collaborative Computational Project , Number 4 , 1994 ) . MolProbity was used during building and refinement for iterative structure improvements ( Davis et al . , 2004 ) . All structural figures were generated with PyMOL ( PyMOL Molecular Graphics System , version 1 . 5 , Schrödinger , LLC ) . The crystallographic parameters , data statistics , and refinement statistics are shown in Supplementary file 2 . Modeling of unknown LTs were accomplished using Phyre 2 ( Kelley et al . , 2015 ) . Videos of the LtgA enzymatic steps were generated in PyMOL and then assembled in photoshop and edited in iMovie . To explore the interactions of LtgA and its PG binding partners , proteins were mixed at equimolar concentrations of 10 µM , incubated on ice for 1 hr , and subjected to gel filtration chromatography on an SD200 10/300 column . Approximately 150–300 µl of each sample was applied to the column in 50 mM HEPES ( pH 7 . 5 ) and 150 mM salt . Peak fractions were then subjected to SDS-PAGE ( 5–15% ) analysis . The peptidoglycan isolated from all four strains ( wild-type , ΔltgA , ΔltgAltgA and ΔltgAltgAΔ30 ) was incubated for 16 hr in the presence of 10 µg of mutanolysin in 12 . 5 mM sodium phosphate buffer ( pH 5 . 6 ) at 37°C ( total reaction volume 150 µl ) . The reaction was stopped by boiling the samples for 3 min , and the supernatant containing the soluble muropeptides was collected after centrifugation at 16 , 000 ×g for 10 min . The supernatant was analyzed by reversed-phase HPLC using a Hypersil GOLD aQ column ( 5 μm particle size , 150 × 4 . 6 mm , flow rate 0 . 5 ml at 52°C , Thermo Fisher Scientific ) with a mobile phase of H2O-0 . 05% trifluoroacetic acid and a 25% acetonitrile gradient over 130 min . Muropeptides of interest were collected and identified by mass spectrometry as previously described ( Williams et al . , 2017; Williams et al . , 2018 ) . Clone 12 is a derivative of strain 8013 , a serogroup C N . meningitidis strain ( Nassif et al . , 1993 ) , and MC58 is a serogroup B strain ( Tettelin et al . , 2000 ) . Bacteria were grown on GCB medium ( Difco ) containing Kellogg's supplements ( Kellogg et al . , 1963 ) . The E . coli strain DH5 ( Hanahan , 1983 ) was used for plasmid preparation and subcloning . Kanamycin , ampicillin and erythromycin were used in E . coli at final concentrations of 50 , 100 and 300 µg/ml , respectively . In N . meningitidis , kanamycin , ampicillin and erythromycin were used at final concentrations of 2 , 20 and 100µg/ml , respectively . The ltgA gene ( 1851 nucleotides according to the genome sequence of the meningococcal strain MC58 ) was chemically synthesized with a deletion of 42 bp ( 14 codons ) between positions 1506 ( codon 502 ) and 1548 ( codon 516 ) ( starting at ATG ) and was cloned into the vector pUC57 to generate the recombinant plasmid pUC57ltgA ( ProteoGenix , Schiltigheim , France ) . The ltgA fragment was amplified using the primer pair NMF1/NMR1 from the plasmid pUC57ltgA and from the strain MC58 . The two fragments were blunt-ended using the Klenow DNA polymerase fragment ( BioLabs ) and subcloned into the BamHI site of the recombinant plasmid pTE-KM ( Taha et al . , 1998 ) . This plasmid contains the pilE gene of clone 12 with the Km cassette , encoding resistance to kanamycin , located immediately downstream of the pilE without modification of pilE expression . Moreover , a unique BamHI site located between the Km cassette and the downstream region at the 3’ end of the pilE gene ( Nassif et al . , 1993; Taha et al . , 1998 ) was used to subclone the two blunt-ended fragments from the plasmid pUC57ltgA and from the strain MC58 to yield the recombinant plasmids pD-ΔltgAltgA and pD-ΔltgAltgAΔ30 , respectively . An internal deletion in the ltgA gene was also constructed by removing the segment between the restriction sites BsmI ( position 21 ) and BalI ( position 1724 ) on pUC57ltgA , and this region was replaced with the ermAM cassette , encoding erythromycin resistance; the construct was checked using the primer pair ERAM1/ERAM3 ( 5’-gcaaacttaagagtgtgttgatag-3’ and 5’-aagcttgccgtctgaatgggacctctttagcttcttgg-3’ , respectively ) ( Taha et al . , 1998 ) . The corresponding recombinant plasmid pDG15-09 was linearized at the EcoRI site of the pUC57 vector and used to transform the clone 12 strain of N . meningitidis . Transformants were selected on standard GCB medium in the presence of 2 µg/ml erythromycin . Integration by homologous recombination into the ltgA gene on the meningococcal chromosome was further confirmed by PCR analysis using the oligonucleotides ERAM1/ERMA3 and NMF1/NMR1 . One transformant was selected for further analysis and named pD-ΔltgA . The two recombinant plasmids pD-ΔltgAltgA and ΔltgAltgAΔ30 were linearized using the ScaI restriction enzyme and used to transform the strain pD-ΔltgA . Transformants were selected on standard GCB medium in the presence of 2 µg/ml erythromycin and 100 µg/ml kanamycin . Integration by homologous recombination into the ltgA gene on the meningococcal chromosome was further monitored by PCR analysis using the oligonucleotides pilE1 , NMF1 , NMR1 and NMF1/NMR1 . One transformant from each transformation was selected for further analysis and named ΔltgAltgA or ΔltgAltgAΔ30 . The strain ΔltgAltgA has the ltgA gene deleted from its locus but harbors the ltgA gene downstream of the pilE site . The ΔltgAltgAΔ30 strain also has the ltgA gene deleted from its locus and contains a downstream pilE gene but harbors the ltgA gene with the region encoding the amino acid residues 501–516 deleted . Strains 8013 expressing mutant lytic transglycosylases ( E481A ) were constructed by transformation with plasmid pRS91 ( Schaub et al . , 2016 ) containing the E481 mutation . Potential transformants were screened by PCR amplification of active site region followed by digestion with Hyp188III . Positive transformants that lacked a Hyp188III site at the active site were confirmed by sequencing . The MC58 Δape1 strain was described in our previous study ( Veyrier et al . , 2013 ) . Briefly , the entire pat operon which consists of patA , patB and ape1 was deleted in MC58 and then this knockout mutant was complemented by the introduction of patA and patB genes . Bacterial cultures were centrifuged 5 min at 5000 rpm and re-suspended in PBS containing 1 µg/mL DAPI and 5 µg/ml FM4-64 FX ( N- ( 3-Triethylammoniumpropyl ) −4- ( 6- ( 4 ( Diethylamino ) Phenyl ) Hexatrienyl ) Pyridinium Dibromide ) probe . The cells were incubated for 10 min at room temperature protected from light , centrifuged and the pellets resuspended in 4% PFA for fixation during 5 min . After fixation , the cells were washed with PBS , and a 10 µL drop of the bacterial suspension was applied onto poly-L-Lysine pre-coated cover glasses ( # 1 . 5 ) . Next , samples were mounted using Prolong Diamond and imaged using Leica SP5 confocal microscope , with a 63X ( 1 . 4 NA ) oil-immersion objective using 405 nm and 514 nm laser lines . Fluorescence was recorded sequentially using hybrid ( HyD ) detectors and images processed using Fiji ( Schindelin et al . , 2012 ) . Neisseria meningitidis was prefixed in 2 . 5% Glutaraldehyde diluted in PHEM ( Pipes , Hepes , EGTA and MgSO4 ) buffer at pH 7 . The cells were prefixed for 1 hr at room temperature , followed by two washes in PHEM buffer . The samples were applied onto the cover glass ( 1 . 5 mm ) pre-coated with poly-L-Lysine . This was followed by a low speed centrifugation to ensure that the cells adhere correctly to the cover slip . The bacteria were post-fixed using 2% osmium tetroxide in PHEM buffer for 30 to 60 min followed by washing with water three times . The specimen was dehydrated using increasing ethanol concentrations of 25% to 100% in increments of 25% . The bacteria were critically point dried using carbon dioxide , coated with gold and examined with the JEOL JSM‐6700F scanning electron microscope . The acetyl esterase activity assays were executed as previously described with minor modifications ( Pfeffer et al . , 2013; Hadi et al . , 2011 ) . Briefly , the reaction utilized 2 mM 4-nitrophenyl acetate as the substrate . The reaction was carried out at 37°C in 50 mM sodium phosphate buffer , pH 6 . 5 in the presence of LtgA , using equimolar amounts of LtgA and Ape1 or Ape1 . The final volume of the reaction was 300 μl . The reaction was initiated with the addition of the substrate 4-nitrophenyl acetate dissolved in 5% v/v ethanol . The release of 4-nitrophenyl was monitored over the time course of an hour in 96 well microtiter plate at an absorbance of 405 nm . To assess the activity of LtgA , PG ( 200 µg ) was incubated in the presence of LtgA , or equimolar amounts of LtgA and Ape1 , in 12 . 5 mM sodium phosphate buffer pH 5 . 6 . Neisseria PG was purified as previously described ( Wheeler et al . , 2014 ) . The reaction mix was initiated by the addition of enzymes and incubated at 37°C for 5 min . Control reactions lacking PG or enzyme/inhibitor were also included . The final reaction volume was 200 µL . Reactions were performed in triplicates . The reaction was stopped by incubating the samples in a heat block at 100°C for 5 min . The soluble 1 , 6-anhydro-muropeptides was collected using centrifugation at 16 , 000 g for 10 min at room temperature . The supernatant was collected and analyzed by reversed-phase HPLC using a Shimadzu LC-20 system with a Hypersil GOLD aQ column ( 5 μm particle size , 250 × 4 . 6 mm , flow rate 0 . 5 mL/mL at 52°C; Thermo Fisher Scientific ( Waltham , MA , USA ) . The mobile phase gradient was 50 mM sodium phosphate pH 4 . 3 to 75 mM sodium phosphate pH 4 . 9 with 15% Methanol over 135 min . A previously published model for meningococcal infection in transgenic mice expressing human transferrin was used ( Szatanik et al . , 2011 ) . Four strains were tested: clone 12 ( wild-type ) , ΔltgAltgA , ΔltgAltgAΔ30 and ΔltgA . Five mice per group were infected by intraperitoneal injection with 500 µl of bacterial suspension of each strain at 1 × 107 CFU/ml . Blood samples were obtained by retro-orbital bleeding after 2 , 6 and 24 hr , and bacterial counts were determined by plating serial dilutions on GCB medium . Protein sequences were aligned using MUSCLE alignment algorithm using UPGMA clustering method in MEGAX ( Kumar et al . , 2018 ) . Using aligned sequences , a maximum likelihood tree was constructed using a Neighbor joining construction method and a JTT protein substitution model in CLC Genomics Workbench 8 . 01 . Robustness was estimated using 500 bootstrap replicates ( values not shown in figures ) . Animal work in this study was carried out at the Institut Pasteur in strict accordance with the European Union Directive 2010/63/EU ( and its revision 86/609/EEC ) on the protection of animals used for scientific purposes . The laboratory at the Institut Pasteur has the administrative authorization for animal experimentation ( Permit Number 75–1554 ) and the protocol was approved by the Institut Pasteur Review Board that is part of the Regional Committee of Ethics of Animal Experiments of Paris Region ( Permit Number: 99–174 ) . All the invasive procedures were performed under anesthesia and all possible efforts were made to minimize animal suffering . Blood samples from infected mice were collected and stored at −80°C . Cytokines ( IL-6 and KC ) were quantified by an enzyme-linked immunosorbent assay ( Quantikine; R and D Systems Europe , Abingdon , Oxon , United Kingdom ) . Bacteria were grown overnight in GC broth with Kellogg's supplements at 37°C and 5% CO2 . Fresh medium was inoculated at an OD600 of 0 . 05 , and growth was measured spectrophotometrically at 1 hr intervals over a period of 24 hr at 37°C and 5% CO2 . When indicated , 2 µg/ml chloramphenicol was added when the OD600 reached 0 . 6 , and incubation was continued at 37°C and 5% CO2 . At different incubation time points , aliquots ( 3 ml ) from each culture were sampled , and the bacteria were collected by centrifugation , lysed by boiling in SDS-containing sample buffer , and analyzed for the presence of LtgA by western blotting using anti-LtgA antibodies . The expression of the outer membrane factor H binding protein ( Fhbp ) was used as an internal control . The binding of the different LtgA proteins to PG was carried out by incubating 100 µg of PG and 10 µg of enzymes suspended in 150 µl of Tris buffer pH 7 . 5 ( 10 mM Tris , 10 mM MgCl2 and 50 mM NaCl ) . After 30 min of rocking at room temperature , 50 µl of the sample was set aside for analysis before centrifugation for 10 min at 20 , 000 xg . The supernatant was discarded , and the insoluble fraction was washed three times . The remaining pellet was boiled for 10 min . Five microliters of the input or unbound and bound fractions was loaded on an SDS-PAGE gel and analyzed by western blotting . Coordinates and structural data have been submitted to the Protein Data Bank under the accession code 6H5F . | Bacteria are surrounded by a tough yet flexible wall that protects the cell and serves as an anchor for several of the cell’s structures . This cell wall contains a large mesh-like molecule called peptidoglycan made of many repeated building blocks . When a bacterial cell divides in two , it needs to make more of this material . Making peptidoglycan involves two different sets of enzymes working together: “polymerases” are the enzymes that link the individual building blocks to peptidoglycan , one after the other; while “lytic transglycosylases” are enzymes that modify the peptidoglycan to create space for the addition of new building blocks and for assemblies of proteins that must span the cell wall . Lytic transglycosylases are known to assemble with other proteins and enzymes to form the cell’s peptidoglycan-modifying machinery , but it was not clear exactly what purpose they serve within these “enzyme complexes” . It was also unclear whether these enzymes would be good targets for new antibiotics . To help answer these questions , Williams et al . looked at a lytic transglycoslyase called LtgA . This enzyme is originally from Neisseria meningitidis , a bacterium that can cause meningitis and life-threatening sepsis in humans . Williams et al . discovered that part of the enzyme’s active site – the region of an enzyme where the chemical reaction takes – can switch from an ordered helix to a disordered , flexible loop . Bacteria were then genetically engineered to make a version of the enzyme that lacked this helix . These bacteria had weaker cell walls and were deformed; they were also less able to grow and divide , both in the laboratory and in a mouse model of infection . Further analysis showed that the deletion of the helix from the enzyme resulted in the peptidoglycan being modified much more than normal , which could likely explain their reduced virulence . Williams et al . also found that deleting the helix from LtgA interfered with the activity of a protein that interacts with this enzyme , called Ape1 , which also contributed to the fragility of the cell wall . This shows that lytic transglycosylases assembled into enzyme complexes can alter the activities of other proteins in the complex . Together these findings show that researchers could target one enzyme in a complex in bacteria , and disrupt the activity of other proteins in that complex . This highlights the possibility of considering enzyme complexes as useful targets for new drugs , which is important considering the current problem of antibiotic resistance . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2020 | Defective lytic transglycosylase disrupts cell morphogenesis by hindering cell wall de-O-acetylation in Neisseria meningitidis |
Midbrain dopamine neurons have been proposed to signal reward prediction errors as defined in temporal difference ( TD ) learning algorithms . While these models have been extremely powerful in interpreting dopamine activity , they typically do not use value derived through inference in computing errors . This is important because much real world behavior – and thus many opportunities for error-driven learning – is based on such predictions . Here , we show that error-signaling rat dopamine neurons respond to the inferred , model-based value of cues that have not been paired with reward and do so in the same framework as they track the putative cached value of cues previously paired with reward . This suggests that dopamine neurons access a wider variety of information than contemplated by standard TD models and that , while their firing conforms to predictions of TD models in some cases , they may not be restricted to signaling errors from TD predictions .
To test this , we trained 14 rats with recording electrodes implanted in the ventral tegmental area ( VTA ) in a sensory-preconditioning task . In the first phase , rats learned to associate two pairs of environmental cues ( A->B; C->D ) in the absence of reward . As there was no reward , rats showed no significant responding at the food cup and no differences in responding during the different cues ( ANOVA , F3 , 55 = 0 . 7 , p=0 . 52; Figure 1A ) . In the second phase , rats learned that the second cue of one pair ( B ) predicted reward and the other ( D ) did not; learning was reflected in an increase in responding at the food cup during presentation of B ( ANOVA , main effect of cue: F 1 , 163 = 280 . 1 , p<0 . 001 , main effect of session: F 5 , 163 = 9 . 7 , interaction: F 5 , 163 = 10 . 81 , p<0 . 001; Figure 1B ) . Finally , in the third phase , the rats were presented again with the four cues , first a reminder of cue B and D’s reward contingency followed by an unrewarded probe test of responding to cues A and C . As expected , the rats responded at the food cup significantly more during presentation of A , the cue that predicted B , than during presentation of C , the cue that predicted D ( ANOVA , main effect of cue: F 1 , 167 = 8 . 7 , p<0 . 001 , main effect of trial: F 5 , 167 = 6 . 08 , p<0 . 001 , interaction: F 5 , 167 = 2 . 07 , p=0 . 07; Figure 1C ) . 10 . 7554/eLife . 13665 . 003Figure 1 . Rats infer the value of cues during sensory preconditioning . Panels illustrate the task design and show the percentage of time spent in the food cup during presentation of the cues during each of the three phases of training . In the 'preconditioning' phase ( A ) rats learn to associate auditory cues in the absence of reinforcement; during this phase there is minimal food cup responding ( ANOVA , F ( 3 , 55 ) = 0 . 7 , p = 0 . 52 ) . In subsequent 'conditioning' ( B ) , rats learn to associate one of the cues ( B ) with reward; conditioned responding at the food cup during B increases across sessions ( ANOVA , main effect of cue: F ( 1 , 163 ) = 280 . 1 , p<0 . 001 , main effect of session: F ( 5 , 163 ) = 9 . 7 , interaction: F ( 5 , 163 ) = 10 . 81 , p<0 . 001 ) . In a final 'probe' test ( C ) , rats are presented with each of the 4 auditory cues; conditioned responding at the food cup is maintained to B and is also now evident during presentation of A , the cue that had been paired with B in the preconditioning phase ( ANOVA , main effect of cue: F ( 1 , 167 ) = 8 . 7 , p<0 . 001 , main effect of trial: F ( 5 , 167 ) = 6 . 08 , p<0 . 001 , interaction: F ( 5 , 167 ) = 2 . 07 , p=0 . 07 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 003 Single unit activity was recorded in the VTA throughout training . To identify putative dopamine neurons , we used a recently developed , optogenetically-validated strategy that classifies VTA neurons on the basis of their response dynamics during Pavlovian conditioning . In published work ( Cohen et al . , 2012; Eshel et al . , 2015 ) , this strategy identified VTA dopamine neurons ( i . e . neurons expressing Cre under the control of the promoter for the dopamine transporter ) with near perfect fidelity . Here we applied this same analysis to the mean normalized responses of all VTA neurons recorded during conditioning and reminder sessions ( n = 632; Figure 2A ) . We extracted the major modes of variation among the neurons with principal components analysis ( PCA; Figure 2B ) and then performed hierarchical clustering on those PCs ( Figure 2C ) . This analysis successfully extracted the 3 previously described VTA response types from our data ( Figure 2D ) : neurons with sustained excitation to cues and reward ( putative GABAergic ) , neurons with phasic excitation to cue onset and reward onset ( putative dopaminergic ) , and neurons with sustained inhibition to cue and reward ( unknown ) . We then assessed the responses of the putative dopamine neurons ( n = 304 ) to the cue and reward over the course of conditioning . Consistent with their classification , we found changes in firing during conditioning that were in accord with signaling of reward prediction errors . Specifically , early in conditioning , these neurons’ maximal response occurred just after reward delivery ( Figure 2E , top trace ) , whereas late in conditioning , the maximal response occurred just after onset of the cue predicting that reward ( Figure 2E , lower traces ) . As a result , the difference in activity at the time of cue onset versus reward increased significantly from the start to the end of conditioning ( r 302 = 0 . 24 , p<0 . 01; Figure 2F ) , consistent with signaling of TD prediction errors ( Glimcher , 2011; Schultz , 2002 ) . 10 . 7554/eLife . 13665 . 004Figure 2 . VTA dopamine neurons exhibit firing to a reward-paired cue that is consistent with TD error signaling . We recorded 632 neurons across all days of conditioning and the final reminder session . ( A ) Normalized responses ( AUC ) are displayed for each neuron , sorted by the classification algorithm applied by Cohen , Uchida and colleagues ( Cohen et al . , 2012 ) . The first three principal components ( PCs ) were extracted , to find the major modes of this population’s response ( B ) , then hierarchical agglomerative clustering was used on those PCs to identify similar neural responses; groups identified are highlighted in color ( C ) ; The mean group response of each of the populations identified are displayed ( D ) ; in accordance with previous results ( Cohen et al . , 2012 ) we found populations undergoing sustained excitation , phasic excitation , and sustained inhibition . Consistent with identification as putative dopamine neurons , the average ( AUC ) response to cue B from the phasic group on each day of conditioning exhibited a peak response that was highest to reward early in conditioning and migrated to earlier cue onset across conditioning ( E–F , r ( 302 ) = 0 . 24 , p<0 . 01 ) . This change in firing is in accordance with signaling of a TD error . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 004 Having established that putative dopamine neurons identified in this manner exhibit firing during conditioning consistent with signaling of TD errors , we next examined activity in neurons recorded in just the probe test . We again identified these neurons by their pattern of firing to the reward predictive cue ( n = 102; Figure 3A–D ) . As before , this analysis identified a group of cells with strong phasic responses to B , the cue that had been directly paired with reward ( n = 52 ) . While this response generalized somewhat to D , the control cue that had been presented without reward during conditioning sessions , these neurons fired significantly more during the first second of B than to D ( t 51 = 4 . 40 , p<0 . 001 , black versus gray lines , respectively , with shading for SEM , Figure 3E ) . 10 . 7554/eLife . 13665 . 005Figure 3 . VTA dopamine neurons exhibit firing to a pre-conditioned cue that is not consistent with TD error signaling . We recorded 102 neurons during the probe test . AUC normalized neural responses were classified with a hierarchical clustering as in Figure 2 ( A–D ) in order to identify putative dopamine neurons ( n = 52 ) . In addition , we also identified 4 neurons based on traditional waveform criteria . While the classified putative dopamine neurons showed firing to all cues , they exhibited the largest responses at the onset of B , the reward-paired cue ( significantly above responding to D , t ( 51 ) = 4 . 40 , p<0 . 001 ) , and to A , the cue that had been paired with B in the preconditioning phase ( significantly above responding to control cue C , t ( 51 ) = 5 . 02 , p<0 . 001 ) ( E–F ) . Further , the activity elicited by these two cues was strongly correlated ( F ) , suggesting that dopamine neurons code errors elicited by these two types of cues in a common framework ( correlation between B–D and A–C , r ( 50 ) = 0 . 63 , p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 00510 . 7554/eLife . 13665 . 006Figure 3—figure supplement 1 . Neural responses from phasic and tonic wide-waveform neurons . ( A ) raster plot of 18 trials of cue responses , resorted according to cue , for phasic responding wide waveform neuron . ( B ) baseline subtracted mean responses of panel ( A ) for cues B and D . ( C ) baseline subtracted mean responses to of panel ( A ) for cues A and C ( D ) raster plot of 18 trials of cue responses , resorted according to cue , for tonic excited wide waveform neuron . ( E ) baseline subtracted mean responses of panel ( D ) for cues B and D . ( F ) baseline subtracted mean responses to of panel D for cues A and C . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 00610 . 7554/eLife . 13665 . 007Figure 3—figure supplement 2 . Neural responses from 39 neurons classified as tonically excited by cue B . ( A ) baseline subtracted , mean responses of all neurons to cues B and D , +/- SEM ( B ) baseline subtracted , mean responses of all neurons to cues A and C , +/- SEM ( C ) histogram of differences in neural responding to cached value ( B-D ) for all tonically excited neurons for the first second of cue response; there was no significant difference ( t ( 38 ) = 0 . 37 , p = 0 . 71 ) between responses to cue B and D ( D ) histogram of differences in neural responding to inferred value ( A–C ) for all tonically excited neurons for the first second of cue response; there was a significant difference between early responses to cue A and C ( t ( 38 ) = 2 . 9 , p<0 . 01 ) , ( E ) histogram of differences in neural responding to cached value ( B–D ) for all tonically excited neurons for the final nine seconds of cue response; neurons fired significantly more to cue B than D ( t ( 38 ) = 6 . 3 , p>0 . 001 ) ( F ) histogram of differences in neural responding to inferred value ( A–C ) for all tonically excited neurons for the last nine seconds of cue response; there was a smaller but significant difference between responses to cue A and C ( t ( 38 ) = 2 . 4 , p<0 . 05 ) ( G ) scatter of individual responses to cached vs inferred value ( i . e . data from panel C vs panel D ) ; while there was a positive relationship , the correlation was not significant ( r ( 37 ) =0 . 26 , p=0 . 11 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 00710 . 7554/eLife . 13665 . 008Figure 3—figure supplement 3 . Neural responses from 11 neurons classified as tonically inhibited by cue B . ( A ) baseline subtracted , mean responses of all neurons to cues B and D , +/- SEM ( B ) baseline subtracted , mean responses of all neurons to cues A and C , +/- SEM ( C ) histogram of differences in neural responding to cached value ( B–D ) for all tonically inhibited neurons for the first second of cue response; there was no significant difference ( t ( 10 ) = -1 . 56 , p=0 . 15 ) between responses to cue B and D ( D ) histogram of differences in neural responding to inferred value ( A–C ) for all tonically inhibited neurons for the first second of cue response; there was no significant difference ( t ( 10 ) = 0 . 99 , p = 0 . 34 ) between responses to cue A and C . ( E ) histogram of differences in neural responding to cached value ( B–D ) for all tonically inhibited neurons for the last nine seconds of cue response; there was no significant difference ( t ( 10 ) = -1 . 6 , p = 0 . 14 ) between responses to cue B and D ( F ) histogram of differences in neural responding to inferred value ( A–C ) for all tonically inhibited neurons for the last nine second of cue response; there was no significant difference ( t ( 10 ) = 0 . 03; p = 0 . 98 ) between responses to cue A and C ( G ) scatter of individual responses to cached vs inferred value ( i . e . data from panel C vs panel D ) ; while there was a positive relationship , the correlation was not significant ( r ( 9 ) =0 . 46 p=0 . 15 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13665 . 008 However , in addition to this expected pattern of firing , these cells also had strong phasic responses to both preconditioned cues ( blue and red traces , A and C , respectively , with shading for SEM , Figure 3E ) . While the common element of these responses could reflect novelty or salience , since these cues had not been presented for a number of days , or perhaps generalization from conditioning to B , the actual phasic neural response was significantly stronger for A , the cue that predicted the reward-paired cue , than for C , the preconditioned control cue ( t 51 = 5 . 02 , p<0 . 001 ) . This difference cannot be explained on the basis of novelty , salience , or generalization , since A and C were treated similarly . Nor can it be explained by direct experience with reward , because A was never paired with reward , and it was only paired with B before conditioning . Thus , the phasic response in these putative dopamine neurons appeared to be influenced by inferred value of cue A . Interestingly , the neural response was perhaps somewhat better at discriminating A from C ( Figure 3E , bottom panel ) than B from D ( Figure 3E , top panel ) , perhaps reflecting the differences in training between A and C , which were only presented a few times in unrewarded sessions , versus B and D , which were presented many times across several days of conditioning . Despite this , the influence of the inferred value of cue A on the firing of these neurons during the first second of their cue response was strongly and significantly correlated with the influence of value on these neurons firing at the onset of B , the cue directly paired with reward ( r 50 = 0 . 63 , p<0 . 001; Figure 3F ) . Notably this was also true for a handful of neurons ( n = 4 ) that exhibited the classic wide , polyphasic waveforms traditionally used to identify dopamine neurons ( Figure 3F , filled circles , see Figure 3—figure supplement 1 for PSTH’s ) . This relationship in the initial phasic response to the cues did not reach significance in the other two neural subtypes identified by the clustering analysis ( see Figure 3—figure supplements 2 and 3 for analyses of tonically-modulated neurons ) . Beyond their phasic responses at the start of the cues and to reward , the putative dopamine neurons also exhibited another notable feature: the average response of these neurons throughout cues A and C was above baseline , and this sustained firing was significantly higher to cue A than C ( Figure 3E , final 9s of cues , t51 = 2 . 56 , p<0 . 05 ) . This elevated firing may be a sign of dopamine’s reported ability to anticipate proximity to reward or to signal state value ( Howe et al . , 2013; Hamid et al . , 2016 ) , if our rats’ expectation of reward delivery is based on knowing that progression to the offset of cue A should lead to the subsequent presentation of cue B and then reward . Importantly , in our design , reward is presented during B rather than at its termination . This would explain why this pattern of sustained firing is not present throughout cue B ( Figure 3E , t51= 1 . 09 , p=0 . 278 ) . Interestingly this pattern of sustained and differential firing to A ( vs C ) and not to B ( vs D ) in the putative dopamine neurons is the mirror image of firing in neurons classified as tonically excited , which showed relatively modest changes in sustained firing to A and C and much larger increases in firing to B ( see Figure 3—figure supplement 2 ) . This relationship would be consistent with recent proposals that these neurons , thought to be GABAergic ( Cohen et al . , 2012 ) , exert tonic inhibition to suppress the firing of dopamine neurons ( Eshel et al . , 2015 ) . Here we report that VTA dopamine neurons , identified based either on traditional waveform criteria or through an optogenetically-validated clustering analysis of their response properties to a conditioned cue , exhibited phasic cue-evoked responses that were influenced by inferred value . These responses were observed even though the critical cue had no prior history of direct pairing with any rewarding event . In addition , they were greater than responses to a control cue that was treated similarly and thus had similar levels of salience or novelty or generalized value , all variables that have been proposed to explain phasic activity in other settings that appeared to be at odds with the standard explanation of phasic dopamine activity ( Kakade and Dayan , 2002; Bromberg-Martin et al . , 2010; Matsumoto and Hikosaka , 2009 ) . These data show that the phasic activity of dopamine neurons can reflect information about value that is not contemplated by TD models , at least as they have been applied to understand the phasic firing of these neurons ( Glimcher , 2011; Schultz , 2002; Niv and Schoenbaum , 2008; Clark et al . , 2012 ) . Our finding is consistent with a number of recent reports , suggesting that dopamine neurons are likely to access more complex information than is available to standard TD models . For example , dopamine neurons in the rat VTA utilize input from the orbitofrontal cortex to disambiguate states that are not easily distinguished via external information in order to more accurately calculate prediction errors ( Takahashi et al . , 2011 ) . While this result does not require the use of inference in calculating errors , merely access to state information , it suggests that dopamine neurons have access to a major source of this information , given the central role of the orbitofrontal cortex in inference-based behavior ( Stalnaker et al . , 2015 ) . The phasic activity of dopamine neurons has also been shown to track the value of one cue after changes in the value of an associated cue ( Bromberg-Martin et al . , 2010b ) . Again these data suggest that dopamine neurons have access to higher order information , which could be described as inference . Indeed these authors describe their results in terms of inference; however , as they note in their discussion ( Bromberg-Martin et al . , 2010b , final paragraph of discussion ) , the inference seen in their task may differ from that shown here in that it does not require access to model-based information , but could instead be based on direct , 'cached' value from earlier training sessions . Finally elevated dopamine has also been found using microdialysis during an aversive version of the sensory preconditioning task used here ( Young et al . , 1998 ) . However the use of an aversive paradigm , a measurement technique with low temporal resolution , and the lack of control conditions to confirm signaling of cached value errors make it difficult to apply these results to address the very specific proposal that phasic changes in the firing of dopamine neurons signal TD prediction errors in appetitive paradigms . Our study addresses the limitations of these best available prior reports . We are recording phasic activity of dopamine neurons at their source . We have identified dopamine neurons by two different classification schemes , an old one that has been used repeatedly across labs and species to identify error-signaling dopamine neurons ( Mirenowicz and Schultz , 1994; Waelti et al . , 2001; Tobler et al . , 2003; Pan et al . , 2005; Roesch et al . , 2007; Jo et al . , 2013; Morris et al . , 2006; Jin and Costa , 2010 ) , as well as a new , optogenetically-validated approach that has identified error signals in mice ( Cohen et al . , 2012 ) and is favored by those that dislike the use of waveform criteria ( Margolis et al . , 2006 ) . We used a carefully designed behavioral preparation in which our critical cue of interest has no prior history of association with reward , thus unique firing to this cue cannot be explained as any sort of cached value ( see ‘A comment on the basis of responding to the preconditioned cue’ in the behavioral Methods for further information ) . Further this appetitive task includes two important control cues: one designed to rule out explanations based on generalization and salience ( cue C ) and another designed to reveal cached value prediction errors ( cue B ) . The inclusion of this cue in particular is important because it allows us to assess the relationship between traditional error signals and any influence of inferred value on the firing of the dopamine neurons . The close relationship in the firing of the dopamine neurons to B , the cue directly paired with reward , and A , the cue that predicts reward only through B , suggests that whatever is ultimately signaled when a cue with inferred , model based value is presented may be similar to what the same neurons signal in response to the unexpected appearance of a cue that has been directly paired with reward . While this might be explained as error signaling to B , calculated from TD models , and error signaling to A , calculated from something beyond TD models , this solution is cumbersome , particularly given that inferred and experienced value are actually confounded for a cue directly paired with reward ( a fact illustrated by the normal efficacy of reinforcer devaluation at changing conditioned responding [Holland and Straub , 1979] ) . A more parsimonious explanation is that dopamine neurons , unlike standard TD models , have access to a wide variety of information when computing expected value . And that while their firing may conform to what is expected for errors calculated from TD models in some special cases , they may not be signaling TD derived errors . Such a suggestion aligns nicely with recent proposals that dopamine neurons signal errors based on changes in economic utility ( Schultz et al . , 2015 ) , and it would be consistent with data presented in abstract form suggesting that cue-evoked dopamine release in nucleus accumbens is sensitive to devaluation of a paired reward ( Martinez et al . , 2008 ) , though it contradicts data only just published from a similar study in which cue-evoked release was not immediately altered when reward value was manipulated via salt depletion ( Cone et al . , 2016 ) . This variability in correspondence between our unit data and evidence from studies of dopamine efflux in accumbens may reflect to the different dynamics of the two processes or it may indicate some specificity with regard to the information content of the dopaminergic afferents in accumbens versus other areas . Finally it is important to explicitly note that the general proposal that phasic changes in dopamine are a TD error signal incorporates two very separate sets of predictions . One set , most relevant to the single unit correlates that form the basis of this hypothesis , concerns the information used to construct the error signals . This is obviously the part of the question we have addressed in the current study . That is , do dopaminergic errors reflect only model-free information derived from TD systems or do they also incorporate the predictions of non-TD , model-based systems ? We believe our data favor the latter position . The second set of predictions , not addressed by our study , concerns what the dopaminergic errors do downstream . Do they act only to stamp in the so-called cached values that are acquired through learning in TD models or do they act more broadly to facilitate increases in the strength of associative representations in a way that is orthogonal to distinctions between the systems , model-free or model based , in which those representations reside ? The latter role would be more in accord with earlier learning theory accounts that viewed prediction errors as acting on the strength of associative representations ( Glimcher , 2011; Bush and Mosteller , 1951; Rescorla et al . , 1972 ) . Importantly the answer to the second question is formally separate from the answer to the first . In other words , phasic changes in dopamine may reflect model-based information and yet only act to support model-free , cached-value learning . Or phasic changes in dopamine could act more broadly , supporting both model-free and model-based learning , even if they only reflected value predictions from the former system . Notably the jury remains out on what sort of learning the brief phasic changes in dopamine thought to signal prediction errors serve to support . In support of dopamine’s role in supporting model-based learning , prediction errors observed in ventral striatal target regions seem to reflect both model based and model free information ( Daw et al . , 2011 ) . Further studies have shown that elevated dopamine levels , either observed ( Deserno et al . , 2015 ) or directly manipulated ( Wunderlich et al . , 2012; Sharp et al . , 2016 ) , bias subjects towards making model-based decisions , as do changes to dopaminergic gene expression ( Doll et al . , 2016 ) . While suggestive , these studies do not directly distinguish the effects of phasic changes in dopamine neuron firing or release from the effects of slower tonic changes . Such tonic changes may play a very different role from the phasic error signals observed in single unit activity ( Hamid et al . , 2016; Niv et al . , 2007 ) . Further these studies do not isolate the effects of errors themselves , independent from other confounding variables . Such isolation and specificity can be achieved using Cre-driver lines in rodent species , and as noted earlier , there is now strong evidence from such studies that brief , phasic changes in dopamine neuron firing can act like a prediction error , at least in some downstream targets and behavioral paradigms ( Steinberg et al . , 2013; Chang et al . , 2016 ) . However the behavior supported by the artificially induced prediction errors in these experiments may be either model-based or model-free ( or a mixture ) . A more definitive answer to this question will require these approaches to be married to paradigms that distinguish these two types of learning .
14 adult Long-Evans rats ( 10 male , 4 female weighing 275–325 g on arrival ) were individually housed and given ad libitum access to food and water , except during behavioral training and testing , which which they received 15 min of ad-lib water access following each training session . Rats were maintained on a 12-hr light/dark cycle and trained and tested during the light cycle . Experiments were performed at the National Institute on Drug Abuse Intramural Research Program , in accordance with NIH guidelines . The number of subjects was chosen to have sufficient power to assess learning on the final test-day ( Jones et al . , 2012 ) , and to gather a sufficient number of isolated neurons ( >100 ) for subsequent analysis on the final test day . Behavioral training and testing were conducted in standard behavior boxes with commercially-available equipment ( Coulbourn Instruments , Allentown , PA ) . A recessed dipper was placed in the center of the right wall approximately 2 cm above the floor . The dipper was mounted outside the behavior chamber and delivered 40 ul of flavored milk ( Nestle ) per dipper elevation . Auditory cues ( tone , siren , 2 Hz clicker , white noise ) calibrated to ~65 dB were used during the behavioral testing . Rats underwent surgery for implantation of chronic recording electrode arrays . Rats were anesthetized with isoflurane and placed in a standard stereotaxic device . The scalp was excised , and holes were bored in the skull for the insertion of ground screws and electrodes . Multi-electrode bundles [16 nichrome microwires attached to a microdrive] were inserted 0 . 5 mm above dorsal VTA [anteroposterior ( AP ) 5 . 4 mm and mediolateral ( ML ) 0 . 8 mm relative to bregma ( Paxinos and Watson , 1998 ) ; and dorsoventral ( DV ) 7 . 0 mm from dura] . In 3 rats , microwire electrodes were also implanted 0 . 5 mm above ipsilateral orbitofrontal cortex [AP 3 . 2 mm and ML 3 . 0 mm relative to bregma ( Paxinos and Watson , 1998 ) ; and DV 4 . 0 mm from the dura] , and in 2 other rats , microwire electrodes were also implanted 0 . 5 mm above ipsilateral ventral striatum [AP 1 . 0 mm and ML 3 . 0 mm relative to bregma ( Paxinos and Watson , 1998 ) ; and DV 6 . 0 mm from the dura] . Once in place , the assemblies were cemented to the skull using dental acrylic . Six rats also received infusions of 1 . 0 ul of AAV5-DIO-HMD4 into central VTA [anteroposterior ( AP ) 5 . 4 mm and mediolateral ( ML ) 0 . 8 mm relative to bregma , and 8 . 1 mm below dura]; there were no effects of this treatment on any of the results we have reported . Rats began sensory preconditioning 2 weeks after electrode implantation . The sensory preconditioning procedure consisted of three phases , of similar design to a prior study ( Jones et al . , 2012 ) . Neural signals were collected from the VTA during each behavioral session . Differential recordings were fed into a parallel processor capable of digitizing 16-to-32 signals at 40 kHz simultaneously ( Plexon ) . Discriminable action potentials of <3:1 signal/noise ratio were isolated on-line from each signal using an amplitude criterion in cooperation with a template algorithm . Discriminations were checked continuously throughout each session . Time-stamped records of stimulus onset and neuronal spikes were saved digitally , as were all sampled spike waveforms and the discrimination file . Off-line re-analysis incorporating 3D cluster-cutting techniques confirmed and corrected on-line discriminations . Except where explicitly noted , all neurons identified via off-line sorting were included in each analysis . Raw data were processed with Matlab to extract food cup entries and spike-timing relative to cue-onset . Entries were converted to a response measure: the percentage of time rats spent with their head in the food cup during cue presentation as measured by an infrared photo beam positioned at the front of the food cup . Spike times were binned and analyzed as specified below . In comparing cue-evoked to reward-evoked activity , bins spanning the first 500ms of each period were analyzed . In comparing response differences evoked by different cues , bins spanning the first 1 s of cue-evoked activity were analyzed . For all statistical tests , an alpha level of 0 . 05 was used . As per prior reports ( Cohen et al . , 2012 ) , we normalized the firing rate of individual neurons by comparing the histogram of spike counts during each bin of spiking activity ( 100 ms , test bins from each trial for a cue , at a particular time post-stimulus ) against a histogram of baseline ( 100 ms ) bins , from all trials for that cue . The ROC in question is calculated by normalizing all test and baseline bin counts , such that the minimum bin count was 0 and the maximal bin count was 1 sliding a discrimination threshold across each histogram of bins , from 0 to 1 in 0 . 01 steps , such that fraction of test bins identified above the threshold was a ‘true positive’ rate and the fraction of baseline bins above the threshold was a ‘false negative’ rate for an ROC curve . The area under this curve was then estimated by trapezoidal numerical estimation , with an auROC below 0 . 5 being indicative of inhibition , and an auROC above 0 . 5 being indicative of excitation above baseline . In order to isolate VTA neuron response types shown to be indicative of putative dopaminergic and GABAergic genetic identities ( Cohen et al . , 2012 ) , we took the auROC normalized responses of neurons during their response to the cue predictive of reward ( cue B ) , and performed a simple classification to separate neural responses . We first performed principal components analysis on a matrix of neural responses during cue B and reward presentation ( neuron-by-time ) to simplify the neural dynamics to the 3 most descriptive ways in which neurons differed . We then classified this description of the neural population ( first 3 principal components ) with a simple unsupervised hierarchical clustering algorithm , finding the similarity ( Euclidean distance ) between all pairs of neurons in principal components space , and iteratively grouping the neurons them into larger and larger clusters on the basis of their similarity ( i . e . agglomerative complete-linkage clustering ) . A distance-criterion was then set to extract exactly 3 clusters from this hierarchical tree . Neurons were screened for wide waveform and amplitude characteristics , calculated on their mean action-potential across a recording session . Neurons were identified as dopaminergic if their negative half-width exceeded a standard criterion ( 450 µs ) and the ratio of ( max ) positive to ( min ) negative voltage deflections was greater than zero ( Mirenowicz and Schultz , 1994; Takahashi et al . , 2011; Roesch et al . , 2007; Jo et al . , 2013 ) . Four such neurons were identified as wide waveform across the probe sessions . After the final recording session , rats were euthanized and perfused first with PBS and then 4% formalin in PBS . Electrolytic lesions ( 1 mA for 10 s ) made just before perfusion were examined in fixed , 0 . 05 mm coronal slices stained with cresyl violet . Anatomical localization for each recording was verified on the basis of histology , stereotaxic coordinates of initial positioning , and recording notes . | Learning is driven by discrepancies between what we think is going to happen and what actually happens . These discrepancies , or ‘prediction errors’ , trigger changes in the brain that support learning . These errors are signaled by neurons in the midbrain – called dopamine neurons – that fire rapidly in response to unexpectedly good events , and thereby instruct other parts of the brain to learn about the factors that occurred before the event . These events can be rewards , such as food , or cues that have predicted rewards in the past . Yet we often anticipate , or infer , rewards even if we have not experienced them directly in a given situation . This inference reflects our ability to mentally simulate likely outcomes or consequences of our actions in new situations based upon , but going beyond , our previous experiences . These inferred predictions of reward can alter error-based learning just like predictions based upon direct experience; but do inferred reward predictions also alter the error signals from dopamine neurons ? Sadacca et al . tested this question by exposing rats to cues while recording the activity of dopamine neurons from the rats’ midbrains . In some cases , the cues directly predicted rewards based on the rats’ previous experience; in other cases , the cues predicted rewards only indirectly and based on inference . Sadacca et al . found that the dopamine neurons fired in similar ways in response to the cues in both of these situations . This result is consistent with the proposal that dopamine neurons use both types of information to calculate errors in predictions . These findings provide a mechanism by which dopamine neurons could support a much broader and more complex range of learning than previously thought . | [
"Abstract",
"Introduction",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2016 | Midbrain dopamine neurons compute inferred and cached value prediction errors in a common framework |
Hotspot mutations of Ras drive cell transformation and tumorigenesis . Less frequent mutations in Ras are poorly characterized for their oncogenic potential . Yet insight into their mechanism of action may point to novel opportunities to target Ras . Here , we show that several cancer-associated mutations in the switch III region moderately increase Ras activity in all isoforms . Mutants are biochemically inconspicuous , while their clustering into nanoscale signaling complexes on the plasma membrane , termed nanocluster , is augmented . Nanoclustering dictates downstream effector recruitment , MAPK-activity , and tumorigenic cell proliferation . Our results describe an unprecedented mechanism of signaling protein activation in cancer .
The small GTPase Ras is dynamically anchored to cellular membranes . Its principal steady-state localization is the plasma membrane , from where most of Ras signaling emerges ( Hancock . , 2003 ) . Ras hyperactivation is a hallmark of cancer and has long been considered to be pharmacologically intractable ( Cox et al . , 2014 ) . However , recent progress encourages the pursuit of specific Ras protein inhibitors ( Spiegel et al . , 2014 ) . Since the seminal publication of Scheffzeck et al . , it is well understood that hot-spot mutations in codons 12 , 13 , and 61 ( Scheffzek et al . , 1997 ) , which account for >99% of Ras mutations ( Prior et al . , 2012 ) , primarily compromise its GAP-dependent hydrolase activity . These oncogenic lesions render Ras transforming and tumorigenic , as they leave Ras constitutively GTP bound and therefore ready to overdrive in particular the Ras/MAPK pathway , which critically controls cell division and other oncogenic processes . A different mutational spectrum is found in RASopathies , developmental syndromes that are characterized by facio-cutaneous malformations , heart developmental and neurodevelopmental defects , as well as a predisposition to certain cancers ( Schubbert et al . , 2007 ) . These syndromes are also characterized by a hyperactivation of the Ras/MAPK-signaling pathway . For example , in Noonan syndrome , NRAS or KRAS can be mutated at various positions along their coding sequences in the germline . The exact molecular and cellular mechanisms that lead to the observed phenotypes are still largely unclear ( Prior et al . , 2012 ) . For non hot-spot mutations in Ras that coincide with the known nucleotide binding regions , the G1–G5 boxes , mechanistic explanations for aberrant activities have been demonstrated or proposed ( Schubbert et al . , 2007; Gremer et al . , 2011; Prior et al . , 2012; Cirstea et al . , 2013 ) . Whether and how additional mutations across the remainder of the coding sequence of Ras affect its pathogenic activity is largely unknown . Ras activity emerges in the plasma membrane , where 20–50% of Ras proteins are organized into isoform-specific , dynamic proteo-lipid complexes that contain 6–8 Ras proteins , termed nanocluster ( Abankwa et al . , 2007 ) . The tight packing of this signaling protein increases its concentration locally and thus enables more efficient effector recruitment ( Rotblat et al . , 2010; Guzmán et al . , 2014b ) . It was proposed that nanoclustering is a basic systems-level design principle for the generation of high-fidelity signal transduction ( Tian et al . , 2007 ) . Essentially only three regulators ( galectin-1 [Gal-1] , galectin-3 , and nucleophosmin ) of Ras nanoclustering , so called nanocluster scaffolds , are known . The lectin Gal-1 is the best characterized nanocluster scaffold , which increases H-ras-GTP nanoclustering and effector recruitment , effectively by stabilizing immobile H-ras-GTP nanocluster ( Rotblat et al . , 2010 ) . We previously revealed another aspect of Ras membrane organization , showing that a novel switch III in Ras is somehow coupled to the reorientation of H-ras on the membrane ( Figure 1—figure supplement 1 ) . Mutations in the switch III and the structural elements of H-ras that stabilize its reorientation ( helix α4 and the C-terminal hypervariable region [hvr] ) systematically modulate Ras signaling ( Gorfe et al . , 2007; Abankwa et al . , 2008b , 2010 ) . More recently , we addressed the mechanistic basis of this activity modulation for computational modeling-derived mutations on helix α4 and the hvr: these alter engagement of the nanocluster modulator Gal-1 and thus H-ras nanoclustering . As a consequence of this up-concentration , effector recruitment and subsequent downstream signaling are increased ( Guzmán et al . , 2014b ) . Here , we report that cancer-associated mutations in the switch III region of the three major Ras oncoproteins , H- , N- , and K-ras , increase Ras activity by a novel disease mechanism , namely signaling protein nanocluster augmentation . We find that these mutations do not alter basic biochemical functions of Ras in solution . Instead , a strict correlation between increased recruitment of the effector to Ras and augmented nanoclustering of Ras on cellular membranes is found . Upregulated effector engagement is directly reflected in the elevated cellular Ras activity , and significantly impacts on the tumorigenic potential . Our results reveal a new mechanism of mutational signaling pathway hyperactivation in a pathophysiological setting and suggest Ras nanoclusters as direct drug targets .
H-ras exists in a nucleotide-dependent conformational equilibrium on the membrane ( Gorfe et al . , 2007; Abankwa et al . , 2008b ) . The two delimiting conformers are stabilized by either helix α4 or the hvr ( Figure 1—figure supplement 1 ) . Conformer reorientation on the membrane was associated with a novel switch III region , which is formed by the β2-β3-loop and helix α5 . However , formal proof for their mechanistic connection is still missing . We previously found that mutations in the hvr or on helix α4 ( left and right tables on top in Figure 1A ) alter the activity of GTP-H-ras , probably by stabilizing preferred conformers ( red and blue GTP-H-ras conformers , respectively , in Figure 1A ) similar to the nucleotide-dependent ones ( Gorfe et al . , 2007; Abankwa et al . , 2008b ) . More recent evidence from these helix α4 and hvr GTP-H-ras mutants suggests that the conformational state couples to nanoscale Ras-signaling hubs in the membrane , termed nanocluster ( middle in Figure 1A ) ( Guzmán et al . , 2014b ) . Nanoclustering then critically determines the recruitment rate of the effector Raf from the cytoplasm to membrane bound Ras and therefore the initiating event of the MAPK-signaling cascade ( bottom in Figure 1A ) . 10 . 7554/eLife . 08905 . 003Figure 1 . Intramolecular switch III-conformer coupling suggests altered nanoclustering in switch III mutants . ( A ) Schematic representation of conformer-nanocluster coupling and its effect on H-ras signaling . For GTP-Ras a conformational equilibrium is assumed , which can be shifted by mutations as indicated by matching colors ( top tables , show mutations by region in Ras ) . Mutations in the table increase ( red ) or decrease/neutralize ( blue ) H-ras activity . For mutations boxed in bold this was previously shown to proceed by increasing or decreasing H-ras nanoclustering , respectively . As a consequence , the effector recruitment and thus downstream signaling are correspondingly modulated . Note , that a direct effect of the Ras conformation on effector engagement is not excluded . The green scenario represents the normal activity , which is predicted for the combinations of activity increasing and decreasing/neutralizing mutations ( green scenario in centre ) , if switch III couples to H-ras reorientation . ( B ) Schematic representation of the RBD-recruitment FRET assay . Ras activity was measured by mRFP-C-Raf-RBD recruitment to mGFP-tagged mutants of Ras in intact cells using FRET . ( C ) RBD-recruitment FRET data of indicated H-ras mutants transiently expressed in BHK cells . Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . The color code of the bars is as in ( A ) and indicates if the mutations increased , decreased/neutralized or normalized RBD-recruitment as compared to H-rasG12V . Statistical analysis of differences vs H-rasG12V was performed as described in ‘Materials and methods’ ( NS , non-significant; *** , p < 0 . 001 ) . ( D ) FRET in ( C ) was measured by fluorescence lifetime microscopy of transfected BHK cells as indicated . Color look up table to the right shows fluorescence lifetimes . See also Figure 1—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 00310 . 7554/eLife . 08905 . 004Figure 1—figure supplement 1 . Nucleotide-dependent H-ras conformers on the membrane . Computational modeling-derived structures of the predominant H-ras-GTP- ( left ) and H-ras-GDP- ( right ) conformers on the membrane . Conformer stabilizing residues R128 , R135 on helix α4 and R169 , K170 in the hypervariable region ( hvr ) , as well as switch III region residues of helix α5 ( R161 ) and β2-β3-loop ( D47 , E49 ) are indicated . Color-coding as in Figure 1A . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 004 In order to validate that the switch III region is coupled intramolecularly to the reorientation of H-ras , we combined activating mutations in the switch III region with inactivating ones on helix α4 or inactivating switch III mutations with activating ones in the hvr ( center table on top , Figure 1A ) . We predicted that in the case of intramolecular ( allosteric ) coupling these mutations would at least partially neutralize each other , as we have previously observed for mutant H-rasG12V-R128A , R135A , R169A , K170A ( Abankwa et al . , 2008b , 2010 ) , which combines the inactivating/neutralizing mutations R128A , R135A on helix α4 and activating ones of the hvr . Mutation R128A , R135A decreases H-rasG12V nanoclustering ( Guzmán et al . , 2014b ) , however , on its own it has only a small decreasing or neutral effect on H-ras activity ( Abankwa et al . , 2008b , 2010 and Figure 1C ) . To measure the specific activity of H-ras mutants in mammalian cell lines , we used our well-established effector-recruitment FRET assay ( Abankwa et al . , 2008b , 2010; Guzmán et al . , 2014b ) ( Figure 1B ) . We coexpressed mGFP-tagged H-ras mutants and the mRFP-tagged Ras binding domain of C-Raf ( in the following in short RBD ) in BHK cells and monitored the change of FRET using fluorescence lifetime imaging microscopy ( FLIM ) ( Figure 1C , D ) . In agreement with our model of switch III coupling to the orientation of Ras on the membrane ( Abankwa et al . , 2008a , 2008b ) , activating mutations D47A , E49A in the switch III region were neutralized by inactivating mutations R128A , R135A on helix α4 , while the inactivating mutation R161A in switch III was neutralized by activating R169A , K170A in the hvr ( Figure 1C ) . These results therefore agree with an allosteric coupling of the switch III region to the reorientation of H-ras on the membrane . Furthermore , they imply that mutations in the switch III could alter Ras nanoclustering , possibly by a conformation-associated mechanism , which we are however not going to focus on in the following , but discuss in detail at the end . We next explored whether altered nanoclustering also underlies the activity changes that we previously observed for computational modeling-derived switch III mutant H-rasG12V-D47A , E49A ( Abankwa et al . , 2008b ) . We compared the nanoclustering of this hyperactive mutant to that of the parent H-rasG12V by statistical analysis of the distribution of immunogold labeled H-ras in electron microscopic images of plasma membrane sheets from BHK cells ( Plowman et al . , 2005 ) . We observed a significant increase in nanoclustering of this modeling-derived mutant , suggesting that increased signaling originates from augmented nanoclustering ( Figure 2A ) . In order to corroborate these data , we used our recently established Gal-1-dose dependent nanoclustering-response assay ( Guzmán et al . , 2014b ) . This assay measures the dependence of H-ras-mutant nanoclustering on different cellular levels of the nanocluster scaffold Gal-1 . Due to the tight packing of mGFP- and mCherry-tagged Ras proteins in nanoclusters FRET emerges , thus serving as a read-out for nanoclustering ( Figure 2B ) . As compared to H-rasG12V , nanoclustering-FRET of the modeling-derived switch III mutant H-rasG12V-D47A , E49A was significantly increased at all Gal-1 levels ( Figure 2C , D ) ; this was also reflected by the increased complexation of this mutant with Gal-1 in the cells ( Figure 2—figure supplement 1A ) , similar to what we previously observed with a hvr-mutant ( Abankwa et al . , 2010 ) . Taken together , these data confirmed that nanoclustering is increased in H-rasG12V-D47A , E49A . 10 . 7554/eLife . 08905 . 005Figure 2 . Computational modeling-derived switch III mutations D47A , E49A in H-ras increase nanoclustering and RBD-recruitment . ( A ) Electron microscopic nanoclustering analysis of mGFP-H-rasG12V and mGFP-H-rasG12V-D47A , E49A in BHK cells . Normalized univariate K-functions , where maximal L ( r ) -r values above the 99% CI for complete spatial randomness indicate clustering at that value of r ( number of membrane sheets analyzed per condition , n = 17 ) . ( B ) Schematic representation of nanoclustering-FRET analysis , where mGFP-tagged and mCherry-tagged Ras constructs were co-expressed in cells . ( C ) The nanoclustering-FRET response of H-rasG12V-D47A , E49A and its parent construct in dependence of the dose of the nanocluster scaffold Gal-1 in BHK cells . ( D ) Representative nanoclustering-FRET fluorescence lifetime images of BHK cells expressing FRET-pairs ( or donor only , left column ) of constructs indicated on the left , under Gal-1 conditions as annotated on the top . Color look up table to the right shows fluorescence lifetimes . ( E ) RBD-recruitment FRET data of H-rasG12V-D47A , E49A and its parent construct at three different Gal-1 doses analyzed using FRET-imaging of transiently transfected BHK cells . ( C , E ) Numbers in bars give numbers of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis vs parent RasG12V was performed as described in ‘Materials and methods’ ( *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 ) . ( F ) Western blot analysis of C-Raf , MEK , and ERK phosphorylation and C-Raf in BHK cells transiently expressing mGFP-H-rasG12V-D47A , E49K , its parent or an empty vector control . Equal expression of Ras constructs can be seen in the mGFP row , equal loading in the Actin row . ( G ) In vitro RBD binding to mant-GTPγS loaded wild-type ( wt ) H-ras or H-ras-D47A , E49A measured with a fluorescence anisotropy assay . Details on the fitting function are in the ‘Materials and methods’ . Data are averages ±SEM of three repeats . ( H ) RBD-pulldown experiments in BHK cells transiently expressing indicated H-ras mutants or wt H-ras . Top panel shows level of active Ras after EGF-stimulation ( 100 ng/ml ) . Bottom panel shows GAP sensitivity of wt and mutant H-ras proteins . See also Figure 2—figure supplement 1 and Figure 2—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 00510 . 7554/eLife . 08905 . 006Figure 2—figure supplement 1 . The computational modeling-derived switch III H-ras mutant exhibits stronger Gal-1 complexation and remains sensitive to GAP-mediated hydrolysis . ( A ) Schematic representation of the Gal-1-complexation FRET assay in which complexation of mRFP-Gal-1 with mGFP-tagged mutants of Ras was measured in intact BHK cells using FRET ( left ) . Gal-1-complexation FRET analysis of H-rasG12V-D47A , E49A and its parent construct . Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis of differences vs H-rasG12V was performed as described in ‘Materials and methods’ ( *p < 0 . 05 ) ( right ) . ( B ) Relative amounts of phosphorylated C-Raf , MEK and ERK in BHK cells transiently expressing vector control , mGFP-H-rasG12V or mGFP-H-rasG12V-D47A , E49A determined by western blotting from three independent repeats . Error bars represent the standard error of the mean ( ±SEM ) . Quantification of band intensities and statistical analysis was performed as described in ‘Materials and methods’ ( *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 ) . ( C ) RBD-pulldown experiment quantification of the active , GTP-bound fraction of H-ras-D47A , E49A in BHK cells . +EGF denotes stimulation with 100 ng/ml EGF . −EGF serum starved cells . +GAP incubation with GAP domain of NF1 , to assay for GAP-sensitivity . The graphs represent the averages of active H-ras-D47A , E49A normalized to wt H-ras + EGF-stimulation from three independent experiments . Blue vertical line annotates the activity of wt H-ras when stimulated with EGF . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant; **p < 0 . 01 , *p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 00610 . 7554/eLife . 08905 . 007Figure 2—figure supplement 2 . SOS-mediated nucleotide exchange kinetics , as well as mantGTPγS dissociation constants of wt H-ras and mutants . SOS-induced Eu3+-GTP association ( A ) and dissociation ( B ) kinetics with wt H-ras ( black ) , H-ras-D47A , E49A ( red ) , H-ras-G48R ( blue ) , and H-ras-G48R , D92N ( orange ) monitored using the quenching resonance energy transfer ( QRET ) technique . Dots represent data points obtained from individual reactions , with a total of 360 individual data points for each H-ras . ( C ) Fluorescence anisotropy binding data for the derivation of dissociation constants ( Kd ) between mantGTPγS and wt H-ras ( black ) , H-ras-D47A , E49A ( red ) , H-ras-G48R ( blue ) , and H-ras-G48R , D92N ( orange ) . Details are described in the ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 007 We next examined , whether the nanoclustering response is also reflected by the recruitment of the effector C-Raf , as predicted by our recent computational simulations of nanoclustering ( Guzmán et al . , 2014b ) . In line with this simulation , RBD-recruitment to the structure modeling-derived switch III H-ras mutant was significantly increased at all Gal-1 levels , as compared to the parent H-rasG12V ( Figure 2E ) . Correspondingly , Raf , MEK , and ERK phosphorylation was significantly increased when H-rasG12V-D47A , E49A was expressed , suggesting that increased nanoclustering was propagated all the way through the MAPK-signaling pathway ( Figure 2F; Figure 2—figure supplement 1B ) . Importantly , enhanced effector-recruitment ( Figure 2E ) , which initiates the signal cascade was not due to an overall increase of the affinity to the Ras mutant , as revealed in anisotropy binding experiments of purified GTPγS-H-ras-D47A , E49A to the RBD in vitro ( Figure 2G , Supplementary file 1 ) . Also GAP-mediated hydrolysis ( Figure 2H , Figure 2—figure supplement 1C ) was effectively unaltered as compared to wt H-ras . A relatively small decrease in the SOS-mediated GTP-analog off-rate and consistently an increase in EGF-stimulated activation however suggested that this mutant may be more efficiently activated ( Figure 2—figure supplement 2 , Supplementary file 1 ) . We conclude that increased effector recruitment ( Figure 2E ) and signaling ( Figure 2F ) of H-rasG12V with the structure modeling-derived mutation D47A , E49A in the switch III region is due to augmented Ras nanoclustering . The computational modeling-derived switch III residues D47 and E49 are phylogenetically highly conserved ( Figure 3—figure supplement 1 ) , consistent with a biologically significant role of their structural context . We hypothesized that if our link between the switch III and nanoclustering was physiologically relevant , genetic diseases with hyperactivated Ras , such as many cancers , should have exploited it to upregulate Ras-signaling activity . We therefore analyzed more than 140 , 000 cancer sample entries from public cancer genomic databases ( COSMIC , cBioPortal and ICGC ) for point mutations in the orientation-switch III region of HRAS , NRAS , and KRAS genes; specifically , in the β2-β3-loop and helix α5 ( switch III ) , as well as the orientation stabilizing elements , the helix α4 and the hvr . We found that approximately 12% of the distinct mutations along the Ras sequences occur in these switch III regions . However , the overall incidence is very low with a total number of 15 , 20 , and 28 reported cases for H- , N- , and K-ras , respectively ( Supplementary file 2 ) . Despite the low number of cases , we analyzed the isoform-specific mutations by region and by cancer type ( Figure 3 , Supplementary file 2 ) . For H-ras colorectal as well as hematopoietic and lymphoid tissue samples showed the highest number of cases with orientation-switch III mutations ( 3 each ) , while for N-ras , we found colorectal and skin tumor samples with a slightly higher occurrence of 5 cases each reported . In agreement with the overall highest mutation frequency of K-ras in cancer ( Prior et al . , 2012 ) , the highest total number of mutations was found in this Ras isoform , with the highest incidence numbers in colorectal and endometrial samples . In total , it appeared that orientation-switch III mutations occur most frequently in colorectal cancer samples . 10 . 7554/eLife . 08905 . 008Figure 3 . Orientation-switch III mutations in oncogenic Ras isoforms that are reported in cancer genome databases . Cancer-associated point mutations ( missense , nonsense , and silent ) in the orientation-switch III regions as reported in COSMIC , cBioPortal , and ICGC databases . Schematic representations of the linear protein structures of ( A ) H-ras , ( B ) N-ras , and ( C ) K-ras . Critical functional regions in the structures are annotated; those of the orientation-switch III by ovals . Approximate position of hot-spot mutation residues ( G12 , G13 , and Q61 ) is marked with red triangles . Mutations from cancer patient samples are boxed , RASopathy mutations are boxed and in italics , other annotated mutations are from cancer cell lines ( Prior et al . , 2012 ) . Mutations that are studied here are highlighted in yellow . See also Figure 3—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 00810 . 7554/eLife . 08905 . 009Figure 3—figure supplement 1 . Phylogenetic analysis of the Ras switch III region . ( A ) ClustalW sequence alignment of 18 Ras proteins from different species as annotated on the left . Consensus sequence is shaded in dark and switch III region elements β2-β3-loop ( residues 40–56 ) and helix α5 ( residues 152–166 ) are highlighted with a black frame . ( B ) Unrooted phylogenetic trees were built from these proteins for the β2-β3-loop and ( C ) helix α5 . ( B , C ) Numbers on internal branches indicate the percentage of 1000 bootstrap trials that support the branch ( numbers <70% are omitted ) . Residues outside the region of interest were excluded using the position-masking tool for both phylogenetic analyses . The analysis revealed that switch III residues are highly conserved among all analyzed Ras proteins . The human Ras isoforms cluster together with other Ras proteins from the Bilateria species ( D . rerio , X . laevis , S . mansoni , M . edulis , D . melanogaster , and C . elegans ) . Ras proteins from the Fungi kingdom ( S . pombe and T . hirsuta ) and Cnidaria phylum ( H . vulgaris ) species show a divergent pattern that corresponds to the speciation events that occurred in the course of evolution ( ∼1298 Ma for Bilateria—Cnidaria and ∼1513 Ma for Animalia—Fungi speciation event ) ( Hedges et al . , 2004 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 009 When analyzed by affected region , it became apparent that certain parts of the orientation-switch III are more frequently hit in a given isoform . In H-ras helix α4 , in N-ras the β2-β3-loop , and in K-ras the helix α5 are most frequently affected by point mutations ( Figure 3 , Supplementary file 2 ) . To the best of our knowledge , it is currently unknown whether and by which mechanism these mutations affect tumorigenesis . Based on our mechanistic insight , we hypothesized that these cancer-associated orientation-switch III mutations increase Ras signaling by augmenting its nanoclustering . Only nanoclustering of GTP-loaded Ras is relevant for its signaling output ( Tian et al . , 2007 ) . We therefore clamped all mutations in the GTP-state , by co-introduction of the G12V mutation . This also allowed us to compare results with those of the computational modeling-derived mutants ( Figures 1 , 2 ) . We then ensured by a comprehensive set of biochemical experiments that the same mutations in the wild type background did not alter essential activities of Ras , such as GTP-loading , GTP-hydrolysis or effector binding , which classically increase Ras activity . Given that the switch III region is most frequently mutated across all Ras isoforms , we focused on patient samples with mutations in this region ( Figure 3 ) . In H-ras only one case of a switch III region mutation is described in an atypical spitzoid tumor ( AST ) sample . Curiously , this mutation G48R is found together with a second D92N-mutation in H-ras . Confocal colocalization analysis confirmed that the localization of both mutants in BHK cells was unaltered as compared to parent H-rasG12V ( Figure 4A ) . Assuming that the G48R , D92N mutations were gain of function , we combined these cancer-associated mutations with the inactivating mutation R128A , R135A on helix α4 . This new mutant showed normalized H-ras activity ( Figure 4—figure supplement 1A ) , comparable to the modeling-derived orientation-switch III mutants ( Figure 1D ) . This suggested that the cancer-associated mutation is also coupled to the reorientation mechanism of membrane bound H-ras and may thus also impact on its nanoclustering . Indeed , electron microscopic analysis of plasma membrane sheets derived from BHK cells expressing the single or double-mutant revealed a strong and significant increase in nanoclustering of both mutants as compared to the parent H-rasG12V ( Figure 4B ) . This was confirmed by nanoclustering-FRET analysis in cells , where consistent with the increased Gal-1 complexation of the G48R-mutant ( Figure 4—figure supplement 1B ) , the Gal-1-dose-dependent nanoclustering response was significantly increased for both the G48R and G48R , D92N mutants at all Gal-1 levels ( Figure 4C ) . Moreover , the Gal-1-dose-dependent RBD recruitment response of both mutants was increased in cells ( Figure 4D ) , while RBD-binding in vitro remained indistinguishable from the parent Ras ( Figure 4E , Supplementary file 1 ) . The correlation of the RBD- and nanoclustering-response pattern again suggested that augmented nanoclustering increased H-ras activity by enhancing effector recruitment . Together with the absence of any RBD-binding defect , this led us to the conclusion that introduction of mutations G48R or G48R , D92N enhances effector recruitment due to increased nanoclustering . 10 . 7554/eLife . 08905 . 010Figure 4 . Increased nanoclustering and effector recruitment by a tumor-derived switch III mutation in H-ras . ( A ) Representative confocal images of mCFP-H-rasG12V co-localization with the switch III mutants mCherry-H-rasG12V-G48R ( top ) or mCherry-H-rasG12V-G48R , D92N ( bottom ) in BHK cells . The colocalization Manders' overlap coefficient ( R ) is marked on the merged images . Scale bar is 10 μm . ( B ) Electron microscopic nanoclustering analysis of mGFP-tagged H-ras mutants in BHK cells . Normalized univariate K-functions , where maximal L ( r ) -r values above the 99% CI for complete spatial randomness indicate clustering at that value of r ( number of membrane sheets analyzed per condition , n > 17 ) . ( C ) The nanoclustering-FRET response of the tumor-derived H-rasG12V-G48R , D92N mutant , its switch III mutation-only derivative H-rasG12V-G48R and its parent construct in dependence of the dose of the nanocluster scaffold Gal-1 in BHK cells . ( D ) RBD-recruitment FRET analysis of H-rasG12V with or without mutations G48R or G48R , D92N at three different Gal-1 doses in BHK cells . ( C , D ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis vs parent RasG12V was performed as described in the ‘Materials and methods’ ( NS , non-significant; **p < 0 . 01; ***p < 0 . 001 ) . ( E ) In vitro RBD binding to mant-GTPγS loaded wild-type ( wt ) H-ras or H-ras carrying mutations G48R or G48R , D92N measured with a fluorescence anisotropy assay . Binding constants are given in Supplementary file 1 . Note the orange curve covers the black . Details on the fitting function are in the ‘Materials and methods’ . Data are averages ±SEM of three repeats . ( F ) RBD-pulldown experiments in BHK cells transiently expressing indicated H-ras mutants or wild type ( wt ) H-ras . Top panel shows level of active Ras after EGF-stimulation ( 100 ng/ml ) . Bottom panel shows GAP-sensitivity of wt and mutant H-ras proteins . Note that control lanes marked G12V and wt are identical to those in Figure 2H , as these H-ras samples were always run on the same gel . See also Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 01010 . 7554/eLife . 08905 . 011Figure 4—figure supplement 1 . Cancer-associated H-ras switch III mutations have no relevant effect on its biochemical properties . ( A ) RBD-recruitment FRET data of indicated H-ras mutants transiently expressed in BHK cells . ( B ) Gal-1-complexation FRET analysis of H-rasG12V-G48R and its parent construct . ( A , B ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant; ***p < 0 . 001 ) . ( C ) RBD-pulldown experiment quantification of the active , GTP-bound forms of H-ras mutants . +EGF denotes stimulation with 100 ng/ml EGF . −EGF serum starved cells . +GAP incubation with GAP domain of NF1 , to assay for GAP sensitivity . The graphs represent the averages of active H-ras-G48R and H-ras-G48R , D92N normalized to wt H-ras + EGF-stimulation from three independent experiments . Blue vertical line annotates the activity of wt H-ras when stimulated with EGF . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 011 As observed for the modeling-derived switch III mutant , basic biochemical parameters of H-ras-G48R or H-ras-G48R , D92N , such as GAP-mediated GTP hydrolysis ( Figure 4F , Figure 4—figure supplement 1C ) , as well as GTP affinity ( Supplementary file 1 ) remained unaltered . However , mutations G48R , D92N , but not G48R alone , showed a ∼1 . 4-fold elevation of SOS-mediated nucleotide exchange ( Supplementary file 1 ) consistent with an increased activation upon EGF-stimulation ( Figure 4F , Figure 4—figure supplement 1C ) . Therefore , in this cancer-derived H-ras mutation a combination of weak biochemical alterations and a significantly increased nanoclustering , which both result in an activating phenotype of the G48R , D92N mutation in H-ras , may explain the occurrence of these two H-ras mutations in the tumor . For unknown reasons , both N- and K-ras are more frequently mutated and associated with more aggressive cancers than H-ras . N-ras is highly associated with skin and hematological cancers , while K-ras is highly mutated in pancreatic and intestinal cancers ( Prior et al . , 2012 ) . We therefore in the following concentrated on these isoforms to investigate the direct involvement of switch III mutations in tumor growth . In N-ras , the T50I mutation was found in a malignant melanoma and was recently also described to be associated with Noonan syndrome ( Cirstea et al . , 2010 ) , a RASopathy . The additional two mutations , C51Y and E49K , were found in hematologic and skin malignancies , respectively . Interestingly , the highly conserved E49 is also mutated in K-ras in colorectal cancer ( Supplementary file 2 ) . Moreover , our data from the computational modeling-derived H-ras switch III mutant D47A , E49A already suggested that this residue might be involved in nanocluster augmentation . As observed for H-ras switch III mutations , none of the three N-ras mutations altered the subcellular distribution of N-ras ( Figure 5A ) . We next assessed nanoclustering and effector recruitment of these mutants in BHK cells . Gal-1 is not known to impact on N-ras nanoclustering and binding to it remained unaltered by the mutations in N-ras ( Figure 5—figure supplement 1A ) . We therefore investigated the N-ras mutants only at endogenous Gal-1 concentrations . While mutation T50I had no significant effect , both mutation E49K and C51Y significantly increased nanoclustering ( Figure 5B , C ) and ensuing Ras effector recruitment as compared to parent N-rasG12V ( Figure 5D ) . 10 . 7554/eLife . 08905 . 012Figure 5 . Tumor-derived N-ras switch III mutations increase cell proliferation and transforming activity by increased nanoclustering . ( A ) Representative confocal images of mCFP-N-rasG12V co-localization with the switch III mutants mCherry-N-rasG12V-T50I , mCherry-N-rasG12V-E49K , or mCherry-N-rasG12V-C51Y in BHK cells . The Manders' coefficient ( R ) that quantifies co-localization is marked on the merged images . Scale bar is 10 μm . ( B ) Electron microscopic nanoclustering analysis of mGFP-tagged N-ras mutants in BHK cells . Normalized univariate K-functions , where maximal L ( r ) -r values above the 99% CI for complete spatial randomness indicate clustering at that value of r ( number of membrane sheets analyzed per condition , n > 17 ) . ( C ) Nanoclustering-FRET analysis of cancer-associated N-rasG12V-T50I , N-rasG12V-E49K and N-rasG12V-C51Y as compared to their parent construct in BHK cells . ( D ) RBD-recruitment FRET analysis of N-rasG12V-T50I , N-rasG12V-E49K , and N-rasG12V-C51Y as compared to their parent construct in BHK cells . ( E ) RBD-pulldown experiments in BHK cells transiently expressing indicated N-ras mutants or wild-type ( wt ) N-ras . Top panel shows level of active N-ras after EGF-stimulation ( 100 ng/ml ) . Bottom panel shows GAP-sensitivity of wt N-ras and mutant N-ras proteins . ( F ) MAPK-dependent PC12-differentiation assay . PC12 cells transiently expressed indicated mGFP-tagged N-rasG12V with or without switch III mutations , or the control tH ( minimal membrane anchor of H-ras ) for 48 hr . Subsequently neurite length was assessed ( right ) . Examples of confocal images of PC-12 cells expressing indicated constructs used for neurite length quantification ( left ) . Scale bar is 15 μm . ( C , D , F ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis vs parent RasG12V was performed as described in the ‘Materials and methods’ ( NS , non-significant; *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 ) . ( G ) Proliferation of NIH/3T3 cells stably expressing N-rasG12V , N-rasG12V-C51Y , or a control construct . Control is transduced with the GFP reporter only . Error bars represent the standard error of the mean ( ±SEM ) . ( H ) Transformation of NIH/3T3 cells stably expressing N-rasG12V as compared to N-rasG12V-C51Y or a vector-only control . Foci formed after 7 days of growth were stained with crystal violet and mean ( ±SEM ) foci areas from three biological repeats were quantified ( right ) . Representative images of crystal violet-stained foci of transduced NIH/3T3 cells ( left ) . ( I ) Anchorage-independent growth of NIH/3T3 cells stably expressing N-rasG12V as compared to N-rasG12V-C51Y or a vector-only control . Colonies were imaged after 14 days ( left ) , and the colony area was determined ( right ) . Three independent biological repeats were performed and each repeat was done in triplicate . Error bars represent the standard error of the mean ( ±SEM ) . See also Figure 5—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 01210 . 7554/eLife . 08905 . 013Figure 5—figure supplement 1 . Cancer-associated mutations in switch III of N-ras do not alter its biochemical properties and Gal-1 complexation . ( A ) Gal-1-complexation FRET data of indicated N-ras mutants and parent construct transiently expressed in BHK cells . ( B ) RBD-recruitment FRET analysis of indicated N-ras mutant and parent constructs transiently expressed in BHK cells with ( + ) or without ( − ) 5 μM compactin treatment for 48 hr . Compactin treatment relocalized N-ras mutant/RBD complexes to the cytoplasm , allowing us to assess binding in solution . ( A , B ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant ) . ( C ) RBD-pulldown experiment quantification of the active , GTP-bound forms of wt and mutant N-ras . +EGF denotes stimulation with 100 ng/ml EGF; −EGF serum starved cells; +GAP incubation with GAP domain of NF1 , to assay for GAP-sensitivity . The graphs represent the averages of active N-ras mutants normalized to wt N-ras + EGF-stimulation from two independent experiments . Blue vertical line annotates the activity of wt N-ras when stimulated with EGF . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant ) . ( D ) Western blot analysis of phosphorylated MEK , ERK , and AKT and total MEK , ERK , and AKT in BHK cells transiently expressing EGFP only , mCit-N-ras-wt , mCit-N-ras-T50I , mCit-N-ras-E49K , or mCit-N-ras-C51Y . Cells were serum starved and collected without EGF stimulation ( −EGF ) or after stimulation with 100 ng/ml EGF for 10 min ( +EGF ) . Equal expression of Ras constructs can be seen in the GFP row , equal loading in the β-actin row . ( E ) Validation of comparable protein expression levels of HA-tagged N-ras mutants stably expressed in NIH/3T3 cells . Anti-HA-tag Western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 013 In order to assess RBD-binding to cytoplasmic Ras , that is , in solution , we treated mammalian cells expressing Ras and RBD constructs with compactin , as compactin blocks prenylation and therefore membrane anchorage of Ras . We previously showed that these compactin-treatment experiments provide the same information as binding experiments with purified proteins ( Guzmán et al . , 2014b ) . All mutants bound the RBD indistinguishably from N-ras without switch III mutations ( Figure 5—figure supplement 1B ) . Moreover , both hyperactive mutants N-ras-E49K and N-ras-C51Y , as well as N-ras-T50I remained sensitive to GAP-mediated hydrolysis and did not display increased background activation of MEK , ERK , or AKT in the absence of serum ( Figure 5E , Figure 5—figure supplement 1C , D ) . Thus , basic biochemical parameters were largely preserved , confirming that increased RBD-recruitment followed augmented nanoclustering of N-ras with switch III mutations E49K or C51Y . It was previously observed that RASopathy-associated Ras mutants do not lead to a clear increase of the downstream pathway ( Gremer et al . , 2011; Cirstea et al . , 2013 ) . Similar to these observations , neither GTP-loading ( Figure 5E , Figure 5—figure supplement 1C ) nor phosphorylation of MEK , ERK , and AKT ( Figure 5—figure supplement 1D ) was clearly increased after EGF-stimulation . We therefore resorted to more complex read-outs of Ras activity . Differentiation of rat adrenal pheochromocytoma ( PC12 ) cells is accompanied by neurite outgrowth , which serves as a well established and sensitive measure for Ras/MAPK activity ( Gorfe et al . , 2007 ) . PC12 cells transfected with mGFP-tagged mutant N-rasG12V-E49K or N-rasG12V-C51Y showed significantly increased neurite length ( Figure 5F ) . The most active mutant N-rasG12V-C51Y was then further analyzed for its effect on the tumorigenic growth potential of cells . NIH/3T3 cells stably expressing this mutant ( Figure 5—figure supplement 1E ) showed significantly increased cell proliferation ( Figure 5G ) , focus formation ( Figure 5H ) , and anchorage-independent growth in soft agar ( Figure 5I ) , which typically correlates with the tumorigenic potential in vivo . In summary , cancer-derived mutations E49K and C51Y do not affect relevant biochemical functions of N-ras , but instead increase its nanoclustering potential to enhance effector recruitment and subsequent Ras/MAPK-signaling output . Our most active N-ras mutant furthermore revealed that augmented nanoclustering ultimately increases cell proliferation and tumorigenic growth of transformed cells . The KRAS gene is the developmentally most significant Ras isoform . Its important developmental role is furthermore underscored by the fact that activating germline mutations that are associated with the K-ras4B isoform ( henceforth K-ras ) lead to relatively severe RASopathy phenotypes ( Carta et al . , 2006 ) . In addition to our N-rasT50I mutant , we therefore added the Noonan syndrome-associated mutation V152G on helix α5 of the switch III of K-ras to our otherwise cancer mutation-focused analysis . These two mutations represented mechanistically unresolved switch III mutations in RASopathies . Like for the N-ras Noonan syndrome mutant , subcellular distribution remained unchanged ( Figure 6A ) and no increase in nanoclustering ( Figure 6B , C ) , effector recruitment ( Figure 6D ) or PC12 differentiation ( Figure 6F ) was found . Also the binding of the mutant to the RBD in solution was unchanged compared to the parent K-ras ( Figure 6—figure supplement 1B ) , and its GAP sensitivity and downstream signaling activity under serum starvation were unaltered ( Figure 6E , Figure 6—figure supplement 1C , D ) . 10 . 7554/eLife . 08905 . 014Figure 6 . K-ras with colorectal cancer-associated mutation R164Q displays increased nanoclustering to drive oncogenic transformation . ( A ) Colocalization of mCFP-K-rasG12V and switch III mutants mCherry-K-rasG12V-V152G and mCherry-K-rasG12V-R164Q in BHK cells . The Manders' coefficient ( R ) that quantifies colocalization is marked on the merged images . Scale bar is 10 μm . ( B ) Electron microscopic nanoclustering analysis of BHK cells expressing indicated mGFP-tagged K-ras mutants . Normalized univariate K-functions , where maximal L ( r ) -r values above the 99% CI for complete spatial randomness indicate clustering at that value of r ( number of membrane sheets analyzed per condition , n > 15 ) . ( C ) The nanoclustering-FRET response of K-rasG12V-V152G , K-rasG12V-R164Q as compared to their parent construct in BHK cells . ( D ) RBD-recruitment FRET analysis of K-rasG12V-V152G , K-rasG12V-R164Q as compared to their parent construct in BHK cells . ( E ) RBD-pulldown experiments in BHK cells transiently expressing indicated K-ras mutants or ( wild-type ) wt K-ras . Top panel shows level of active K-ras after EGF-stimulation ( 100 ng/ml ) . Bottom panel shows GAP-sensitivity of wt and mutant K-ras proteins . ( F ) MAPK-dependent PC12 differentiation assay . Cells transiently expressed indicated mGFP-tagged K-rasG12V with or without mutations V152G and R164Q , or the control tH for 48 hr . Subsequently neurite length was assessed ( left ) . Examples of confocal images of PC-12 cells expressing indicated constructs used for quantification ( right ) . Scale bar is 15 μm . ( C , D , F ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in the ‘Materials and methods’ ( NS , non-significant; *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001 ) . ( G ) Proliferation of NIH/3T3 cells stably expressing K-rasG12V , K-rasG12V-R164Q , or a control construct for 72 hr . Control is transduced with the GFP reporter only . Error bars represent the standard error of the mean ( ±SEM ) . ( H ) Transformation assay of NIH/3T3 cells virally transduced to stably express-indicated constructs and a GFP reporter . Foci formed after 7 days of growth were stained with crystal violet and mean ( ±SEM ) foci areas from three biological repeats were quantified ( right ) . Representative images of crystal violet stained foci of transduced NIH/3T3 cells ( left ) . ( I ) Anchorage-independent growth of NIH/3T3 cells stably expressing K-rasG12V as compared to K-rasG12V-R164Q or a vector-only control . Colonies were imaged after 14 days ( left ) , and the colony area was determined ( right ) . Three independent biological repeats were performed , and each repeat was done in triplicate . Error bars represent the standard error of the mean ( ±SEM ) . See also Figure 6—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 01410 . 7554/eLife . 08905 . 015Figure 6—figure supplement 1 . K-ras with mutation R164Q has unaltered Gal-1-complexation and biochemical properties . ( A ) Gal-1-complexation FRET analysis of K-ras mutants and its parent construct transiently expressed in BHK cells . ( B ) RBD-recruitment FRET analysis of K-ras mutants and its parent construct transiently expressed in BHK cells with ( + ) or without ( − ) 5 μM compactin treatment for 48 hr . Compactin treatment relocalized K-ras mutant/RBD complexes to the cytoplasm , allowing us to assess binding in solution . ( A , B ) Numbers in bars give number of analyzed cells from three independent experiments . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant , ***p < 0 . 001 ) . ( C ) RBD-pulldown experiment quantification of the active , GTP-bound forms of wt and mutant K-ras . +EGF denotes stimulation with 100 ng/ml EGF . −EGF serum-starved cells . +GAP incubation with GAP domain of NF1 , to assay for GAP sensitivity . The graphs represent the averages of active K-ras mutants normalized to wt K-ras + EGF-stimulation from three independent experiments . Blue vertical line annotates the activity of wt K-ras when stimulated with EGF . Error bars represent the standard error of the mean ( ±SEM ) . Statistical analysis was performed as described in ‘Materials and methods’ ( NS , non-significant , ***p < 0 . 001 ) . ( D ) Western blot analysis of phosphorylated MEK , ERK , and AKT and total MEK , ERK , and AKT in BHK cells transiently expressing EGFP only , mCFP-K-ras-wt , mCFP-K-ras-V152G , or mCFP-K-ras-R164Q . Cells were serum starved and collected without EGF stimulation ( −EGF ) or after stimulation with 100 ng/ml EGF for 10 min ( +EGF ) . Equal expression of Ras constructs can be seen in the GFP row , equal loading in the β-actin row . ( E ) Validation of comparable protein expression levels of HA-tagged K-ras mutants stably expressed in NIH/3T3 cells . Anti-HA-tag Western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 015 We then turned to investigate the most frequent switch III mutation , R164Q in K-ras , which was found in colorectal and endometrial cancers . A previous study by others showed that this helix α5 mutation alone ( i . e . , in the wt background ) did not possess any detectable increase in its transforming potential , yet gene expression analysis suggested a mild activation phenotype ( Smith et al . , 2010 ) . Addition of R164Q did not alter the subcellular distribution of mGFP-tagged K-rasG12V ( Figure 6A ) in BHK cells . Like for the N-ras mutants , binding to Gal-1 was unaltered ( Figure 6—figure supplement 1A ) , which again led us to study the activity only at endogenous Gal-1 levels . Nanoclustering analysis of K-rasG12V-R164Q by electron microscopic- ( Figure 6B ) and by nanoclustering-FRET-analysis ( Figure 6C ) in BHK cells both confirmed significantly increased nanoclustering as compared to the parent K-ras . In agreement with our nanocluster-activity model ( Figure 1A ) , augmented nanoclustering was strictly followed by an increased recruitment of the effector fragment RBD in BHK cells ( Figure 6D ) . Importantly , in solution the effector fragment again was bound equally to the mutant and to the parent K-ras ( Figure 6—figure supplement 1B ) . Together with the normal GAP-sensitivity of K-ras-R164Q ( Figure 6E , Figure 6—figure supplement 1C ) , these results support that the mutation neither affected the affinity to the RBD of C-Raf nor GAP-mediated Ras inactivation . As for the two N-ras nanocluster mutations , we observed increased Ras/MAPK activity of K-rasG12V-R164Q by significantly increased PC12 cell differentiation ( Figure 6F ) but not on the phospho-protein level ( Figure 6—figure supplement 1D ) . In agreement with this hyperactivation phenotype , NIH/3T3 cells expressing the mutant ( Figure 6—figure supplement 1E ) proliferated faster ( Figure 6G ) , were more transformed ( Figure 6H ) , and grew more anchorage independently in soft agar ( Figure 6I ) than the K-rasG12V parent . Thus mutation R164Q in K-ras increases cell proliferation and tumorigenicity by its increased nanoclustering .
Here , we have described an unprecedented mechanism for the activation of a signaling protein in cancer . We showed that rare cancer-associated mutations in the switch III regions of H- , N- , and K-ras can increase effector engagement , while not changing the affinity to the effector . Instead the increase of the nanoscale concentration of Ras in nanocluster hyperactivates Ras to promote tumorigenic growth ( Figure 7 ) . 10 . 7554/eLife . 08905 . 016Figure 7 . Summary table of Ras switch III mutant properties studied in this work . The table summarizes major experimental results obtained in different assays that were used to characterize switch III mutants of H-ras ( highlighted in green ) , N-ras ( violet ) , and K-ras4B ( blue ) . Note that in addition extensive biochemical characterization data of the H-ras mutants can be found in Supplementary file 1 . Black circle dot indicates that no significant change was observed . Black arrows represent significant increase ( up ) or decrease ( down ) of quantified parameters as compared to the parent RasG12V control . The percentage of these changes is given in addition . The following columns ( italics ) report on: Conformation—predicted H-ras mutant conformation , according to Figure 1A; GAP—sensitivity of mutants to GAP-stimulated GTP hydrolysis; RBD + soluble Ras—binding of the C-Raf-RBD to Ras mutants in solution , which was either measured by fluorescence anisotropy with purified proteins or in BHK cells treated with compactin using FRET ( annotated with * ) ; Localization—change in colocalization of mutants relative to their parent constructs in BHK cells determined by confocal microscopy; Nanoclustering-EM—changes in Ras nanoclustering obtained by electron microscopy or -FRET in BHK cells; RBD + membrane Ras—binding of the C-Raf-RBD to Ras mutants in BHK cells measured by FRET; PC12—results of PC12-cell neurite outgrowth assay; last three columns report on Ras-mutant transformed NIH/3T3 cell proliferation , focus formation ( transformation ) and anchorage-independent growth ( soft agar ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08905 . 016 Our data at the outset suggested a connection between the H-ras conformers and nanoclustering . How could the conformation couple to nanoclustering ? To address this question , we will here try to synthesize the most recent work on the structure and conformation of Ras in particular in the lipid environment . We will furthermore relate this to the assembly of switch III-mutated Ras into dimers and oligomers in nanoscale clusters in the membrane and ultimately to effector activation . Molecular dynamics simulations suggest that Ras isoforms change their conformation via different intramolecular routes ( Gorfe et al . , 2008 ) . An important shared feature is that the identical effector lobe of Ras , which contains the switch I , II , and III regions actuates the C-terminal helix α5 . This might provide the mechanistic explanation of how the nucleotide exchange-induced structural change is converted into different Ras isoform-specific conformational equilibria on the membrane ( Abankwa et al . , 2010 ) . A milestone work by Mazhab-Jafari et al . has recently provided the first structural data on membrane bound K-ras4B ( Mazhab-Jafari et al . , 2015 ) . Their data suggest at least two conformational states , with the GTP-state displaying an occluded effector-binding site . Conversely , in the GDP-state the G-domain contacts the membrane via helix α4 and α5 thus exposing the effector interaction site . Also for N-ras low-resolution spectroscopic data support a nucleotide-dependent conformational change on the membrane ( Kapoor et al . , 2012a , 2012b ) . Importantly , the K-ras conformational reorientation resembles the one that is here assumed for H-ras . Interestingly , an inverse role of helix α4 in stabilizing GDP- and GTP-conformations of H- and K-ras is implied , which matches to experimental data that we have provided earlier ( Abankwa et al . , 2010 ) . Therefore H- and K-ras reorientation may work just in the opposite fashion . However , the effect of the conformational change can be several fold . On the one hand , Ras dimer formation and oligomerization into nanocluster might become altered . Evidence for dimers of H-ras and N-ras in the membrane ( Lin et al . , 2014; Güldenhaupt et al . , 2012 ) and K-ras even in solution ( at micromolar concentrations ) has recently been provided ( Muratcioglu et al . , 2015; Nan et al . , 2015 ) . The structural data of K-ras dimers suggest that K-ras might preferentially dimerize via specific surfaces in both the GTP- and GDP-state ( Muratcioglu et al . , 2015 ) . Interestingly , the more stable β-dimer , which interfaces basically via the effector lobe of Ras , could prevent effector access . Similarly , coarse grain simulations of the clustering behavior of membrane-bound H-ras in two different conformations ( basically corresponding to those in Figure 1—figure supplement 1 ) revealed that depending on the conformational state , preferred interfaces of the Ras proteins could be utilized ( Li and Gorfe , 2013 ) . On the other hand , a mutation-induced conformational change may directly affect effector engagement , as has been proposed for H-ras ( Abankwa et al . , 2008a ) and shown for K-ras on nanodiscs ( Mazhab-Jafari et al . , 2015 ) . However , the mechanistic basis of how Ras activation is propagated is further complicated by the fact that we still lack detailed structural insight into for example , the precise activation of effectors , such as Raf . Ras-GTP nanoclusters are immobile , suggesting that a transient signaling machinery is built up , which is also coupled or trapped by the cytoskeleton ( Kusumi et al . , 2012 ) . The ( isoform- or mutation- ) specific GTP-Ras conformational states on the membrane may differentially engage Raf-isoforms thus allosterically enabling specific Ras-Raf-isoform coupling ( Abankwa et al . , 2010 ) . Here , the Ras-isoform-specific ( acidic ) membrane environment needs to be considered , as it is significant to retain clustered Raf ( Plowman et al . , 2008; Cho et al . , 2012 ) . Nanocluster formation can in addition be facilitated by nanocluster scaffolds , such as Gal-1 . Gal-1 is typically dimeric suggesting that in nanoclusters , dimeric Ras–Raf units may exist . Of note , the lipid anchors of Ras alone are sufficient to drive higher order clustering of both Ras and Raf ( Nan et al . , 2013 ) . Additional support for a role of stoichiometric Ras–Raf complexes in nanocluster is lent by the fact that Raf dimerization inducing compounds can increase Ras nanoclustering ( Cho et al . , 2012 ) . Therefore , dimeric signaling units may be both the outcome , as well as the driver of multimeric nanocluster formation . This nanoclustering promoting activity of Raf-inhibitors might underlie the paradoxical hyperactivation of Ras signaling in Ras mutant tumors . Intriguingly , these Raf inhibitors act Ras isoform specifically on K- and N-ras-nanoclustering , suggesting that additional factors break the promiscuity of ‘dimer inducers’ . Given that membrane anchored ( nanoclustered ) Raf is constitutively active ( Nan et al . , 2013 ) , facilitation of this effector clustering process is critical for downstream signaling output . In the light of the complexity of the events that lead from Ras activation to conformational changes , oligomerization and effector recruitment , we here refrained from making any conformational predictions for the cancer-associated mutations of N- and K-ras . These would have to be experimentally established by high-resolution methods . We therefore limited ourselves to nanoclustering as an activity predicting observable . If Ras dimerization or multimerization into nanocluster is so crucial for signal propagation , why is it not more frequently exploited in cancer ? First of all , next to mutations in Ras that we have just described , one would have to take mutations and expression changes in largely unknown nanocluster scaffolds into account . Secondly , based on our data Ras activity is typically up-tuned 2–3 fold by nanoclustering ( Guzmán et al . , 2014b ) . By contrast , GEF activation switches Ras on in an ‘all-or-nothing’ fashion ( and GAPs off in the same manner ) , changing the enzymatic activity of Ras by 100 or even >50 , 000-fold ( Eccleston et al . , 1993; Ford et al . , 2006 ) . This would approximate the relative difference in Ras activity induced by hot-spot mutations and is much higher than activity changes by switch III-nanocluster mutations . Correspondingly , one could expect that this relation be to some extent reflected in the relative abundance ( ∼1/2300 ) of switch III mutations in cancer . Currently it is unclear , whether there is a particular relevance for switch III mutations in certain cancer types , as the frequencies are too low for any conclusive statement . Likewise , it is unclear whether these mutations represent remnants of early mutational events in tumors or stem from certain niches of it . Given the large activity difference , it is obvious that switch III mutated clones would easily be outcompeted by Ras hot-spot-mutated clones in a tumor . These features combined with the fact that switch III mutations may be associated with conformational changes in Ras that alter effector engagement , suggest that these mutations belong to the recently proposed class of ‘latent driver’ mutations ( Nussinov and Tsai , 2015 ) . Latent drivers are non-hotspot , lower frequency mutations , which appear like passenger mutations in that they do not cause malignancy . However , it is speculated that in a certain context they could provide a growth advantage . In addition to cancer samples with switch III mutations , we also investigated two mutations that are found in RASopathies . It was previously speculated that the Noonan syndrome-associated mutation N-ras-T50I might be linked to the orientation-switch III mechanism ( Cirstea et al . , 2010 ) . However , our data on this mutation do not support a gain-of-function downstream of GTP-loaded Ras ( Figure 5C–F ) , neither could we observe any change in EGF-dependent Ras activation ( Figure 5E ) . Likewise , we did not observe any changes by mutation V152G in K-ras ( Figure 6 ) . Thus RASopathy-associated switch III mutations do not show changes in nanoclustering , but may have unknown alterations ( N-rasT50I , Figure 5 and [Cirstea et al . , 2010]; K-rasV152G , Figure 6; K-ras-E153V and K-ras-F156L [Gremer et al . , 2011; Mazhab-Jafari et al . , 2015] ) . Given the evidence for nanoclustering of a number of other signaling proteins , such as GPI-anchored proteins ( Goswami et al . , 2008 ) , heterotrimeric G proteins ( Abankwa and Vogel , 2007 ) , Rho- and Rab-small GTPases ( Köhnke et al . , 2012 ) , as well as Src-family kinases ( Najumudeen et al . , 2013 ) , it is timely to consider this mode of activity regulation as a new fundamental building block in the cellular signaling architecture .
Plasmids pmGFP-H-rasG12V , pmGFP-H-RasG12V-D47A , E49A , pmGFP-H-rasG12V-R169A , K170A , pmGFP-H-rasG12V-R128A , R135A , pmCFP-K-rasG12V , pmCit-N-rasG12V , pmRFP-C-Raf-RBD , and mRFP-Gal-1 were described in Abankwa et al . ( 2008 ) , ( 2010 ) . Plasmid pcDNA3-asGal-1 is an antisense construct and was used to deplete Gal-1 , while plasmid pcDNA3-Gal-1 was used to overexpress Gal-1 as described in Paz et al . ( 2001 ) . Plasmid pmGFP-tH was described elsewhere ( Zhou et al . , 2012 ) . Site-directed mutagenesis was done by GenScript USA Inc . ( Piscataway , NJ , USA ) to introduce the specific mutations . To generate plasmids pmGFP-H-rasG12V- R128A , R135A , D47A , E49A mutations , R128A and R135A were introduced to the plasmid pmGFP-H-rasG12V-D47A , E49A and for pmGFP-H-ras-R161A , R169A , K170A , mutation R161A to the plasmid pmGFP-H-ras-R169A , K170A . For pmGFP-H-ras ( wt ) plasmid pmGFP-H-rasG12V was backmutated to G12G . Plasmids pQE-A1 and pQE-A1-H-ras ( wt ) were described in Guzmán et al . ( 2014b ) . To generate plasmids pmGFP-H-ras-G48R , pmGFP-H-rasG12V-G48R , pQE-A1-H-ras-G48R , D92N , pmGFP-H-rasG12V-G48R , D92N , mutations G48R alone or together with D92N were introduced to the plasmids pmGFP-H-rasG12V , pmGFP-H-ras ( wt ) , and pQE-A1-H-ras ( wt ) , respectively . H-ras-D47A , E49A and H-ras-G48R from pmGFP-H-ras-D47A , E49A and pmGFP-H-rasG48R were subcloned into the KpnI and BglII restriction sites of the pQE-A1 plasmid . Plasmids pGEX-4T1-huNF1-333 encoding the catalytic domain of human neurofibromin GAP , NF1-333 ( comprising residues 1198–1530 ) , and pGEX-2T-huC-RAF-RBD for expression of GST-C-Raf-RBD were described previously ( Herrmann et al . , 1995; Gremer et al . , 2011 ) . Plasmid pmCit-N-rasG12V was first backmutated to generate pmCit-N-ras ( wt ) and both plasmids then served as a template to which mutations T50I , E49K , and C51Y were introduced in order to generate pmCit-N-ras-T50I , pmCit-N-rasG12V-T50I , pmCit-N-ras-E49K , pmCit-N-rasG12V-E49K , pmCit-N-ras-C51Y , and pmCit-N-rasG12V-C51Y . Plasmids pmCFP-K-ras-R164Q , pmCFP-K-rasG12V-R164Q , pmCFP-K-ras-V152G , and pmCFP-K-rasG12V-V152G were constructed by site-directed mutagenesis on the plasmids pmCFP-K-rasG12V and pmCFP-K-ras ( wt ) , which was first generated by backmutating the pmCFP-K-rasG12V ( by GenScript Inc . ) . To generate pmGFP-K-rasG12V , pmGFP-K-rasG12V-R164Q , pmGFP-K-rasG12V-V152G , pmGFP-N-rasG12V , pmGFP-N-rasG12V-T50I , pmGFP-N-rasG12V-E49K , and pmGFP-N-rasG12V-C51Y , mGFP from pmGFP-H-rasG12V was subcloned into the NheI and BsrGI restriction sites of the corresponding pmCit-N-ras and pmCFP-K-ras constructs . To generate pmCherry-H-rasG12V , pmCherry-H-rasG12V-D47A , E49A , pmCherry-H-rasG12V-G48R , pmCherry-H-rasG12V-G48R , D92N , pmCherry-K-rasG12V , pmCherry-K-rasG12V-R164Q , pmCherry-K-rasG12V-V152G , pmCherry-N-rasG12V , pmCherry-N-rasG12V-T50I , pmCherry-N-rasG12V-E49K , and pmCherry-N-rasG12V-C51Y , mCherry from pmCherry-C1 plasmid ( Clontech Laboratories Inc . , Mountain View , CA , USA ) was subcloned into the NheI and BsrGI restriction sites of the corresponding pmGFP-H-ras , pmCit-N-ras , and pmCFP-K-ras constructs . To generate pmCFP-H-rasG12V and pmCFP-N-rasG12V , mCFP from pmCFP-K-rasG12V was subcloned into the NheI and BsrGI restriction sites of the corresponding pmGFP-H-rasG12V and pmCit-N-rasG12V . All fluorescent protein-tagged Ras constructs also contained a HA-tag , which is located between the sequence of the fluorescent tag and that of Ras . All constructs were verified by sequencing ( GATC Biotech , Cologne , Germany and GenScript USA Inc . ) . Baby Hamster Kidney ( BHK ) and NIH/3T3 cells were grown in Dulbecco's modified Eagle medium ( DMEM ) supplemented with 10% FBS ( not heat inactivated in case of NIH/3T3 cells ) , L-glutamine , penicillin ( 100 U/ml ) , and streptomycin ( 100 μg/ml ) . PC-12 cells were grown on plates coated with 50 μg/ml of rat tail collagen I ( Termo Fisher Scientific Gibco , Waltham , MA , USA ) in Roswell Park Memorial Institute ( RPMI ) 1640 medium supplemented with 10% horse serum , 5% FBS , L-glutamine , penicillin ( 100 U/ml ) , and streptomycin ( 100 μg/ml ) . All cell types were typically grown to a confluency of 80% ( 8 × 107 cells/ml ) and then sub-cultured every 2–3 days . BHK and PC-12 cell lines were obtained from ATCC cell repository ( LGC Standards GmbH , Wesel , Germany ) . Cells were cultured for not more than 10 passages . NIH/3T3 cell line was obtained from cell line collection of VTT Technical Research Centre of Finland . Immuno electron microscopy spatial mapping was performed as described previously ( Prior et al . , 2003a , 2003b ) . Apical plasma membrane sheets were prepared from BHK cells transiently expressing mGFP-tagged Ras mutants , fixed with 4% PFA , 0 . 1% glutaraldehyde and labeled with 4 . 5 nm ( diameter ) gold nano-particles coupled to anti-GFP antibody . Digital images of the immunogold-labeled plasma membrane sheets were taken at 100 , 000× magnification in an electron microscope Jeol JEM-1400 ( JEOL USA , Inc . , Peabody , MA , USA ) . Intact 1 μm2 areas of the plasma membrane sheet were identified using ImageJ and the ( x , y ) coordinates of the gold particles were determined as described . A minimum of 17 plasma membrane sheets were imaged and analyzed for each Ras mutant . A bootstrap test constructed as previously described is then used to evaluate the statistical significance of differences between replicated point patterns ( Plowman et al . , 2005 ) . BHK cells were transfected using Fugene 6 transfection reagent ( Promega Biotech AB , Madison , WI , USA ) with the donor alone ( mGFP-tagged Ras constructs ) in control samples , or mGFP-tagged Ras together with the acceptor . Acceptors were mCherry-tagged Ras constructs in nanoclustering-FRET experiments , mRFP-C-Raf-RBD in RBD-recruitment FRET experiments or mRFP-Gal-1 in cellular Gal-1 complexation experiments . To modulate cellular Gal-1 levels co-transfections with pcDNA3-asGal-1 or pcDNA3-Gal-1 were done , vectors which express antisense and sense Gal-1 sequences to deplete and increase Gal-1 levels in cells , respectively . Treatment with 5 μM compactin was done 4 hr after transfection . After 48 hr , cells were fixed with 4% PFA and mounted with Mowiol 4–88 ( Sigma–Aldrich , St . Louis , MO , USA ) . The mGFP fluorescence lifetime was measured using a fluorescence lifetime imaging attachment ( Lambert Instruments , Groningen , Netherlands ) on an inverted microscope ( Zeiss AXIO Observer D1 , Jena , Germany ) . For the sample excitation sinusoidally modulated 3 W , 497 nm LED at 40 MHz under epi-illumination was used . Cells were imaged using the 63× , NA 1 . 4 oil objective with the GFP filter set ( excitation: BP 470/40 , beam splitter: FT 495 , emission: BP 525/50 ) . The phase and modulation were determined using the manufacturer's software from images acquired at 12 phase settings . Fluorescein at 0 . 01 mM , pH 9 was used as a lifetime reference standard . Per condition , the fluorescence lifetime was measured typically for >40 cells from three biological repeats . From the obtained fluorescence lifetimes the apparent FRET efficiency was calculated as described in Guzmán et al . ( 2014b ) . For a comprehensive analysis of cancer-associated mutations in the orientation-switch III region , point mutation data for Ras isoforms detected either in clinical tumors or cancer cell lines were downloaded from public databases: cBioPortal to access data from TCGA and other cancer genomic studies ( http://www . cbioportal . org/ ) , COSMIC v68 ( http://cancer . sanger . ac . uk/ ) and ICGC data portal ( https://dcc . icgc . org/ ) . Frame-shifts , insertions , deletions , and stop codon mutations in the Ras isoforms were removed from the analyses . Point mutations , both synonymous and non-synonymous , were categorized into the 4 regions based on the following criteria: β2-β3-loop ( residues of Ras 40–56 ) , helix α4 ( 122–138 ) , helix α5 ( 152–166 ) , hvr ( 166–188/9 ) . For those samples that were redundant across the databases , the union of the observed point mutations was reported for the co-mutated genes . For analysis of phylogenetic conservation of computationally and experimentally suggested switch III residues D47 and E49 , the NCBI's Batch-Conserved Domain interface ( http://www . ncbi . nlm . nih . gov/Structure/bwrpsb/bwrpsb . cgi ) was used to identify proteins similar to that of human H- , N- , and K-ras isoforms . The top 17 hits were further analyzed in the MacVector software ( MacVector Inc . , North Carolina , USA ) . Sequences were first aligned using the ClustalW algorithm and the phylogenetic trees of the switch III elements β2-β3-loop ( residues of human Ras 40–56 ) and helix α5 ( residues 152–166 ) were reconstructed using unrooted neighbor-joining method . The bootstrap probability was calculated from consensus of 1000 bootstrap samples . The co-evolutionary distances of β2-β3-loop and helix α5 regions were compared using the Mirror tree web server ( http://csbg . cnb . csic . es/mtserver/ ) and displayed as the correlation coefficient with assigned statistical significance p values . To analyze the activity of mutant H-rasG12V-D47A , E49A , BHK cells were transiently transfected with vector control ( pC1 vector , Promega Biotech AB ) , mGPF-H-rasG12V or mGFP-H-rasG12V-D47A , E49A using Lipofectamine ( Thermo Fischer Scientific Invitrogen ) . After 16 hr , cells were serum starved for 2 hr . Whole cell lysates were separated by SDS-PAGE and blotted using Cell Signalling Technology ( Boston , MA , USA ) antibodies against pC-Raf ( Ser338 ) ( 9427S ) , pMEK ( 9154S ) , pERK ( 4695S ) , total C-raf , GFP and actin . Band intensities were imaged using FluorChem Q System ( ProteinSimple , San Jose , CA , USA ) , quantified by densitometry and normalized to the average intensity values of all measured conditions . Averages were calculated from three independent biological repeats . To analyze the activity of the K-ras and N-ras switch III mutants , BHK cells were transiently transfected with vector control ( pEGFP-N1 vector , Clontech laboratories Inc . ) , pmCFP-K-ras ( wt ) , pmCFP-K-ras-V152G , pmCFP-K-ras-R164Q , pmCit-N-ras ( wt ) pmCit-N-ras-E49K , pmCit-N-rasT50I , and pmCit-N-rasC51Y using JetPRIME transfection reagent ( Polyplus-transfection , New York , NY , USA ) . After 24 hr , cells were serum starved for 5 hr and stimulated with 100 ng/ml EGF for 10 min . Whole cell lysates were first separated using SDS-PAGE and blotted using primary antibodies for MEK1/2 ( 9126 ) , pMEK1/2 ( 9121 ) , ppERK1/2 ( 9101 ) , ERK1/2 ( 9102 ) , AKT ( 9272 , Cell Signalling Technology ) , pAKT1 ( MAB7419 , R&D Systems , Minneapolis , MN , USA ) , GFP ( 3999-100 , BioVision , Inc . , Milpitas , CA , USA ) , and β-actin ( A1978 , Sigma–Aldrich ) . Membranes were then probed with secondary peroxidase-conjugated IgG antibodies ( sc-2370 , sc-2954 , Santa Cruz Biotechnology , Inc . , Dallas , TX , USA ) . Chemiluminescence was detected using ChemiDoc MP System ( Bio-Rad , Hercules , CA , USA ) . Wild-type and mutant H-ras as well as C-Raf-RBD proteins were prepared from Escherichia coli using the pQE-expression system ( Qiagen , Hilden , Germany ) as described ( Guzmán et al . , 2014b ) . The C-Raf-RBD and GAP-domain of neurofibromin comprising residues 1198–1530 of human NF1 ( NF1–333 ) were produced as glutathione S-transferase ( GST ) fusion proteins in BL21 strain of E . coli from plasmids pGEX-4T1-Ntev-NF1-333 and pGEX-2T-C-Raf-RBD . All proteins were purified as described previously ( Guzmán et al . , 2014b ) . Fluorescence anisotropy was used to study the affinity between mant-GTPγS and H-ras mutants . Measurements were conducted at room temperature in a buffer containing 25 mM Hepes pH 7 . 2 , 100 mM NaCl , and 5 mM EDTA . The reaction was monitored by the anisotropy of 100 nM mant-GTPγS ( BP 340/30 nm excitation filter and BP 485/20 nm emission filter ) in the presence of H-ras mutants ( from 0 nM to 3000 nM ) using a Synergy H1 hybrid fluorescence plate-reader ( BioTek , Winooski , VT , USA ) equipped with a polarization filter cube . Data processing was done as described using the concentration of H-ras as X and the concentration of mant-nucleotide as Lt ( Guzmán et al . , 2014b ) . Fluorescence anisotropy was used to study the affinity between the Ras binding domain of C-Raf ( RBD ) and H-ras mutants in solution . The experimental part and the data processing were done as described in Guzmán et al . ( 2014b ) . The GEF ( histidine tagged SOScat , amino acids 564–1049 ) mediated Ras nucleotide exchange kinetics and data analysis were performed as described ( Kopra et al . , 2014 ) . In brief , the homogeneous Eu3+-GTP exchange assays were performed using the quenching resonance energy transfer ( QRET ) technique . In association experiments , H-Ras ( 200 nM ) , Eu3+-GTP ( 10 nM ) , Quench II ( 22 μM ) ( QRET Technologies , Turku , Finland ) were incubated for 5 min before Eu3+-GTP association was triggered by SOS ( 1:1 H-Ras:SOS ratio ) addition . In the Eu3+-GTP dissociation assay , the Eu3+-GTP association to H-Ras ( 200 nM ) was first performed in the presence of SOS ( 200 nM ) . After 20 min of incubation the Eu3+-GTP dissociation was induced with 100 μM GTP . All the measurements were performed using a Victor2 1420 multilabel counter ( Perkin Elmer Life and Analytical Sciences , Turku , Finland ) to record time-resolved luminescence data ( emission 615 nm , excitation 340 nm , delay 400 μs and window time 400 µs ) . The kinetic data were analyzed using GraphPad Prism 6 software from GraphPad Software Inc . ( La Jolla , CA , USA ) . GST-RBD pulldown experiments were performed essentially as described ( Gremer et al . , 2011 ) . In brief , BHK cells were transiently transfected with the respective plasmids ( mGFP-tagged H-ras or mCFP-tagged K-ras or mCit-tagged N-ras constructs ) using JetPRIME transfection reagent ( Polyplus-transfection , New York , NY , USA ) . 24 hr after transfection cells were serum starved for 4 hr and half of them stimulated with 100 ng/ml EGF for 5 min . Ras-GTP levels were determined using GST-RBD , the GST-fusion of the Ras binding domain of C-Raf , to pulldown active GTP-bound Ras from cellular lysates by glutathione beads ( Pierce , Thermo Fisher Scientific , Waltham , MA , USA ) . Cells were lysed in lysis buffer ( Tris/HCl pH 7 . 5 , 100 mM NaCl , 2 mM MgCl2 , 1% Nonidet-40 , 10% glycerol , EDTA-free inhibitor cocktail ) . In order to assess GAP-sensitivity of Ras mutants , the cell lysates were incubated in the presence and in the absence of 10 µg purified catalytic domain of NF1 ( NF1-333 ) for 30 min at +4°C and then mixed with glutathione beads coupled to GST-RBD . The samples were washed four times with lysis buffer and subjected to SDS-PAGE ( 15% polyacrylamide ) . Bound Ras proteins were detected by western blotting using monoclonal antibodies against the HA-tag present in Ras constructs ( HAtag antibody , C29F4 , Cell Signaling Technology ) . To compare the subcellular distribution of Ras-proteins in BHK cells , co-expressing mCherry-tagged mutants with their mCFP-tagged parent Ras constructs confocal imaging using Zeiss LSM 780 was done ( 63× , NA 1 . 2 water immersion objective , mCFP excitation at 405 nm and mCherry excitation at 543 nm ) . To determine co-localization , the overlap Manders' coefficient ( R ) was calculated using the ‘Manders' coefficient’ ImageJ plugin developed by Tony Collins , Wayne Rasband and Kevin Baler ( Wright Cell Imaging Facility , Toronto , Canada ) . PC-12 differentiation is a sensitive measure of MAPK-activity ( Cowley et al . , 1994; Vaudry et al . , 2002 ) . PC12 cells were seeded at a density of 104 cells per well of a 4 well Lab-Tek II Chambered Coverglass ( Thermo Fischer Scientific Nunc ) coated with 0 . 1% of rat tail collagen I ( Termo Fischer Scientific Gibco ) in 30% ethanol . After 24 hr , cells were transfected with GFP-tagged Ras constructs or control construct mGFP-tH ( tH representing the minimal membrane anchor of H-ras [Plowman et al . , 2005] ) , using JetPRIME transfection reagent ( Polyplus-transfection ) . GFP-positive cells that exhibited neurites with a length of at least two times the diameter of the cell body were imaged using the confocal microscope Zeiss LSM 510 META , 40× , NA 1 . 4 oil immersion objective , excited at 488 nm . Neurite length was determined using the ImageJ plugin ‘NeuronJ’ ( Meijering et al . , 2004 ) . For each experimental condition at least 25 cells from three biological repeats were analyzed . To generate pooled clones of NIH/3T3 cells stably expressing N-rasG12V , N-rasG12V-C51Y , K-rasG12V , and K-rasG12V-R164Q , cells were transduced with lentiviral particles for expression of HA-tagged Ras-proteins under CMV promoter and GFP-Puromycin marker under RSV promoter ( AMS Biotechnology , Abingdon , UK ) . As a control , cells stably expressing only GFP-Puromycin marker were used . Selection with puromycin ( 1 μg/ml ) started 72 hr after infection . After selection , cells were expanded and used in assays as indicated . Western blotting with anti-HA-tag antibody ( Cell Signalling , C29F4 ) was performed to verify stable and comparable protein expression levels of Ras mutants . In cell proliferation , dose response , focus formation and anchorage-independent growth assays NIH/3T3 cells stably expressing indicated Ras constructs were used . In proliferation assays , 500 cells were seeded per well in a 96-well plate and grown for 72 hr . At intervals of 24 hr , 15 µl AlamarBlue ( Thermo Fisher Scientific Invitrogen ) was added to each well , and after 3 hr of incubation , fluorescence intensity was measured at an excitation of 570 nm and an emission of 590 nm using a Synergy H1 Hybrid Reader ( BioTek ) . The fluorescence values at each timepoint were normalized to the value at 0 hr and plotted . The error bars represent ±SEM . For each experimental condition , data are averages of data points that were acquired in hexaduplicate in each of three biological repeats . In a focus formation assay , 1 , 500 cells were seeded per well of a 6-well plate . After 7 days of growth , cells were fixed with 4% PFA , stained with 0 . 5% crystal violet in 10% ethanol for 15 min and washed with PBS to remove excess stain . The average colony area percentage from three independent biological repeats was calculated using the ‘ColonyArea’ ImageJ plugin ( Guzmán et al . , 2014a ) . In the anchorage-independent colony formation assay , which correlate typically with in vivo tumorigenicity ( Shin et al . , 1975 ) . 50 , 000 cells were resuspended in growth medium containing 0 . 4% agarose ( 4% Agarose Gel , Termo Fisher Scientific Gibco; top layer ) and plated on a bottom layer containing growth medium and 1 . 2% agarose in a 6-well plate . After 10–14 days of growth , cells were fixed in methanol/acetone ( 1:1 ) and washed with PBS . Colonies were imaged using a Zeiss SteREO Lumar V12 stereomicroscope . Analysis was done using the ImageJ software . First , the background was subtracted using the rolling ball function with a radius of 50 μm , then auto-tresholding was applied to separate the colonies . Area percentage was calculated using the ImageJ built-in function ‘Analyze Particles’ with exclusion of particles smaller than 500 μm2 that are not considered colonies . Statistical differences were determined by two-way ANOVA test . For all experimental data , statistical differences were determined using an analysis of variance ( ANOVA ) complemented by Tukey's honest significant difference test ( Tukey's HSD ) . The software R version 2 . 15 . 2 ( R Development Core Team , Vienna , Austria ) was used to perform these analyses . Statistical significance levels are annotated as NS = non significant , that is , p > 0 . 05 , *p < 0 . 05 , **p < 0 . 01 , ***p < 0 . 001 . | Cancer is a disease that develops when cells within the body acquire genetic mutations that allow them to grow and divide rapidly . Many human cancers have mutations in a gene that encodes a protein called Ras , which promotes cell growth and division by controlling the activities of other proteins . Ras congregates at the membrane that surrounds the cell and can assemble into clusters ( called nanoclusters ) that each contain six to eight Ras proteins . The tight packing of the proteins in these nanoclusters increases the amount of Ras in the membrane locally , which allows Ras to interact with other proteins more efficiently to promote growth and cell division . In normal cells , other proteins control when Ras is active . However , in many cancer cells , Ras is active all the time due to mutations that occur in three ‘hotspots’ within its gene . Other mutations in the gene that encodes Ras are also found in cancer cells , but these are less common and it is not clear how they alter the activity of the protein . Here , Solman et al . used microscopy and biochemical techniques to study the effects of some of the less common mutations on Ras activity in human cells . The experiments show that several mutations that alter a region of Ras called the ‘switch III region’ moderately increase the activity of Ras . The mutations probably alter the way that Ras sits in the membrane , which in turn changes the way it interacts with other proteins and the membrane so that more Ras nanoclusters form . Solman et al . 's findings reveal a new way that Ras can be activated in cancer cells . The next challenge is to develop drugs that block the formation of Ras nanoclusters and to find out if they have the potential to be used to treat cancer . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | Specific cancer-associated mutations in the switch III region of Ras increase tumorigenicity by nanocluster augmentation |
CRISPR/Cas9 can be used for precise genetic knock-in of epitope tags into endogenous genes , simplifying experimental analysis of protein function . However , Cas9-assisted epitope tagging in primary mammalian cell cultures is often inefficient and reliant on plasmid-based selection strategies . Here , we demonstrate improved knock-in efficiencies of diverse tags ( V5 , 3XFLAG , Myc , HA ) using co-delivery of Cas9 protein pre-complexed with two-part synthetic modified RNAs ( annealed crRNA:tracrRNA ) and single-stranded oligodeoxynucleotide ( ssODN ) repair templates . Knock-in efficiencies of ~5–30% , were achieved without selection in embryonic stem ( ES ) cells , neural stem ( NS ) cells , and brain-tumor-derived stem cells . Biallelic-tagged clonal lines were readily derived and used to define Olig2 chromatin-bound interacting partners . Using our novel web-based design tool , we established a 96-well format pipeline that enabled V5-tagging of 60 different transcription factors . This efficient , selection-free and scalable epitope tagging pipeline enables systematic surveys of protein expression levels , subcellular localization , and interactors across diverse mammalian stem cells .
Defining the protein levels , subcellular localisation and biochemical interactions for the >20 , 000 protein-coding genes in the mammalian genome remains a formidable challenge . Ideally , these would be explored across diverse primary cells – rather than genetically transformed and corrupted cell lines . Large-scale projects , such as the human protein atlas , have performed systematic characterisation of antibodies against native proteins ( Thul et al . , 2017 ) . However , there are inherent difficulties in discovering , validating and distributing high-quality antibodies that cover all species and key applications ( e . g . immunoblotting , immunoprecipitation , ChIP-Seq and immunocytochemistry ) . A complementary strategy is to use epitope tagging: the fusion of small peptide-coding sequences to a protein of interest ( e . g . 3XFLAG , HA , V5 , and Myc ) ( Jarvik and Telmer , 1998 ) . In contrast to plasmid-driven cDNA overexpression , which creates artificially high-protein levels , the knock-in of small epitope tags to endogenous genes provides physiologically relevant levels . A small set of pre-validated tag-specific antibodies is then used across diverse downstream experimental applications . This has been a key tool in yeast studies , resulting in global characterization of core protein complexes and their extensive interaction networks ( Gavin et al . , 2006; Krogan et al . , 2006 ) . However , to date this approach has not been widely adopted in mammalian cells , primarily due to the poor efficiency of homologous recombination ( HR ) . Repurposed programmable nucleases now provide a potential solution . Formation of a site-specific DNA double-strand break ( DSBs ) , will massively enhance HR-mediated repair . This has become straightforward with the discovery and application of clustered regularly interspaced short palindromic repeats ( CRISPR ) and CRISPR-associated ( Cas ) proteins as designer site-specific nucleases . CRISPR/Cas9 was uncovered as a form of microbial adaptive immunity ( Bhaya et al . , 2011 ) that has been repurposed for genome editing in mammalian cells ( Cong et al . , 2013; Doudna and Charpentier , 2014 ) . Cas9 is an RNA-guided endonuclease that binds complementary DNA via formation of a 20 bp RNA:DNA heteroduplex . Stable binding of Cas9/gRNA complex at the target site leads to activation of nuclease domains and formation of a double-stranded DNA break ( DSB ) ( Jinek et al . , 2014 ) . DSBs are predominantly repaired through the error-prone non-homologous end joining ( NHEJ ) pathway , resulting in insertion or deletion mutations ( indels ) ( Sander and Joung , 2014 ) . Alternatively , at lower frequencies HR-mediated repair will occur . Successful knock-in of tags , therefore , requires co-delivery of three ingredients to the mammalian cells: the Cas9 protein , a guide RNA , and a donor repair template ( i . e . single or double-stranded DNA encoding the tag or reporter with homology arms ) . These CRISPR components are typically delivered via transient plasmid transfection or viral vectors . Bespoke targeting vector plasmids are usually constructed for delivery of large cargoes or conditional alleles . Selection strategies , such as flow cytometry or use of antibiotic resistance cassettes , are then used to enrich for edited cells . Despite the successes of current approaches ( Dalvai et al . , 2015; Savic et al . , 2015; Mikuni et al . , 2016; Xiong et al . , 2017 ) , it is clear that many bottlenecks restrict widespread applications; ( i ) Production of tailored plasmids for each component can be tedious and time-consuming; ( ii ) Plasmid-based selection strategies introduce additional regulatory elements that can disrupt normal regulatory processes; ( iii ) Plasmid DNA , either Cas9/gRNA expression vectors or targeting vectors , can randomly integrate in the genome causing insertional mutagenesis or increasing risks of off-target cleavage ( Kim et al . , 2014; Liang et al . , 2015 ) ; ( iv ) current strategies are often not readily scalable to enable routine exploration of large numbers of genes; ( v ) recovery of biallelic clonal lines is inefficient . Thus , there remains an unmet need for improved knock-in strategies that can work efficiently in primary mammalian cells , such as pluripotent ( ES/iPS ) , multipotent ( tissue stem cells ) and cancer stem cells . Improved efficiencies of CRISPR editing in mammalian cells have been demonstrated using recombinant Cas9 protein ( Kim et al . , 2014; Ramakrishna et al . , 2014; Bressan et al . , 2017 ) . Cas9 protein is complexed with in vitro-transcribed ( IVT ) RNA , to produce a ribonucleoprotein ( RNP ) complex that can be then delivered into cells . The RNPs are short-lived and cleared by cells within 24–48 hr , reducing the risk of both formation of mosaic clones or off-target cleavage ( Kim et al . , 2014; Lin et al . , 2014; Zuris et al . , 2015; Cameron et al . , 2017 ) . However , RNP-assisted methods have mainly been used for gene knock-out ( Kim et al . , 2014; Liang et al . , 2015 ) , incorporation of point mutations ( Ma et al . , 2017; Rivera-Torres et al . , 2017 ) , or knock-in of restriction enzyme sites ( Lin et al . , 2014; Schumann et al . , 2015 ) . We and others have recently made use of RNP with in vitro-transcribed ( IVT ) sgRNAs for CRIPSR epitope tagging ( Bressan et al . , 2017; Liang et al . , 2017 ) , yet the efficiencies in the absence of selection are highly variable , and this approach cannot be scaled . To avoid IVT , chemically modified ~100 nt-long sgRNAs can be synthesised ( Hendel et al . , 2015 ) ; however , these are prohibitively expensive , limiting applications . Alternatively , a two-part , chemically synthesised , short target-specific crRNA plus a longer generic tracrRNA can be used . This is cheaper ( only the crRNA needs to be resynthesised for new targets ) and has better performance ( Aida et al . , 2015; Anderson et al . , 2015 ) . This ‘dual-RNA’ approach has been advanced further by ‘base , backbone and end’ modifications of cr/tracrRNAs ( Rahdar et al . , 2015; Kelley et al . , 2016 ) and also by the use of shorter and more effective modified cr/tracrRNAs ( Jacobi et al . , 2017 ) . Modified synthetic cr/tracrRNAs are resistant to nuclease digestion , limit cellular immune responses , have greater stability , and therefore provide enhanced targeting efficiency ( Jacobi et al . , 2017 ) . Here , we explored whether RNPs with synthetic modified two-part guide RNA can enhance efficiency of epitope tagging in mammalian stem cells . We find that co-delivery of a Cas9 RNP ( dual-synthetic RNA ) with ~200 bp single-stranded oligodeoxynucleotides ( ssODNs ) supports highly efficient epitope tagging . This is achieved across a variety of stem cell types without any requirement for plasmids , selection steps , flow-cytometry-based enrichment , or IVT reactions . To provide a proof-of-principle , we developed a novel web-based design tool and demonstrate effective tagging in 96-well plate format . We demonstrate one application , by identifying Olig2 protein partners using immunoprecipitation-mass spectrometry ( IP-MS ) .
sgRNAs produced by IVT reactions can vary in quality and quantity and are prone to degradation , either during production and/or following delivery into cells . We therefore explored a synthetic modified two-part guide RNA system ( annealed 36-mer crRNA: 67-mer tracrRNA ) ( Anderson et al . , 2015; Jacobi et al . , 2017 ) . Guide RNAs were designed to cut proximal to the stop codon in the 3’ UTR of Olig2 or Sox2 ( Figure 1A ) . The efficacy of custom synthetic modified RNAs ( csRNAs ) was compared to IVT-generated sgRNAs . RNA was complexed with recombinant Cas9 protein and transfected into an adult mouse neural stem ( NS ) cell line ( ANS4 ) , using an optimised nucleofection program . RNP was delivered together with a ~ 200 bp single-stranded DNA donor encoding the V5 tag , flanked with ~70 nucleotide homology arms ( Figure 1B ) . After 5 days , cells were analysed using immunocytochemistry ( ICC ) for the V5 fusion protein ( Figure 1C ) . The csRNA-based RNP ( csRNP ) gave a >4-fold and >10-fold increase in V5 knock-in efficiency for Olig2 and Sox2 , respectively ( Figure 1D ) . Improved knock-in efficiencies were also obtained using an independent cell line ( glioma-initiating neural stem cells; termed GNS ) ( Figure 1D ) . PCR genotyping and Sanger sequencing confirmed in-frame and error-free insertion of V5 tag sequence at the C-terminus of Olig2 and Sox2 loci ( Figure 1—figure supplement 1A ) . V5-positive cells all displayed the anticipated nuclear localisation and levels with no indication of non-specific expression . Knock-in efficiency might vary when using distinct biochemical tags . We therefore tested a variety of widely used alternative tags ( V5 , HA , 3XFLAG , or Myc ) . Each tag varied substantially in size , and consequently homology arm length ( Table 1 ) . Nevertheless , we observed similar rates of knock-in ( 15–21% ) across all four tags for Sox2 ( Figure 1E; Table 1 ) . An independent adult NS cell culture also gave similar results ( 9–15% knock-in efficiency , Figure 1—figure supplement 1B ) . High knock-in efficiencies were not limited to Sox2 and Olig2 . We found that Sox3 , Sox8 , and Sox9 – three Sox family members that are expressed in NS cells – had knock-in efficiencies of 30% , 14% and 26% , respectively ( Figure 1F; Figure 1—figure supplement 1C and Figure 1—figure supplement 2A ) . Furthermore , we could simultaneously knock-in two different tags ( Sox2: HA tag; and Olig2: V5 tag ) in the same cells using a single transfection ( 4% double-positive cells , Figure 1—figure supplement 2B ) . Altogether , these results indicate that use of the modified two-part synthetic cr/tracrRNA system is more effective than IVT for epitope knock-in in mammalian NS and GNS cells . Using csRNP delivery , we achieved 5–30% knock-in efficiency across distinct cell lines for different genes and using different tags . Notably , this was accomplished without the requirement for flow cytometry or plasmid-based selection strategies . Generation of biallelic-tagged clonal lines could be advantageous for downstream applications , as all target protein will be tagged , enabling improved signal-to-noise ratios in assays . Low efficiency of tagging requires screening of hundreds or thousands of clones that need to be screened and genotyped , limiting downstream applications . Heterozygous lines might also harbour indels on the other non-HR untagged allele ( Merkle et al . , 2015; Bressan et al . , 2017 ) , which may cause inappropriate transcriptional/post-transcriptional regulation . The improved knock-in efficiency using the csRNP method encouraged us that recovery of biallelic clonal lines might be straightforward . Clonal NS cell lines were established from bulk tagged populations following single-cell deposition to 96-well plates . Tagged clones were then identified following replica plating and ICC for V5 tag ( Figure 2A ) . Biallelic clones were scored using PCR primer-pairs flanking the tag sequence ( Figure 2B ) and validated using ICC and Sanger sequencing ( Figure 2C; Figure 2—figure supplement 1 ) . Eighty-nine clonal lines were generated from the Sox2-V5 knock-in cells . Thirteen of these were V5-positive by ICC and 11 were correctly targeted as confirmed by PCR ( Table 2 ) . Of these four had successfully integrated the V5 tag sequence at both alleles ( 30% of V5-positive clones ) ( Table 2 ) . High frequencies of bi-allelic knock-in were also obtained for Sox3-V5 ( 62 . 5% ) . We also derived several biallelic knock-in lines from another independent cell line ( IENS , mouse GNS cells ) : Sox2 , Sox3 , Sox8 and Sox9 ( 7% , 26% , 57% , and 15% biallelic knock-in of V5-positive clones , respectively ) ( Table 2 ) . Thus , biallelic-tagged clonal lines can be readily recovered . To test the general applicability of the csRNP tagging method across other types of stem cells , we compared head-to-head tagging efficiencies between mouse ES cells and NS cells . We initially focussed on four transcription factors ( TFs ) : Sox2 , Sox3 , Ctcf and Pou3f1 , and the chromatin regulator Ezh2; each of these is expressed in both cell types . In each case , we found that mouse ES cells ( E14Tg2a ) were tagged at similar level of efficiency to the NS cells ( knock-in efficiency range 6–11% , Figure 3A , B ) . Non-expressed genes are often difficult to engineer . We therefore tested csRNPs for several neural-affiliated TFs ( Sox9 , Pou3f2 , and Pou3f3 ) that are expressed in NS cells but not ES cells ( Figure 3C ) . V5 insertion was first confirmed by PCR genotyping in the bulk populations and suggested both ES cells and NS cells were effectively tagged ( Figure 3—figure supplement 1 ) . V5-tagged protein was detected by ICC only in NS cells and not in ES cells ( Figure 3C , left and middle panels ) . However , for each of these genes , upon neural lineage differentiation of the ES cells , we noted a proportion of the Nestin-expressing neural rosettes that were V5-positive; 10 . 5% , 20% , and 11 . 8% for Sox9 , Pou3f2 , and Pou3f3 , respectively ( Figure 3C , right panel ) . Thus , non-expressed genes can be successfully tagged in ES cells , without deploying any selection strategies or plasmids . We also assessed csRNP-based tagging in human ES cells . V5 knock-in was successfully demonstrated in human ES cells for SOX2 ( Figure 3D ) . We also tested human GBM-derived cells for OLIG2 , SOX2 , and SOX9 genes ( 15–70% efficiency ) ( Figure 3D ) . These data illustrate the power of the csRNP-mediated efficient tagging of non-expressed genes in stem cells , and subsequent monitoring of the tagged protein in their differentiating progeny . We also conclude that the same csRNP epitope tagging approach and reagents can work effectively across diverse mouse and human pluripotent stem cells , neural stem cells , and cancer-derived stem cells . To further define the parameters influencing the reliability and efficiency of tagging , we attempted V5 epitope knock-in at the C-terminus for all 50 Sox and Fox genes . This set of genes included both expressed and non-expressed genes . Previous studies have indicated that the distance of the DSB to the insertion site influences the frequency of successful HR ( Bialk et al . , 2015; Paquet et al . , 2016; Liang et al . , 2017 ) . We designed two different target crRNAs in the 3’UTR of the gene for each of the 50 target genes; one cutting proximal and the other distal to the stop codon . For each of the TF , cells were transfected with Cas9 RNP containing either of the crRNA and a common matched ssODN to assess if distance of the cut site from the stop codon influenced knock-in efficiencies ( Figure 4A ) . By PCR genotyping , we found that both proximal and distal gRNAs could result in successful tagging in the majority of cases; 30/50 genes ( 60% ) for proximal DSB and 27/50 for the distal DSB ( 54% ) ( Figure 4—figure supplements 1 , 2 and 3 ) . However , this assay is qualitative . To quantitatively score the knock-in efficiency , we performed V5 ICC assay for the seven expressed TFs ( Figure 4B , C ) . Sanger sequencing confirmed targeted insertion of the V5 tag-coding sequence ( Figure 4D ) . Importantly , by comparing the efficiency of tagging for these seven genes , we noted a consistent trend towards increased tagging efficiency for the most proximal cut site . For example , Sox3 showed 18% and 5% knock-in efficiency , for proximal and distal gRNAs , respectively ( Figure 4B ) . For four genes ( Sox9 , Foxj3 , Fok1 and Foxk2 ) , the distal gRNA did not work , whereas the proximal gRNA facilitated high knock-in efficiency ( Figure 4C ) . There were no genes for which the more distal gRNA worked better than the proximal gRNA . These results suggest proximity of DSB to the stop codon influences the efficiency of knock-in . It is often desirable to explore large numbers of proteins within a shared family , complex or pathway . Methods enabling knock-in of many genes in parallel would be valuable . The gene-specific synthetic short crRNAs and matched ssODN repair templates can be obtained from commercial suppliers in 96-well microplates . Indeed , all steps can be performed easily in 96-well format: preparation of the transfection-ready components via automated liquid handling , benchtop incubation/annealing , 96-well transfection , and automated microscopy to acquire images across 96-well plates . We reasoned that scale-up could therefore be relatively straightforward . A major remaining bottleneck , however , is the need for bioinformatics design tools specifically tailored towards epitope tagging applications . Manually extracting gene sequence data , identifying appropriate gRNAs , and design of modified repair ssODNs , would be laborious and error-prone for hundreds of genes . To automate the batch design of crRNAs and ssODNs , we developed ‘Tag-IN’ , a novel web-based tool ( Figure 5 , http://tagin . stembio . org ) . This enables design of appropriate gRNA and ssODNs for both human and mouse species , with flexibility in choice of tag . Our ‘Tag-IN’ incorporates key design rule and parameters – e . g . incorporating ‘Rule Set 2’ for maximised activity ( Doench et al . , 2016 ) , and ‘MIT’ scoring model ( Hsu et al . , 2013 ) to minimise off-target effects . Our tool also considers optimal distance from the insertion site ( stop codon ) , and outputs the matched ssODNs modified with PAM-blocking mutations and appropriate epitope tag sequence . ‘Tag-IN’ also enables batch design; critical for the effective scale up to 96-well format ( Figure 5 ) . Using the Tag-IN tool , we designed crRNAs and matched V5 encoding repair ssODNs against 185 different transcription factors . Each of these crRNAs/ssODNs were also manually verified to ascertain that the tool picked proximal-cutting crRNA and the ssODNs contained appropriate PAM-blocking mutations and homology arms . These mouse genes were selected based on expression in human glioblastoma stem cells . One gRNA was tested for each gene . The RNPs were prepared using a 96-well head liquid handling device ( Felix , CyBio ) and then transfected in parallel into mouse GNS cells ( Figure 5 ) . Five days later , ICC was performed and V5 knock-in efficiencies were quantified across the entire plate using automated plate image capture ( Operetta high-content imaging system , Perkin Elmer ) . Sox2 tagging was used as a positive control in six wells; these gave consistent V5 knock-in efficiency across the plate , confirming no ‘edge-effects’ during the procedure ( 10 . 5 ± 2 . 5% ) ( Figure 6A ) . Remarkably , for the first 96-well plate , 30 out of 90 TFs were positive for V5 ICC with typical knock-in efficiencies ranging from 6% to29% ( Figure 6A , B ) . A second 96-well plate with distinct TFs showed similar knock-in efficiency ( 5%–38% ) , with 31 out of 95 TFs positive for V5 ICC ( Figure 6C , D ) . These are similar efficiencies to those observed in our earlier single transfections ( Figures 1 and 2 ) . V5 ICC confirmed the expected nuclear localisation of these TFs ( Figure 6B ) . Thus , ~30% of genes were successfully tagged at the first attempt with good knock-in efficiency ( Figure 6E ) . Interestingly , two of the V5-positive TFs- Cbfb and Ybx1 , showed nucleocytoplasmic localisation ( Figure 6B , D ) . This illustrates the valuable information regarding protein localisation and levels data that can be quickly obtained . The frequency of successful genes tagged as scored by V5 ICC in these experiments is likely to be an underestimate , as ~20% of the failed TFs are low or non-expressed NS lines ( Pollard lab , unpublished data ) . These results clearly demonstrate the ease with which our method can be scaled to 96-well format epitope tagging . As a proof-of-principle of the applications , we performed V5-immunoprecipitation followed by mass spectrometry ( IP-MS ) , to identify interaction partners of Olig2 in mouse GNS cells using the RIME ( Mohammed et al . , 2016 ) and ChIP-SICAP ( Rafiee et al . , 2016 ) methods . These enable identification of chromatin-bound protein partners – the latter being more stringent for chromatin-bound proteins . For each assay , we used V5 monoclonal antibody conjugated to magnetic beads . RIME analysis showed high enrichment of the bait protein Olig2 ( Figure 7A ) and subunits of SWI/SNF complexes and histone deacetylases ( HDACs ) in the pull-down complexes ( Figure 7—source data 1 ) . Physical interaction of Olig2 and SWI/SNF complex has been previously reported and this interaction is essential for oligodendrocyte differentiation ( Yu et al . , 2013 ) . HDACs have a known functional role in Olig2 function during development . ChIP-SICAP analysis showed strong enrichment of Olig2 and core histone proteins suggesting specific pull-down of chromatin fragments bound by Olig2 TF ( Figure 7B ) . Noteworthy , two other oligodendrocyte lineage transcription factors , Olig1 and Olig3 , were detected among the most enriched proteins . Earlier studies reported that sets of genes regulated by various Olig proteins have a partial overlap ( Ligon et al . , 2007; Meijer et al . , 2012 ) , explaining co-occupation of the same DNA-sites as shown by these data . Furthermore , we detected two other members of the basic helix-loop-helix ( bHLH ) family: Npas3 and Tcf4 . The bHLH transcription factors are known to form heterodimers with other bHLH proteins on chromatin ( Massari and Murre , 2000 ) and the presence of Npas3 and Tcf4 may be explained by their direct physical interaction ( Figure 7B; Figure 7—source data 1 ) . Both Npas3 and Tcf4 have been reported to be involved in CNS development ( Shin and Kim , 2013; Chen et al . , 2016 ) . Interestingly , a recent study reporting interactions of Olig2 , Tcf4 and Npas3 in mouse neural stem cells by FLAG-affinity purification ( Moen et al . , 2017 ) allowed us to correlate interactomes of all the three bHLH proteins with our data . Taken together , our results indicate that Tcf4 and Npas3 co-localise with Olig2 on chromatin , suggesting a functional interaction . In addition to protein interactions of Olig2 on-chromatin , we also analysed the flow-through of the streptavidin enrichment , representing interactions with soluble Olig2 ( Supplementary file 3 ) . As expected , Olig2 , Olig1 , and Olig3 were found among the most highly abundant proteins , indicating high specificity of the immunoprecipitation . Furthermore , composition of identified proteins correlates with previous findings: we detected three known interactors of Olig2 ( Cul3 , Smarca4 and Sox8 ( BioGRID , Intact ) ) and 55 other proteins reported to co-precipitate with Olig2 ( Moen et al . , 2017 ) ( Supplementary file 3 ) . These included many SWI/SNF family members , and chromatin regulators Cbx3 and Chd4 , consistent with RIME . Collectively , our data confirm that V5-tagging can be effectively combined with ChIP-SICAP to identify proteins that co-localise on chromatin or that interact off-chromatin . Insertion of fluorescent protein-encoding sequences ( such as GFP , mCherry ) in frame within gene coding regions enables monitoring of eukaryotic protein localisation in live cells . Additionally , fusion proteins can be used for pull-down assays ( immunoprecipitation and ChIP ) . Encouraged by the facile deployment and efficiencies of the Cas9 RNP method for small epitope tagging , we next asked if the same approach could be used for knock-in of relatively large mCherry-encoding DNA sequence ( ~700 bp ) in NS cells . First , we tested the effect of variable length of homology arms on knock-in efficiency ( Figure 8A ) . For these experiments , double-stranded linear DNA fragments harbouring variable homology arms were used as donor DNA templates ( PCR-amplified from a previously reported promoterless Sox2-mCherry plasmid ) ( Bressan et al . ) . We found that dsDNA with larger homology arms were more effective in mCherry knock-in ( Figure 8B ) , although homology arms as small as 100 bp showed ~0 . 29% cells positive for mCherry in the bulk populations . We next tested mCherry knock-in at Olig2 and Foxg1 loci and achieved 0 . 34% and 0 . 11% efficiency , respectively ( Figure 8C ) . Live-imaging of mCherry-sorted populations for the three genes ( Sox2 , Olig2 , and Foxg1 ) showed the expected localisation and levels of fusion proteins in NS cells ( Figure 8D ) . Together , these results show that our csRNP method can be used for knock-in of fluorescent reporters in NS cells .
An overarching goal in biology is to determine the key functions of each protein encoded in the genome . Epitope tagging of endogenous genes using CRISPR-assisted knock-in provides new opportunities to interrogate protein function , expression , subcellular localisation and interacting partners . In this study , we demonstrated simple and efficient epitope tag knock-in across a large number of genes in mouse and human stem cells . Use of Cas9 recombinant protein is critical . We found that when combined with synthetic modified RNAs significant enhancement in the efficiency is possible . This was achieved using faster and more reliable method – compared to plasmid-based strategies . Importantly , we showed primary mammalian stem cell lines are readily amenable to RNP engineering , without the need for plasmid production , or prior genetic manipulation of the host cell lines . Our method is therefore versatile enough to be implemented by any laboratory in their existing cell lines . All reagents can now be obtained ‘off the shelf’ . Assembly of RNPs with csRNAs is simple and requires only ~30 min hands-on-time; this compared favourably to IVT reactions , which require multiple steps , are time-consuming , and result in variable quality of gRNAs . In our hands , these reagents are stable at −20o C for many months , and is particularly convenient when re-used across a range of cell types . Perhaps more importantly , use of the custom synthetic cr/tracrRNAs with their modified backbone and protected ends , shields the RNA from cellular RNases . We found greatly reduced toxicity – one of the key advantages of the csRNP over IVT sgRNA and plasmids-based delivery methods . We routinely generated clonal knock-in lines in a time-frame of 4–6 weeks . Consistent with a previous report ( Liang et al . , 2015 ) , we found that RNP complexes target both copies of the genes at high frequency , enabling facile isolation of bi-allelic knock-in clones . Although monoallelic knock-in clones are sufficient for pull-down assays , bi-allelic clones are preferred to have more confidence in the interpretation of downstream assays . While our focus has been on neural stem cells and their malignant counterparts ( glioma stem cells ) , these same methods and reagents can work well in other stem cells , such as ES cells . Indeed , our knock-in data using ES cells revealed that non-expressed genes can be efficiently tagged using the same protocol . Under appropriate differentiation cues , induction of these proteins can be monitored using ICC analysis in the ES cell differentiating progeny . We believe this could be a key application – rapidly enabling assessment of proteins across a range of lineage contexts . Despite the excitement , there are also some caveats . Foremost , inevitably there is a risk that C-terminus tagging can potentially compromise protein function , localisation or levels in the edited cells . Not all proteins will be amenable to tagging . If the protein of interest harbours critical C terminal domains , an N-terminus approach could be pursued – or knock-in to other regions such as structurally neutral linkers , if this is known . In this study , we used PAM-strand ( non-complementary strand ) for donor DNA synthesis , which is less likely to be cut by Cas9 complex ( Gasiunas et al . , 2012 ) . We have not compared PAM versus non-PAM donor DNA strands for knock-in efficiency although a recent report suggests that for symmetrical PAM ssODNs , both strands are equally effective for knock-in ( Liang et al . , 2017 ) . Previous findings have reported a higher HDR rate with the guide RNAs that cut near the insertion site ( Bialk et al . , 2015; Paquet et al . , 2016; Liang et al . , 2017 ) . Our results corroborate these findings and reveal that cut-to-insertion-site distance is a critical factor for successful knock-in . Lower knock-in efficiency achieved with the distal-cutting gRNAs could be attributed to ‘partial HDR’ events ( Guo et al . , 2018 ) , which are not detectable in our ICC end-point assays . Although beyond the scope of our present study , whole-genome sequencing of the CRISPR-edited bulk cells can shed light on the extent of ‘partial HDR’ events in epitope knock-in experiments . The average size of CRISPR/Cas9-induced indels in mammalian cells has been reported to be around 1–5 bp ( Paquet et al . , 2016 ) . We therefore recommend designing guide RNAs in the 3’UTR , preferably cutting 8–15 bp downstream of stop codon , to avoid interfering with the stop codon . The ability to scale-up tagging to 96-well format required us to tackle the bottleneck in design . Our newly developed ‘Tag-IN’ tool simplifies gene sequence retrieval and designing crRNA and ssODN for medium-throughput experiments and is available from any web browser ( http://tagin . stembio . org ) . This can be modified for use with alternative design parameters and distinct tags . We demonstrated one of the key applications by single-cell imaging of the levels and localisation for 60 different transcription factors in glioma stem cells . These were successfully tagged without any screening of gRNAs or ssODN; those that failed could be due to the gRNA not working effectively or the ssODN donor being sub-optimal . These could be re-tested with replacements – particularly for the ssODN , other tag can be used that works effectively in some instances ( data not shown ) . Alternatively , the protein maybe expressed at levels too low to detect by ICC , but the tag insertion might still be detected by PCR genotyping . This is likely the case for 10–20% of the genes we explored . The pipeline that emerges is fully scalable and does not require sophisticated tools , expertise , or know-how . There are no significant bottlenecks in either the design , acquisition/production of reagents or delivery into cells . It was possible to generate such data with relatively little labour investment , and the whole pipeline from obtaining the reagents to imaging data was accomplished within three weeks . Increasing the throughput further , we can envision systematic surveys of many hundreds or thousands of genes . Because tag-specific antibodies work universally across different cell types and species , this approach should allow cross-species comparisons and will also complement existing efforts to develop and characterise native antibodies . Furthermore , tagging of DNA-binding proteins will allow us to generate comprehensive genome-wide binding sites via ChIP-Seq for key transcription factors , chromatin-modifying enzymes , and other gene regulatory proteins . Although not a focus of the current study , we also find that csRNP is highly effective for gene knockout studies ( Pollard laboratory , unpublished and ( Jacobi et al . , 2017 ) . We demonstrated proof-of-principle in identification of Olig2 protein partners . This is likely a key application for future studies; simplifying and scaling up the ability to probe protein complexes in a range of cell types and cell states . Using the Olig2-V5 knockin lines we demonstrated on-chromatin partners in mouse glioma cells using optimised RIME and ChIP-SICAP methods , validating previously reported protein interaction partners . These optimised V5 RIME and ChIP-SICAP methods can now be deployed using the same optimised protocol for other mammalian proteins , particular for those with no good quality antibodies – a key advantage of epitope tagging . With the remarkable developments in single-cell mRNA profiling , and plans for systematic RNA and protein atlases of mammalian cell types ( Rozenblatt-Rosen et al . , 2017 ) , there is a greater need than ever to relieve the bottleneck of exploring protein products . Descriptive maps of cell types , while valuable , must be complemented by careful and detail molecular and cellular functional studies . Our findings suggest that epitope using CRISPR/Cas-assisted knock-in is now simple and efficient enough that systematic annotation of many hundreds or thousands of endogenous will be possible in mouse and human stem and progenitor cells .
Mouse and human NS and GNS cell lines were cultured essentially as described previously ( Conti et al . , 2005; Pollard et al . , 2006 ) . Laminin was purchased from Cultrex , R and D Systems . ANS4 and BL6 NS cell lines have been described previously ( Bressan et al . , 2017 ) . Mouse ES ( Tg2a ) cells were cultured in GMEM supplemented with 10% fetal calf serum , 1x non-essential amino acids , 1x glutamine/sodium pyruvate , 1xLIF , 1x Pen/strep and 100 μM of ß-mercaptoethanol . Media was changed every day and cells passaged approximately every other day onto plates pre-coated with 0 . 1% gelatin . Differentiation of ES cell was performed as described previously ( Pollard et al . , 2006 ) with 1 × 104 cells per cm2 being seeded in N2B27 complete media for 7 days , with media being changed every 1–2 days . MasterShef7 ( MS7 ) human embryonic stem cells ( hESCs ) were re-cultured in Essential 8 ( E8 ) medium ( Gibco , A1517001 ) on tissue culture plastic coated with Human Recombinant Laminin-521 ( BioLamina , LN521 ) at 5 μg/ml . Routine passaging was performed by incubating cells for 5 min at 37°C in 0 . 5 mM EDTA in PBS . Single-cell dissociation prior to nucleofection was performed by incubating cells for 10 min at 37°C in accutase . Y-27632 ( Cell Guidance Systems , SM02 ) was included in the culture medium at 10 μM following initial thawing and after plating following nucleofection . New engineered cell reporters described here are available upon request . No standard cell lines were used . We used primary stem cell lines . Human ES cells were provided by the UK stem cell bank and had appropriate contamination testing and authentication . Clonal cell lines were derived from the bulk populations using either single-cell deposition to 96-well plates or by manual colony picking . Single cells were deposited into 96-well plates using BD FACSAria II cell sorter . Depending on the cell lines , we obtained 30–40 colonies per 96-well plate in 2 weeks . For manual colony picking , mouse cells were seeded at clonal density ( 400 cells per 10 cm dish for NS cells , 100 cells per dish for GNS cells ) to a 10 cm dish and incubated in the complete media for 10–12 days for colony formation . From each dish , we picked 25–30 manually with a P20 pipette . Colonies from both methods were later replica plated into 96-well plates and analysed for successful knock-in using immunocytochemistry against the V5 tag . The V5-positive clones were further expanded for DNA extraction ( PCR genotyping ) and cryostorage . For manual design: the 3’UTR sequence and 500 bp sequence upstream of the 3’UTR were retrieved using Biomart tool . The final coding exon and 3’UTR features were manually annotated using SnapGene and the ~200 bp around the stop codon were used as an input for guide RNA designing . We designed guide RNAs using either the web-based tool form Desktop Genetics ( https://www . deskgen . com/landing/ ) or our own bioinformatics ‘Tag-IN’ tool ( below ) . High scoring guide RNAs were picked for synthesis ( i . e those with cut sites in the 3’UTR , preferably within 8–15 bp distance from the stop codon and minimal predicted off-target cleavage ) . For ssODN design , first the PAM-blocking mutations ( NGG >NGC or NGT ) were introduced into the SnapGene sequence and then the epitope-tag coding sequence was inserted before the stop codon . The <200 mer ssODN ultramer was chosen to be the same strand as the guide RNA ( also referred to as the PAM-strand , non-complementary strand or non-targeting strand ) and is comprised of: a 5’ homology arm ( ~70 mer ) , the epitope tag coding sequence , stop codon , and a 3’ homology arm with the PAM-blocking mutations ( ~70 mer ) . For some of the ultramers the PAM-strand synthesis had failed and , therefore , the complementary strand ( non-PAM strand ) was synthesised as a donor DNA . Custom synthetic crRNAs , tracrRNA , and ssODNs were manufactured by Integrated DNA Technologies , USA . The RNA backbone and ends were chemically modified for protection against cellular RNases . The 36-mer crRNA contains a variable gene-specific 20-nt target sequence followed by 16-nt sequence that base-pairs with the tracrRNA . The 67-mer tracrRNA contains the gRNA-scaffold sequence as well as 16-nt sequence complementary to crRNA . The lyophilised crRNA and tracrRNA pellets were resuspended in Duplex buffer ( IDT ) at 100 μM concentration and stored in small aliquots at −80°C . ssODN donor DNAs lyophilised pellets were supplied without modifications and resuspended in Duplex buffer ( IDT ) at 30 μM concentration . DNA template for T7-driven synthesis was prepared by annealing 119-mer , single-stranded , complementary ultramers ( from IDT ) encoding T7 promoter , guide RNA , and gRNA scaffold sequences . 200 ng of the template were used to synthesise sgRNA with the MEGAscript T7 Transcription Kit . The sgRNA was further purified using MEGAclear Transcription Clean-Up Kit and stored at −80°C . BL21 ( DE3 ) cells ( New England Biolabs , C2527 ) were transformed with the plasmid pET28a/Cas9-Cys ( Addgene , Cambridge , USA , plasmid #53261 ) using standard protocols . Cas9 protein expression was induced with 0 . 5 mM IPTG ( Isopropyl β-D-1-thiogalactopyranoside ) ( Fisher , 10715114 ) and the cells were incubated overnight at 20°C . 24 hr later , bacterial pellets were resuspended in 20 ml of lysis buffer ( 20 mM Tris-HCl pH 8 . 0 , 500 mM NaCl , 1 mM TCEP , 5 mM imidazole pH 8 . 0 ) , sonicated and loaded on a HisTrap HP 5 ml column ( GE , 17-5248-01 ) . The Cas9 protein was collected in elution buffer ( 20 mM Tris-HCl pH 8 . 0 , 250 mM NaCl , 10% glycerol , 1 mM TCEP , 250 mM imidazole pH 8 . 0 ) . The fractions containing Cas9 protein were pooled and loaded into a HiPrep 26/10 Desalting Column ( GE , 28-4026-52 ) to equilibrate in Cas9 buffer ( 20 mM HEPES-KOH pH 7 . 5 , 150 mM KCl , 1 mM TCEP ) . The purified Cas9 protein was further concentrated using Vivaspin columns ( Vivaspin20 , 30 000 MWCO PES , Sartorius stedim , VS2021 ) as per the users-guide instructions . Synthetic Alt-R CRISPR/Cas9 crRNAs and tracrRNA were supplied by IDT . We prepared Cas9 RNP complexes immediately before electroporation experiments ( a detailed protocol in a separate Appendix 1 is available ) . Cas9 RNPs with IVT sgRNA were assembled ( 1–3 μg of IVT sgRNA with 5–10 μg of Cas9 protein ) as described previously ( Bressan et al . , 2017 ) . For csRNP preparation , 100 picomoles of each crRNA and tracrRNA were annealed using gradual step-down cooling in the PCR block ( 5 min at 95°C , step cool-down from 95°C to 25°C at ramp rate 0 . 1°C/s , 4°C ( store ) at ramp rate 0 . 5°C/s ) . Ribonucleoprotein ( RNP ) complexes were assembled by adding 10 μg of recombinant Cas9 protein to the annealed cr/tracrRNAs , incubated at room temperature for 10 min and stored on ice until electroporation into cells . 30 picomoles of single-stranded donor DNA were added to RNP complexes just before electroporation to prepare the complete RNP mix . For mCherry knock-in , csRNPs were prepared similarly and 200 ng of PCR products were used as donor DNA templates per reaction . For multiplex epitope tagging , 100 picomoles of cr/tracrRNA of each Sox2 and Olig2 were mixed together with 20 μg of rCas9 protein . We used 4D Amaxa nucleofection system for the delivery of CRISPR ingredients . For NS cells and GNS cells , approximately 1 . 5 × 105 cells were resuspended in 20 μL of Lonza SG cell line buffer and were mixed with the complete RNP mix and electroporated using the DN-100 program ( two consecutive pulses for mouse NS cells ) or using EN-138 program ( one pulse for human GBM-derived cells ) . For embryonic stem cells , approximately 6 × 104 cells in 20 μL of Lonza P3 primary cell buffer were used for each transfection with different programs: one pulse of program CA-120 for mouse ESCs; program CB-150 for human ESCs . After the electroporation , cells were transferred into a 6-well plate and allowed to recover for 3–5 days and later seeded into 96-well plates ( 1–2 × 104 cells per well ) for ICC . For scale up , RNP assembly and delivery were performed as above , except that RNP complexes were prepared a day before and stored at −20°C . Electroporation was performed using the 96-well Shuttle device ( Amaxa , Lonza ) . Immediately after transfection cells were transferred into a 96-deep-well plate and replica plates for immunocytochemistry assay were prepared by dispensing 1 × 104 cells into 96-well plate using CyBi-FeliX Liquid Handling Platform . We performed ICC on 96-well plates 5 days after transfection . Cells were washed once with PBS and fixed using 4% paraformaldehyde for 10 min at room temperature and then permeabilised in PBST ( PBS + 0 . 1% Triton X-100 ) for 20 min . Samples were incubated with blocking solution ( 1% goat serum in PBST ) for 30 min at room temperature to block non-specific binding of the antibodies . Samples were treated overnight with primary antibodies in blocking solution followed by incubation with appropriate secondary antibodies and 4′ , 6-diamidino-2-phenylindole ( DAPI ) . Images were acquired using either a Nikon wide-field fluorescence microscope or a PerkinElmer Operetta high-content imaging system . V5-positive cells were scored using Fiji software . The following primary antibodies were used: V5 tag ( eBioscience , TCM5 #14-6796-82 , 1:1000 ) ; HA Tag ( Cell Signalling , 6E2 #2367 , 1:100 ) ; FLAG tag ( Sigma-Aldrich , #F3165 , 1:2000 ) ; Myc tag ( Cell Signalling , 9B11 #2276 , 1:4000 ) , Alexa Fluor secondary antibodies mostly Alexa Fluor Plus 647 ( Thermo Fisher Scientific , 1:1000 ) . HCS CellMask Green Stain ( Thermo Fisher Scientific , #H32714 ) for nucleocytoplasmic staining was used at 1: 10 , 000 for 20 min at room temperature . Genomic DNA was extracted either using in-house lysis buffer as described previously ( Bressan et al . , 2017 ) ( bulk populations from 96-well plate or using DNeasy Blood and Tissue Kit ( Qiagen , # 69506 , for DNA extraction from clonal lines in a 24-well plate ) . PCR primers flanking the V5 tag were designed online using Primer3Plus to generate 400–600 bp PCR amplicons . PCR genotyping and Sanger sequencing were done as described previously ( Bressan et al . , 2017 ) . DNA samples were analysed using 2 . 5% agarose gels . Approximately 40 million mouse NS cells from three T150 flasks ( 150 cm2 ) were cultured until 70–80% confluence and then dissociated into single-cells using accutase . To fix DNA-protein and protein-protein interactions , the cell pellet was resuspended in 1 . 5% methanol-free formaldehyde ( Pierce ) in 10 mL PBS for 10 min at room temperature . Excess formaldehyde was quenched by adding 125 mM glycine and incubated for 5 min at room temperature . Cells were washed twice with cold PBS and stored at −80°C until further use . ChIP-SICAP experiments were performed as described previously ( Rafiee et al . , 2016 ) . Briefly , chromatin from 40 million formaldehyde-fixed cells was sheared by sonication ( Bioruptor Pico , Diagenode ) down to 150–500 bp fragments , which were used as input for immunoprecipitation with anti V5 antibody ( Abcam , 15828 ) overnight at 4°C . Antibody was captured with Protein-G beads ( LIFE technologies , 10004D ) , the associated DNA was biotinylated by terminal deoxynucleotidyl transferase ( Thermo Fisher , EP0162 ) in the presence of biotin-11-dCTP ( Jena Bioscience , NU-809-BIOX ) . The antibody was eluted from the beads in 7 . 5% SDS with 200 mM DTT and the released DNA-protein complexes were caputred by streptavidin magnetic beads ( NEB , S1420 ) . After subsequent washes with SDS washing buffer ( Tris-CL 1 mM , 1% SDS , 200 mM NaCl , 1 mM EDTA ) , 20% isopropanol and 40% acetonitrile , the beads were boiled in 0 . 1% SDS in 50 mM ammonium bicarbonate and 10 mM DTT at 95°C for 20 min . Proteins were digested overnight with trypsin at 37°C and the resulting peptides were purified with the SP3 protocol as described previously ( Hughes et al . , 2014 ) and analysed using an Orbitrap Fusion LC-MS system . The implementation of our crRNA/ssODN design tool was completed in four stages: extraction of a target genomic sequence from GRCh38 . p5 or GRCm38 . p4 genome builds , retrieval of crRNA sequences matching the pattern N20NGG , scoring and ranking of each crRNA using the ‘Rule Set 2’ ( Doench et al . , 2016 ) and ‘MIT’ scoring models ( Hsu et al . , 2013 ) , and design of each corresponding ssODN sequence . To accommodate genomic sequence extraction , an SQL database of genomic coordinates was built using the Genomic Features package in R . This SQL database was used to retrieve coding DNA sequence ( CDS ) ranges upon user query with a desired Ensembl transcript Id . Given a CDS range , a genomic sequence was then extracted from a corresponding GRCh38 . p5 or GRCm38 . p4 Fasta file . For each target genomic sequence , crRNAs were extracted limited to the pattern N20NGG . crRNAs were then ranked using two scoring models , ‘Rule Set 2’ for assessing crRNA efficiency , and the MIT scoring system for crRNA specificity ( Hsu et al . , 2013; Doench et al . , 2016 ) . The former was utilised as a standalone script retrieved from the ‘sgRNA Designer’ website ( https://portals . broadinstitute . org/gpp/public/analysis-tools/sgrna-design ) and the latter was implemented as documented on the ‘CRISPR Design’ website ( http://crispr . mit . edu/about ) . Off targets for each crRNA were found using the short read aligner tool Bowtie , searching up to three mismatches ( Langmead et al . , 2009 ) . Off targets that then match the PAM pattern NAG and NGG were extracted from the Bowtie output . In addition , crRNAs that cut close to the stop codon ( 8–15 bp in the 3’UTR ) , and within the UTR region , were prioritised . Given a batch request , the top two ranking crRNAs were selected for output . A maximum distance of 30 bp from the stop codon was chosen as an additional threshold for batch processing . To implement ssODN design , user-defined tags were inserted immediately 5I proximal to the stop codon . PAM sequences were changed to minimise potential for Cas9 cleavage of donor sequences . Where the PAM sequence resided in the 3I UTR , our tool modified the NGG PAM to NGC . Intronic or exonic PAM changes instead aimed to produce silent mutations , and where this was not possible , aimed to reduce alterations in function by minimising differences in hydrophobicity and charge . Final ssODN sequences were limited to 200-mer including the tag sequence . We therefore present the ‘Tag-IN’ design tool , a novel crRNA and ssODN design tool aimed at streamlining CRISPR knock-in experimentation ( http://tagin . stembio . org ) . The Olig2 protein interactors were identified using the Rapid immunoprecipitation mass spectrometry of endogenous protein ( RIME ) protocol . The nuclear fraction was resuspended using 1 ml of LB1 ( 50 mM HEPES KOH pH 7 . 5 , 140 mM NaCl , 1 mM EDTA , 10% glycerol , and 0 . 5% NP-40 ) with protease and phosphatase inhibitors . Lysate was cleared by centrifugation at 2000 g for 5 min at 4°C , and the pellet was resuspended with 1 ml of LB2 ( 10 mM Tris-HCL [pH 8 . 0] , 200 mM NaCl , 1 mM EDTA , and 0 . 5 mM EGTA ) , and mixed at 4°C for 5 min . Lysate was cleared by centrifugation and the pellet was resuspended in 0 . 5 ml of LB3 ( 10 mM Tris-HCl [pH 8] , 100 mM NaCl , 1 mM EDTA , 0 . 5 mM EGTA , 0 . 1% Na-deoxycholate , and 0 . 5% N-lauroylsarcosine ) . Samples were sonicated in a waterbath sonicator ( Diagenode Bioruptor ) and cleared by centrifugation . 10 µL V5 trap magnetic beads ( MBL ) were used per sample . IPs , washes and on-bead digests were performed using a Thermo Kingfisher Duo , all steps are at 5°C unless otherwise stated . Beads were transferred into 500 μL of cleared lysate and incubated for 2 hr with mixing . Beads were then transferred for two washes in RIPA buffer and three washes in non-detergent lysis buffer . On-bead digest was performed by transferring the washed beads into 100 μL 2M urea , 100 mM Tris-HCl pH 7 . 5 , 1 mM DTT containing 0 . 3 μg trypsin ( Promega ) per sample , beads were incubated at 27°C for 30 min with mixing to achieve limited proteolysis . The beads were then removed and tryptic digest of the released peptides was allowed to continue for 9 hr at 37°C . Reduced cysteine residues were alkylated by adding iodoacetamide solution to a final concentration of 50 mM and incubated 30 min at room temperature , in the dark . Trypsin activity was inhibited by acidification of samples to a concentration of 1% TFA . Samples were desalted on a C18 Stage tip and eluates were analysed by HPLC coupled to a Q-Exactive mass spectrometer as described previously ( Turriziani et al . , 2014 ) . Peptides and proteins were identified and quantified with the MaxQuant software package ( 1 . 5 . 3 . 8 ) , and label-free quantification was performed by MaxLFQ ( Cox et al . , 2014 ) . The search included variable modifications for oxidation of methionine , protein N-terminal acetylation , and carbamidomethylation as fixed modification . The FDR , determined by searching a reverse database , was set at 0 . 01 for both peptides and proteins . | Genes are often referred to as the blueprints of life . Understanding the role of the genes in human cells is one of the major goals of biology . Recent advances in gene editing technologies , such as CRISPR/Cas9 , mean scientists can now edit or delete precise sections within human genes , similar to how we edit words in a document on a computer . This has made it possible to insert small sequences that encode specific “tags” into genes . This in turn means that when a protein is built following the instructions in the gene , the protein includes the tag too , making it easy to monitor . Tags on proteins can help scientists understand what those proteins do by answering various questions , such as: where is the protein found in the cell ? How much of the protein is there in each cell ? Does this change as the cell matures ? What does the protein interact with ? Yet , more research could be done if the tagging process was made easier , quicker and more efficient . Dewari et al . have now come up with an improved gene editing approach that enabled them to rapidly tag hundreds of proteins all at the same time , with efficiencies that were much higher than expected based on previous approaches . The strategy uses common “off the shelf” reagents that can be designed with a new user-friendly , web-based tool called “Tag-IN” . Dewari et al . focused on optimizing their method in freshly grown stem cells , originally collected from mice and humans . They then went on to show the scalability and efficiency of this approach by tagging 60 different proteins in brain stem cells from mice . Now , rather than being limited to a handful of genes of interest , scientists can explore large families of genes in a variety of mouse and human cells in a much quicker and more comprehensive manner . Also , working with stem cells that can be freshly collected from individuals rather than cells that have been grown in the laboratory for a long time will be more useful for biological and disease studies . In the long-term , more knowledge of how protein-coding genes work in different human cells will benefit patients as new drugs or therapeutic targets are discovered . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology",
"tools",
"and",
"resources"
] | 2018 | An efficient and scalable pipeline for epitope tagging in mammalian stem cells using Cas9 ribonucleoprotein |
A key point to regulate gene expression is at transcription initiation , and activators play a major role . CarD , an essential activator in Mycobacterium tuberculosis , is found in many bacteria , including Thermus species , but absent in Escherichia coli . To delineate the molecular mechanism of CarD , we determined crystal structures of Thermus transcription initiation complexes containing CarD . The structures show CarD interacts with the unique DNA topology presented by the upstream double-stranded/single-stranded DNA junction of the transcription bubble . We confirm that our structures correspond to functional activation complexes , and extend our understanding of the role of a conserved CarD Trp residue that serves as a minor groove wedge , preventing collapse of the transcription bubble to stabilize the transcription initiation complex . Unlike E . coli RNAP , many bacterial RNAPs form unstable promoter complexes , explaining the need for CarD .
Decades of research using Escherichia coli ( Eco ) as a model system inform most of our understanding of how bacteria control transcription initiation . First , dissociable promoter specificity subunits , σ factors , direct the catalytic core of the RNA polymerase ( RNAP ) to promoter DNA sites and play a key role in unwinding the DNA duplex to create the transcription bubble in the RNAP holoenzyme open promoter complex ( RPo ) ( Feklistov et al . , 2014 ) . Second , DNA-binding transcription factors either activate or repress the initiation rate ( Browning and Busby , 2004 ) . The majority of transcription activators characterized to date are dimeric proteins that bind operators upstream of the promoter −35 element and directly contact the RNAP α subunit ( Ebright , 1993 ) , the σ4 domain positioned at the −35 element , or both ( Nickels et al . , 2002; Dove et al . , 2003; Jain et al . , 2004 ) . Activators can accelerate initiation by stabilizing the initial RNAP/promoter complex , by stimulating the isomerization of the initial RNAP/promoter complex to RPo ( i . e . , unwinding the duplex DNA to form the transcription bubble ) , or both ( Li et al . , 1997; Roy et al . , 1998 ) . CarD , first identified as a regulator of ribosomal RNA ( rRNA ) transcription in Mycobacterium tuberculosis ( Mtb ) , is a transcriptional activator widely distributed among bacterial species , including Thermus species ( Stallings et al . , 2009; Srivastava et al . , 2013 ) , but is absent in Eco ( Table 1 ) . CarD is a global regulator ( Srivastava et al . , 2013 ) that is an essential protein in Mtb ( Stallings et al . , 2009 ) , the causative agent of tuberculosis . A deeper understanding of the CarD functional mechanism and its role in the Mtb transcription program is therefore warranted . 10 . 7554/eLife . 08505 . 003Table 1 . Distribution of CarD in bacterial phylaDOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 003Phyla*Clades and colloquial names noted . Select genera within some phyla are also listedCarD presence in phyla# of completed genomes and draft assemblies†Acidobacteria/Fibrobacterdiderm Gram−Yes ( only Acidobacteria ) 24Actinobacteriamonoderm , high G + C Gram+: Streptomyces , MycobacteriaYes932Aquificaediderm Gram−: glidobacteriaYes16Bacteroidetesdiderm Gram−: Green sulfur bacteriaNo468Caldisericadiderm Gram−No2Chlamydiaediderm Gram− Planctobacteria: Chlamydia trachomatisYes21Chlorobididerm Gram−No12Chloroflexididerm Gram−: glydobacteriaNo32Chrysiogenetesdiderm Gram−: DesulfurispirillumNo2Cyanobacteria‡diderm Gram−: glydobacteriaYes103Deferribacteresdiderm Gram−No6Deinococcus–Thermusdiderm Gram−: glydobacteriaYes43Dictyoglomididerm Gram−No2Elusimicrobiadiderm Gram−No3Firmicutesmonoderm low G + C Gram+: Bacillus , ClostridiumYes1149Fusobacteriadiderm Gram−No25Gemmatimonadetesdiderm Gram−No5Lentisphaeraediderm Gram−No2Nitrospiraediderm Gram−No10Planctomycetesdiderm Gram−: planctobacteriaNo22Proteobacteria-αdiderm Gram−: Rickettsia , RhizobiumYes678Proteobacteria-βdiderm Gram−: Bordetella , NeisseriaNo350Proteobacteria-γdiderm Gram−: Escherichia , PseudomonasNo982Proteobacteria-δdiderm Gram−: Desulfovibrio , GeobacterYes142Proteobacteria-εdiderm Gram−: HelicobacterNo78Spirochaetesdi-derm Gram−: Borrelia , TreponemaYes81Synergistetesdiderm Gram−No18TenericutesMonoderm: MycoplasmaNo132Thermodesulfobacteriadiderm Gram−: glidobacteriaYes3Thermotogaediderm Gram−No26Verrucomicrobiadiderm Gram−No37*Phyla list based on the list of prokaryotic names with standing in nomenclature ( LPSN ) ( http://www . bacterio . net/-classifphyla . html ) and the NCBI taxonomy list ( http://www . ncbi . nlm . nih . gov/Taxonomy/Browser/wwwtax . cgi ) . The diverse phylum proteobacteria are divided into subgroups of α , β , γ , δ and ε . †Genomes and draft assemblies sequenced list are shown to illustrate representation of each phylum in the Blast database and gathered from http://www . ncbi . nlm . nih . gov/genomes/MICROBES/microbial_taxtree . html . ‡Phyla containing CarD are highlighted in bold . Method: Using the Blast database search engine ( http://blast . ncbi . nlm . nih . gov/Blast . cgi ? PROGRAM=blastp&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome ) we searched for sequences similar to Tth CarD with restrictions of amino acid length of 120:200 amino acids within each phylum . Crystal structures of Tth ( Srivastava et al . , 2013 ) and Mtb ( Gulten and Sacchettini , 2013 ) CarD reveal an N-terminal domain with a Tudor-like fold ( CarD-RID , RNAP interacting domain ) in common with the Eco transcription repair coupling factor ( TRCF ) -RID ( Deaconescu et al . , 2006; Stallings et al . , 2009; Weiss et al . , 2012 ) , and a helical C-terminal domain ( CarD-CTD ) . Unique among known transcription activators , the CarD-RID interacts with the RNAP β subunit β1-lobe ( Stallings et al . , 2009; Weiss et al . , 2012 ) ( corresponding to the eukaryotic RNAP II Rpb2 protrusion domain; Cramer et al . , 2001 ) , which is near the upstream portion of the transcription bubble in RPo ( Bae et al . , 2015 ) . The disposition of the CarD-CTD with respect to the CarD-RID is widely divergent in Tth and Mtb CarD crystal structures , leading to conflicting models for the CarD activation mechanism ( Gulten and Sacchettini , 2013; Srivastava et al . , 2013 ) . To resolve these ambiguities , we determined crystal structures of Thermus aquaticus ( Taq ) transcription initiation complexes ( RPo ) ( Bae et al . , 2015 ) containing CarD ( Figure 1A , B ) . The structures show that CarD interacts with the unique DNA topology of the upstream double-stranded/single-stranded ( ds/ss ) DNA junction of the transcription bubble . Additional biochemical data confirm that our structures correspond to functional activation complexes , and extend our understanding of the role of a universally conserved CarD Trp residue in stabilizing the unwound transcription bubble , thereby stabilizing the transcription initiation complex . 10 . 7554/eLife . 08505 . 004Figure 1 . Structure of the Thermus CarD/RPo complex . ( A ) Synthetic oligonucleotides used for CarD/RPo crystallization . The numbers above denote the DNA position with respect to the transcription start site ( +1 ) . The DNA sequence is derived from the full con promoter ( Gaal et al . , 2001 ) . The −35 and −10 ( Pribnow box ) elements are shaded yellow , the extended −10 ( Keilty and Rosenberg , 1987 ) and discriminator ( Feklistov et al . , 2006; Haugen et al . , 2006 ) elements purple . The nt-strand DNA ( top strand ) is colored dark grey; the t-strand DNA ( bottom strand ) , light grey; the RNA transcript , red . The colored blocks denote protein/nucleic acid interactions: σA , orange; β , cyan; β′ , pink; CarD , green . CarD interacts exclusively at the upstream junction of the transcription bubble . ( B ) Overall structure of CarD/RPo—two orthogonal views . The nucleic acids are shown as CPK atoms and color-coded as above . Proteins are shown as molecular surfaces . The RNA polymerase ( RNAP ) holoenzyme is color coded as follows: αI , αII , ω , grey; β′ , light pink; Δ1 . 1σA , light orange; β is light cyan except the β1-lobe ( interacting with the CarD-RID , corresponding to RNAP β subunit residues 18–138 and 333–392 ) is light blue . The CarD-RID is magenta , CarD-CTD green . In the right view , the boxed region is magnified in ( C ) . ( C ) Magnified view illustrating the CarD-RID/β1-lobe protein/protein interaction and CarD-CTD ( α3 and α5 ) /DNA interactions at the upstream ds ( −12 ) /ss ( −11 ) junction of the transcription bubble . ( D ) CarD does not alter the transcription bubble . KMnO4 footprints ( t-strand ) of Thermus RNAP holoenzyme on the Mtb AP3 promoter . ( Top ) Sequence of the AP3 promoter ( Hartmann et al . , 1987 ) . T-strand thymidines rendered KmnO4 reactive by RNAP are denoted ( red arrows ) . ( Bottom ) KMnO4 footprints . Lane 1 , no protein added; lanes 2–3 , RNAP holoenzyme − or + CarD ( respectively ) ; lanes 4–7 , the effect of incubating with a competitor promoter trap for the indicated amounts of time . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00410 . 7554/eLife . 08505 . 005Figure 1—figure supplement 1 . Sequences of Mtb rrnAAP3 ( Gonzalez-y-Merchand et al . , 1996 ) and Tth 23S ribosomal RNA ( rRNA ) ( Hartmann et al . , 1987 ) , promoters used in in vitro assays , and full con ( Gaal et al . , 2001 ) used for structural studies . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00510 . 7554/eLife . 08505 . 006Figure 1—figure supplement 2 . Crystal packing interactions in CarD/RPo P43212 crystals . One asymmetric unit of the crystals contains two CarD/RPo complexes , complex A [RNAP ( A ) , cyan; CarD ( A ) , blue] and complex B [RNAP ( B ) , pink; CarD ( B ) , red] . One central asymmetric unit is shown ( proteins as molecular surfaces ) , with neighboring symmetry-related complexes shown as ribbons; only symmetry-related complexes that make crystal packing contacts with the central asymmetric unit are shown . CarD ( A ) makes a crystal packing contact with a symmetry-related CarD ( A ) ( circled in red ) , but CarD ( B ) is not involved in any crystal packing interactions . Nevertheless , the protein/protein and protein/DNA contacts in complex ( A ) and complex ( B ) are essentially identical . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00610 . 7554/eLife . 08505 . 007Figure 1—figure supplement 3 . CarD/β1-lobe structure . ( Top ) View of the CarD/RPo structure , similar to Figure 1B ( Right ) except the RNAP β1-lobe and CarD are shown as backbone ribbons without surfaces . ( Bottom ) The CarD/β1-lobe structure ( 2 . 4 Å-resolution , Table 1 ) shown in the orientation corresponding to the top view . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00710 . 7554/eLife . 08505 . 008Figure 1—figure supplement 4 . Slight movement of CarD-CTD towards DNA when DNA is present . CarD/us-fork and CarD/RPo structures ( four copies , two crystallographically independent copies from each structure ) are shown superimposed by the Cα positions in the β1-lobe . In all these structures in the presence of promoter DNA , the β1-lobe is colored cyan , CarD is colored dark red , and CarD-W86 is shown in CPK format . The CarD/β1-lobe structure is also superimposed by the Cα positions of the β1-lobe ( slate blue ) , with the CarD-RID magenta and the CarD-CTD green . Viewing the structures superimposed this way reveals a rotation of the CarD-CTD of ∼11° towards the DNA ( when promoter DNA is present ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00810 . 7554/eLife . 08505 . 009Figure 1—figure supplement 5 . Data and model quality . Plots relating data quality with model quality using the Pearson correlation coefficient ( CC ) analysis described by Karplus and Diederichs ( 2012 ) . CC1/2 ( red squares ) was determined from the unmerged diffraction data randomly divided in half . Since CC1/2 underestimates the information content of the data ( since it's calculated by dividing the dataset in half ) , CC* was calculated from an analytical relation to estimate the information content of the full data ( Karplus and Diederichs , 2012 ) . CC* provides a statistic that assesses data quality as well and also allows direct comparison of crystallographic model quality and data quality on the same scale through CCwork and CCfree , the standard and cross-validated correlations of the experimental intensities with the intensities calculated from the refined model . A CCwork/CCfree smaller than CC* indicates that the model does not account for all of the signal in the data , meaning it is not overfit . Plotted also are the standard <I>/σI for the diffraction data , as well as the Rwork/Rfree for the refined models . ( Left ) Data for Tth CarD/Taq EΔ1 . 1σA/us-fork ( −12 bp ) at 4 . 4 Å-resolution . ( Right ) Data for Tth CarD/Taq EΔ1 . 1σA RPo ( with 4-nt RNA primer ) at 4 . 3 Å-resolution . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 00910 . 7554/eLife . 08505 . 010Figure 1—figure supplement 6 . CarD does not alter the structure of the transcription bubble . Superimposition of the nucleic acids from the CarD/RPo ( colored as in Figure 1A ) and RPo ( magenta ) ( Bae et al . , 2015 ) structures . The only significant differences occur in the single-stranded t-strand from −11 to −7; this part of the DNA is relatively unconstrained by protein/DNA interactions and has very high B-factors . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 010 Throughout this work , we use three promoter sequences , full con ( Gaal et al . , 2001 ) , Tth 23S ( Hartmann et al . , 1987 ) , and Mtb AP3 ( Gonzalez-y-Merchand et al . , 1996 ) ( Figure 1—figure supplement 1 ) . The full con promoter sequence , derived by in vitro evolution , is likely to be optimized for binding to EσA . We use this sequence only for structural studies where high-affinity , homogeneous complexes are critical for crystallization . AP3 is a native Mtb rRNA promoter and its regulation by Mtb CarD has been well characterized ( Srivastava et al . , 2013; Davis et al . , 2015 ) . In order to biochemically characterize more than one promoter , we also studied 23S , a native Tth rRNA promoter . In promoter-based assays , the effects of Tth or Mtb CarD on each promoter were qualitatively the same . In general , we present the results from Tth 23S since most of the studies used Thermus EσA and CarD . In some cases , it was advantageous to use Mtb AP3 instead and we note the rationales below .
Crystals of CarD transcription activation complexes were prepared by soaking Tth CarD into Taq Δ1 . 1σA-holoenzyme/us-fork ( −12 bp ) or full RPo crystals ( Bae et al . , 2015 ) . Analysis of the diffraction data indicated high occupancy of one CarD molecule bound to each of two RNAP/promoter complexes in the asymmetric unit of the crystal lattice ( Figure 1—figure supplement 2 ) . Docking CarD onto the RNAP was facilitated by a high-resolution crystal structure of a Tth CarD/Taq β1-lobe complex ( 2 . 4 Å-resolution , Table 2 , Figure 1—figure supplement 3 , Figure 1—figure supplement 4 ) . The structures of CarD transcription activation complexes were refined to 4 . 4 and 4 . 3 Å-resolution , respectively ( Table 2 , Figure 1—figure supplement 5 ) . The protein/protein and protein/nucleic acid interactions were essentially identical among all of the four crystallographically independent complexes , so the more complete and higher resolution CarD/RPo structure ( Figure 1A , B , Figure 1—figure supplement 5 , Table 2 ) is described here . Although the CarD bound to one RPo in the crystallographic asymmetric unit made crystal-packing interactions with a symmetry-related CarD , the CarD bound to the second RPo did not participate in any crystal-packing interactions ( Figure 1—figure supplement 2 ) , indicating the architecture and interactions observed here are unlikely to be influenced by crystal packing interactions and likely represent the functional activation complex in solution . 10 . 7554/eLife . 08505 . 011Table 2 . Crystallographic statisticsDOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 011Holo-bubble-CarDHolo-fork-CarDCarD/β1-lobeData collection Space groupP43212P43212I4 Combined datasets461 Cell dimensions a ( Å ) 289 . 84293 . 15149 . 32 b ( Å ) 289 . 84293 . 15149 . 32 c ( Å ) 536 . 34539 . 1352 . 26 Wavelength ( Å ) 1 . 0751 . 0751 . 1 Resolution ( Å ) 39 . 56–4 . 3 ( 4 . 45–4 . 3 ) †49 . 61–4 . 40 ( 4 . 56–4 . 40 ) †49 . 32–2 . 40 ( 2 . 49–2 . 40 ) † Total reflections1 , 204 , 932 ( 93 , 381 ) 2 , 004 , 840 ( 73 , 134 ) 138 , 950 ( 13 , 077 ) Unique reflections153 , 939 ( 12 , 740 ) 148 , 420 ( 10 , 172 ) 22 , 705 ( 2257 ) Multiplicity7 . 8 ( 6 . 2 ) 13 . 5 ( 5 . 0 ) 6 . 1 ( 5 . 8 ) Completeness ( % ) 99 . 6 ( 99 . 2 ) 99 . 9 ( 99 . 6 ) 100% ( 100% ) <I>/σI5 . 06 ( 0 . 65 ) 9 . 10 ( 0 . 41 ) 19 . 13 ( 1 . 66 ) Wilson B-factor165 . 15151 . 3349 . 38 Rpim‡0 . 295 ( 1 . 61 ) 0 . 138 ( 2 . 03 ) 0 . 033 ( 0 . 44 ) CC1/2§0 . 948 ( 0 . 114 ) 0 . 971 ( 0 . 166 ) 0 . 998 ( 0 . 49 ) CC*§0 . 987 ( 0 . 453 ) 0 . 993 ( 0 . 534 ) 1 . 00 ( 0 . 811 ) Twinning operator––−k , −h , −l fraction––0 . 42Anisotropic scaling B-factors# a* , b* ( Å2 ) 16 . 9516 . 01– c* ( Å2 ) −33 . 90−32 . 03–Refinement Rwork/Rfree0 . 2748/0 . 3094 ( 0 . 3916/0 . 4100 ) 0 . 2198/0 . 2639 ( 0 . 3660/0 . 3920 ) 0 . 1629/0 . 1863 ( 0 . 2582/0 . 3036 ) CCwork/CCfree§0 . 928/0 . 890 ( 0 . 261/0 . 267 ) 0 . 921/0 . 891 ( 0 . 318/0 . 262 ) 0 . 870/0 . 498 ( 0 . 498/0 . 437 ) No . atoms60 , 87858 , 9902753 Protein/DNA60 , 87258 , 9842657 Ligand/ion6620 Water0076 Protein residues71977195342 B-factors Protein179 . 52194 . 6660 . 35 Ligand/ion158 . 99139 . 4849 . 77 Water––52 . 81 R . m . s deviations Bond lengths ( Å ) 0 . 0050 . 0040 . 010 Bond angles ( ° ) 0 . 961 . 011 . 35 Clashscore19 . 5814 . 8319 . 72 Ramachandran favored ( % ) 888991 Ramachandran outliers ( % ) 0 . 480 . 570 . 89†Values in parentheses are for highest-resolution shell . ‡ ( Diederichs and Karplus , 1997 ) . § ( Karplus and Diederichs , 2012 ) . #As determined by the UCLA MBI Diffraction Anisotropy Server ( http://services . mbi . ucla . edu/anisoscale/ ) . The relative orientation of the CarD domains ( CarD-RID , CarD-CTD ) seen in the Thermus CarD ( Srivastava et al . , 2013 ) and CarD/β1-lobe ( Figure 1—figure supplement 3 ) structures is only slightly altered in the Thermus CarD/RPo complex: the CarD-CTD is rotated ∼11° ( with respect to the CarD-RID ) to interact with the DNA ( Figure 1—figure supplement 4 ) . By maintaining the CarD-RID/CTD interface seen in all the Tth CarD structures , binding of the CarD-RID to the RNAP β1-lobe ( Figure 1B , C , Figure 1—figure supplement 4 ) ( Stallings et al . , 2009; Weiss et al . , 2012; Gulten and Sacchettini , 2013 ) positions the CarD-CTD to interact directly with the upstream ds/ss junction of the transcription bubble ( Figure 1A–C ) . In the RPo structure , the σ2A and σ3A domains make extensive interactions with the promoter DNA ( −17 to −4 ) from the ( distorted ) major groove side of the DNA , including critical interactions that maintain the upstream ds ( −12 ) /ss ( −11 ) junction of the transcription bubble ( Figure 1A–C ) ( Bae et al . , 2015 ) . CarD does not make significant interactions with σA but interacts with the promoter DNA from −14 to −10 from the opposite , ( distorted ) minor groove side of the DNA ( Figure 1A–C ) such that the σA/DNA interactions and the structure of the transcription bubble in RPo and CarD/RPo are essentially the same ( Figure 1—figure supplement 6 ) . The KMnO4 reactivity of thymine ( T ) bases within the transcription bubble ( Sasse-Dwight and Gralla , 1991; Ross and Gourse , 2009 ) is identical in the presence or absence of CarD ( Figure 1D , lanes 2 and 3 ) , supporting the structural observation that the transcription bubble is the same with or without CarD . Although CarD does not alter the structure of the transcription bubble , it does increase the lifetime of RPo , as measured by the rate of disappearance of the KMnO4 footprint after challenge with an excess of unlabeled competitor promoter ( Figure 1D , lanes 4–7 ) ( Davis et al . , 2015 ) . The N-terminal ends of two CarD-CTD α-helices ( α3 and α5 ) make direct contacts with the promoter DNA ( Figure 1C , Figure 2 ) . The two α-helices are positioned roughly perpendicular to the duplex DNA axis , forming a modest CarD/DNA interaction surface of 380 Å2 . 10 . 7554/eLife . 08505 . 012Figure 2 . CarD-CTD/promoter DNA interactions . ( A ) Stereo view of the refined , B-factor sharpened ( −80 Å2 ) 2Fo − Fc map ( grey mesh , contoured at 1σ ) , with superimposed DNA and CarD . Density for the close approach of the CarD peptide backbone to the −14 ( t ) DNA phosphate backbone and for CarD-W86 are clearly resolved . ( B ) Close up view showing interactions between the N-terminal ends of α3 and α5 of the CarD-CTD with promoter DNA at the upstream ds ( −12 ) /ss ( −11 ) junction of the transcription bubble . Grey dashed lines indicate potential polar interactions between the peptide backbone nitrogen of L124 and the −14 ( t ) phosphate oxygen , and W86 Nε and O2 of T−12 ( nt ) . ( C ) Same view as Figure 2B . Superimposed is the simulated annealing omit map ( dark green mesh , Fo − Fc , contoured at 3σ ) , calculated from a model where CarD-W86 was mutated to Ala . The unbiased difference Fourier density shows that the side chain position is specified in the data . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01210 . 7554/eLife . 08505 . 013Figure 2—figure supplement 1 . Alignment of CarD homologs found in bacteria from 11 diverse phyla/groups . The CarD sequences shown are from the following organisms chosen to represent the preceding phylum/group: Deinococcus–Thermus-Tth HB8 , Actinobacteria–Mtb , Acidobacteria–Candidatus Solibacter usitatus , α-Proteobacteria–Rickettsia belli , Aquificae–Desulfurobacterium thermolithotrophum , Chlamydae–Chlamydae trachomatis , Cyanobacteria–Mastigocoleus testarum , δ-Proteobacteria–Desulfobulbus propionicus , Firmicutes–Bacillus cereus , Spirochaetes–Treponema pallidum and Thermodesulfobacteria–Thermodesulfatator atlanticus . Alignments were performed using the ClustalW algorithm in MegAlign ( DNASTAR ) . Groups of residues considered homologous ( DE ) , ( HKR ) , ( ALMIV ) , ( NQ ) , ( ST ) , ( C ) , ( G ) and ( P ) are shaded blue when occurring in greater than 9/11 sequences . Identical residues occurring in all 11 sequences are shaded black . Histograms above the alignment graphically illustrate residues that are absolutely conserved within each of the 11 sequences and W86 is asterisked . The 100% identical residue is listed immediately below the histogram . A larger alignment of 831 CarD sequences is included ( Source code 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 013 The peptide backbone nitrogen of CarD-L124 , at the N-terminal end of α5 , closely approaches the backbone phosphate oxygen of the template strand ( t-strand ) at the −14 position [−14 ( t ) ] ( Figure 2 ) , possibly forming a hydrogen bond , an interaction likely facilitated by the partial positive charge of the α5 helix dipole ( Hol et al . , 1978 ) . Similar interactions have been observed in other DNA-binding proteins , such as helix-turn-helix proteins ( Harrison and Aggarwal , 1990 ) and the nucleosome core particle ( Luger et al . , 1997 ) . W86 is conserved among greater than 95% of CarD proteins ( Figure 2—figure supplement 1; Source code 1 ) and was shown to be important for CarD function as an activator ( Srivastava et al . , 2013 ) . The bulky , hydrophobic planar side chain of W86 , located at the N-terminal end of α3 , wedges into the splayed minor groove at the upstream edge of the transcription bubble ( Figure 2 ) . Despite the relatively low resolution of our analysis ( Table 2 ) , CarD-W86 was clearly resolved in electron density maps ( Figure 2A ) . The positioning of CarD-W86 was further supported by an unbiased simulated annealing omit Fo − Fc map calculated from coordinates in which CarD-W86 had been mutated to Ala ( Figure 2C ) . Previous work showed that substitution of the bulky CarD-W86 side chain by Ala ( Tth CarD-W86A or Mtb CarD-W85A ) greatly reduced the activation efficiency of both Tth and Mtb CarD ( Srivastava et al . , 2013 ) . To further evaluate the role of W86 in CarD function , we tested the activation efficiency of CarD harboring substitutions of W86 to other hydrophobic residues ( A , F , Y , L , I and V ) in an in vitro abortive transcription assay on the Tth 23S promoter ( Figure 3A , Figure 3—figure supplement 1 ) . All of the mutants tested showed impaired activity compared to wild-type CarD . A , F , and Y substitutions showed partial activation , while substitutions with branched-chain residues ( I , L , V ) showed no activation ( I , V ) or even a reduction of transcription compared to wild type CarD ( L ) . Structural modeling suggests the branched-chain residues would clash with the DNA and interfere with CarD function . 10 . 7554/eLife . 08505 . 014Figure 3 . Function of CarD-W86 . ( A ) The effect of CarD-W86 substitutions on activation of abortive initiation ( UpG dinucleotide + α-32P-CTP ) on the Tth rrnA-23S promoter ( normalized with respect to no CarD ) . Error bars denote the standard error from a minimum of three experiments . ( B ) The effect of promoter −12 base pair substitutions on activation of abortive initiation ( GpU dinucleotide + α-32P-UTP ) by CarD on the Mtb rrnA-AP3 promoter . Error bars denote standard errors . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01410 . 7554/eLife . 08505 . 015Figure 3—figure supplement 1 . Complete gels for the abortive initiation assays shown in ( A ) Figure 3A and ( B ) Figure 3B . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 015 The position of the W86 side chain Nε allows it to interact with T−12 ( nt ) O2 ( Figure 2B ) . Since this mode of Trp/Thymine interaction is not common among DNA-binding proteins ( Lejeune et al . , 2005 ) , we mutated the promoter −12 T/A base pair to C/G , G/C and A/T , and compared CarD activation at each of the three mutant promoters with its effect at the wild type promoter ( we used Mtb AP3 for this analysis since Tth EσA was more active on this promoter than on Tth 23S , allowing us to analyze the weak activity of the mutant promoters ) . The −12 T/A base pair is a conserved part of the promoter −10 element ( Shultzaberger et al . , 2007 ) , and , as expected , transcription activity from each of the mutant promoters was reduced at least threefold ( Figure 3B , Figure 3—figure supplement 1 ) ( Moyle et al . , 1991 ) . In addition , CarD activation on each mutated promoter was substantially lower than on the wild type promoter ( Figure 3B ) , suggesting that the observed interaction between W86 and T−12 ( nt ) contributes to CarD activity . Note that the CarD W86F substitution results in an approximately twofold loss in CarD fold activation ( Figure 3A , threefold activation for wild-type CarD vs 1 . 5-fold for W86F ) , as does substitution of the promoter −12 bp by anything other than the wild-type T/A bp ( Figure 3B ) . The Phe side chain at CarD position 86 would be expected to fulfill the stacking and steric roles of CarD-W86 effectively , but would not be able to participate in the putative H-bond with the T−12 ( nt ) O2 atom . We tentatively suggest that the reduced activation efficiency of the CarD-W86F mutant is primarily due to the loss of the minor groove polar interaction with T−12 ( nt ) . A crystal structure of Mtb CarD in complex with an Mtb RNAP β-subunit fragment that includes the β1-lobe shows a relative orientation of the CarD-RID/CarD-CTD domains very different from the one in our Tth CarD structures , despite high sequence and structural similarity within the domains ( Gulten and Sacchettini , 2013; Srivastava et al . , 2013 ) . In the Mtb structure , the CarD-CTD is rotated ∼140° relative to the CarD-RID ( Figure 4A ) . Structural modeling in the context of RPo positions the Mtb CarD-CTD and the functionally important W85 away from the promoter DNA ( Figure 4A , B ) . To determine the functional conformation of CarD , we introduced a disulfide to lock the conformation of Mtb CarD into the one observed in the Tth CarD structures . In the seven crystallographically independent copies of Tth CarD ( PDB IDs 4L5G and structures reported here ) ( Srivastava et al . , 2013 ) , the average distance between the α-carbons of CarD-RID-P13 and CarD-CTD-G100 is 5 . 7 ± 0 . 8 Å , and among the four copies determined in the presence of promoter DNA , an even tighter distribution is observed , 5 . 2 ± 0 . 1 Å ( Figure 4B , right ) . On the other hand , the corresponding positions in the Mtb CarD structure ( P12/G99 ) are 24 Å apart ( Figure 4B , left ) ( Gulten and Sacchettini , 2013 ) . Cys substitions at these positions are predicted to form a disulfide bond under oxidizing conditions in the Tth CarD conformation ( thus locking the domain orientation ) , but not the Mtb CarD conformation ( Figure 4B ) . We engineered the P12/G99 Cys substitutions in Mtb CarD ( Mtb CarD2C; wild-type Mtb CarD is devoid of Cys residues ) . Non-reducing SDS polyacrylamide gel electrophoresis and liquid chromatography mass spectrometry confirmed that under oxidizing conditions , the CarD-RID and CarD-CTD were disulfide crosslinked in greater than 98% of CarD2C , while under reducing conditions , no disulfide bond was present in >99% of CarD2C ( Figure 4C ) . We tested the function of oxidized ( crosslinked ) and reduced CarD2C using a mycobacterial transcription system ( Srivastava et al . , 2013; Davis et al . , 2015 ) on the Mtb AP3 promoter . Under oxidizing conditions , the cross-linked , conformationally locked CarD2C activated transcription as well as wild type CarD ( Figure 4D , 0 mM dithiothreitol [DTT] , Figure 4—figure supplement 1 ) . The observation that under reducing conditions , CarD2C was somewhat impaired in transcription activation ( Figures 4D , 5 mM DTT ) is explained by the fact that the CarD positions corresponding to Mtb CarD P13 and G99 are conserved ( Srivastava et al . , 2013; Figure 2—figure supplement 1 , Source code 1 ) ; on this basis substitution of these positions would be expected to impair uncrosslinked CarD2C function . We conclude that the Tth CarD structures , with CarD-CTD W86 positioned to interact with the upstream edge of the transcription bubble ( Figure 2 ) , represents the functional conformation of CarD . 10 . 7554/eLife . 08505 . 016Figure 4 . Inter-domain crosslinking confirms the functional conformation of CarD . ( A ) View of the Thermus CarD/RPo complex . RNAP holoenzyme and nucleic acids are shown as in Figure 1B; Tth CarD is shown as an α-carbon ribbon ( Tth CarD-RID , magenta; Tth CarD-CTD , green ) but with W86 shown in CPK format and colored dark green . Also shown is Mtb CarD from the Mtb CarD/β1-β2-lobe structure ( 4KBM; Mtb CarD-RID , dark red; Mtb CarD-CTD , brown , but with W85 colored dark brown ) , superimposed by alignment of 145 Cα atoms from the β1-lobe ( 1 . 39 Å rmsd ) . The boxed region is magnified in ( B ) . ( B ) ( Left ) Magnified view showing the modeled Mtb CarD in the context of RPo . The α-carbons of CarD-RID-P12 and CarD-CTD-G99 , shown as red spheres , are ∼24 Å apart ( red dashed line ) . A disulfide bond between these two positions in Mtb CarD2C ( P12C/G99C substitutions ) would disallow this conformation of CarD . ( Right ) Magnified view of the Thermus CarD/RPo complex . CarD-RID-P13 and CarD-CTD-G100 are ∼5 . 2 Å apart ( red dashed line ) . A disulfide bond between the corresponding two positions in Mtb CarD2C would lock this DNA-interacting conformation of CarD . ( C ) Purification of disulfide crosslinked ( lanes 5 , 6 ) and reduced ( lanes 7 , 8 ) CarD2C . Non-reducing SDS-PAGE illustrates that CarD2C is oxidized ( crosslinked ) in the absence of reducing agent dithiothreitol ( DTT ) and is reduced ( uncrosslinked ) in the presence of DTT . Samples were excised from gels and LC-MS was used to confirm oxidation states . ( D ) Effect of oxidation state on Mtb CarD2C activation of abortive transcription on the Mtb AP3 promoter ( GpU dinucleotide + α-32P-UTP ) . Conformationally locked ( no DTT ) Mtb CarD2C exhibits wild type activation activity . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01610 . 7554/eLife . 08505 . 017Figure 4—figure supplement 1 . Complete gel for the abortive initiation assay shown in Figure 4D . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01710 . 7554/eLife . 08505 . 018Figure 5 . CarD increases the lifetime of Thermus RPo . ( A ) Sequences of Mtb rrnAAP3 ( Gonzalez-y-Merchand et al . , 1996 ) and Tth 23S rRNA ( Hartmann et al . , 1987 ) promoters used in in vitro assays . ( B , C ) Lifetimes of promoter complexes measured by abortive transcription . At the top of each panel , [32P]-labeled abortive transcript production at times after addition of a large excess of competitor promoter DNA trap was monitored by polyacrylamide gel electrophoresis and autoradiography . On the bottom , transcript production was quantified by phosphorimagery and plotted . The lines indicate single-exponential decay curves fit to the data points . The calculated decay half-lives ( t1/2 ) are shown to the right of the gel images . Assays were performed on the following templates: ( B ) Tth rrnA-23S promoter ( UpG dinucleotide + α-32P-CTP ) . ( C ) Mtb rrnA-AP3 promoter ( GpU dinucleotide + α-32P-UTP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01810 . 7554/eLife . 08505 . 019Figure 5—figure supplement 1 . Complete gels for the abortive initiation assays shown in Figure 5B . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 01910 . 7554/eLife . 08505 . 020Figure 5—figure supplement 2 . Complete gels for the abortive initiation assays shown in Figure 5C . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 020 CarD may stabilize RPo by forming favorable interactions with the upstream edge of the unwound transcription bubble ( Figures 1C , 2B ) . We tested the lifetime of competitor-resistant RPo challenged with a competitive promoter trap ( Davis et al . , 2015 ) using the abortive initiation assay on both the 23S and AP3 promoters ( Figure 5A ) . Tth CarD increased the half-life ( t1/2 ) of the Thermus RPo ∼threefold on each promoter ( Figure 5B , Figure 5C , Figure 5—figure supplement 1 , Figure 5—figure supplement 2 ) while Eco RNAP did not dissociate significantly from either promoter over the lifetime of the experiments ( t1/2 >> 120 min; Figure 5B ) ( Davis et al . , 2015 ) . The Tth CarD W86A substitution diminished or abolished the ability of CarD to increase t1/2 on the 23S and AP3 promoter , respectively ( Figure 5B , C ) . Clearly , dissociation of RPo and transcription bubble collapse ( rewinding ) are closely linked . We hypothesized that CarD may increase the lifetime of RPo by preventing transcription bubble collapse . To test this hypothesis , we determined the effect of CarD on the lifetime of promoter complexes on a synthetic promoter template based on the 23S sequence and compared it with the same synthetic template but with a non-complementary transcription bubble ( from −11 to +2 ) unable to collapse ( Figure 6A ) . On the duplex template ( 23S_DS ) , CarD increased the t1/2 more than fivefold ( Figure 6B , Figure 6—figure supplement 1 ) . On the bubble template ( Figure 6A , 23S_Bub ) in the absence of CarD , the t1/2 was also increased more than fivefold , indicating that the relatively short lifetime of Tth RPo on the 23S promoter is due largely to bubble collapse ( Figure 6B ) . Addition of CarD to the bubble template had no effect on the level of transcription and did not affect RPo lifetime ( Figure 6B ) . We thus conclude that a primary function of Tth CarD , like Mtb CarD ( Davis et al . , 2015 ) , is to stabilize RPo by preventing collapse of the transcription bubble . 10 . 7554/eLife . 08505 . 021Figure 6 . CarD increases the lifetime of Thermus RPo by preventing transcription bubble collapse . ( A ) Synthetic duplex ( 23S_DS ) and artificial bubble ( 23S_Bub ) promoters used in in vitro assays . ( B ) Lifetimes of promoter complexes formed on synthetic templates measured by abortive transcription ( UpG dinucleotide + α-32P-UTP ) . ( Left ) [32P]-labeled abortive transcript production at times after addition of a large excess of competitor promoter DNA trap was monitored by polyacrylamide gel electrophoresis and autoradiography . ( Right ) transcript production was quantified by phosphorimagery and plotted . The lines indicate single-exponential decay curves fit to the data points . The calculated decay half-lives ( t1/2 ) are shown to the right of the gel images . Assays were performed on the synthetic double-stranded ( 23S_DS ) and bubble ( 23S_Bub ) templates . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 02110 . 7554/eLife . 08505 . 022Figure 6—figure supplement 1 . Complete gels for the abortive initiation assays shown in Figure 6B . DOI: http://dx . doi . org/10 . 7554/eLife . 08505 . 022
CarD is an essential transcription activator in Mtb that is also widely distributed among bacterial species , including Thermus species but not found in Eco ( Stallings et al . , 2009; Srivastava et al . , 2013; Table 1 ) . In the absence of a structure of a mycobacterial transcription initiation complex , we present here the structure of Tth CarD with a Taq transcription initiaton complex ( Figure 1B ) . The structural results , combined with supporting biochemical studies , establish that the CarD-RID makes a protein/protein interaction with the RNAP β1-lobe , thereby positioning the CarD-CTD and a conserved Trp residue to interact with the upstream edge of the transcription bubble , using a wedge mechanism to prevent collapse of the transcription bubble ( Figures 1C , 2B ) . This is a previously unseen mechanism of activation by a transcription factor . Specifically: ( 1 ) CarD does not induce any major changes on the holoenzyme nor the transcription bubble ( Figure 1D ) ; ( 2 ) CarD contacts with the DNA are mostly confined to the backbone phosphates , with the exception of the conserved Trp ( W86 ) that serves as a wedge at the upstream edge of the bubble , which may be stabilized by a hydrogen bond with the conserved T−12 ( t ) ( Figure 2B ) . We show that this W-wedge residue and its interaction with T−12 ( t ) is important for full CarD function ( Figure 3 ) ; ( 3 ) We show that Tth CarD functions similarly to Mtb CarD to increase the lifetime of RPo by preventing collapse of the transcription bubble ( Figure 5 , Figure 6 ) ; ( 4 ) We show that the mode of CarD interaction with RNAP and with the promoter DNA revealed by our structures represents the functionally relevant conformation ( Figure 4 ) , resolving conflicting models . Eco has served as a model organism for the study of many cellular processes over the last few decades , including transcription . Eco RNAP forms unusually stable RPo and Eco lacks CarD , while RNAPs shown to form relatively unstable RPo come from bacteria that harbor CarD ( Bacillus subtilis , Whipple and Sonenshein , 1992; Artsimovitch et al . , 2000; Mtb , Davis et al . , 2015; Myxococcus xanthus , Gallego-Garcia et al . , 2014; Taq , Miropolskaya et al . , 2012; Tth , Xue et al . , 2000 ) . Moreover , CarD is a global regulator , found at most σA promoters throughout the Mycobacterium smegmatis genome ( Srivastava et al . , 2013 ) , and is essential in the two mycobacterial species where it has been tested ( Stallings et al . , 2009 ) . These observations suggest that CarD boosts transcription at most ( if not all ) promoters by acting as a basal transcription factor , required to compensate for otherwise rapidly dissociating RNAP/promoter complexes . While the interaction with T−12 ( t ) ( Figure 2B , Figure 3B ) may modulate the effect of CarD in a promoter-specific manner , this is likely to be a minor effect in vivo since T−12 ( t ) is present at most σA promoters ( Shultzaberger et al . , 2007 ) , and CarD can nevertheless activate transcription from promoters that lack a −12 T/A base pair ( although less effectively; Figure 3B ) . The structural and biochemical studies of Tth CarD/RPo complexes presented here reveal how the widely distributed transcription factor CarD interacts with RPo to activate initiation . The CarD-RID/RNAP β1-lobe protein/protein interaction positions the CarD-CTD to interact with the upstream edge of the transcription bubble in a functionally relevant pose that does not allosterically alter the structure of the transcription bubble nor RNAP holoenzyme/promoter interactions ( Figure 1D , Figure 1—figure supplement 6 ) ; instead CarD supports pre-existing RNAP holoenzyme/promoter DNA interactions in RPo . The mode of CarD/DNA interaction is incompatible with duplex B-form DNA; the normal minor groove is too narrow to accommodate the end of CarD-α3 and CarD-W86 ( Srivastava et al . , 2013 ) . This is consistent with a kinetic analysis of CarD function that concluded CarD stabilizes RPo by increasing the rate of isomerization to RPo and decreasing the rate of bubble collapse , but has little effect on the formation of the closed RNAP/promoter complex ( Rammohan et al . , 2015 ) . The CarD contacts with the DNA occur mostly through the backbone phosphates , except for highly conserved CarD-W86 , which wedges between the splayed DNA strands at the upstream edge of the transcription bubble and may form a hydrogen bond with T−12 ( t ) O2 presented in the highly distorted minor groove ( Figure 2B ) . The minor groove W-wedge increases the lifetime of RPo by preventing transcription bubble collapse ( Figure 6 ) . While we believe this is the dominant mode of action for CarD , CarD may affect other steps of the initiation process as well . This previously unseen mode of transcription activation may be absent in Eco ( the focus of most mechanistic transcription studies ) since Eco RNAP forms relatively stable complexes on most promoters ( Figure 5B ) ( Davis et al . , 2015 ) .
Crystals of Taq Δ1 . 1σA-holoenzyme/promoter complexes were grown as described ( Bae et al . , 2015 ) . Tth CarD ( prepared as described previously; Srivastava et al . , 2013 ) , in 1 mM in 20 mM Tris-HCl , pH 8 . 0 , 0 . 2 M NaCl , was added directly to the hanging drops containing RPo crystals to a final concentration of 100 μM . After 1 day of incubation , the crystals were cryo-protected and frozen as described ( Bae et al . , 2015 ) . X-ray diffraction data were collected at Brookhaven National Laboratory National Synchrotron Light Source ( NSLS ) beamline X29 . Data were integrated and scaled using HKL2000 ( Otwinowski and Minor , 1997 ) . The diffraction data were anisotropic . To compensate , isotropy was approximated by applying a positive b factor along a* and b* and a negative b factor along c* ( Table 1 ) , as implemented by the UCLA MBI Diffraction Anisotropy Server ( http://services . mbi . ucla . edu/anisoscale/ ) ( Strong et al . , 2006 ) , resulting in enhanced map features ( Figure 2A , C ) . Initial Fourier difference maps , calculated after rigid body refinement ( Adams et al . , 2010 ) starting with the appropriate RNAP-holoenzyme/promoter complex structure ( Bae et al . , 2015 ) , revealed clear density corresponding to CarD . CarD was docked into the maps with the aid of a 2 . 4 Å-resolution structure of a Tth CarD/Taq β1-lobe complex ( PDB ID 4XAX , Figure 1—figure supplement 3 , Table 2 , see below ) . The models were improved in further steps of refinement: ( 1 ) rigid body refinement of 20 individual mobile domains in RNAP and 2 domains of CarD ( CarD-RID and CarD-CTD ) ( Adams et al . , 2010 ) ; ( 2 ) deformable elastic network refinement ( Schröder et al . , 2010 ) with noncrystallographic symmetry restraints using CNS 1 . 3 ( Brunger et al . , 1998 ) performed on the Structural Biology Grid portal ( O'Donovan et al . , 2012 ) ; ( 3 ) iterative cycles of manual building with COOT ( Emsley and Cowtan , 2004 ) and refinement with PHENIX ( Adams et al . , 2010 ) . The PDBePISA server ( http://www . ebi . ac . uk/pdbe/pisa/ ) was used to calculate intermolecular buried surface areas ( Krissinel and Henrick , 2007 ) . We follow the criteria of Karplus and Diederichs ( 2012 ) , as explained in the accompanying paper ( Bae et al . , 2015 ) . In the final 2Fo − Fc electron density maps , the CarD-W86 side chain was clearly resolved ( Figure 2A ) . To confirm the side chain position , we produced an unbiased difference Fourier map using a simulated annealing omit procedure . The CarD-W86 side chain was removed from the structural model by mutation to Ala , and the mutated models were subjected to simulated annealing refinement ( 2500 K ) using PHENIX ( Adams et al . , 2010 ) ( Figure 2C ) . The plasmids pET21a Taqβ1 ( Westblade et al . , 2010 ) and pETsumoTthCarD ( Srivastava et al . , 2013 ) were separately transformed into Eco BL21 ( DE3 ) cells ( EMD Millipore , Billerica , MA , United States ) and transformants were grown at 37°C in Luria–Bertani media containing ampicillin ( 200 µg/ml ) and kanamycin ( 50 µg/ml ) . At an A600nm between 0 . 6–0 . 8 , the cultures were supplemented with isopropyl-β , D-thiogalactopyranoside ( 0 . 5 mM final concentration ) to induce protein expression for 4 hr at 30°C . The cells were then spun down by centrifugation and resuspended in buffer A ( 20 mM Tris-HCl , pH 8 . 0 at 4°C , 500 mM NaCl , 5 mM imidazole , 5% ( vol/vol ) glycerol , 1 mM β-mercaptoethanol ) . The cells were lysed using a continuous-flow homogenizer ( Avestin Inc . , Ottawa , ON , Canada ) and then centrifuged to remove insoluble debris . The clarified cell lysate containing overexpressed ( His ) 10Sumo-Tth CarD was first applied to a Ni2+-charged HiTrap column ( GE Healthcare Bio-Sciences , Marlborough , MA , United States ) that was equilibrated with buffer A , followed by a wash with five column volumes ( cv ) of buffer A + 25 mM imidazole . Subsequently , clarified cell lysate containing overexpressed Taq β1 was injected into the same column to form a complex with the immobilized ( His ) 10Sumo Tth CarD . The column was washed with five column volumes ( cv ) of buffer A + 25 mM imidazole and 5 cv buffer A + 40 mM imidazole . The complex bound to the column was eluted with buffer A + 250 mM imidazole . After overnight cleavage with UlpI protease ( GE Healthcare ) to remove the ( His ) 10Sumo-tag from Tth CarD and dialysis against buffer A + 25 mM imidazole , a subtractive Ni2+-chelating chromatographic step removed uncleaved ( His ) 10Sumo-Tth CarD and the cleaved ( His ) 10Sumo-tag . The sample was concentrated and injected on a Superdex 75 gel filtration column ( GE Healthcare ) that was equilibrated with GF buffer ( 50 mM MES-OH , pH 6 . 5 , 500 mM NaCl , 5% ( vol/vol ) glycerol ) . Fractions containing purified Tth CarD/Taq β1 complex were pooled and concentrated to 15 mg/ml by centrifugal filtration . Sodium dodecyl sulfate polyacrylamide gel electrophoresis and Coomassie blue staining were used to analyze the purity of the complex . Crystals were grown by hanging-drop vapor diffusion by mixing 1 μl of protein solution ( 15 mg/ml in GF buffer ) with 1 μl of crystallization solution ( 1 . 5 M ammonium sulfate , 0 . 1 M sodium acetate , pH 5 . 0 , 25% ( vol/vol ) ethylene glycol ) and incubating over a well containing crystallization solution at 22°C . Large crystals ( 0 . 5 mm ) grew within 1 day . The crystals were directly frozen in liquid nitrogen for data collection . X-ray diffraction data were collected at Brookhaven National Laboratory NSLS beamline X29 . Data were integrated and scaled using HKL2000 ( Otwinowski and Minor , 1997 ) ( Table 1 ) . Initial electron density maps were calculated by molecular replacement using Phaser ( McCoy et al . , 2007 ) from starting models of the Taq β1-lobe ( 2 . 9 Å-resolution; 3MLQ; Westblade et al . , 2010 ) and Tth CarD ( 2 . 4 Å-resolution; 4L5G; Srivastava et al . , 2013 ) . One CarD/β1-lobe complex was clearly identified in the asymmetric unit . The model was first adjusted manually using COOT ( Emsley and Cowtan , 2004 ) , then further refined using the Autobuild feature of PHENIX ( Adams et al . , 2010 ) . At this point , the model fit well to the electron density but the Rfree and R factors remained relatively high ( >0 . 3 ) . Twinning was identified by Xtriage in PHENIX ( twinning operators −k , −h , −l; twinning fraction 0 . 42 ) . The final model was obtained after twinning refinement using PHENIX . To prepare the promoter DNAs , fragment −86 to +70 of pUC57-MtbrrnAP3 ( −60 to +15 of the endogenous promoter sequence ) was prepared as described ( Davis et al . , 2015 ) . Fragment −171 to +69 of pRLG6768-Tthrrn23S ( −68 to +15 of the endogenous promoter sequence ) ( Vrentas et al . , 2008 ) was prepared similarly to AP3 . These fragments ( AP3 and 23S ) served as templates for all transcription assays unless otherwise noted . AP3 −12T substitutions were synthesized ( GenScript , Piscataway , NJ , United States ) and placed into pUC57 and prepared as described for AP3 ( Davis et al . , 2015 ) . Artificial bubble and double-stranded templates of 23S ( −60 to +20 ) were synthesized as oligonucleotides and gel purified ( IDT; Figure 6A ) . The purified oligonucleotides were annealed and used as templates for assays . KMnO4 footprinting on the Mtb rrnAP3 promoter ( Figure 1D ) was performed as described ( Davis et al . , 2015 ) except reactions were at 65°C with 100 mM NaCl . Abortive initiation assays ( Figure 3 , Figure 3—figure supplement 1 , Figure 4D , Figure 4—figure supplement 1 , Figure 5B , Figure 5C , Figure 5—figure supplement 1 , Figure 5—figure supplement 2 , Figure 6B , Figure 6—figure supplement 1 ) were performed as previously described ( Srivastava et al . , 2013; Davis et al . , 2015 ) with the following adaptations for the Thermus transcription system . Briefly , reactions were performed in transcription buffer ( 10 mM Tris-HCl , pH 8 . 0 , 1 mM MgCl2 , 0 . 1 mM DTT , 50 μg/ml BSA ) with 100 mM NaCl for the AP3 promoter or 100 mM K-glutamate for the 23S promoter , at 65°C . Core RNAP ( 200 nM ) and σA ( 1 μM ) were combined and incubated at 65°C for 5 min to form holoenzyme . CarD ( 2 μM , when used ) was then added to the holoenzyme and incubated for an additional 5 min . Next , promoter DNA ( 10 nM ) was added and RPo was allowed to form for 15 min at 65°C . Abortive transcription was initiated by the addition of an NTP mix containing the initiating dinucleotide ( 250 μM , GpU for AP3 , UpG for 23S; TriLINK ) , the next NTP ( α-32P-labeled , UTP for AP3 , CTP for 23S; 1 . 25 μCi , with 50 μM of the same unlabeled NTP ) and 2 μM of FC-bubble competitor DNA when used ( Figure 1D ) ( Davis et al . , 2015 ) . After 10 min , transcription was quenched and analyzed as previously described ( Davis et al . , 2015 ) . For half-life assays , competitor was first added and NTP substrates were added at different times as indicated ( Figures 5B , C , 6B ) . Single amino acid substitutions of CarD W86 were generated using site-directed mutagenesis ( Stratagene-Agilent Technologies , Santa Clara , CA , United States ) and purified using the same procedure as wild-type CarD ( Srivastava et al . , 2013 ) . Mtb CarD2C ( P12C/G99C ) was also made using site-directed mutagenesis but was subjected to two additional purification steps . Tandem Q-sepharose column chromatography ( GE Healthcare ) was used to remove inter-molecular cross-linked CarD . Sample was first applied on a 5 ml column and eluted using a NaCl gradient from 200 mM to 1 M over 20 column volumes ( cv ) . The purest fractions were combined and reapplied to a second 5 ml Q column and eluted using a NaCl gradient from 100 mM to 1 M over 40 cv . This purification yielded >95% intra-molecular cross-linked CarD as verified by non-reducing SDS-PAGE ( Figure 3C ) and liquid chromatography-mass spectromety-MS analysis ( The Rockefeller University Proteomics Resource Center ) . Transcription assays with Mtb CarD were performed similarly to the Thermus assays in the same transcription buffer but at 37°C with 10 mM K-Glutamate rather than 100 mM NaCl . Transcription at reducing conditions included 5 mM DTT , at oxidizing conditions no DTT was present . The structure factor files and X-ray crystallographic coordinates have been deposited in the Protein Data Bank under ID codes 4XLS ( Tth CarD/Taq holoenzyme/us-fork ( − 12 bp ) complex ) , 4XLR ( Tth CarD/Taq RPo ) , and 4XAX ( Tth CarD/Taq β1-lobe ) . | Inside cells , molecules of double-stranded DNA encode the instructions needed to make proteins . To make a protein , the two strands of DNA that make up a gene are separated and one strand acts as a template to make molecules of messenger ribonucleic acid ( or mRNA for short ) . This process is called transcription . The mRNA is then used as a template to assemble the protein . An enzyme called RNA polymerase carries out transcription and is found in all cells ranging from bacteria to humans and other animals . Bacteria have the simplest form of RNA polymerase and provide an excellent system to study how it controls transcription . It is made up of several proteins that work together to make RNA using DNA as a template . However , it requires the help of another protein called sigma factor to direct it to regions of DNA called promoters , which are just before the start of the gene . When RNA polymerase and the sigma factor interact the resulting group of proteins is known as the RNA polymerase ‘holoenzyme’ . Transcription takes place in several stages . To start with , the RNA polymerase holoenzyme locates and binds to promoter DNA . Next , it separates the two strands of DNA and exposes a portion of the template strand . At this point , the DNA and the holoenzyme are said to be in an ‘open promoter complex’ and the section of promoter DNA that is within it is known as a ‘transcription bubble’ . Another protein called CarD helps to speed up transcription but it is not clear how this stage of the process works . Bae et al . have now used X-ray crystallography to reveal the structure of CarD bound to the RNA polymerase holoenyzme and a DNA promoter . The structures show that one part of CarD interacts with the DNA at the start of the transcription bubble , and another part binds to the RNA polymerase . CarD fits between the two strands of DNA in the promoter , like a wedge , to keep the strands apart . Therefore , CarD stabilizes the open promoter complex and prevents the transcription bubble from collapsing . These findings reveal a previously unseen mechanism involved in activating transcription and will guide further experiments probing the role of CarD in living cells . Another study by Bae , Feklistov et al . —which involves some of the same researchers as this study—reveals that the sigma factor also binds to DNA at the start of the transcription bubble . The general principles outlined by these studies may help to identify other proteins that regulate transcription . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2015 | CarD uses a minor groove wedge mechanism to stabilize the RNA polymerase open promoter complex |
Vaccines induce memory B-cells that provide high affinity secondary antibody responses to identical antigens . Memory B-cells can also re-instigate affinity maturation , but how this happens against antigenic variants is poorly understood despite its potential impact on driving broadly protective immunity against pathogens such as Influenza and Dengue . We immunised mice sequentially with identical or variant Dengue-virus envelope proteins and analysed antibody and germinal-centre ( GC ) responses . Variant protein boosts induced GCs with a higher proportion of IgM+ B cells . The most variant protein re-stimulated GCs with the highest proportion of IgM+ cells with the most diverse , least mutated V-genes and with a slower but efficient serum antibody response . Recombinant antibodies from GC B-cells showed a higher affinity for the variant antigen than antibodies from a primary response , confirming a memory origin . This reveals a new process of antibody memory , that IgM memory cells with fewer mutations participate in secondary responses to variant antigens , demonstrating how the hierarchical structure of B-cell memory is used and indicating the potential and limits of cross-reactive antibody based immunity .
Antibody-based immunity is underpinned by memory B-cells that have undergone antibody somatic hyper-mutation ( SHM ) and selection for improved antigen binding in germinal centres ( GCs ) ( MacLennan et al . , 1997 ) . Re-challenge with the same antigen stimulates a rapid , higher affinity , secondary antibody response . Protective immunity to highly mutable viruses , like Dengue and Influenza , can be induced by vaccination but the high level of variation often leads to immune escape ( Nabel and Fauci , 2010 ) , leading to a focus on generating vaccine responses against conserved antigenic regions ( Wu et al . , 2010; Corti et al . , 2011; Wang et al . , 2015 ) . Memory B-cells of IgM and IgG isotypes can also re-instigate GCs after secondary exposure ( Dogan et al . , 2009; Pape et al . , 2011; McHeyzer-Williams et al . , 2015 ) , but how this happens against variant antigens is poorly understood despite its potential impact on driving the most broadly protective immunity . Several studies suggest diversity in the memory B-cell population , showing that cells can express IgM or IgG ( Dogan et al . , 2009; Pape et al . , 2011 ) , be mutated or non-mutated ( Kaji et al . , 2012 ) and have low affinities ( Smith et al . , 1997 ) , but still persist in GCs ( Kuraoka et al . , 2016 ) . It has long been speculated that this diversity may facilitate the recognition of antigenic variants ( Herzenberg et al . , 1980; Pape et al . , 2011; Kaji et al . , 2012 ) which could stimulate secondary GCs derived from less mutated , naïve-like , memory B-cells that still had an advantage over naive B-cells due to their increased numbers , pre-selected V-genes and lower activation thresholds ( Good and Tangye , 2007; Good et al . , 2009 ) . By sequentially immunizing mice with the same or different Dengue-virus envelope proteins , and analyzing serum antibodies and GC B-cells , we provide evidence that supports the hypothesis that less developed memory B-cells are used in secondary responses to variant antigens .
We chose Dengue-3 envelope protein ( E3 ) for all priming immunisations . Boost immunisations were performed 38 days later with identical E3 protein or variant E2 or E4 proteins which have 68% and 63% overall sequence identity with E3 , respectively . The cross reactivity of E3-primed mouse serum IgG correlated with sequence identity ( Figure 1A ) , and overall cross-reactivity also correlated ( Figure 1B ) . Boosting with homotypic E3 antigen induced a rapid antibody memory response with anti-E3 titres rising rapidly to day 7 , and not increasing further ( Figure 1D ) . E-protein boosted antisera was not reactive with an irrelevant His-tagged protein ( PR8 HA ) ( Figure 1C ) . Heterotypic boosting with E2 induced a rapid and significant increase in anti-E3 titre , as might be expected if cross-reactive memory antibodies against the priming E3 antigen were recalled ( Figure 1D ) , that did not increase further by day 17 . E4 boosting induced a modest but not statistically significant increase in the anti-E3 titre , even by day 17 , showing the E4 variant boost had not induced a significant anti-E3 antibody memory response , or the induced antibodies had a low affinity for E3 ( see discussion ) . The anti-E2 titre induced by the E2 boost increased about 120-fold by day 7 ( Figure 1E ) , and did not increase further by day 17 , further indicating that E2 boosting induced a rapid memory-like serum IgG response against E2 derived from cross-reactive E3 primed memory B-cells . Conversely the anti-E4 titre , induced by E4 boosting , rose significantly but to a lower level , about 20-fold , by day 7 ( Figure 1F ) and showed a further rise by day 17 . A boost alone did not induce a detectable antibody titre however , ( ‘BO’ , Figure 1D ) suggesting a role for memory B-cells of some type and/or cross-reactive T-cell memory , facilitating the E4 boost response . E3 and E2 boosting induced early GC B-cell levels similarly by day 7 , to 4 . 5–5 . 5% of total lymphocytes , which then reduced by two-thirds by day 17 ( Figure 2B ) . E4 boosting induced GC B-cell levels about a third as high , which then reduced similarly by about 60% at day 17 , remaining 4-fold higher than controls . Analysis of the proportion of IgM+ GC B-cells showed a highly significant trend at day 7 after boosting , with the proportion of IgM+ GC B-cells correlating with increasingly variant challenge ( Figure 2C ) . This trend continued to day 17 . The proportion of IgM+ B cells was also consistent between individuals in an experimental group ( Figure 2D ) . Overall levels of VH mutations increased in all groups from day 7 to day 17 ( Figure 2E ) , consistent with secondary affinity maturation . Sequences are available in Supplementary file 1 . There were lower levels of SHM in IgM+ GC B-cells 7 days after the variant boosts , particularly with the most variant protein E4 , compared to the homotypic E3 boost ( Figure 2F ) . Boosting with variant proteins , therefore , induced early GCs with increased proportions of IgM B-cells that had fewer VH mutations . Analysis of the VH clonality of GC B-cells after E-protein boosts showed that almost every VH sequence was from a distinct B-cell clone ( Figure 2G ) . These data also showed that the two variant boosts elicited different repertoires of VH . 40% of the VH sequences sampled at day 7 from E2 boosted mice were either VH14-3 or the closely related VH14-4 ( black dots , Figure 2G ) , suggestive of a secondary response more focused on a particular epitope ( see discussion ) . Some of these VH were also present in the homotypic E3 boost day 7 samples , but neither were detected at day 7 after E4 boosting ( Figure 2G ) . E2-variant boosting induced an immediate and significant increase in avidity by day 7 ( Figure 3A ) which did not detectably change until perhaps day 32 , although data variability is high . A modest but significant increase in serum affinity , however , was detected by day 17 , with a further increase detected by day 32 ( Figure 3C ) . We interpret this to mean that a relatively small portion of serum IgG underwent affinity maturation by day 17 in response to the E2 boost and was not detectable by the Urea avidity assay due to high variability and the high pre-existing IgG titres ( Figure 1E ) , or other limitations of the Urea assay ( Alexander et al . , 2015 ) . Boosting with the E4 variant elicited slower increases in relative affinity and avidity , only detectable by day 32 , but by then representing an equivalent , if not greater , increase compared to that induced by E2 ( Figure 3B and D ) . Memory T-cells are necessary for memory B-cell responses against haptens and viral proteins ( Aiba et al . , 2010; Hebeis et al . , 2004 ) . We found no evidence that the memory T-cell response to re-stimulation by variant E-proteins was any different from re-stimulation by E3 ( Figure 3E ) . These data imply that a deficiency in T-cell recognition of these antigens cannot explain the differences in response to E2 and E4 challenge , and supports the idea that either T-cell receptors can recognize antigenic peptides from regions with around 50% sequence difference ( see discussion ) or , more likely , B-cells present peptides from different , more conserved regions than those their antibodies bind to . For comparison with the E4 boost response , we performed primary immunisations with E4 and analysed serum antibodies and GC B-cells at day 7 and day 17 . Serum levels of anti-E4 IgG rose to a moderate level by day 17 ( mean EPT = 3 . 6 , Figure 4A ) , being less than seen after E4 boosting ( Figure 1F ) . GC B-cell levels rose to a mean of 0 . 8% lymphocytes at day 7 after E4 priming , half as much as after the E4 boost , then fell similarly to the post boost samples by around 60% by day 17 ( Figure 4B ) . As with the E-boost GCs , the proportion of IgM+ GC B-cells fell over time ( Figure 4C ) and levels of VH mutation in all B-cells and IgM+ B cells increased ( Figure 4D and E ) . The median level of VH mutation in IgM+ GC B-cells at day 7 after E4 priming is less ( =2 ) than after E4 boosting ( =3 ) suggesting , not conclusively , that GC Bells at day 7 after E4 boosting are memory derived . Antibody titres were insufficient to do a relative affinity competition ELISA and no 7M Urea-resistant IgG was detected 7 or 17 days after E4 priming ( data not shown ) . If E4 boost induced B-cells are memory derived the antibodies should show evidence of pre-selection by the E3 prime . We made 48 recombinant antibodies ( rAbs ) , 38 of which were IgM ( supplementary file 2 ) , 24 from E4 primed mice ( day 7 and day 17 ) and 24 from E4 boosted mice ( day 7 ) . Figure 4F and Supplementary file 2 , show the results from the initial screen of all rAbs against E4 , indicating that the efficiency of detection of positive binding ( deemed as O . D . > 0 . 1 , useful for subsequent titration ) was quite low but consistent with the 30–50% binding frequency of GC rAbs previously observed ( Kuraoka et al . , 2016 ) , except for E4 prime day 7 , which has only 2/13 rAbs binding strongly enough to be titrated . This might be expected of antibodies from a day 7 primary response GC , and indicated they were overall of lower affinity . Other rAbs from this group showed evidence of weak binding ( supplementary file 2 ) , indicating that the rAb cloning efficiency for this group was not reduced and only the two strongest binders were above the ELISA titration threshold . All but one of the positive binding rAbs were IgM . Figure 4G shows the ELISA titration and Figure 4H the derived endpoint titres , which we are using as a proxy of affinity . A more strongly binding IgM rAb from E4 boost day 7 , B5 , and the only positive binding IgG1 rAb , G6 , are indicated on Figure 4H . The positive-binding rAbs from E4 prime day 17 show a higher affinity than those from prime day 7 , consistent with affinity maturation . Six of the seven positive-binding IgM rAbs from E4 boost day 7 show a higher affinity than the two strongest binding IgM rAbs from E4 prime day 7 . This is consistent with pre-selection by the E3 prime immunization , and also considering the higher proportion of rAbs with an anti-E4 O . D . > 0 . 1 , implies the GC B-cells expressing these antibodies are memory derived . rAb affinities were generally low , which might be expected of IgMs particularly in early GCs . We estimated the Kd of rAbs B5 and G6 ( an IgG1 ) as around 150 nm and 1 μm respectively ( see Materials and methods ) . Other rAbs would be in the super-micromolar range . Figure 4I shows the cross reactivity of rAbs with E3 . Binding to E3 correlates with binding to E4 , but because of the generally low rAb affinities we suggest that the antibodies cannot discriminate between similar epitopes . The higher affinity of E4 boost rAbs B5 and G6 , and binding to E3 , suggest they may have genuine specificity for E3 , thus consistent with their derivation from anti-E3 memory . That rAb B5 is an IgM with only one VH ( and one Vkappa ) mutation , provides further support for the proposal of this study .
The most variant protein we boosted with , E4 , stimulated GCs with the highest proportion of IgM+ cells and with the lowest levels of VH gene mutation , greater VH-gene diversity , and a slower , more specific , serum IgG response that resulted in equivalent if not higher affinity , compared to the heterotypic E2 boost . This response was higher than the primary response to E4 . IgM rAbs cloned from E4 boost day 7 GC showed a higher affinity for E4 than those from E4 primed day 7 GC , implying they were memory derived . This demonstrates that IgM memory cells with fewer mutations , from ‘lower’ levels of the memory compartment , participate in secondary responses to variant antigens , and further challenges the hypothesis that highly mutated , class-switched cells elicited by homotypic antigen boosting are a ‘mirror’ of the antibody memory compartment ( Weiss and Rajewsky , 1990 ) . The slower nature of the E4 boost serum response also suggests a lower level of immediate differentiation of memory cells into AFCs than seen with for example the homotypic or E2 response , and is consistent with reduced numbers of high affinity class-switched memory cells recognizing E4 . The serum antibody response to the closer variant , E2 , was more rapid , more cross-reactive and evidenced some earlier affinity maturation . These observations are consistent with a response derived more from the ‘higher’ layers of the E3 specific memory compartment . The IgM +cells induced by E2 boosting have more mutations than after E4 boosting , indicating they are memory derived . As there are higher proportions of these IgM+ GC B-cells , with fewer mutations relative to the homotypic E3 boost , this provides further support for the hypothesis that IgM+ B cells with fewer mutations furnish memory responses to variant antigens Naïve B-cells may contribute to the IgM+ GC B-cells we observe after E4 boosting , although the higher affinities of the rAbs from this group suggest many are memory derived . Also , the slightly higher median level of VH mutation and the higher levels of IgM+ GC B-cells after E4 boosting ( 2x ) compared to priming , suggest IgM +memory B-cells are involved in the boost response consistent with the well established presence of IgM +memory cells with few or no mutations ( Dogan et al . , 2009; Pape et al . , 2011; Kaji et al . , 2012 ) and the known lower activation threshold of memory B-cells in response to antigen ( Good and Tangye , 2007 ) . Whilst E3 specific memory cells may be expected to increase the anti-E3 titre when stimulated by a cross-reactive E4 boost , the small but not significant effect we observe ( Figure 1D ) is consistent with the lowest affinity , least mutated , E3-specific memory cells being stimulated by an E4 boost . Antibodies from such cells may , therefore not add much to the already high , affinity matured , anti-E3 titre induced by E3 priming . The 14-fold higher anti-E4 titre at day 7 after boost ( Figure 1F ) versus day 17 after prime ( Figure 4A ) also argues for a significant contribution from B-cell memory . The fusion-loop epitope in domain 2 of the dengue envelope protein is 100% conserved between strains and in humans , antibodies against this are prevalent in cross-reactive secondary responses ( Lai et al . , 2013; Chaudhury et al . , 2017 ) . The E2 boost response is consistent with this effect , especially considering the restricted clonality seen in VH sequences , but the low anti-E3 titre induced by E4 is not . A recent study ( Chaudhury et al . , 2017 ) showed that the mouse response to recombinant E-protein is predominantly focused on domain 3 of the protein , and so cross reactivity with the fusion loop epitope ( domain 2 ) should be less dominant . While E2 and E4 are 68% and 63% overall identical to E3 , in domain 3 , a focus of mouse antibodies , they are 62% and 51% identical , a bigger difference in differences , helping explain the responses we observe here .
Female 8–11 week old BALB/c mice were purchased from Charles River , U . K . Primary immunisations were intra-peritoneal ( IP ) with 25 μg recombinant Dengue envelope protein ( Biorbyt , UK ) precipitated in alum with 2 × 107 heat-killed B . pertussis . Secondary immunisations were IP with 25 μg recombinant Dengue envelope protein ( Biorbyt ) dissolved in phosphate-buffered saline ( PBS ) . At designated time points mice were anaesthetized and bled for collection of serum and then humanely sacrificed for collection of spleen cells . Dengue envelope ( E ) proteins were C-terminal His-tagged and expressed in E-coli prior to purification . Dengue proteins were tested for endotoxin by LAL assay ( Fisher Scientific , UK ) and contained it at a low level: E2 , 5 . 4EU/μg; E3 , 2 . 5EU/μg; E4 , 3 . 1EU/μg . Endotoxin in this range does not give a detectable physiological response in mice ( Copeland et al . , 2005 ) . ELISA plates ( Nunc Maxisorp , Fisher Scientific , UK ) were coated overnight at 4°C with 1 μg/ml protein in 0 . 1M bicarbonate buffer pH 9 . 3 . Plates were washed three times in PBS/0 . 05% Tween-20 ( Sigma , UK ) ( PBST ) and blocked for 30mins at room temperature with PBST/2% bovine serum albumin ( BSA , Sigma ) . Plates were then washed three times and incubated with serum dilutions in PBST/1 . 0% BSA for two hours at room temperature . After three washes plates were incubated with alkaline-phosphatase conjugated goat anti-mouse IgG ( Sigma ) for one hour at room-temperature , washed three times and developed with pNPP substrate ( Sigma ) for one hour . Absorbance was measured at 405 nm . For the initial rAb screen , rAbs were incubated at 100μgml−1 in PBST/1 . 0% BSA for 2 hr at room temperature on plates coated with E4 and blocked as above , and subsequently treated as above except with use of anti-human IgG second layer ( Sigma ) . Background binding to plates was determined using binding of non-specific polyclonal human IgG at 100μgml−1 , because the rAbs were expressed as chimeric constructs with human constant regions , and this was subtracted from the rAb O . D . Positive binding rAbs were deemed to be those with O . D . > 0 . 1 that could be subject to an ELISA endpoint titration . For the ELISA titration and endpoint analysis , doubling dilutions of positive binding rAbs , and polyclonal IgG background subtraction control , were used starting at 100μgml−1 . Endpoint titre was set at O . D . = 0 . 1 and calculated using interpolation on Graphpad Prism . The assay was repeated using E3 coated plates to determine the rAB cross reactivity . The affinity ( Kd ) of rAbs B5 and G6 ( the two strongest binding rAbs ) was estimated from the inflection point of the ELISA titration curve as indicating 50% maximal binding , and on the assumption that at these higher antibody concentrations binding of rAB to immobilized antigen will have a minor effect on concentration of unbound rAb . We estimated the B5 inflection point to be at approximately 25ugml−1 ( =approx . 150 nM ) and the G6 inflection point to be just above 100ugml−1 ( =approx . 1 uM ) ELISA plates were coated as above with target protein , then washed , blocked and washed as above except the blocking was done at 37°C for one hour . Mouse serum samples were diluted in PBST/1% BSA to twice the concentration of the maximum dilution that gave an absorbance at 405nm = 1 . 0 in ELISA to the target protein . Serial six-fold dilutions of competitor protein were made in PBST/1% BSA , such that the highest concentration of competitor was 2 . 4 μg in 30 μl . 30 μl of diluted serum was mixed with 30 μl of each competitor protein dilution and incubated in a polypropylene 96-well plate at 37°C for 1 hr . Serum/competing antigen mixture ( 50 μl ) was then added to each well of the target antigen coated plate and incubated at 37°C for one hour . Plates were washed as above and then 50 μl of alkaline–phosphatase conjugated anti-mouse IgG ( Sigma ) was added to each well followed by incubation at 37°C for one hour . Plates were washed as above and incubated with 75 μl per well of p-nitrophenyl phosphate substrate ( Sigma ) for one hour at room temperature . Absorbance was measured at 405 nm . All individual serum dilutions were also reacted in the absence of competitor , against BSA coated wells , following the same incubation protocol . These background values were subtracted from the competition ELISA values obtained above . The readings were then normalized so that the samples with the maximum competitor dilution gave a value of 1 . 0 Adapted from Puschnik et al . , 2013 . Assay plates were coated with antigen and blocked as for the ELISA protocol . 1/200 dilutions of serum in PBST/1% BSA were incubated on plates for 2 hr at room temperature . Wells were washed once with PBST , incubated for 10 min at room temperature with PBST or PBST/7M Urea , washed a further two times with PBST and then treated as for standard ELISA . The avidity index was calculated by dividing readings from 7M Urea treatment by readings from PBST-only treatment , after subtraction of background absorbance . Whole spleen cell-suspensions were red-cell depleted with Pharm-Lyse ( BD Biosciences , UK ) and incubated with anti-CD16/32 monoclonal antibody ( Fc-block , BD Biosciences ) for 15 min at 4°C . Cells were then stained with APC anti-B220 , BV421 anti-CD38 , PE anti-CD95/Fas ( all BD ) and FITC anti-IgM ( eBioscience , Thermofisher Scientific , UK ) for 45 min at 4°C . After washing , cells were re-suspended in PBS 5% FCS ( Gibco , Thermofisher Scientific ) and analysed or single-cell sorted on a FACS Aria II ( BD ) . Single GC B-cells were sorted into half a 96 well PCR plate ( less three control wells ) containing10μl of chilled 10 mM Tris pH 8 . 0 , 1 U/μl RNAsin ( Promega , UK ) and placed on dry ice then at −80°C . One-Step RT-PCR ( Qiagen , UK ) was performed according to manufacturers instructions , by adding 15 μl RT-PCR master mix , using first-round primer sets described in Tiller et al . ( 2009 ) , with heavy-chain and kappa-chain primers , for 50 cycles , annealing at 53 . 6°C . Heavy-chain second-round PCRs were performed using 2 μl first-round product and the nested/semi-nested primer sets from Tiller et al . ( 2009 ) , with Hot Star Taq polymerase ( Qiagen ) for 50 cycles annealing at 56°C . Second round PCR product ( 4 μl ) was analysed on a 1 . 2% agarose gel . Successful PCRs were then Sanger sequenced . For this study the sequencing primer was the pan VH primer 5’MsVHE ( Tiller et al . , 2009 ) which leaves part of the 5’ of FR1 unsequenced . For this reason the FR1 sequence was not included in the analysis . VH sequence identification and SHM analysis was done using the IMGT V-Quest online platform . VH sequences are in Supplementary file 1 . Further cloning , construction and expression of antibodies as chimeric IgG1 rAbs was done according to Tiller et al . ( 2009 ) . Briefly , second round PCRs of in-frame VH and VK sequences were repeated with V-gene specific primers that included a restriction site for sub cloning ( Tiller et al . , 2009 ) . These PCR products were purified ( Qiagen ) , restriction digested , purified ( Qiagen ) and ligated ( instant sticky-end ligase , NEB , UK ) into the appropriate expression vector containing either human IgG1 or Kappa constant regions , prior to transformation into E . Coli NEB5-alpha ( NEB ) . Expression constructs in transformed colonies were verified by sequence analysis prior to preparation of plasmid mini-preps ( Qiagen ) . 293A cells were split and grown to 80% confluence in DMEM with ultra-low IgG FCS ( PAN Biotech , Germany ) in 150 mm plates prior to replacement of medium with 20 ml Panserin 293A serum free medium ( PAN Biotech ) . 15 ug each of matched VH and VKappa constructs were added to 2 ml saline with 90 ug PEI , briefly vortexed and rested for 10mins . Transfection solution was added to plates and mixed gently . After 3 days medium was collected , centrifuged at 800 g for 10mins to clear debris , and further medium added . After a further 3 days medium was collected , cleared of debris as before and pooled . 100 ul protein-G sepharose ( GE Healthcare , UK ) was added to supernatants and incubated with rocking overnight at 4°C . Protein G sepharose was collected by centrifugation at 800 g for 10 mins and transferred in PBS to a PBS equilibrated spin column ( Bio-Rad , UK ) . After 3 rounds of washing with 800 ul of PBS , rAbs were eluted in two 200 ul passes of 0 . 1M Glycine ( pH2 . 9 ) into a tube with 40 ul of 1M Tris pH 8 . 0 , 0 . 5% Sodium Azide . Antibody concentrations were determined by O . D . on a Nanodrop instrument ( Thermo ) and corrected for an extinction co-efficient of 1 . 36 . Spleens were harvested from female BALB/c AnCrl mice 39 days after challenge . Splenocytes ( 5 × 105 ) were cultured in triplicate with the indicated concentration of E-protein in X-VIVO 15 medium . Cells were cultured for 96 hr and 0 . 5 μCi of [3H] thymidine was added to wells for 16 hr before measurement with a 1450 MicroBeta counter ( Wallac , Perkin Elmer , UK ) . For statistical analysis sample sizes were chosen to address group size reductions that observe the ARRIVE guidelines . Cages of three mice were randomly allocated to treatment groups . These group treatments were independently biologically replicated to give a sample size of 6 . Where statistical analysis was applied , data points were analysed with Levene’s test for equality of variance and where violated they were subject to a two-tailed Students t-test for unequal variance , otherwise the two-tailed t-test for equal variance . | Many devastating infectious diseases are caused by viruses that change over time . When a vaccine exists , it usually protects against a particular strain of virus , but often fails to defend against new versions of the microbe . This is why the flu vaccine has to be ‘updated’ every year , for example . Vaccines rely on the memory of our immune system . When a virus enters the body , a group of immune cells known as B cells gets activated . Certain B cells can recognise the invader and produce specific proteins , the antibodies , which can target and kill the invader . During the infection some of these B cells become ‘memory B cells’ , having gone through a maturation process that hones their ability to specifically recognize this particular microbe . If the same virus enters the organism again , the memory B cells rapidly identify it and produce a quicker and more efficient immune response than during the first attack . This is how vaccines work . However memory B cells may not be able to recognize a previous intruder if it has changed too much . The memory B cell population is diverse . Some cells are fully mature and can quickly recognize the original virus . But others have not finished their maturation process: these cells are less focussed , and cannot target the original microbe with the same exact precision shown by mature memory cells . For almost forty years it was thought that this reduced focus might make the immature cells better at identifying new versions of the original attacker , but up until now , it was not clear what these memory cells could do . Here Burton , Tennant et al . injected a group of mice with proteins from the Dengue virus , which prompted an immune reaction . After several weeks , the animals received either the same proteins again , or proteins that were different . Compared to the fully mature cells , the immature memory B cells were much better at recognizing the variants of the proteins , and these cells then multiplied and mounted an immune response . Without the original protein injection , the response without the immature memory B cells was not as efficient . The body therefore has a pool of memory B cells that can recognise a wider range of virus protein variants than the ones that caused the first immune reaction . Understanding the role of immature memory B cells in immunity could help design vaccines that protect against several strains or fast-evolving viruses . This could have the potential to reduce the severity of diseases that affect hundreds of millions of people every year . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"immunology",
"and",
"inflammation"
] | 2018 | Variant proteins stimulate more IgM+ GC B-cells revealing a mechanism of cross-reactive recognition by antibody memory |
To improve chemical cross-linking of proteins coupled with mass spectrometry ( CXMS ) , we developed a lysine-targeted enrichable cross-linker containing a biotin tag for affinity purification , a chemical cleavage site to separate cross-linked peptides away from biotin after enrichment , and a spacer arm that can be labeled with stable isotopes for quantitation . By locating the flexible proteins on the surface of 70S ribosome , we show that this trifunctional cross-linker is effective at attaining structural information not easily attainable by crystallography and electron microscopy . From a crude Rrp46 immunoprecipitate , it helped identify two direct binding partners of Rrp46 and 15 protein-protein interactions ( PPIs ) among the co-immunoprecipitated exosome subunits . Applying it to E . coli and C . elegans lysates , we identified 3130 and 893 inter-linked lysine pairs , representing 677 and 121 PPIs . Using a quantitative CXMS workflow we demonstrate that it can reveal changes in the reactivity of lysine residues due to protein-nucleic acid interaction .
Proteins execute diverse functions by interacting with multiple protein partners in different complexes . The study of protein complex structures and protein-protein interactions is critical for understanding their functions . Recently , chemical cross-linking of proteins coupled with mass spectrometry analysis ( CXMS ) has emerged as a powerful tool for the analysis of such structures and interactions ( Sinz , 2006; Leitner et al . , 2010; Petrotchenko and Borchers , 2010; Singh et al . , 2010; Rappsilber , 2011; Bruce , 2012 ) . CXMS methods are less time-consuming and less demanding of sample purity than are traditional methods; this technology has thus been increasing in popularity . Recent progress in the development of analytical instruments , cross-linking reagents , and software has catapulted CXMS from obscurity to prominence , as witnessed by an explosion of successful applications ( Bohn et al . , 2010; Chen et al . , 2010; Kao et al . , 2011; Lauber and Reilly , 2011; Herzog et al . , 2012; Jennebach et al . , 2012; Kalisman et al . , 2012; Kao et al . , 2012; Leitner et al . , 2012; Bui et al . , 2013; Murakami et al . , 2013; Tosi et al . , 2013 ) . However , CXMS is still limited by sample complexity and by low abundances of cross-linked peptides . Extensive fractionation is often required to reduce the complexity of samples that contain macromolecular complexes ( Chen et al . , 2010; Lauber and Reilly , 2011; Jennebach et al . , 2012; Kalisman et al . , 2012; Kao et al . , 2012; Murakami et al . , 2013; Tosi et al . , 2013 ) . The identification of cross-linked peptides in more heterogeneous samples such as crude immunoprecipitates and whole-cell lysates is even more difficult ( Rinner et al . , 2008; Luo et al . , 2012; Yang et al . , 2012; Liu et al . , 2015 ) . Given the sparsity of cross-linked peptides in samples , it would be beneficial to purify them from complex mixtures using affinity tags after cross-linking . However , despite increased efforts to develop chemical cross-linkers with enrichment functions ( Luo et al . , 2012; Trester-Zedlitz et al . , 2003; Fujii et al . , 2004; Chowdhury et al . , 2006; Chu et al . , 2006; Chowdhury et al . , 2009; Kang et al . , 2009; Nessen et al . , 2009; Yan et al . , 2009; Vellucci et al . , 2010; Petrotchenko et al . , 2011; Sohn et al . , 2012; Kaake et al . , 2014 ) , few such agents have been shown to improve identification capabilities in complex samples . Two exceptions include Azide-A-DSBSO , which is used with biarylazacyclooctynone ( Kaake et al . , 2014 ) , and the protein interaction reporter ( PIR ) ( Chavez et al . , 2013; Weisbrod et al . , 2013 ) . However , special instrument control is recommended for their application ( Chavez et al . , 2013; Weisbrod et al . , 2013 ) . In this work , we developed a series of chemical cross-linkers with a modular design as pioneered previously ( Trester-Zedlitz et al . , 2003 ) . They each contain a biotin tag for affinity purification and a cleavage site that can be used to release cross-linked peptides from streptavidin beads . We selected the cross-linker with the best performance and developed a robust enrichment protocol with >97% enrichment efficiency . We termed it Lysine-targeted enrichable cross-linker ( Leiker ) . Using our previously developed pLink identification software ( Yang et al . , 2012 ) , we here demonstrate that the use of Leiker effectively facilitates CXMS analysis in a variety of sample types , from purified complexes , crude immunoprecipitates , to highly complex whole-cell lysates . Quantification of cross-linker modified peptides has the potential to detect protein conformational changes and changes in molecular interactions , though these methods are not mature . To address this potentially critical application of our technology , we synthesized stable isotope-labeled Leiker . Also , we established an automated data analysis workflow for the relative quantitation of light and heavy Leiker cross-links . As a proof of concept , we carried out a quantitative CXMS analysis of an RNA-binding protein L7Ae . Using deuterium-labeled Leiker , we found that for the three L7Ae lysine residues that are buried upon RNA binding , their mono-links decreased dramatically in the presence of RNA , exactly as expected . We further extended the application of quantitative CXMS to a highly complex system consisting of log-phase and stationary-phase E . coli cells and identified a growth phase specific protein interaction .
We aimed to develop a cross-linker similar to the widely used BS3 but that had two major advantages: first , a biotin tag for affinity purification of cross-linked peptides , and second , a cleavage site to release cross-linked peptides after enrichment on streptavidin beads without carrying the biotin group; biotin can interfere with subsequent LC-MS/MS analysis . After experimenting with different designs of Leiker ( Figure 1 , Figure 1—figure supplements 1–6 , and Appendix ) , we found that bAL1 and bAL2 worked the best and there was no difference in performance between these two ( Figure 1—figure supplement 5 ) . Hereafter , Leiker refers to either bAL1 or bAL2 . In this study , bAL2 was used in most of the experiments and a bAL2-based CXMS workflow is illustrated in Figure 2 . Both bAL1 and bAL2 feature a one-piece design with an azobenzene-based chemical cleavage site ( Yang et al . , 2010 ) and a 9 . 3-Å carbon chain that connects two sulfo-NHS esters . This spacer arm is shorter than that of BS3 ( 11 . 4 Å ) , so it may confer a higher specificity to Leiker in capturing protein-protein interactions . Inter- , loop- , and mono-linked peptides generated by either all produce a reporter ion of m/z 122 . 0606 in higher-energy collisional dissociation ( HCD ) spectra ( Figure 2D ) . It can be used to verify the identification of Leiker-cross-linked peptides . For quantitative CXMS analysis , we synthesized isotope-labeled bAL2 in which six hydrogen atoms in the spacer arm were replaced with deuterium ( Figure 1 and Figure 1—figure supplement 6 ) . The six-dalton difference was sufficient to separate peptides cross-linked by [d0]-Leiker from the same peptides cross-linked by [d6]-Leiker . 10 . 7554/eLife . 12509 . 003Figure 1 . Chemical structures of different designs of Leiker . The top panel shows four designs of two-piece Leiker with a photo-cleavage site ( sulfo-PL , PL , and PEG-PL ) or an azobenzene-based cleavage site ( AL ) . Biotin is attached via click chemistry by reacting with bio-aizde . The bottom panel shows two unlabeled ( bAL1 , bAL2 ) and deuterium-labeled ( [d6]-bAL2 ) one-piece Leiker molecules . The biotin moiety is colored magenta . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00310 . 7554/eLife . 12509 . 004Figure 1—figure supplement 1 . Optimization of protein-to-cross-linker ratio ( w/w ) for ( A ) sulfo-PL , ( B ) AL , ( C ) bAL1 , and ( D ) bAL2 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00410 . 7554/eLife . 12509 . 005Figure 1—figure supplement 2 . Evaluation of azobenzene-based chemical cleavage . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00510 . 7554/eLife . 12509 . 006Figure 1—figure supplement 3 . The one-piece Leiker ( bAL1 ) outperformed the two-piece Leiker ( AL ) in the CXMS analysis of a mixture of ten standard proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00610 . 7554/eLife . 12509 . 007Figure 1—figure supplement 4 . Evaluation of the two piece Azo-Leiker ( AL ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00710 . 7554/eLife . 12509 . 008Figure 1—figure supplement 5 . bAL1 and bAL2 performed similarly . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00810 . 7554/eLife . 12509 . 009Figure 1—figure supplement 6 . MS1 spectra of ( A ) [d0]-bAL2 and ( B ) [d6]-bAL2 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 00910 . 7554/eLife . 12509 . 010Figure 2 . Scheme of the Leiker-based CXMS workflow . ( A ) Leiker contains a biotin moiety ( magenta ) , a cleavage site ( arrows ) , and six hydrogen atoms that are accessible to isotope labeling ( asterisks ) . ( B ) The workflow for purification of Leiker-linked peptides . ( C ) Three types of Leiker-linked peptides . ( D ) Leiker-linked peptides generate a reporter ion of 122 . 06 m/z in HCD , as shown in the spectrum of an inter-linked peptide NYQEAKDAFLGSFLYEYSR-LAKEYEATLEECCAK ( +4 charged , MH+ 4433 . 0553 ) , in which C denotes carbamidomethylated cysteine . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 010 To assess to what extent Leiker could improve the identification of low-abundance cross-linked peptides from a complex background , a mixture of ten standard proteins ( Figure 3—source data 1 ) , consisting of RNase A , lysozyme , PUD-1/PUD-2 heterodimer , GST , aldolase , BSA , lactoferrin , β-galactosidase , mouse monoclonal antibody , and myosin , was treated with Leiker , digested with trypsin , and then diluted or not with a tryptic digest of non-cross-linked E . coli lysates at different ratios ( 1:1 , 1:10 , and 1:100 , w/w ) . The peptide mixture was then incubated with streptavidin agarose . After extensive washes , cross-linked peptides were released using the Na2S2O4 elution buffer . BS3 was used in parallel as a control . As is shown in Figure 3A the number of BS3-linked peptide pairs identified decreased dramatically , from 109 in the undiluted sample to only one in the 100-fold diluted sample . The number of Leiker-linked peptide pairs identified after enrichment was in no way affected by increasing background complexity , with >160 inter-links detected in each sample . To be noted , these inter-linked peptide pairs , or inter-links for abbreviation , can result from either intra-protein or inter-protein cross-linking ( illustrated in Figure 2B ) . Strikingly , cross-linking products , including inter- , loop- and mono-links ( Figure 2C ) constituted over 97% of all peptides identified post enrichment ( Figure 3A ) . Of the Leiker-linked lysine pairs that can be mapped to the pdb structures ( Figure 3—source data 1 ) , 82% have Cα – Cα distance ≤22 Å and 93% have Cα – Cα distance ≤30 Å ( FDR < 5% , E-value < 0 . 01 ) , which is comparable to BS3 ( Figure 3—figure supplement 1 ) . This result demonstrated that Leiker enables effective enrichment of cross-linked peptides . 10 . 7554/eLife . 12509 . 011Figure 3 . Evaluating the performance of Leiker . ( A ) Leiker allowed near 100% enrichment of target peptides from a cross-linked ten-protein mixture diluted with increasing amounts of non-cross-linked E . coli lysates . Dark blue , inter-links; light blue , mono-links; green , loop-links; grey , regular peptides not modified by Leiker . ( B ) Number of cross-link identifications from E . coli lysates treated with Leiker or BS3 . Shown in the left and right panels are the identified spectra and peptides , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01110 . 7554/eLife . 12509 . 012Figure 3—source data 1 . Ten standard proteins used to evaluate Leiker , mixed at equal amounts by mass . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01210 . 7554/eLife . 12509 . 013Figure 3—source data 2 . Summary of identified spectra from the ten-protein mixture . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01310 . 7554/eLife . 12509 . 014Figure 3—figure supplement 1 . Distance distributions of cross-linked lysine pairs in the undiluted ten-protein mixture . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 014 The ten standard proteins also allowed us to assess the specificity of Leiker . Because Leiker has more functional groups than BS3 does , a concern arises that Leiker may produce more cross-linking artifacts . Cross-links between non-interacting proteins are surely artifacts , which include all the inter-protein cross-links identified from the ten-protein mixture except those between the light-chain and the heavy-chain of myosin , between the light-chain and the heavy-chain of an IgG antibody , and between PUD-1 and PUD-2 , which form a heterodimer . We found that the percentage of artifactual cross-links is 3% for both Leiker and BS3 ( Figure 3—source data 2 ) , fitting with the filtering criteria that were applied ( FDR cutoff 0 . 05 followed by E-value cutoff 0 . 01 ) . The results demonstrate that Leiker is as specific as BS3 . Further , we cross-linked highly complex E . coli lysates with either Leiker or BS3 for a side-by-side comparison . After enrichment and a single reverse phase LC-MS/MS analysis , Leiker yielded at least a fourfold increase in the number of inter-links identified ( Figure 3B ) . Next , we applied Leiker to real-world samples , starting with purified E . coli 70S ribosome , a 2 . 5 MDa ribonucleoprotein ( RNP ) complex consisting of more than 50 proteins . A total of 222 inter-linked lysine pairs were identified with high confidence , including 95 inter-molecular and 127 intra-molecular cross-links ( Figure 4—source data 1 ) . This is three times as many as in a previous study ( Lauber and Reilly , 2011 ) . Of the 95 cross-links connecting two lysine residues that are both present in the crystal structure of a 70S ribosome ( Fischer et al . , 2015 ) ( PDB code: 5AFI ) , 75% are compatible with the crystal structure with a Cα-Cα distance ≤22 Å , which is the length of the spacer arm of Leiker plus two lysine side chains . Among the subset of intra-molecular cross-links , 84% have Cα-Cα distances ≤22 Å; among the subset of inter-molecular cross-links , 50% have Cα-Cα distances ≤22 Å and 73% have Cα-Cα distances ≤30 Å , which could be a reasonable cutoff considering conformation flexibility of proteins in solution ( Figure 4—source data 1 and Figure 4—figure supplement 1 ) . One particular ribosomal protein L9 is a good example to illustrate conformational flexibility and the dynamic nature of interactions between proteins or protein complexes . A large b-factor in the crystal structure has suggested that L9 is highly mobile . It has been observed to adopt an extended , rod-like conformation in the crystal structure ( Schuwirth et al . , 2005 ) and a strikingly different bent conformation in the solution structure of the ribosome determined using cryo-EM ( Fischer et al . , 2015; Seidelt et al . , 2009 ) . Bending of L9 was echoed in this study , as reflected in the cross-links bridging L9 and L2 and the cross-links bridging the two termini of L9 ( Figure 4—figure supplement 2 ) . Three additional cross-links involving L9 have Cα-Cα distances >50 Å if measured within a ribosomal particle ( Figure 4—source data 1 ) . We propose these apparently long distance cross-links , which are similar to the ones observed in a previous CXMS study ( Lauber and Reilly , 2011 ) , reflect interactions between ribosomal particles . L9 locates at the interface between ribosomal particles in higher-order configurations ( e . g . polysome ) ( Brandt et al . , 2009 ) . Dimerization or oligomerization of 70S ribosomes in the absence of mRNA was also observed using negative staining EM from highly purified non-cross-linked 70S ribosomes ( Figure 4—figure supplement 3 ) . The peripheral regions of the ribosome are critical for protein translation and regulation ( Savelsbergh et al . , 2000; Valle et al . , 2003; Kothe et al . , 2004 ) . However , despite of extensive studies on the ribosome structures , these peripheral parts are still largely missing because they are either too dynamic or refractory to crystallography . For E . coli ribosomal proteins currently lacking well-defined coordinates in the 70S crystal structures , Leiker-based CXMS provided remarkably more linkages than other ribosomal proteins . The top four proteins with the most inter-molecular cross-links identified are S1 , L1 , L7/12 , and L31 , all of which are mobile components in the peripheral regions and often invisible in the crystal structure ( Figure 4A and Figure 4—source data 2 ) . S1 is the largest ribosomal protein , which binds to mRNA and initiates translation , but has no high-resolution structures available either alone or in the context of the 70S ribosome ( Fischer et al . , 2015; Lauber et al . , 2012 ) . A previous CXMS analysis of the 30S subunit revealed interaction between S1 and a region near the 3’ end of 16S rRNA ( Lauber et al . , 2012 ) , but it is unknown whether or not S1 interacts with the 50S subunit . Our analysis of the 70S ribosome revealed extensive contacts between the C-terminal mRNA binding domain of S1 and L1 in the 50S subunit ( Figure 4—figure supplement 4A ) . Since the two of them localize to a region where both tRNA and mRNA leave the ribosome , the observation of six cross-links between them hints that there might be a coordination between deacylated tRNA release and mRNA exit from the ribosome . The 30S proteins that were found to interact with S1 in this study were largely consistent with those identified in the previous study ( Lauber et al . , 2012 ) . In particular , four cross-links were identified between the N-terminal peptide of S1 ( M1-K14 ) and the N-terminal peptide of S2 ( M1-K11 ) ( Figure 4—source data 1 and Figure 4—figure supplement 4A ) . This result agrees perfectly with a recent structural finding on the direct interaction between S1 and S2 ( Byrgazov et al . , 2015 ) . L1 had the highest number of cross-links with S1 , followed by cross-links with L33 , L5 , L9 , S13 , and S2 ( Figure 4—figure supplement 4B ) . The proximity of L1 to L33 and L5 implicates a rotated conformation of L1 in the sample ( Figure 4—source data 3 ) , which was repeatedly observed in various structures of the 70S ribosome in different functional states ( Valle et al . , 2003 ) . Furthermore , beyond the expected interactions between L7/12 and L6 , L10 , or L11 ( Diaconu et al . , 2005 ) , we also found novel interactions between L7/12 and L19 or S3 ( Figure 4A and Figure 4—figure supplement 4C ) . These findings suggest that the highly flexible L7/12 stalk might be able to contact the 30S subunit , given the predicted large length of this dynamic stalk ( Diaconu et al . , 2005 ) . Nine cross-links between E . coli L31 and L5 placed L31 in the central protuberance region ( Figure 4—figure supplement 4D ) , which is supported by the crystal structure of T . thermophilus 70S ribosome ( Voorhees et al . , 2009 ) and the newly revealed structure of 70S ribosome ( Fischer et al . , 2015 ) . Together , these results demonstrate that Leiker-based CXMS analysis can provide structural information that is highly complementary to crystallography and cryo-EM , especially for the flexible or dynamic regions that cannot be deduced using traditional methods . 10 . 7554/eLife . 12509 . 015Figure 4 . Leiker-based CXMS analyses of large protein assemblies . ( A ) Analysis of a purified E . coli 70S ribosome revealed the locations of highly dynamic periphery ribosomal proteins S1 , L1 , and L7/12 that were refractory to crystallography and cryo-EM analysis . Cross-links to S1 , L1 , and L7/12 are colored red , blue , and yellow , respectively , and the cross-linked residues on these three proteins are numbered according to the Uniprot sequences . ( B ) Analysis of a crude immunoprecipitate of the yeast exosome complex . Dashed blue and grey lines denote 50 compatible and 22 incompatible cross-links , respectively , according to the structure of the RNA-bound 11-subunit exosome complex ( PDB code: 4IFD ) . Rrp44 , green; Rrp40 , orange; Rrp4 , violet; Rrp42 , gold; other exosome subunits , yellow; RNA , black . Known and candidate exosome regulators revealed by Leiker-cross-links are shown along the periphery and highlighted in green and yellow circles , respectively . ( C ) Connectivity maps of the ten-subunit exosome core complex based on the inter-molecular cross-links identified in the current IP-CXMS experiments or on previous yeast two-hybrid ( Y2H ) studies ( Stark et al . , 2006; Uetz et al . , 2000; Oliveira et al . , 2002; Luz et al . , 2007; Yu et al . , 2008 ) . Blue solid lines: experimentally identified putative direct protein-protein interactions; grey dashed lines: theoretical cross-links according to the crystal structure; Cα-Cα distance cutoff ≤30 Å . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01510 . 7554/eLife . 12509 . 016Figure 4—source data 1 . CXMS analysis of E . coli 70S ribosomes . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01610 . 7554/eLife . 12509 . 017Figure 4—source data 2 . Number of cross-linked lysine pairs classified by ribosomal proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01710 . 7554/eLife . 12509 . 018Figure 4—source data 3 . Identified cross-linked lysine pairs involving L1 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01810 . 7554/eLife . 12509 . 019Figure 4—source data 4 . CXMS analysis of the Saccharomyces cerevisiae exosome complex . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 01910 . 7554/eLife . 12509 . 020Figure 4—figure supplement 1 . Distance distribution of the inter-molecular and intra-molecular cross-links identified in 70S ribosomes . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02010 . 7554/eLife . 12509 . 021Figure 4—figure supplement 2 . Alignment of L9 and L2 from the crystal structure ( L9 , orange; L2 , wheat; PDB code: 2AW4 ) and their counterparts from the cryo-EM reconstruction ( L9 , blue; L2 , lightblue; PDB code: 5AFI ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02110 . 7554/eLife . 12509 . 022Figure 4—figure supplement 3 . Negative staining of non-cross-linked E . coli 70S ribosome . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02210 . 7554/eLife . 12509 . 023Figure 4—figure supplement 4 . Connectivity maps of cross-links involving ( A ) S1 , ( B ) L1 , ( C ) L7/12 , and ( D ) L31 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02310 . 7554/eLife . 12509 . 024Figure 4—figure supplement 5 . Silver-stained SDS-PAGE gel of the crude immunoprecipitate of TAP-tagged Rrp46 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02410 . 7554/eLife . 12509 . 025Figure 4—figure supplement 6 . Number of identified inter-linked peptide pairs from decreasing amount of Leiker-cross-linked exosome immunoprecipitate ( FDR < 0 . 05 , E-value < 0 . 01 ) . After enrichment , 30% ( orange ) or 60% ( blue ) of each sample was analyzed by LC-MS/MS . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 025 Combining CXMS and immunoprecipitation ( IP ) has great potential for the detection of binding partners in close proximity among co-immunoprecipitated proteins; such method may be widely adopted in biology laboratories . Much progress has been made recently in this area by the use of a modified anti-GFP single-chain antibody that cannot be cross-linked so that GFP-tagged protein complexes can be cross-linked on beads and separated away from the antibody for CXMS analysis ( Shi et al . , 2015 ) . For highly heterogeneous IP samples , however , cross-linked peptides can be inundated by non-cross-linked peptides even if the antibody is removed from the background . As a test , we prepared a crude immunoprecipitate of a TAP-tagged yeast exosome subunit Rrp46 ( Figure 4—figure supplement 5 ) , from which 740 proteins were identified at 0 . 1% protein FDR . The immunoprecipitated proteins were eluted off IgG beads and cross-linked with Leiker . To evaluate the sensitivity of the method , we varied the amount of immunoprecipitates from 40 μg to 3 μg of proteins and found that the number of inter-link identifications did not change much as the input decreased from 40 to 20 μg ( Figure 4—figure supplement 6 ) . From three experiments starting with 40 μg of proteins , a total of 195 cross-linked lysine pairs ( 43 inter-molecular and 152 intra-molecular ) were identified ( Figure 4B and Figure 4—source data 4 ) . Thanks to cross-linking , not only did we identify all ten exosome core subunits , but also 15 putative direct protein-protein interactions amongst the core subunits , which generated a connectivity map more complete than the one from yeast two-hybrid experiments ( Stark et al . , 2006; Uetz et al . , 2000; Oliveira et al . , 2002; Luz et al . , 2007; Yu et al . , 2008 ) and showed that among the co-immunoprecipitated proteins , Rrp41 and Rrp45 directly bind to the bait protein Rrp46 ( Figure 4C ) . Of the cross-links identified , 69% were compatible with the crystal structure of an RNA-bound 11-subunit exosome complex ( Makino et al . , 2013 ) ( PDB code: 4IFD ) . Among the cross-links that disagreed with the RNA-bound structure , 68% involved the catalytic subunit Rrp44 , which has a large rotation relative to the rest of the exosome core between the RNA-bound and the RNA-free states ( Makino et al . , 2013; Liu et al . , 2014 ) . The crude Rrp46 immunoprecipitate should mainly contain apo exosome , because magnesium was included in the buffer to activate the nuclease activity of exosome . Therefore , the presence ( in the crystal structure ) or absence ( in our exosome preparation ) of bound RNA is likely to be the primary reason behind most of the seemingly inconsistent inter-molecular cross-links . To fulfill different functions in multiple biological processes ( Houseley and Tollervey , 2009 ) , the core exosome complex must recruit additional regulators , of which only a few are known . Here we found two known ( Mpp6 [Milligan et al . , 2008] and Ski7 Araki et al . , 2001 ) and four potential exosome regulators through nine cross-links with core exosome subunits ( Figure 4B and Figure 4—source data 4 ) . These cross-links revealed residues in close proximity . Ski7 was found to cross-link with Rrp4 via K111 , which fits well with previous co-IP results obtained by using different fragments of Ski7 ( Araki et al . , 2001 ) and a recently published CXMS study of the yeast exosome ( Shi et al . , 2015 ) . Among the newly identified candidate regulators , the translation initiation factor Tif1 stood out; it had interactions with the Rrp4 and Rrp44 exosome subunits ( Figure 4B ) . Translation has been implicated in RNA quality control ( Shoemaker and Green , 2012 ) . The linkages identified here support the hypothesis that exosome complexes ‘stand by’ the translation machinery and recognize and degrade aberrant mRNA molecules . We further tested Leiker for the purpose of mapping protein-protein interaction networks using E . coli and C . elegans lysates . E . coli whole-cell lysates are commonly used for evaluating CXMS methods ( Rinner et al . , 2008; Yang et al . , 2012; Weisbrod et al . , 2013 ) . In three independent experiments , Leiker-treated , trypsin digested E . coli lysates were fractionated on a high pH reverse phase column , and cross-linked peptides were enriched from each of the 10 or 11 fractions ( Figure 5—figure supplement 1 ) . After filtering the data by requiring FDR < 0 . 05 , E-value < 0 . 01 , spectral count ≥ 3 , we identified a total of 2003 non-redundant inter-linked lysine pairs including 1386 ( 69% ) intra-molecular and 617 ( 32% ) inter-molecular cross-links ( Figure 5—source data 1 and Figure 5—figure supplement 2 ) . Protein structure information is available in the PDB database for 984 intra-molecular cross-links identified with Leiker , and is consistent with 80% of them , indicating the high quality of the results ( Figure 5—source data 1 ) . Of note , the inter-molecular cross-links represent 436 pairs of protein-protein interactions , and 25% of the cross-links are supported by the combined network of the bacteriome . org database ( Peregrín-Alvarez et al . , 2009 ) Figure 5—source data 1 ) . Most of the inter-molecular cross-links suggest novel protein-protein interactions . Based on the Leiker cross-links , we constructed a protein-protein interaction network and extracted the most highly connected module ( Bader and Hogue , 2003; Saito et al . , 2012 ) ( Figure 5A ) . This 12-protein module consists of 9 ribosomal proteins and two DNA-binding proteins ( the Hu heterodimer DBHA/DBHB ) organized around a translation elongation factor Tu ( EF-Tu ) . Evidently , it is enriched with proteins that function in translation , suggesting that DBHA/DBHB also plays a role in this process . Indeed , previous studies reported that a small fraction of this Hu heterodimer is bound to ribosomes ( Rouvière-Yaniv and Kjeldgaard , 1979 ) and that this protein can enhance or repress translation of the mRNA molecules that it binds to ( Balandina et al . , 2001 ) . In contrast , the most connected module obtained from the previously identified BS3 cross-links ( Yang et al . , 2012 ) comprised only three ribosomal proteins ( Figure 5A ) . These results indicate the potential of Leiker in generating comprehensive protein-protein interaction networks using CXMS . Since ribosomal proteins dominated E . coli whole-cell lysates , we prepared samples in which ribosomes were removed by centrifugation through a layer of sucrose cushion . Analysis of the ribo-free samples ( two repeats ) with Leiker identified 1971 inter-links , 1127 of which were not identified in the whole-cell lysates ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 3 ) ( Figure 5B , Figure 5—figure supplement 2 , and Figure 5—source data 2 ) . Together , we identified a total of 3130 non-redundant cross-linked lysine pairs from E . coli . This allowed us to construct a network comprising 677 protein-protein interactions ( Figure 5—figure supplement 3A ) . 10 . 7554/eLife . 12509 . 026Figure 5 . CXMS analyses of E . coli and C . elegans lysates . ( A ) The best protein-protein interaction cluster extracted from the Leiker-identified or BS3-identified ( Yang et al . , 2012 ) inter-links from E . coli whole-cell lysates . Node size represents the degree of connectivity of the indicated protein in the network . Line width represents the spectral counts of every inter-molecular cross-link . The line color is set to blue when the two peptides of an inter-link are both attributed to unique proteins , to grey if either could be assigned to multiple proteins . All the lines connected to EF-Tu1 are grey because EF-Tu1 differs from EF-Tu2 by only one amino acid . ( B ) Comparison of the identified inter-links in E . coli whole-cell lysates and ribo-free lysates ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 3 ) . ( C and D ) Comparison of the number of Leiker-identified inter-links and that of BS3-identified inter-links ( Yang et al . , 2012 ) from C . elegans ( C ) and E . coli ( D ) whole-cell lysates ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02610 . 7554/eLife . 12509 . 027Figure 5—source data 1 . CXMS analysis of E . coli whole-cell lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02710 . 7554/eLife . 12509 . 028Figure 5—source data 2 . CXMS analysis of E . coli ribo-free lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02810 . 7554/eLife . 12509 . 029Figure 5—source data 3 . CXMS analysis of C . elegans whole-cell lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 02910 . 7554/eLife . 12509 . 030Figure 5—source data 4 . CXMS analysis of C . elegans mitochondrial proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03010 . 7554/eLife . 12509 . 031Figure 5—figure supplement 1 . Fractionation of digested , Leiker-treated E . coli lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03110 . 7554/eLife . 12509 . 032Figure 5—figure supplement 2 . Overlap of cross-linked lysine pairs between biological replicates of E . coli lysates ( FDR < 0 . 05 , E-value < 0 . 01 , and spectral count ≥ 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03210 . 7554/eLife . 12509 . 033Figure 5—figure supplement 3 . Protein-protein interaction networks constructed from the cross-links identified in ( A ) E . coli and ( B ) C . elegans . The labeling scheme is the same as described in Figure 5A except for the node color . For E . coli , node color is set to orange if the protein was only identified in the whole-cell lysates , to yellow only identified in the ribo-free lysates , or to green if identified in both . There are 626 proteins in the E . coli network and 155 proteins in the C . elegans network . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 033 Applying Leiker to an even more complex lysate from C . elegans , which has a similar number of protein coding genes as human ( ~20 , 000 ) , we identified 459 inter-links ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 3 ) ( Figure 5—source data 3 ) . We also analyzed a C . elegans mitochondrial fraction and identified 547 inter-linked lysine pairs ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 3 ) , of which 434 were not detected in the whole-worm lysate ( Figure 5—source data 4 ) . Together , we identified 893 non-redundant cross-linked lysine pairs from C . elegans and constructed protein-protein interactions between 155 proteins ( Figure 5—figure supplement 3B ) . In order to compare with previous studies , we also applied a less stringent cutoff ( 5% FDR , E-value < 0 . 01 , spectral count ≥ 1 ) to the data sets of E . coli and C . elegans whole-cell lysates . This allowed us to determine that the number of C . elegans cross-links identified in this study was 23 times as many as the previous record ( Figure 5C ) ( Yang et al . , 2012 ) . The number of E . coli cross-links identified in this study is four times greater than the number of PIR-identified inter-links ( Chavez et al . , 2013 ) and eight times greater than the number of BS3-identified inter-links ( Yang et al . , 2012 ) . Half of the BS3-identified cross-links ( Yang et al . , 2012 ) were recapitulated in this study ( Figure 5D ) . Relative quantification of cross-linker modified peptides can reveal changes in protein conformation and/or interactions between a protein and another molecule ( e . g . nucleic acid , ligand , or protein ) . To apply Leiker in quantitative CXMS , we synthesized deuterium-labeled Leiker ( [d6]-bAL2 ) in addition to the unlabeled version ( [d0]-bAL2 ) . Few software tools reported to date directly support quantitative CXMS ( Fischer et al . , 2013; Walzthoeni et al . , 2015 ) . We therefore modified the quantification software pQuant ( Liu et al . , 2014 ) and established an automated data analysis workflow for quantitative CXMS ( Figure 6 and Materials and methods ) . As a proof-of-principle experiment , we compared the RNA-free and H/ACA RNA-bound states of a Pyrococcus furiosus ribosomal protein L7Ae ( Rozhdestvensky et al . , 2003; Li and Ye , 2006 ) . We treated RNA-free L7Ae with [d0]-bAL2 and the assembled L7Ae-RNA complex with [d6]-bAL2 in the forward reaction , and switched the isotope labels in the reverse reaction ( Figure 7A ) . An equal amount of BSA protein was included in each sample to control for possible difference in cross-linking efficiency between [d0]- and [d6]-Leiker . We expected that the formation of the protein-RNA complex would block the access of Leiker to lysine residues at the binding interface , which would be manifested as large abundance decrease of mono- or inter-linked peptides at these sites . 10 . 7554/eLife . 12509 . 034Figure 6 . Workflow for quantification of cross-linked peptides using pQuant . For each identified cross-link spectrum , an extracted ion chromatogram ( EIC ) is constructed for each isotopic peak of the [d0]- and [d6]-labeled precursor . The [d6]/[d0] ratios can be calculated based on the monoisotopic peak , the most intense peak , or the least interfered peak of each isotopic cluster as specified by users . The accuracy of the ratio calculation was evaluated with the confidence score σ ( range: 0–1 , from the most to the least reliable ) . If a cross-link have ratios with σ < 0 . 5 , the median of these ratios is assigned to this cross-link . The cross-link ratios of the proteins of interest are normalized to the median ratio of all BSA cross-links . For each cross-link , the median [state1]/[state2] ratio of three independent forward labeling experiments is plotted against the median ratio of three independent reverse labeling experiments . Cross-links that are only present in state1 or state2 due to a dramatic conformational change cannot be quantified as described above because the ratios would be zero or infinite and their σ values would be 1 . Therefore , if a cross-link does not have a valid ratio after automatic quantification , the EICs were manually inspected to determine if it was an all-or-none change . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03410 . 7554/eLife . 12509 . 035Figure 7 . Quantitative CXMS analysis of the L7Ae-RNA complex . ( A ) Reciprocal labeling of RNA-free ( F ) and RNA-bound ( B ) L7Ae with [d0]/[d6]-Leiker . ( B ) Abundance ratios of mono-links ( F/B ) in the forward ( F[d0]/B[d6] ) and the reverse labeling experiment ( F[d6]/B[d0] ) . Each circle represents a mono-linked lysine residue and is colored red if it has a ratio greater than five in both labeling schemes . ( C ) The three lysine residues affected by RNA binding are highlighted in the structure model ( PDB code: 2HVY ) . The number below each such lysine residue indicates the buried surface area ( Å2 ) upon RNA binding . ( D ) Extracted ion chromatograms ( left ) and representative MS1 spectra ( right ) of a K42 mono-link . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03510 . 7554/eLife . 12509 . 036Figure 7—source data 1 . Quantitative CXMS analysis of L7Ae with or without the H/ACA RNA . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03610 . 7554/eLife . 12509 . 037Figure 7—figure supplement 1 . Extracted ion chromatograms ( left ) and representative MS1 spectra ( right ) of a mono-linked peptide corresponding to ( A ) K35 and ( B ) K84 . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 037 Mono-linked peptides are usually neglected in CXMS , but they are valuable because they indicate that the modified lysine residues are exposed to solvent . Mono-links at all 15 lysine residues and the N-terminus of L7Ae were reliably quantified ( Figure 6 ) in both forward and reverse labeling experiments . Three mono-links at K35 , K42 , and K84 consistently had significantly higher abundance ( >5 fold ) in the RNA-free state ( F ) than in the RNA-bound state ( B ) ( Figure 7B–D , Figure 7—figure supplement 1 and Figure 7—source data 1 ) . None of inter-links passed the quantification criteria described above . These results suggest that the three lysine residues are buried upon RNA binding , either due to direct protein-RNA binding or indirect protein conformational changes induced by RNA binding . This is in perfect agreement with the crystal structure ( Li and Ye , 2006 ) ( PDB code: 2HVY ) , which shows that K35 , K42 , and K84 all bind to the RNA , each with a buried area greater than 20 Å2 ( Figure 7C ) . Lastly , we applied quantitative CXMS to E . coli lysates . The log phase and the stationary phase cell lysates were cross-linked , respectively , with [d0]- and [d6]-bAL2 in the forward labeling experiment , or with [d6]- and [d0]-bAL2 in the reverse labeling experiment . After a single enrichment step without pre-fractionation , a total of 161 inter-linked lysine pairs were quantified in both the forward and the reverse labeling experiments , and most of them had similar [log phase]/[stationary phase] ratios in the two experiments ( Figure 8 and Figure 8—source data 1 ) . Noticeably , the cross-link between YqjD and ElaB increased at least 10 times in the stationary phase compared to the log phase . These two paralogous proteins are associated with the inner membrane of E . coli cells through their C-terminal transmembrane motifs and both bind to stationary phase ribosomes , probably through their N-terminal regions ( Yoshida et al . , 2012 ) . It is suggested that YqjD binding to ribosomes inhibits translation ( Yoshida et al . , 2012 ) . Association of YqjD and ElaB has been detected but the sites of interaction are not known ( Hu et al . , 2009 ) . Here , our results not only confirm previous findings , but also provide new insights that YqjD and ElaB form a heterodimer through their central regions , presumably as a stronger , divalent anchoring site for ribosomes to inhibit protein translation in the stationary phase . 10 . 7554/eLife . 12509 . 038Figure 8 . Quantitative CXMS analysis of E . coli lysates . Abundance ratios of ( A ) inter-linked lysine pairs and ( B ) mono-linked sites in the forward ( [log phase]d0/[stationary phase]d6 ) and the reverse labeling experiment ( [log phase]d6/[stationary phase]d0 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 03810 . 7554/eLife . 12509 . 039Figure 8—source data 1 . Quantitative CXMS analysis of E . coli lysates . DOI: http://dx . doi . org/10 . 7554/eLife . 12509 . 039
In this study , we developed an MS-friendly and isotope-encodable cross-linker called Leiker that enables the efficient enrichment of cross-linked peptides through biotin-based immobilization and azobenzene-based chemical cleavage . With an enrichment efficiency of 97% or more , Leiker yields a fourfold increase in the number of identified cross-linked peptide pairs from complex samples . Also established is a workflow for quantitative CXMS based on deuterium-labeled Leiker . In theory , a comprehensive network of putative direct protein-protein interactions could be obtained by applying Leiker to lysates . However , the interaction networks obtained as such are limited , because the cross-links identified are dominated by those from highly abundant proteins , for example , EF-Tu and ribosomal proteins in E . coli . This can be overcome with subcellular fractionation , which can separate abundant proteins from less abundant ones . We increased the number of unique inter-link identifications by more than 50% ( from 2003 to 3130 ) by simply removing ribosomes from the E . coli lysates ( Figure 5B ) . This is also obvious by contrasting the CXMS results of the whole-worm lysate and the mitochondrial fraction of C . elegans , from which 459 and 547 inter-linked lysine pairs were detected , respectively , with an overlap of only 113 . We anticipate that extensive protein fractionation coupled with Leiker-assisted CXMS will pave the way towards constructing comprehensive interactomes for different model organisms , and next-generation cross-link identification software of higher sensitivity will also help . Further , with the advantage of heavy isotope labeling for quantification in addition to the enrichment function , Leiker shows promise for use in differential interactome analysis ( Ideker and Krogan , 2012 ) . When we examined the cross-links identified from E . coli against the protein structures deposited in the PDB database , we noted that the intra-molecular cross-links in both the whole-cell lysates and the ribo-free samples had similar rates of structural compatibility ( 80% and 84% , respectively ) . This shows that the quality of our Leiker-based CXMS data is high . Interestingly , the inter-molecular cross-links detected from the ribo-free samples had a much higher rate of structural compatibility ( 69% ) than those detected in the whole-cell lysates ( 12% ) . Given that 92% of the inter-links with existing structural information in the whole-cell lysate involved at least one ribosomal protein and many were between ribosomal proteins , we think that most of the apparently incompatible inter-molecular cross-links seen in the whole-cell lysates likely result from cross-linking of adjacent ribosomal particles . Previous cross-linking studies have typically treated mono-linked peptides as by-products , and have ignored them . This is regrettable , as they carry structural information about proteins and always outnumber inter-links ( Figure 3 ) . Leiker also generates abundant mono-links . In this study , we demonstrate that mono-links are highly valuable in mapping RNA-binding lysine residues . As the positively charged lysine residue is frequently involved in binding the negatively charged phosphate backbone of DNA and RNA , relative quantification of lysine mono-links would be particularly suited for mapping the DNA or RNA binding surface on a protein . We suggest that mono-link data should be used in routine practice .
Acetonitrile , methanol , formic acid , ammonium bicarbonate , and acetone were purchased from J . T . Baker ( Center Valley , PA ) . Dimethylsulfoxide ( DMSO ) , HEPES , urea , thiourea , and other general chemicals were purchased from Sigma-Aldrich ( St . Louis , MO ) . Trypsin and Lys-C were purchased from Promega ( Wisconsin , WI ) . Bis ( sulfosuccinimidyl ) suberate ( BS3 ) , streptavidin agarose resin , and high capacity streptavidin agarose resin were purchased from Pierce ( Rockford , IL ) . Dynabeads M-280 streptavidin was purchased from Invitrogen ( Carlsbad , CA ) . RNase A , lysozyme , aldolase , BSA , lactoferrin , β-galactosidase , and myosin were obtained from Sigma-Aldrich . Recombinant GST containing an N-terminal His tag was expressed in E . coli BL21 cells from the pDYH24 plasmid and purified with glutathione sepharose ( GE Healthcare , Piscataway , NJ ) . PUD-1/PUD-2 heterodimers were purified on a HisTrap column followed by gel filtration . Stock solutions of the ten standard proteins were individually buffer exchanged into 20 mM HEPES , pH8 . 0 by ultrafiltration , and then mixed to make a total protein mixture with a 2 µg/µl protein concentration . Purification of 70S ribosomes from E . coli cells was performed as previously described ( Guo et al . , 2011 ) . E . coli cells ( DH5α ) were grown in 2 L LB medium to an OD600=0 . 8 . Cells were collected by centrifugation , washed with 100 mL lysis buffer ( 50 mM HEPES-KOH , pH 7 . 5 , 500 mM KCl , 12 mM MgCl2 , 1 mM DTT , 1 mM PMSF ) and resuspended in 100 mL of lysis buffer . Cells were then disrupted with an Ultrasonic Cell Disruptor . The lysate was clarified at 13 , 000 rpm for 1 hr at 4°C in a JA 25 . 50 motor ( Beckman Coulter , UK ) . The supernatant was layered on a sucrose cushion ( 50 mM HEPES-KOH , pH 7 . 5 , 500 mM KCl , 12 mM MgCl2 , 33% sucrose ) and centrifuged at 30 , 000 rpm for 18 hr in a 70Ti rotor ( Beckman Coulter ) at 4°C . The supernatant was collected as the ribo-free lysate . The pellet was resolved with a buffer containing 50 mM HEPES-KOH , pH 7 . 5 , 500 mM KCl , and 12 mM MgCl2 . The crude ribosomes were then layered on a 10–50% sucrose gradient ( 50 mM HEPES-KOH , pH 7 . 5 , 500 mM KCl , 12 mM MgCl2 , 10% to 50% sucrose ) and centrifuged at 28 , 000 rpm for 5 hr in an SW28 rotor ( Beckman Coulter ) at 4°C . The gradient was scanned at 260 nm and fractionated in an ISCO gradient collector . The fractions of 70S ribosomes were pooled and concentrated with Amicon Ultra centrifugation filters ( Millipore , China ) with a buffer containing 50 mM HEPES-KOH , pH 7 . 5 , 500 mM KCl , and 12 mM MgCl2 . The yeast exosome complex was immunoprecipitated with IgG beads as described previously ( Liu et al . , 2014 ) , with the following modifications: a gentle wash buffer ( 150 mM NaCl ) was applied and the mono-Q anion exchange step was not performed . These modifications were made in order to maintain the interaction of the proteins in the sample . Eluted proteins were exchanged into 20 mM HEPES , pH 8 . 0 , 150 mM NaCl . E . coli OP50 lysates and C . elegans N2 lysates were prepared following a protocol from Bing et al . ( Yang et al . , 2012; Zhao et al . , 2015 ) . Mitochondria were isolated from the wild-type N2 worms as described previously ( Shen et al . , 2014 ) and lysed by incubation in 100 mM HEPES pH 8 . 0 , 1% NP-40 , 10 mM CaCl2 at 4°C for 30 min . The Pyrococcus furiosus L7Ae and the H/ACA RNA were prepared as described previously ( Li and Ye , 2006 ) . The buffer was exchanged to 50 mM HEPES , pH 7 . 6 , 1 M NaCl . E . coli ( MG1665 ) cells were grown at 37°C in 500 mL M9 minimal medium from a 1 mL overnight culture . Log phase cells were harvested after 11 hr at OD600 0 . 7; stationary phase cells were harvested after 26 hr at OD600 2 . 3 . Cell lysates were prepared in 50 mM HEPES pH 8 . 0 , 150 mM NaCl using a FastPrep system ( MP Biomedicals , Santa Ana , CA ) using two volumes of glass beads at 6 . 5 m/s , 20 s per pulse for four pulses , with 5 min of cooling on ice between pulses . The lysates were cleared by centrifugation at top speed in a tabletop microfuge for 30 min . Protein concentrations were determined using the bicinchoninic acid assay . At room temperate ( RT ) , protein pellets were dissolved ( assisted by sonication ) in 8 M urea , 20 mM methylamine ( to reduce carbamylation ) , 100 mM Tris , pH 8 . 5 , reduced with 5 mM TCEP for 20 min and alkylated with 10 mM iodoacetamide for 15 min in the dark . Then , the samples were diluted with 3 volumes of 100 mM Tris , pH 8 . 5 and digested with trypsin at 1/50 ( w/w ) enzyme/substrate ratio at 37°C for 16–18 hr . The optimal protein-to-cross-linker mass ratio was determined by a titration experiment . 1 µl of cross-linker at increasing concentrations ( 2 . 5 µg/µl , 5 µg/µl , 10 µg/µl , 20 µg/µl , 40 µg/µl ) in DMSO was incubated with 20 µl of 2 µg/µl of the ten-protein mixture at RT for 1 hr to make 16:1 , 8:1 , 4:1 , 2:1 , and 1:1 protein-to-cross-linker mass ratios , respectively . The reactions were quenched with 20 mM NH4HCO3 at RT for 20 min . Cross-linking products were analyzed by SDS-PAGE . The 4:1 ratio was ultimately chosen for both the one-piece and the two-piece Leiker . Higher dosages were avoided to minimize excessive cross-linking . For comparison of the one-piece and two-piece Leikers , 50 µl of the 2 µg/µl ten-protein mixture was incubated with 0 . 5 µl of 50 µg/µl AL or bAL1 at RT for 1 hr . The reactions were quenched as described above . For AL , the solution was mixed with 350 µl of 8 M urea , 100 mM Tris , pH 8 . 5 , and filtered with an Amicon Ultra-0 . 5 10-kD filter device ( Millipore ) . Excess cross-linker molecules were removed by two additional washes with urea . Click chemistry was subsequently performed on the membrane . In a 100 µl reaction , 28 nmol of azide-biotin was added ( an amount equal to the starting amount of the alkyne group of AL ) , followed by the addition of 2 mM CuSO4 , 2 mM TCEP , and 200 µM TBTA . Samples were gently rotated and incubated at RT for 2 hr . The excess free azide-biotin was then removed by washes with urea in the filter device . Finally , the proteins were collected by centrifugation with the filter device placed upside down inside the tube . Recovered proteins were transferred to a new 1 . 5 mL tube , precipitated at -20°C with four volumes of pre-cooled acetone for at least 30 min , and digested with trypsin . The bAL1 samples were processed in the same way except that the reaction mixture was precipitated directly without going through the 10-kD filter device . The AL- and bAL1-cross-linked peptides were enriched in parallel . The tryptic digests , without formic acid ( FA ) acidification , were directly mixed with an equal volume of 20 mM HEPES , pH 8 . 0 and incubated with 40 µl pre-washed high capacity streptavidin agarose for 2 hr . Then , the beads were washed three times with 20 mM HEPES , pH 8 . 0 , 1 M KCl , once with H2O , three times with 10% acetonitrile ( ACN ) , and another three times with H2O , each time with 1 mL buffer or H2O , with 5-min rotation . Supernatants were removed carefully with a 1 mL syringe needle connected to a vacuum pump . Loss of beads was avoided by keeping the beveled surface of the needle tip in contact with the wall of the tube . After the extensive washes , the peptides were released by incubating the beads with 5× bed volumes of cleavage buffer ( 300 mM Na2S2O4 in 6 M urea , 2 M thiourea , 10 mM HEPES , pH8 . 2 ) ( Yang et al . , 2010 ) at 37°C for 30 min , with end-to-end rotation . Recovered peptides were acidified with 5% FA and subsequently desalted on home-made C18 desalting columns , followed by elution with 70% ACN/0 . 1% FA . Eluates were vacuum dried and reconstituted in 0 . 1% FA for mass spectrometry analyses . The color of the beads could be used to monitor the entire enrichment process: a bright yellow color indicated the binding of Leiker-linked peptides; a return to a white color occurred when the cleavage reaction was successful . Comparison of bAL1 and bAL2 was carried out in two samples . For the first comparison , 50 µg of the ten-protein mixture was cross-linked with bAL1 or bAL2 at 4:1 protein-to-cross-linker mass ratio and then digested with trypsin . After mixing with the tryptic digest of an E . coli lysate containing 500 µg of total proteins , the digested Leiker-linked peptides were affinity purified with 20 µl of high-capacity streptavidin agarose . For the second comparison , 30 µg of ribosome was treated with bAL1 or bAL2 at 8:1 , 4:1 , or 2:1 protein-to-cross-linker mass ratios , digested , and enriched using 20 µl of high-capacity streptavidin agarose . For the serial dilution experiment ( Figure 3 ) , 200 µg of the ten-protein mixture was treated with 50 µg of bAL1 at RT for 1 hr . After quenching , the proteins were precipitated and digested with trypsin . Four equal aliquots of this digest were either not diluted to serve as a control ( 1:0 ) or diluted with the tryptic digest of a non-cross-linked E . coli lysate at 1:1 , 1:10 , or 1:100 ( w/w ) ratio . Each mixture was enriched with 200 µl of pre-washed streptavidin agarose . 30 µg of ribosome was treated with bAL2 at 8:1 , 4:1 , or 2:1 protein-to-cross-linker mass ratios . 40 µl of the exosome complex sample ( 1 µg/µl ) was incubated with 0 . 25 µl of 40 µg/µl bAL2 at RT for 1 hr . 20 µl of high-capacity streptavidin agarose was used to enrich Leiker-linked peptides in each sample . 70S ribosomes were negatively stained with 0 . 2% uranyl acetate . Carbon coated grids were first glow-discharged to increase the surface hydrophilicity using a Harrick Plasma cleaner . 4 µL aliquots of 70S ribosomes ( ~10 nM ) were placed on grids for about 1 min , and excessive liquid was absorbed by filter paper . After that 0 . 2% uranyl acetate was applied on the grid for about 1 min and absorbed using filter paper . The grids were air-dried and examined using an FEI Tecnai Spirit BioTwin microscope ( FEI , Hillsboro , OR ) ( 120 KV ) at 49 , 000× magnification . E . coli or C . elegans lysates prepared as described previously ( Yang et al . , 2012; Zhao et al . , 2015 ) ( 1 mg of total proteins ) were treated with 250 µg bAL1 at RT for 1 hr , in 300 µl reactions; NH4HCO3 was added to quench the reactions . Proteins were precipitated and digested with trypsin . After centrifugation in a bench top centrifuge at top speed for 30 min and filtering with a 50-kD cutoff filter , the digested peptides were brought to a volume of 3 mL with 2% ACN , 20 mM HEPES , pH 8 . 2; the pH was adjusted to 10 . 0 with ammonia prior to high-pH reverse phase separation on an Xtimate column ( 10×250 mm ) packed with 5 μm C18 resin ( Welch Materials , China ) at a flow rate of 2 mL/min . A 70 min gradient was applied as follows: 0-6% B in 10 min , 6-40% B in 40 min , 40-100% B in 10 min , 100% B for 10 min ( A = 4% ACN , 5 mM NH4COOH , pH 10 , B = 80% ACN , 5 mM NH4COOH , pH 10 ) . A total of 39 two-min fractions were collected , and then combined into 9–11 fractions of similar shades of color judging by naked eyes . These pooled samples were evaporated to 200–300 µl volumes before Leiker-linked peptides were enriched with 50 µl of high-capacity streptavidin beads from each sample . For the ribo-free lysates , 3 mg of proteins were cross-linked with 0 . 75 mg bAL2 at RT for 1 hr , and subjected to tryptic digestion and fractionation as described above . C . elegans mitochondria were prepared as described previously ( Shen et al . , 2014 ) , and the CXMS analysis was performed as described above except with two differences: 3 . 2 mg of total proteins was used as the starting material and the collected fractions were pooled into 5 fractions . In the forward and reverse labeling experiments , 0 . 7 nmol of RNA-free L7Ae was treated with [d0]-bAL2 and [d6]-bAL2 , respectively; an equal amount of L7Ae was pre-incubated with 1 nmol of the 65 nt H/ACA RNA at 4°C for 30 min and then treated with [d6]-bAL2 and [d0]-bAL2 , respectively . An equal amount of BSA was spiked into each cross-linking reaction . A 4:1 protein-to-cross-linker ratio ( w/w ) was used for each reaction . The cross-linking reactions were quenched with ammonium bicarbonate after 1 hr at RT . The paired [d0]- and [d6]-bAL2 samples were combined and subjected to acetone precipitation and trypsin digestion . In the forward labeling experiment , the log phase and the stationary phase cell lysates ( 100 µg proteins each ) were cross-linked with 50 µg of [d0]-bAL2 and 50 µg of [d6]-bAL2 , respectively , with 1 µg of BSA spiked into each sample . After 1 hr at RT , the two reactions were quenched , mixed , precipitated with acetone , and digested with trypsin . The reverse labeling experiment was conducted in the same way except that the log phase lysate was cross-linked with [d6]-bAL2 and the stationary phase lysate was cross-linked with [d0]-bAL2 . All protein samples were analyzed with an EASY-nLC 1000 system ( Thermo Fisher Scientific , Waltham , MA ) interfaced with a Q-Exactive mass spectrometer ( Thermo Fisher Scientific ) . A two-column setup was used , consisting of a pre-column ( 100 μm×4 cm , 3 μm C18 ) with a frit at each end and an analytical column ( 75 μm×10 cm , 1 . 8 μm C18 ) with a 5 µm tip . For the Leiker-cross-linked samples after enrichment , typically one third of a reconstituted sample was injected and separated with a 65 min linear gradient at a flow rate of 300 nl/min as follows: 0–5% B in 2 min , 5–28% B in 41 min , 28–80% in 10 min , 80% for 12 min ( A = 0 . 1% FA , B = 100% ACN , 0 . 1% FA ) . Slight modifications to the separation method were made for different samples . A 120 min gradient was used with a more gradual ramp to 28% buffer B . The Q-Exactive mass spectrometer was operated in data-dependent mode with one full MS scan at R = 70000 ( m/z = 200 ) , followed by ten HCD MS/MS scans at R = 17 , 500 ( m/z = 200 ) , NCE = 27 , with an isolation width of 2 m/z . The AGC targets for the MS1 and MS2 scans were 3e6 and 1e5 , respectively , and the maximum injection times for MS1 and MS2 were both 60 ms . For cross-linked samples , precursors of the +1 , +2 , +7 or above , or unassigned charge states were rejected; exclusion of isotopes was disabled; dynamic exclusion was set to 30 s . For accurate mass analysis , 20 µg/ml of [d0]-bAL2 or [d6]-bAL2 in methanol was sprayed directly into a LTQ Orbitrap XL mass spectrometer ( Thermo Fisher Scientific ) operated in the negative mode with a spray voltage of 0 . 8 kV and a scan mass range of 150–1000 m/z . The Xcalibur raw data was converted to ms2 files using RawExtract ( McDonald et al . , 2004 ) . Cross-linked peptides were identified using pLink software as described previously ( Yang et al . , 2012 ) , with the following modifications Cross-linker was set to AL , bAL1 , bAL2 , [d6]-bAL2 , or BS3; The minimum peptide length was 5 amino acids for lysate samples; oxidation on Met was set as a variable modification . For the ten-protein mixture and ribosome complexes , the search databases consisted of the sequences of all of the proteins in question . The sequences were downloaded from NCBI or Uniprot . Prior to the CXMS analysis of the exosome complex , LC-MS/MS analyses of digested , uncross-linked samples were carried out to identify the proteins present in the samples . For protein identification , the precursors of +1 or unassigned charge states were rejected; MS2 spectra were searched against a S . cerevisiae protein database ( downloaded from Uniprot on 2013-04-03 ) using ProLuCID2 ( Xu et al . , 2006 ) and filtered using DTASelect 2 . 0 ( Tabb et al . , 2002 ) with a spectral false identification rate ≤1% and a minimum of two identified peptides for each protein . A restricted database containing only the identified proteins ( 740 in total ) was generated using Contrast 2 . 0 ( Tabb et al . , 2002 ) . MS2 spectra from the cross-linked samples were then searched against this small database using pLink . For the CXMS analysis of E . coli whole-cell lysates and ribo-free lysates , the sequences of the entire proteome of the K12 strain were downloaded from Uniprot on 2014-07-31 and used for searching . For the CXMS analysis of C . elegans lysates , a database consisting of proteins identified from N2 C . elegans lysates generated with ProLuCID2 was used for searching ( unpublished ) . For the CXMS analysis of C . elegans mitochondrial proteins , a restricted database was constructed in a similar way as for the exosome complex . pQuant ( Liu et al . , 2014 ) was used to determine the heavy-to-light ratio ( H/L ) of each cross-link . The regression model Y = aX +e is used to calculate peptide ratios . The optimal value of a is solved using the least-squares method as a^=∑XjYj/∑XjXj , and the estimated standard error of a^ is σ^= ( K−1∙∑ ( Yj−a^Xj ) 2/∑Xj2 ) 1/2 . is then normalized to the interval of [0 , 1] , and is named confidence score . If the value of σ^ is zero ( the highest confidence ) , there is no interference signal; if the value is one ( the lowest confidence ) , the peptide signals are inundated by interference signals . For each identified cross-link spectrum , an extracted ion chromatogram ( EIC ) was constructed for each isotopic peak of the light- and heavy-labeled precursor . The H/L ratios can be calculated based on the monoisotopic peak , the most intense peak , or the least interfered peak of each isotopic cluster as specified by users . For L7Ae , all options yielded similar results and we selected the monoisotopic peak . For the highly complex samples of the log phase versus stationary phase E . coli , the option of the least interfered peak performed the best . For each cross-link , every identified spectrum ( E-value < 0 . 001 ) will lead to a H/L ratio and a confidence score σ , because pQuant conducts the quantitation independently starting from each identified MS/MS spectrum . In most cases , the H/L ratios obtained for the same precursor ion are close , but sometimes the ratios may differ due to multiple reasons including local interference signals or a sudden decrease followed by recovery in signal intensity in the chromatograms , all of which can affect the calling of the start and the end of a chromatogram peak . Ratios with σ values above or equal to 0 . 5 were discarded . The median H/L ratio obtained from the remaining spectra was assigned to a cross-linked lysine pair or a mono-linked lysine residue as the final quantification value . If a cross-link had no assigned ratio value ( i . e . , none of its ratios had a σ value less than 0 . 5 ) , we manually evaluated the reconstructed ion chromatograms to assess abundance changes . All the ratios were normalized against the median value of all the H/L ratios belonging to the spiked-in BSA . | Proteins fold into structures that are determined by the order of the amino acids that they are built from . These structures enable the protein to carry out its role , which often involves interacting with other proteins . Chemical cross-linking coupled with mass spectrometry ( CXMS ) is a powerful method used to study protein structure and how proteins interact , with a benefit of stabilizing and capturing brief interactions . CXMS uses a chemical compound called a linker that has two arms , each of which can bind specific amino acids in a protein or in multiple proteins . Only when the regions are close to each other can they be “cross-linked” in this way . After cross-linking , the proteins are cut into small pieces known as peptides . The cross-linked peptides are then separated from the non cross-linked ones and characterized . Although CXMS is a popular method , there are aspects about it that limit its use . It does not work well on complex samples that contain lots of different proteins , as it is difficult to separate the cross-linked peptides from the overwhelming amounts of non cross-linked peptides . Also , although it can be used to detect changes in the shape of a protein , which are often crucial to the protein's role , the method has not been smoothed out . Tan , Li et al . have now developed a new cross-linker called Leiker that addresses these limitations . Leiker cross-links the amino acid lysine to another lysine , and contains a molecular tag that allows cross-linked peptides to be efficiently purified away from non cross-linked peptides . As part of a streamlined workflow to detect changes in the shape of a protein , Leiker also contains a region that can be labeled . Analysing a bacterial ribosome , which contains more than 50 proteins , showed that Leiker-based CXMS could detect many more protein interactions than previous studies had . These included interactions that changed too rapidly to be studied by other structural methods . Tan , Li et al . then applied Leiker-based CXMS to the entire contents of bacterial cells at different stages of growth , and identified a protein interaction that is only found in growing cells . In future , Leiker will be useful for analyzing the structure of large protein complexes , probing changes in protein structure , and mapping the interactions between proteins in complex mixtures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources"
] | 2016 | Trifunctional cross-linker for mapping protein-protein interaction networks and comparing protein conformational states |
The small signaling molecule auxin controls numerous developmental processes in land plants , acting mostly by regulating gene expression . Auxin response proteins are represented by large families of diverse functions , but neither their origin nor their evolution is understood . Here , we use a deep phylogenomics approach to reconstruct both the origin and the evolutionary trajectory of all nuclear auxin response protein families . We found that , while all subdomains are ancient , a complete auxin response mechanism is limited to land plants . Functional phylogenomics predicts defined steps in the evolution of response system properties , and comparative transcriptomics across six ancient lineages revealed how these innovations shaped a sophisticated response mechanism . Genetic analysis in a basal land plant revealed unexpected contributions of ancient non-canonical proteins in auxin response as well as auxin-unrelated function of core transcription factors . Our study provides a functional evolutionary framework for understanding diverse functions of the auxin signal .
Auxins are a group of structurally related chemical compounds that control a multitude of growth and developmental processes in plants . The most common , naturally occurring auxin is indole-3-acetic acid ( IAA ) , but synthetic analogs such as 2 , 4-dichlorophenoxy acetic acid ( 2 , 4-D ) have largely overlapping biological activities ( Woodward and Bartel , 2005 ) . While auxins have been shown to trigger rapid cellular events such as membrane hyperpolarization ( Bates and Goldsmith , 1983; Etherton , 1970 ) , calcium influx ( Monshausen et al . , 2011; Schenck et al . , 2010 ) , and changes in endocytosis ( Paciorek et al . , 2005; Robert et al . , 2010 ) , its activity in controlling growth and development appear to be mainly mediated by changes in gene expression via a nuclear auxin pathway ( NAP ) . Perturbation of this gene regulatory pathway interferes with most , if not all , developmental responses ( Weijers and Wagner , 2016 ) . Indeed , in the moss Physcomitrella patens , it was shown that a complete knock-out mutant of this pathway does not show any transcriptional response to auxin ( Lavy et al . , 2016 ) . The NAP encompasses three dedicated protein families ( Figure 1A , B ) . Various auxins , including IAA and 2 , 4-D , are perceived by a co-receptor complex consisting of TRANSPORT INHIBITOR RESPONSE 1/AUXIN SIGNALING F-BOX ( TIR1/AFB ) and AUXIN/INDOLE-3-ACETIC ACID ( Aux/IAA ) proteins ( Dharmasiri et al . , 2005; Kepinski and Leyser , 2005; Tan et al . , 2007 ) . Subsequent ubiquitination of the Aux/IAA proteins causes their degradation in the 26S proteasome ( Gray et al . , 2001 ) . When not degraded , Aux/IAA proteins bind to and inhibit DNA-binding transcription factors , the AUXIN RESPONSE FACTORS ( ARF ) ( Kim et al . , 1997 ) . Thus , auxin de-represses ARFs , allowing these to activate or repress their direct target genes ( Ulmasov et al . , 1999 ) . A central question in plant biology is how this simple transcriptional system with only three dedicated components can generate a multitude of local auxin responses to support various developmental functions . In flowering plants such as Arabidopsis thaliana , it is likely that the size of TIR1/AFB ( six members ) , Aux/IAA ( 29 members ) and ARF ( 23 members ) gene families allows combinatorial assembly of distinct , local auxin response pathways . Given that diversity in auxin responses follows from diversification in its response proteins , it is still unclear how NAP complexity evolved from simpler ancestral states . Furthermore , while intuitive , a key question is whether increased NAP complexity indeed enabled more complex and diverse auxin responses during plant evolution . A third important question is where , when , and from what precursors the NAP originated . Eukaryotic photosynthetic organisms diverged into three groups , Glaucophyta , Rhodophyta ( red algae ) , and Viridiplantae more than 1 . 5 billion years ago ( Yoon et al . , 2004 ) . Viridiplantae are further classified into chlorophyte algae and streptophytes , which include charophyte algae and land plants . Bryophytes represent the earliest diverging land plants and consist of three groups: hornworts , liverworts and mosses . After the split from bryophytes , ancestral vascular plants changed their life cycle from haploid-dominant to diploid-dominant and established a vascular system and root architecture , forming the group of lycophytes and euphyllophytes ( ferns , gymnosperms and angiosperms ) . The presence of a functional NAP with reduced genetic redundancy has been reported in model bryophytes ( Flores-Sandoval et al . , 2015; Kato et al . , 2015; Prigge et al . , 2010; Rensing et al . , 2008 ) , whereas the presence of endogenous auxin is also reported in wide range of algal species ( Žižková et al . , 2017 ) . Thus , a prediction is that the auxin response system may predate land plants , and that complexity evolved after the divergence of ancestral vascular plants from bryophytes . A key challenge is to identify the origin of the NAP system , as well as to reconstruct the steps in the evolution of its complexity . However , only little genome data are currently available from non-flowering land plants ( Rensing , 2017 ) , which makes such inferences extremely challenging . In addition , studies using only selected model species bear the risk of generalizing observations from non-representative genomes , due to species-specific gene-duplication , -loss , and -diversification . Therefore , it is necessary to analyze multiple species to understand evolutionary trends . Here , we describe a deep phylogenomic analysis of NAP components using a large transcriptome dataset with more than 1000 plant species including many algae . This extensive dataset enabled us to reconstruct the ancestral states of auxin response gene families at key nodes in plant evolution . We infer plausible origins and evolutionary patterns for each auxin response gene family and predict auxin response properties at evolutionary nodes . Using comparative RNA-seq of six species , we tested and extended these predictions . Finally , we used a genetic strategy in a bryophyte to demonstrate surprising non-contributions of an ancient ARF class as well as contribution of deeply conserved non-canonical NAP components to auxin signaling . Our work provides a deep view into early steps in the origin , evolution and design principles of the multi-functional auxin response system .
To reconstruct origin and early diversification in auxin response gene families , we designed a strategy ( Figure 1—figure supplement 1 ) that uses a large transcriptome dataset ( OneKP ) including multiple species for each major branch in plant species phylogeny ( Matasci et al . , 2014 ) . The depth and quality of each individual RNA-seq-derived transcriptome is limited , and a further caveat of transcriptome-based gene identifications is that the number of genes may be underestimated if a gene is not expressed under the sampling conditions or in the sampled tissue . However , the availability of transcriptomes from multiple tissue samples of multiple related species , should allow deduction of the ancestral state that defines the gene complement at each evolutionary node . It should be stressed that this number represents the ancestral state at a given node , and species-specific gene duplications and gene losses will have modified the gene complement in individual species . Given our focus on early events in nuclear auxin response evolution , we have used all available transcriptomes of red algae , green algae , bryophytes , lycophytes , ferns , and gymnosperms from the OneKP dataset ( Supplementary file 1 ) . We also included all available angiosperm species in the Chloranthales , Magnoliids and ANA grade , as well as several species in both monocots and dicots ( Supplementary file 1 ) . For reference and quality control purposes , we included genome-based sequences from well annotated model species . Each of the three auxin response protein types ( ARFs , Aux/IAAs , and TIR1/AFBs ) are multi-domain proteins and we initially focused on the origin of these proteins . Therefore , we asked where domains , or parts thereof , were found , and at what node the multi-domain proteins first appear . ARF proteins carry an N-terminal DNA-binding domain ( DBD ) which consists of a composite dimerization domain ( DD; made up of two separate subdomains [DD1 and DD2] that fold into a single unit ) , a B3-type DNA-interaction domain , and an ancillary domain ( AD ) of unknown function ( Figure 1C; Boer et al . , 2014 ) . In land plants , the DD and AD are only found in the ARF family . The C-terminal Phox and Bem 1 ( PB1 ) domain is shared among ARF and Aux/IAA proteins and mediates homo- and hetero-oligomerization ( Korasick et al . , 2014; Nanao et al . , 2014 ) . Finally , ARFs contain a less well-defined Middle Region ( MR ) separating the PB1 and DBD ( Figure 1C ) . In red algae , we found proteins containing an N-terminal portion of DD1 , DD2 , and AD , lacking a B3 or PB1 domain , but instead flanked by a C-terminal bromodomain ( BROMO; InterPro ID: IPR001487; Figure 1C ) . The DD1 and DD2 motifs in red algae are spaced by 20–30 conserved amino acids , which is much shorter than the B3 domain ( ~120 amino acids; Supplementary file 2 ) . In chlorophytes , we found a protein with only AD , flanked by a DNA-binding AT-rich interaction domain ( ARID; InterPro ID: IPR001606; Figure 1C ) . Furthermore , we found separate proteins that either represented a B3 or a PB1 domain ( Figure 1C ) . Thus , all ARF subdomains had been established before the split of the streptophytes , but not combined in a single protein . In contrast , we discovered full-length ARF-like proteins containing a DBD with a B3 domain inserted between DD and AD in charophytes ( Figure 1C and Figure 1—figure supplement 2 ) . Land plant ARFs can be grouped into three classes , A , B and C ( Finet et al . , 2013 ) . Based on transactivation assays , class A and B ARFs are classified as transcriptional activators and repressors , respectively ( Kato et al . , 2015; Ulmasov et al . , 1999 ) . Class C-ARFs are generally recognized as transcriptional repressors based on the amino acid composition of MR , but this has not yet been fully supported by experimental evidence ( Kato et al . , 2018 ) . Phylogenetic analysis revealed that the ARF-like proteins in charophytes fall in two sister clades and likely represent separate precursors of class C-ARFs ( proto-C-ARFs ) and A/B-ARFs ( proto-A/B-ARFs ) of land plants ( Figure 2 and Figure 1—figure supplement 2 ) . Interestingly , we found the PB1 domain only in proto-C-ARFs , which could , however , be due to sparse sampling in some charophyte lineages ( Figure 1—figure supplement 2 ) . To understand if the proto-ARFs share conserved , functionally important residues , we generated homology models based on available DBD crystal structures of A . thaliana ARF1 and ARF5 ( Boer et al . , 2014 ) . As no class C-ARF structure is known , we first modeled the A . thaliana ARF10 DBD to compare with proto-C-ARFs . Next , homology models for proto-ARFs in Spirogyra pratensis ( SpARF; proto-C-ARF ) and Mesotaenium caldariorum ( McARF; proto-A/B-ARF ) were generated . We also included all three ARFs of the bryophyte M . polymorpha ( MpARF1-3 ) representing each major class , and compared all models to A . thaliana ARF structures . This analysis revealed that all proto-ARFs likely share a conserved structural topology ( Figure 3A ) . Strikingly , all DNA-binding residues follow the spatial restraints needed for DNA binding in all ARFs tested , suggesting a conserved mode of DNA binding . On the other hand , dimerization residues are conserved only in the ( proto- ) A/B-ARFs ( McARF , MpARF1 , and MpARF2 ) but not in the ( proto- ) C-ARFs ( SpARF , MpARF3 , and ARF10 ) . These results clearly demonstrate that canonical ARF proteins were established and differentiated into two classes in charophyte algae . In addition to the proteins with canonical ARF-like structure , we found a group of charophyte proteins consisting of an AP2 DNA-binding domain along with B3 and PB1 domains ( Figure 1C ) . Land plants also have a protein family containing AP2 domain in their N-terminus , followed by a B3 domain . These proteins are called REALATED TO ABI3 AND VP1 ( RAV ) . Interestingly , land plant RAV proteins do not have a PB1 domain and it is known that the B3 domain of RAV and ARF binds different DNA sequences ( Boer et al . , 2014; Matías-Hernández et al . , 2014 ) . The B3 domain of RAV-like proteins in charophytes is much more similar to RAV’s than to ARF’s in land plants and phylogenetic analysis showed that the RAV-like proteins of charophytes position along with RAV family in land plants ( Figure 2B and Figure 1—figure supplements 2 and 3 ) . Thus , we classify these proteins as proto-RAV . In the charophyte green algae , the two classes of proto-ARFs and proto-RAVs are found in various combinations in each species ( Figure 2A ) . While sequencing depth may be insufficient to detect all proto-ARFs and proto-RAVs , there does not appear to be a conserved pattern in the order of appearance and retention of these genes . We next considered the origin of the Aux/IAA proteins . These proteins contain two functional small domains in addition to a C-terminal PB1 domain ( Figure 1B , C ) . The N-terminal domain I recruits the TOPLESS ( TPL ) transcriptional co-repressor ( Szemenyei et al . , 2008 ) . Domain II mediates the auxin-dependent interaction with TIR1/AFB and thus acts as a degron ( Dharmasiri et al . , 2005; Gray et al . , 2001; Kepinski and Leyser , 2005 ) . Because domain I and II are too small for reliable BLAST searches , we used the PB1 domain to identify potential family members . No PB1-containing proteins were identified in red algae , while we found proteins with a PB1 domain but no DBD in chlorophytes ( Figure 1C ) . Phylogenetic analysis based on the PB1 domain indicated these are neither closely related to RAV , nor to Aux/IAA and ARF families ( Figure 2B and Figure 2—figure supplement 1 ) . PB1 domain-containing proteins that lack a DBD were also found in many of the charophyte algae ( Figures 1C and 3B and Figure 2—figure supplement 1 ) . Most of them were placed along with proto-RAV in phylogenetic tree , but the sequences from Coleochaetae irregularis were placed along with the Aux/IAA in land plants that is separate from the PB1 of both ARFs and proto-RAV proteins ( Figures 2B and 3B and Figure 2—figure supplement 1 ) . Even though the N-terminal part of the PB1 domain is not as conserved as the C-terminal part , several critical residues were found to be conserved in Aux/IAA-like sequences ( Figure 3B ) . These results indicate that the PB1 domain of land plant ARFs and Aux/IAAs had separate precursors in charophytes . We could , however , not detect domain I or II in Aux/IAA-like genes of charophyte algae , even when scrutinizing individual sequences . We thus conclude that Aux/IAA proteins with all three functional domains are limited to land plants . Finally , we explored the origin of the TIR1/AFB auxin co-receptor that consist of an N-terminal F-box domain that anchors the protein to the other subunits in the SCF E3 ubiquitin ligase complex , and a C-terminal leucine-rich repeat ( LRR ) domain that contains the auxin-binding pocket . Auxin acts as a molecular glue to stabilize the interaction between TIR1/AFBs and Aux/IAAs ( Tan et al . , 2007 ) . The closest homolog of the TIR1/AFB proteins in A . thaliana is CORONATINE INSENSITIVE 1 ( COI1 ) , which functions as a receptor of the jasmonic acid ( JA ) phytohormone ( Katsir et al . , 2008 ) . In our homology search , we could not identify any proteins showing homology to either TIR1/AFB or COI1 in red algae and chlorophytes ( Figures 1C and 2A ) . We did find many proteins showing homology to TIR1/AFB and COI1 in the transcriptomes of charophyte algae ( Figures 1C and 2A ) . Phylogenetic analysis indicated that some of these proteins form a sister group to both TIR1/AFB and COI1 in land plants and none of the charophyte proteins are specifically grouped into either TIR1/AFB or COI1 clades ( Figure 4 and Figure 4—figure supplement 2 ) , suggesting that charophytes had an ancestor that gave rise to both auxin and JA receptors . To infer whether the TIR1/AFB/COI1-like proteins of charophytes function as receptors for auxin or JA , we generated homology models of the TIR1/AFB/COI1-like protein from C . irregularis and the bryophyte M . polymorpha MpTIR1 and MpCOI1 , using the A . thaliana TIR1 and COI1 crystal structures ( Sheard et al . , 2010; Tan et al . , 2007 ) as templates for modeling . Even though the secondary structure of the C . irregularis protein was highly similar to that of land plant TIR1 and COI1 ( Supplementary file 2 ) , at the level of amino acid sequence , the protein did not resemble either TIR1/AFB or COI1 . Out of 40 residues conserved in either TIR1/AFB’s or COI1’s , only 7 and 11 residues are identical to TIR1/AFBs and COI1s , respectively ( Supplementary file 2; black stars ) . Notably , most of the hormone-contacting residues ( 11 out of 12 ) are different from both TIR1/AFB and COI1 ( Figure 3C and Supplementary file 2 ) . These results suggest that the charophyte TIR1/AFB/COI1 precursor may not act as an auxin or JA receptor , and we conclude that dedicated receptors for auxin and JA were established only in land plants . Taken together , our analyses suggest that the components of NAP were established in the common ancestor of land plants by combining pre-existing components and that the system evolved to regulate pre-existing transcription factors . All three gene families have evolved to considerable size and diversity in angiosperms , and this diversity is thought to underlie multifunctionality of auxin as a hormone . We next aimed to reconstruct the evolutionary history of auxin response components across all land plant lineages . Consistent with previous descriptions ( Finet et al . , 2013 ) , our phylogenetic analysis showed that all land plant ARFs are divided into three phylogenetic lineages ( Figure 4 and Figure 1—figure supplement 2 ) . Within the class C lineage , we did not find any duplications in the ancestors of non-angiosperm species . The split that generated A . thaliana ARF10/16 and ARF17 likely occurred early in angiosperm evolution , while the PB1 domain was lost in the ARF17 group ( Figure 4 and Figure 1—figure supplement 2 ) . The class A-ARF is represented by a single copy in bryophytes and lycophytes . We found that a subset of genes lacking the DBD diverged from class A-ARFs in early land plants , is missing in hornworts and has been retained in liverworts , mosses and lycophytes ( non-canonical ARF , ncARF; Figures 3B and 4 and Figure 2—figure supplement 1 ) . A further gene duplication event in the ancestor of euphyllophytes gave rise to two class A sub-families corresponding to A . thaliana ARF5/7/19 and ARF6/8 , respectively . In the ancestor of seed plants a gene duplication caused differentiation between the A . thaliana ARF5 and ARF7/19 subfamilies ( Figure 4 and Figure 1—figure supplement 2 ) . Finally , two gene duplication events in the ancestral angiosperms led to ARF6 and ARF8 and to a paralogue of ARF7/19 , which was lost in A . thaliana ( Figure 4 and Figure 1—figure supplement 2 ) . Class B-ARFs are represented by a single gene in the ancestor of liverworts , mosses , lycophytes , and ferns . However , no hornwort species appears to contain class B-ARFs ( Figure 4 and Figure 1—figure supplement 2 ) . Gene duplications in the ancestral gymnosperms gave rise to three class B-ARF copies , one representing A . thaliana ARF3/4 , another leading to A . thaliana ARF2 and the third generating the remainder of the class B-ARFs in A . thaliana ( Figure 4 and Figure 1—figure supplement 2 ) . Notably , the reported lack of the PB1 domain in ARF3 ( Finet et al . , 2013 ) is an independent loss in the common ancestor of monocots and eudicots ( Figure 1—figure supplement 2 ) . Our data indicated that an ancestral Aux/IAA gene lacking domain I and II had been established during the evolution of charophytes , while ‘true’ Aux/IAAs with all functional domains are found only in land plants ( Figure 1C ) . In addition to one copy of ‘true’ Aux/IAA , we found another set of deeply conserved non-canonical Aux/IAA-like sequences that lack the domain I and II ( non-canonical Aux/IAA , ncIAA; Figures 2B , 3B and 4 , Figure 2—figure supplement 1 , and Figure 4—figure supplement 1 ) . Strikingly , while the Aux/IAAs have diversified through gene duplications , the ncIAA is found only in a single copy in all evolutionary nodes examined here and is represented by IAA33 in A . thaliana . In the ancestor of euphyllophytes , gene duplication events gave rise to three Aux/IAAs , which were retained in the ancestral seed plants ( Figure 4 and Figure 4—figure supplement 1 ) . Common ancestor of angiosperms have 11 Aux/IAA proteins , which is more than triple the number found in gymnosperms ( Figure 4 and Figure 4—figure supplement 1 ) . Finally , in addition to the ancient ncIAA generated in a first duplication event , several independent later events generated non-canonical family members lacking domains . For example , the lack of domain II in IAA20 , IAA30 , IAA31 , IAA32 , and IAA34 of A . thaliana appears to be an independent loss in their respective lineages in the core angiosperms ( Figure 4—figure supplement 1 ) . Our data indicated that ancestral charophyte green algae had one common ancestor for both auxin ( TIR1/AFB ) and JA ( COI1 ) F-box co-receptors , and following duplication in the ancestor of all land plants , developed into two independent receptors ( Figure 4 and Figure 4—figure supplement 2 ) . The common ancestor of bryophytes and lycophytes had a single orthologue of A . thaliana TIR1/AFBs . Gene duplication events in the ancestor of euphyllophytes gave rise to three subgroups; one leading to TIR1/AFB1-3 , one leading to AFB4/5 and another which is widely present in many species including the angiosperms , but has been lost in some monocots and dicots including A . thaliana ( Figure 4 and Figure 4—figure supplement 2 ) . Thus , our analysis of the patterns of diversification in the ARF , Aux/IAA and TIR1/AFB families identifies the auxin response complement at each evolutionary node , and in addition reveals deeply conserved non-canonical family members . Notably , many changes occurred in the composition of NAP from the common ancestor of lycophytes to euphyllophytes , which may have led to complex auxin response . The complements of auxin response components identified from phylogenomic analysis allow for clear predictions of which species possess a functional transcriptional auxin response system . Based on our predictions , only land plants should be able to respond . In addition , it is intuitive that the number of components in auxin response will relate to the complexity of response , but as yet there is no experimental basis for such relationship . To experimentally address the competence of species to respond to auxin , and to explore the relationship between auxin response components and the qualitative and quantitative aspects of auxin response , we performed comparative transcriptome analysis . We selected six species that belong to different ancient lineages and that each have a different complement of auxin response components ( Figure 5A ) . We used the charophyte algae Klebsormidium nitens and Spirogyra pratensis , the hornwort Anthoceros agrestis , the liverwort Marchantia polymorpha , the moss Physcomitrella patens , and the fern Ceratopteris richardii . To detect only early transcriptional responses , we treated plants with auxin for 1 hr , and performed RNA-seq followed by de novo transcriptome assembly and differential gene expression analysis . To avoid inactivation of the natural auxin IAA by conjugation or transport , we treated with 10 μM of the synthetic auxin 2 , 4-dichlorophenoxyacetic acid ( 2 , 4-D ) . This compound was shown to behave like IAA in the context of the NAP ( Tan et al . , 2007 ) . Importantly , 68–90% of the differentially expressed genes ( DEG ) from de novo assemblies in K . nitens , M . polymorpha and P . patens matched with genome-based differential gene expression performed in parallel ( Figure 5—source data 1 ) , thus validating our approach . Transcriptome analysis after prolonged auxin treatment in P . patens had identified a large set of auxin-responsive genes ( Lavy et al . , 2016 ) . Indeed , we found 105 and 1090 genes to be auxin-regulated in M . polymorpha and P . patens , respectively ( Figure 5A ) . Likewise , we found 159 and 413 genes to be auxin-regulated in A . agrestis and C . richardii ( Figure 5A ) . Unexpectedly , despite lacking Aux/IAA and dedicated TIR1/AFB genes , both charophyte algae species showed a strong transcriptional response to 2 , 4-D treatment . A total of 1094 and 1681 genes were differentially expressed in K . nitens and S . pratensis , respectively ( Figure 5A ) . Thus , there is a clear transcriptional response to 1 hr of 2 , 4-D treatment in all species analyzed , yet the number of genes is different , with an exceptionally large number of responsive genes in charophytes . We next determined if the number of DEG correlates with gene number in each transcriptome assembly ( Figure 5—source data 2 ) , and found that differences in DEG among species cannot be explained by total gene number . We next addressed whether there were differences in the characteristics of regulation . Both charophyte species showed a high percentage of gene repression . Only 37% and 33% of DEG were activated in K . nitens and S . pratensis , respectively ( Figure 5A ) . In contrast , the distribution of fold change amplitude values differed between the two charophytes where S . pratensis showed a general shift toward larger amplitudes of regulation ( Figure 5A ) . Even though the complement of auxin response proteins are different , all three bryophytes showed a similar pattern: 36–53% of DEG were activated , with very few genes showing an amplitude over 2-fold up- or down-regulation ( Figure 5A ) . In contrast , 82% of DEG were activated in C . richardii . We also found that there was a notable difference in the distribution of fold-change values , with a larger fraction of genes being more strongly activated ( maximum 28 fold; Figure 5A ) . We found that the number of auxin-responsive genes is positively correlated with the number of ARFs in land plants as seen in the expanded number of ARFs and DEG in P . patens and C . richardii . A switch to gene activation is not correlated with the number of ARFs , but rather with a duplication in the class A-ARFs in the ancestor of euphyllophytes and/or increase of Aux/IAA and TIR1/AFB . The increase in amplitude of auxin-dependent gene regulation in C . richardii could be a consequence of higher activation upon treatment , increased repression in the absence of auxin , or both . To determine its basis , we compared normalized expression values for the 20 top-most auxin activated , and the 20 least auxin activated genes in all species ( Figure 5B ) . This revealed that the increased amplitude of the top-most activated genes in C . richardii is not correlated with increased expression in the presence of auxin , but rather caused by reduced expression in its absence . This quantitative property of the auxin response system is correlated with the increased numbers of Aux/IAA genes . Given that the mechanism of auxin response is ancient and conserved among all land plants , a key question is whether responses in different species involve regulation of a shared set of genes . To address this question , we performed tBLASTx searches among all DEG in our comparative transcriptome data and visualized the network of their similarities ( Figure 6—figure supplements 1 and 2 ) . Even though BLAST filtering is not sufficient to distinguish orthology groups in large families such as kinases , we could identify several gene families to be commonly regulated by auxin in different land plants species . Classical primary auxin-responsive genes—the Aux/IAA , GH3 and SAUR families—were shown to be auxin responsive in many angiosperm species ( Abel and Theologis , 1996 ) . We found different bryophyte species to show auxin-dependence in only some of these three gene families ( Figure 6A ) , yet no species showed regulation of all three gene families . In contrast , C . richardii displayed auxin-dependence of members of all three gene families ( Figure 6A ) . Given that the Aux/IAA and GH3 proteins themselves regulate auxin levels or response , this indicates that a robust feedback mechanism involving all these gene families did not exist prior to the emergence of vascular plants , and bryophytes might have different feedback mechanism . In addition , we identified the members of class II homeodomain-leucine zipper ( C2HDZ ) and WIP families to be commonly activated by auxin in all land plants in our RNA-seq ( note that no WIP gene was identified in the A . agrestis assembly ) . Indeed , qPCR analysis confirmed auxin-activation of C2HDZ ( Figure 6B ) . We also identified the members of auxin biosynthesis gene YUCCA ( YUC ) family to be commonly down-regulated among multiple land plant species ( except A . agrestis ) , and qPCR analysis demonstrated this to be true in A . agrestis , as well ( Figure 6B ) . It is known that some members of C2HDZ , WIP , and YUC families in A . thaliana are also up- or down-regulated by auxin , respectively ( Crawford et al . , 2015; Sawa et al . , 2002; Takato et al . , 2017 ) . While homologues of C2HDZ were detected in the charophyte assemblies , none was regulated by auxin , which supports the different nature of the auxin response system in these species . In summary , land plants share a deeply conserved set of auxin up- and down-regulated genes . Our phylogenomic analysis identified several components that are deeply conserved , yet whose contributions to auxin response are unknown: two deeply conserved non-canonical auxin signaling components lack important domains ( ncIAA and ncARF ) , while class C-ARFs diverged from all other ARFs in green algae prior to establishment of the NAP . To investigate the biological roles of these genes , we chose the liverwort M . polymorpha , the only genetically tractable model plant encoding ncIAA , ncARF and C-ARF genes . We first addressed ncIAA and ncARF function and performed CRISPR/Cas9-mediated mutagenesis ( Sugano et al . , 2014 ) to obtain two different alleles for each gene which presumably cause a loss-of-function by frame shift mutation ( nciaa-6 , nciaa-10 , ncarf-2 , ncarf-10; Figure 7A , Figure 7—figure supplement 1A , B , E ) . To investigate whether ncIAA and ncARF are involved in auxin response , we grew mutants on auxin-containing medium . Exogenously supplied auxin causes severe inhibition of thallus growth and increased formation of rhizoids in wild-type ( Figure 7B; Ishizaki et al . , 2012 ) . nciaa mutants showed auxin response similar to wild-type , while growth inhibition was strongly suppressed in ncarf mutants although rhizoid formation was still promoted by auxin ( Figure 7B ) . We next selected two auxin-up-regulated genes ( EXP and WIP ) and one auxin-down-regulated gene ( YUC2; Eklund et al . , 2015 ) , and examined their expression in all mutants by qPCR analysis ( Figure 7C ) . In nciaa mutants , the expression of auxin-up-regulated genes responded similarly to the wild-type , while the expression of the auxin-repressed YUC2 gene was significantly reduced in the absence of auxin , but similarly repressed by auxin . In ncarf mutants , the basal expression of auxin-upregulated genes was similar to WT , while the expression after auxin treatment was significantly reduced in the mutants . The expression of YUC2 was reduced in mock condition and auxin treatment did not change the expression . Thus , in M . polymorpha , ncIAA may have a function in gene expression , but is not critical for auxin response itself . On the other hand , ncARF represents a novel positive regulator of both auxin-dependent gene activation and repression . Finally , we focused on C-ARF function . While partial mutants have been reported in A . thaliana , no plants completely lacking C-ARF have been described . We used CRISPR/Cas9 gene editing to generate a series of loss-of-function mutants in MpARF3 , the single C-ARF of M . polymorpha ( arf3ge1-1 , arf3ge1-2 , arf3ge2-1; Figure 7A and Figure 7—figure supplement 1C , D ) . All three arf3 mutants showed dramatic defects in development , notably in vegetative propagules ( gemmae ) which arrested before maturation , consistent with ubiquitous ARF3 protein accumulation in these structures ( Figure 7D–F , and Figure 7—figure supplement 1G ) . A previous study reported characterization of mutants in the class A-ARF in M . polymorpha ( arf1-4 ) and showed that ARF1 is important for auxin response ( Kato et al . , 2017 ) . arf1-4 produces narrower and twisted thallus which is distinct from flat thallus of arf3 mutants . In addition , previous studies also showed that gemma development was regulated by Aux/IAA and the class A-ARF ( Kato et al . , 2015; Kato et al . , 2017 ) , and we hence tested if transcriptional responses to auxin were altered in arf3 mutants . Strikingly , all auxin-responsive genes we tested showed similar responses in WT and arf3 mutants , while arf1 mutants showed no auxin responses ( Figure 7C ) . This result suggests that , class C-ARF in M . polymorpha have different target genes from A-ARF and may not be critical for auxin-dependent gene regulation .
Phylogenetic analysis and domain structural analysis provided many insights into the origin of NAP and its evolutionary trajectory . All subdomains of dedicated auxin-response proteins were recovered in transcriptomes from red algae and chlorophytes , but the multidomain protein appears only in the charophyte and land plant lineage . These findings show that proto-ARF transcription factor was established during the evolution of ancestral charophytes by combining existing domains . However , given that no defined Aux/IAA and TIR1/AFB could be identified in charophytes , a complete nuclear auxin response system is limited to land plants . Ancestors of TIR1/AFB and COI1 co-receptors could be identified in charophytes , but detailed residue analysis suggested these to be neither auxin- nor JA receptor . Thus , duplication of this gene , as well as multiple mutations in the LRR domain , must have preceded the deployment of these proteins as co-receptors . Auxin-dependence of ARFs is mediated by auxin-triggered degradation of Aux/IAA proteins , bridging ARF and TIR1/AFB proteins through two protein domains: the ARF-interacting PB1 domain and the TIR1/AFB-interacting domain II . We did find charophyte PB1-containing proteins that form a sister clade of land plant Aux/IAA . However , domain II was not detected in these proteins . Along with innovations in the proto-TIR1/AFB/COI1 protein , gain of a minimal degron motif in the Aux/IAA precursor likely completed the auxin response system in the early ancestor of land plants . Whether proto-TIR1/AFB/COI1 interact with Aux/IAA-like protein via an unknown ligand would be an interesting question for future analysis . Despite the lack of defined Aux/IAA and TIR1/AFB auxin co-receptor , the charophytes K . nitens and S . pratensis showed an extensive transcriptional response to exogenously supplied 2 , 4-D within 1 hr . A recent independent study showed IAA-dependent gene expression in K . nitens upon prolonged treatment with higher concentrations ( 100 μM for 10 hr to 7 days; Ohtaka et al . , 2017 ) . While S . pratensis has a proto-C-ARF , K . nitens does not appear to have proto-ARFs . Thus , by definition this response system must be different from the land plant auxin response system . Indeed , the charophyte orthologue of core land plant auxin responsive genes ( C2HDZ ) did not respond to 2 , 4-D and IAA . There was little , if any , overlap between auxin-responsive transcripts in the two charophytes , and in qPCR experiments on individual genes we noticed a high variability between experiments ( not shown ) . Thus , it appears that charophytes do respond to auxin-like molecules , but this response may not be robust , or it may strongly depend on growth conditions . Auxin resembles indole and tryptophan , and it is possible that the response to auxin observed is in fact a metabolic response to nutrient availability . Presence of endogenous IAA is observed in a wide range of algal species including charophytes , chlorophytes , rhodophytes , chromista , and cyanobacteria ( Žižková et al . , 2017 ) . Moreover , non-photosynthetic bacteria and fungi produce IAA and use it for communication with plants and algae ( Amin et al . , 2015; Fu et al . , 2015 ) , and thus it is likely that a response mechanism independent of the NAP exists in these species . Our data clearly indicate that ARF transcription factors were established in common ancestor of charophyte green algae and land plants . Structural homology models suggest that all the important residues for DNA-binding are conserved in proto-ARFs , suggesting that these should bind the same target DNA sequences . This should be assessed by biochemical experiments in the future . Given that there is a core set of auxin-regulated genes shared in all land plants , an intriguing possibility is that proto-ARFs already regulated this core set of genes that only became auxin-dependent upon establishment of TIR1/AFB and Aux/IAA proteins . Identification of the transcriptional targets of these proto-ARFs should help address this question . In any event , proto-ARFs—as well as critical residues for DNA binding—have been retained in many algal genomes for hundreds of millions of years , which suggests that they perform a biologically relevant function . Whether this function is related to the processes that auxin controls in land plants is an open question . Interestingly , our phylogenetic analysis indicated that the split between class C- and class A/B-ARFs occurred in charophytes before the establishment of Aux/IAA-TIR1/AFB co-receptor , and by extension likely before proto-ARFs were auxin-dependent . This suggests that class C-ARFs are fundamentally different from class A/B-ARFs . Indeed , genetic analysis in M . polymorpha revealed that its C-ARF likely does not act in auxin-dependent gene regulation . Several studies in A . thaliana showed that C-ARFs are involved in auxin response but the proposed role was different between studies ( Ding and Friml , 2010; Liu et al . , 2010; Mallory et al . , 2005; Wang et al . , 2005; Yang et al . , 2013 ) . In addition , C-ARFs of A . thaliana generally have weak affinity to Aux/IAA proteins ( Piya et al . , 2014 ) . To clarify the function of this ancient ARF subfamily , auxin-responsiveness of C-ARF proteins and relationship with A- or B-ARFs should be investigated in different species . A surprising outcome of the phylogenomic analysis was the discovery of two deeply conserved non-canonical proteins: ncIAA and ncARF . Charophytes have an Aux/IAA-like protein containing a PB1 domain , but lacking domain II , which is critical for auxin perception . This protein could regulate the function of proto-ARF ( or proto-RAV ) , but not in an auxin-dependent manner . While the canonical Aux/IAA gave rise to a large gene family , the ncIAA clade represented by a single member in every evolutionary node . The retention of a single ncIAA gene across plants suggests a fundamental function . Unfortunately , our mutant analysis in M . polymorpha could not reveal the function of ncIAA in auxin response and development in vegetative phase . ncIAA might have a function only in other developmental stages , or under specific stress conditions or environmental signals . No mutant in the Arabidopsis IAA33 gene has yet been reported , and perhaps such a mutant will help understand the ancient function of this protein . This work revealed that a class A-ARF-derived ncARF subfamily lacking a DBD is evolutionarily conserved among liverworts , mosses , and lycophytes . Mutant analysis using M . polymorpha clearly showed that ncARF functions as positive regulator in transcriptional auxin responses . There are two hypothetical models for ncARF function . ( 1 ) ncARF protects canonical ARFs from AUX/IAA-mediated inactivation through the interaction of PB1 domain . ( 2 ) ncARF interacts with target gene loci by interaction with canonical ARFs and help activate expression by recruiting co-factors . Irrespective of the mechanism of ncIAA and ncARF function , future models of auxin response will need to incorporate these conserved components . Through comparative transcriptomics , we infer that the number of DNA-binding ARF transcription factors scales with the number of auxin-regulated genes . Both P . patens and C . richardii have an expanded set of ARFs and display substantially more auxin-responsive genes than A . agrestis and M . polymorpha . It is likely that later duplications in the ARF family in the seed plants led to the thousands of auxin-responsive genes in these species ( Paponov et al . , 2008 ) . Another key evolutionary change is the transition from mostly gene repression to gene activation . We infer that this transition occurred in a common ancestor of euphyllophytes , and transcriptome analysis in A . thaliana and O . sativa shows this pattern persists in angiosperms ( Jain and Khurana , 2009; Paponov et al . , 2008 ) . There is a defining difference between bryophyte and euphyllophyte ARF families—a persisting duplication in the class A-ARFs . We hypothesize that the euphyllophytes duplication created an ARF copy that is more potent , or perhaps even specialized for gene activation . However , we cannot exclude the possibility that the difference in endogenous auxin levels or tissue complexity among species may results in different sensitivity to auxin treatment . The comparative transcriptomics also adds an interesting twist to our understanding of the functional distinction among ARF classes . Class A-ARFs are considered activators , and class B-ARFs repressors , perhaps through competing with class A-ARFs ( Lavy et al . , 2016; Ulmasov et al . , 1999 ) . Despite a complete lack of class B-ARFs , the hornwort A . agrestis showed comparable auxin-dependent gene repression to the other bryophytes , suggesting that auxin-dependent gene repression may not be mediated by class B-ARFs . Based on these findings , the role of class B-ARFs in auxin response may need to be reconsidered . A remarkable difference between bryophyte and euphyllophyte auxin-dependent transcriptomes is the appearance of genes with a large amplitude of regulation in the latter . Many auxin-responsive genes that were first identified in angiosperms such as A . thaliana have very high amplitudes ( Lee et al . , 2009 ) , but this appears to be a later innovation in the response system . The high amplitude is caused by more effective repression of gene activity in the no-auxin state , a property that is likely mediated by Aux/IAA proteins . Indeed , ferns have a much larger set of Aux/IAA proteins , as do all seed plants , and we propose that expansion of the Aux/IAA family enabled plants to articulate a clear distinction between on and off states in auxin response . In summary , this analysis reveals several design principles of the auxin response system .
Male M . polymorpha strain Takaragaike-1 ( Tak-1 ) was used as wild type and cultured as described previously ( Kato et al . , 2015 ) . K . nitens ( NIES-2285 ) , P . patens ( Gransden ) , and A . agrestis ( Oxford; Szövényi et al . , 2015 ) were cultured on BCD medium ( Cove et al . , 2009 ) solidified with 1% agar under the same condition with M . polymorpha . S . pratensis ( UTEX928 ) was cultured on Guillard’s Woods Hole medium ( Nichols , 1973 ) , pH7 . 9 containing 1% agar under white light with a 16 hr light/8 hr dark cycle at 22°C . C . richardii ( Hn-n ) was cultured on C-fern medium ( Plackett et al . , 2015 ) under continuous white light at 28°C . Data access to 1000 plant transcriptomes was provided by the OneKP consortium ( www . onekp . com; Matasci et al . , 2014 ) . All the transcriptome assemblies of the species from red algae , green algae , bryophytes , lycophyes , monilophytes , gymnosperms and basal angiosperms that were safely identified as non-contaminated has been used for this analysis ( Supplementary file 1 ) . CDS and protein sequences encoding all the orthologous genes in the three ( ARF , Aux/IAA amd TIR1/AFB ) gene families from M . polymorpha , P . patens , Amborella trichopoda , Oryza sativa , Zea mays , Solanum lycopersicum and A . thaliana were obtained from Phytozome ver11 ( phytozome . jgi . doe . gov/pz/portal . html ) . Aux/IAA genes from Picea abies were obtained from Spruce Genome Project ( www . congenie . org ) . K . nitens genome information was accessed from Klebsormidium nitens NIES-2285 genome project ( Hori et al . , 2014 ) . BLAST database for all the selected species were generated using 'makeblastdb' module in BLAST +v2 . 2 . 28 ( https://blast . ncbi . nlm . nih . gov ) . Protein sequences from A . thaliana , M . polymorpha and P . patens were used to query each database independently for each gene family using tBLASTn . All the scaffolds with the BLAST hits were extracted from the respective transcriptomes and further translated using TransDecoder ( ver2 . 0 . 1; http://transdecoder . github . io ) . This provided the CDS and protein sequences of all the scaffolds of the BLAST hits to any of the query sequences . The protein sequences were run through the InterProScan database ( ver5 . 19–58 . 0; http://www . ebi . ac . uk/interpro/ ) to look for conserved domains . Filtered sequences were further tested by BLASTx search against A . thaliana proteome to confirm orthology inferences . Some PB1 domain sequences in Chlorophytes that showed low similarity to A . thaliana proteins were also compared with M . polymorpha sequences to ascertain orthology . MAFFT ( ver7 . 123b; Katoh and Standley , 2013 ) iterative refinement algorithm ( E-INS-i ) was used to align the CDS sequences . Alignment positions with more than 50% gaps were removed using the Phyutility program ( ver2 . 2 . 6; http://blackrim . org/programs/phyutility/ ) before the phylogeny construction . PartitionFinder ( ver1 . 1 . 1; Lanfear et al . , 2012 ) was used to identify the most suitable evolutionary model for all the three gene families using the complete trimmed alignments on all the domains . Maximum likelihood algorithm implemented in RAxML ( ver8 . 1 . 20; Stamatakis , 2014 ) with General Time Reversible ( GTR ) model of evolution under GAMMA rate distribution with bootstopping criterion ( up to a maximum of 1000 bootstraps ) was used for the phylogenetic analysis . Obtained trees were visualized using the iTOL ( ver3; http://itol . embl . de/ ) phylogeny visualization program . Phylogenetic trees were cleaned up manually for misplaced sequences as well as for clades with long branch attraction . M . polymorpha gemmae or thallus explant without meristem and A . agrestis small thalli were planted on the medium covered with nylon mesh ( 100 μm pore ) and grown for 10 days . P . patens protonematal tissues were grown on the medium covered with cellophane for 10 days . Sterilized spores of C . richardii were grown for 2 weeks after which fertilization was performed by adding 5 ml of water on the plate . Seven days after fertilization , prothalli carrying sporophytic leaves were transferred on the medium covered with nylon mesh and grown for a further 7 days , after which sporophytes contained 3–4 leaves . After growing , plants with mesh or cellophane were submerged into liquid medium and cultured for 1 day . After pre-cultivation , 2 , 4-D was added to a final concentration of 10 μM and plants were incubated for 1 hr . Excess liquid medium were removed with paper towels and plants were frozen in liquid nitrogen . K . nitens and S . pratensis were streaked on solid medium and grown for 2 weeks . Algal cells were collected into 40 ml of liquid medium and cultured for 1 day with shaking at ~120 rpm . Then 2 , 4-D was added so that final concentration became 10 μM , followed by incubation for 1 hr with shaking . After auxin treatment , algal cells were collected using filter paper and frozen in liquid nitrogen . Frozen plant sample were grinded into fine powder with mortar and pestle . RNA from K . nitens , S . pratensis , M . polymorpha , and P . patens were extracted using Trizol Reagent ( Thermo Fisher Scientific; Waltham , Massachusetts ) and RNeasy Plant Mini Kit ( QIAGEN; Venlo , the Netherlands ) . RNA from A . agrestis and C . richardii were extracted using Spectrum Plant Total RNA Kit ( Sigma-Aldrich ) . Total RNA was treated with RNase-free DNase I set ( QIAGEN ) and purified with RNeasy MinElute Clean Up Kit ( QIAGEN ) . RNA-seq library construction with TruSeq kit ( Illumina; San Diego , California ) and 100 bp paired-end sequencing with Hiseq4000 ( Illumina ) were performed by BGI Tech Solutions ( Hong Kong ) . cDNA was synthesized with iScript cDNA Synthesis Kit ( Bio-Rad; Hercules , California ) . Quantitative PCR was performed using iQ SYBR Green Supermix ( Bio-Rad ) and CFX384 Touch Real-Time PCR Detection System . A two-step cycle consisting of denaturation at 95°C for 10 s followed by hybridization/elongation at 60°C for 30 s , was repeated 40 times and then followed by a dissociation step . Three technical and biological replicates were performed for each condition . PCR efficiencies were calculated using CFX Manager ( Bio-Rad ) software in accordance with the manufacturer’s instructions . For Marchantia polymorpha , relative expression values were normalized by the expression of EF1α ( Saint-Marcoux et al . , 2015 ) . All primers used for the analysis are listed in Supplementary file 3 . Obtained raw fastq reads were checked for quality control using FastQC ( www . bioinformatics . babraham . ac . uk/projects/fastqc ) . De novo transcriptome assemblies for all six species were generated using Trinity ( http://trinityrnaseq . github . io ) with default settings . To avoid any possible contamination from sequencing method and to improve the data quality , raw reads from land plants were mapped against charophyte de novo assemblies using Bowtie2 ( http://bowtie-bio . sourceforge . net/bowtie2/index . shtml ) in default settings and all the perfectly mapped pairs were removed , after which new assemblies were generated from pure raw read data for each species . In a similar way , contamination was removed in charophytes by mapping them against land plant transcriptome assemblies . Once the pure de novo transcriptome assemblies were generated , again Bowtie2 was used to map individual sample to the respective transcriptome assemblies using default parameters . Further , to improve the read count estimation and reduce the redundancy in Trinity transcripts , Corset ( Davidson and Oshlack , 2014 ) was implemented to estimate raw read counts using the Bowtie2 mapped alignment data . The obtained raw read counts were normalized and differentially expressed genes ( Padj <0 . 01 ) were identified using DEseq2 ( Love et al . , 2014 ) implemented in R Bioconductor package . All the RNAseq raw reads were deposited in NCBI Short Read Archive ( SRA ) under the BioProjectID: PRJNA397394 ( www . ncbi . nlm . nih . gov/bioproject/397394 ) . All other protein alignments mentioned in the manuscript were generated using ClustalOmega ( http://www . ebi . ac . uk/Tools/msa/clustalo/ ) . Visualization of the alignments were generated using Espript ( espript . ibcp . fr ) . Homology models were generated using Modeller v9 . 17 ( https://salilab . org/modeller/ ) . Modeled 3D structures were visualized using PyMol v1 . 7 . 4 ( The PyMOL Molecular Graphics System , Schrödinger , LLC ) . All the up-regulated genes’ nucleotide sequences from the six species were aligned against the same sequences using tBLASTx to find the similar ( orthologous ) genes among various species . From these results , the BLAST hits with E-value less than 0 . 001 with a length of at least 30 amino acids were considered for further analysis . Moreover , these sequences were also searched for orthologues in A . thaliana proteome using BLASTx . Both the similarities among the six species and the orthologous gene information from A . thaliana were loaded into Cytoscape ( www . cytoscape . org ) to visualize the network of similar gene families . A similar procedure was performed for finding the commonly downregulated gene families . To generate the entry clones carrying sgRNA cassette , pairs of oligo DNAs ( HK001/HK002 or HK003/HK004 for ARF3 , HK162/HK163 for ncARF , HK168/HK169 for ncIAA ) were annealed and cloned into pMpGE_En03 ( Addgene; Cambridge , Massachusetts ) using BsaI site . The sequence of oligo DNAs are listed in Supplementary file 3 . Resultant sgRNA cassette were transferred into pMpGE_010 ( Addgene ) by LR reaction using Gateway LR Clonase II Enzyme Mix ( Thermo Fisher Scientific ) . Transformation into Tak-1 was performed as described previously ( Kubota et al . , 2013 ) using Agrobacterium strain GV3101:pMp90 . For genotyping , genomic DNA was extracted by simplified CTAB ( cetyltrimethylammonium bromide ) method ( http://moss . nibb . ac . jp/protocol . html ) . Genomic region including target site of sgRNA was amplified with PCR using the primer set HK079/HK131 ( ARF3 ) , HK172/HK173 ( ncARF ) and HK174/HK175 ( ncIAA ) , and sequenced . All primers used in this study are listed in Supplementary file 3 . MpARF3 promoter fragment including 5’ UTR and 3 kb up stream region was amplified with PCR using the primer set HK111/HK026 and cloned into pMpGWB307 ( Ishizaki et al . , 2015 ) using XbaI site ( pJL002 ) . Genomic CDS of MpARF3 without stop codon was amplified with PCR using the primer set HK027/028 and subcloned into pENTR/D-TOPO vector ( Thermo Fisher Scientific ) . Mutation which confers resistant to sgRNA was introduced by PCR using primer set HK137/138 . Then mutated CDS fragment was transferred into pJL002 by by LR reaction and fused with promoter and Citrine tag ( pHKDW103 ) . All primers used in this study are listed in Supplementary file 3 . Resultant vector was transformed into arf3ge2-1 mutant thallus as described previously . Citrine signal and bright field images were captured using a Leica SP5-II confocal laser scanning microscope system , with excitation at 514 nm and detection at 520–600 nm . Raw read data of RNA-seq can be accessed in NCBI Short Read Archive ( ID: PRJNA397394 ) . | Across all kingdoms of life , signaling molecules like hormones , for example , control many aspects of the lives of organisms , including how they grow and develop . Cells have dedicated proteins that can recognize the signaling molecules , relay the information , and respond to the signal , for example by switching genes on or off . Such response systems usually consist of multiple components , and , throughout evolution , these response components have regularly been copied such that many species have multiple different versions of each one . Auxin is a plant hormone that controls virtually all growth and developmental processes in plants , including many yield traits in crops . However , no one knows why it is involved in so many processes . This is partly because it is not clear how the response system for this central signaling molecule was first born , or how it has increased in its complexity . To address this , Mutte , Kato et al . explored the genetic information of more than a thousand plant species , including algae , which span more than 700 million years of evolution . Their analysis showed that all auxin response components were assembled from pieces of much older genes , but that they first came together when plants conquered land . Indeed , the auxin response appears to have developed on top of a pre-existing genetic regulator that is still present in modern-day algae . Mutte , Kato et al . then used experiments to show how stepwise increases in the number and types of auxin response components have shaped sophisticated , complex responses in land plants , and to demonstrate how ancient components control auxin response . Together these findings provide a framework for understanding the many functions of auxin in plants , and how this came to be . They also show how complexity can be accomplished in a signal response pathway , and how diversity evolves in gene families . Similar studies on other response systems in plants and beyond are likely to help reveal common principles of hormone response evolution and diversification of gene regulation systems . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"evolutionary",
"biology"
] | 2018 | Origin and evolution of the nuclear auxin response system |
The effects of land use on soil invertebrates – an important ecosystem component – are poorly understood . We investigated land-use impacts on a comprehensive range of soil invertebrates across New Zealand , measured using DNA metabarcoding and six biodiversity metrics . Rarity and phylogenetic rarity – direct measures of the number of species or the portion of a phylogeny unique to a site – showed stronger , more consistent responses across taxa to land use than widely used metrics of species richness , effective species numbers , and phylogenetic diversity . Overall , phylogenetic rarity explained the highest proportion of land use-related variance . Rarity declined from natural forest to planted forest , grassland , and perennial cropland for most soil invertebrate taxa , demonstrating pervasive impacts of agricultural land use on soil invertebrate communities . Commonly used diversity metrics may underestimate the impacts of land use on soil invertebrates , whereas rarity provides clearer and more consistent evidence of these impacts .
Land-use changes through deforestation , agricultural development , and urbanisation have caused worldwide impacts on the biodiversity of terrestrial communities and ecosystems ( Dirzo et al . , 2014; Newbold et al . , 2015 ) . Invertebrates are the most diverse and abundant component of animal biodiversity worldwide and are major contributors of terrestrial ecosystem services such as pollination , soil formation , and nutrient cycling ( Lavelle et al . , 2006; Wagg et al . , 2014; Yang and Gratton , 2014 ) . Long-term declines in the richness and biomass of insects and other terrestrial invertebrates are predicted to have major impacts on food webs and ecosystem functions ( Eisenhauer et al . , 2019; Hallmann et al . , 2017; Potts et al . , 2010 ) . Despite this , most invertebrate species remain undescribed , and there is an incomplete understanding of land-use effects on invertebrate biodiversity , particularly for those that reside in soils ( Cameron et al . , 2018; Eisenhauer et al . , 2019 ) . Biodiversity loss is typically measured as reductions in species richness ( i . e . , total number of species; e . g . Forister et al . , 2010; George et al . , 2019; Newbold et al . , 2015 ) . Despite widespread concern about biodiversity loss , evidence for impacts of anthropogenic land use on terrestrial invertebrate species richness is mixed , with studies often detecting richness declines for some taxa or groups but not others ( Allan et al . , 2014; Attwood et al . , 2008; Blaum et al . , 2009 ) . Among the few studies that have examined land-use impacts on below-ground invertebrate communities , one detected negative impacts of long-term disturbance on soil invertebrate richness ( Callaham et al . , 2006 ) , another detected increasing alpha diversity and homogenisation of soil invertebrates with increasing grassland intensification ( Gossner et al . , 2016 ) ; while others detected inconsistent richness patterns among different soil invertebrate taxa across land uses ( George et al . , 2019; Tsiafouli et al . , 2015; Wood et al . , 2017 ) . These inconsistent patterns make it difficult to draw general conclusions about the impacts of land use on soil invertebrate biodiversity ( Allan et al . , 2014 ) , and make the use of individual taxa as bioindicators problematic ( Gerlach et al . , 2013 ) . Inconsistent patterns in biodiversity measurement may reflect limitations of the diversity index used . In particular , species richness provides no indication of the distribution , taxonomy or function of species or communities ( Fleishman et al . , 2006; Hillebrand et al . , 2018 ) , potentially overlooking the nature and extent of land-use impacts on soil invertebrate communities . In contrast , rarity ( sometimes termed ‘endemism richness’; Kier et al . , 2009 ) measures the extent to which species are widely distributed generalists or limited to particular sites or land-use types . Rarity may thus indicate homogenising effects of land use on communities ( McKinney and Lockwood , 1999; Smart et al . , 2006 ) , and the conservation value of sites ( Kier and Barthlott , 2001 ) . Furthermore , rare species can contribute disproportionately to ecosystem functioning ( Dee et al . , 2019; Leitão et al . , 2016; Lyons et al . , 2005; Mouillot et al . , 2013 ) . Rarity may therefore more accurately reflect the impacts of land use on soil invertebrate communities than species richness . Rarity and other diversity metrics can also be placed in a phylogenetic context . Phylogenetic diversity reflects the evolutionary history and taxonomic range of communities and associated traits and functions ( Faith , 1992 ) , thus providing robust information for conservation assessment purposes ( Faith , 1992; Forest et al . , 2007; González-Orozco et al . , 2015; Mishler et al . , 2014 ) . Phylogenetic diversity can also act as a proxy for functional diversity , albeit imperfectly ( Mazel et al . , 2018; Srivastava et al . , 2012; Winter et al . , 2013 ) . Phylogenetic rarity , calculated as the portion of a phylogeny that is unique to a region or habitat ( Mishler et al . , 2014; Rosauer et al . , 2009 ) , combines elements of both rarity and phylogenetic diversity; high phylogenetic rarity implies that a community contains a taxonomically distinct assemblage of species and associated ecosystem functions . Mean pairwise distance , meanwhile , measures the phylogenetic relatedness of species within a community , which may reflect land-use driven filtering or competitive exclusion processes ( Webb et al . , 2002 ) . The additional information represented by rarity and phylogenetic biodiversity metrics suggests that land-use related patterns based on these values may be clearer and more consistent among soil invertebrate taxa than those based on species richness and other non-phylogenetic diversity measures . Furthermore , rarity and phylogenetic rarity may be more sensitive indicators of land-use impacts on soil invertebrate communities than richness or phylogenetic diversity , because the former metrics reflect the distribution of species and lineages whereas the latter do not . These possibilities remain untested . Here we present a comprehensive analysis of soil invertebrate biodiversity across different land-use types at a national spatial scale . We use modern DNA metabarcoding methods to measure invertebrate responses , as this enables the rapid and detailed identification of large numbers of invertebrate specimens from multiple taxonomic groups simultaneously ( Drummond et al . , 2015; George et al . , 2019; Wood et al . , 2017; Yu et al . , 2012 ) and allows more efficient calculation of biodiversity metrics than previously possible . We analysed the invertebrate faunas in soil samples collected from 75 sites distributed across five different major land-use categories ( natural forest , planted forest , low-producing and high-producing grassland , and perennial cropland ) throughout New Zealand . Based on these data , we calculated six different biodiversity metrics: species richness , effective species numbers , rarity , phylogenetic diversity , phylogenetic rarity , and mean pairwise distance; as well as standardised effect size ( SES ) values for the latter phylogenetic metrics . We used these metrics to assess the impacts of land use on a comprehensive range of soil invertebrate taxa . We tested the following hypotheses: 1 ) all soil invertebrate taxa show the same biodiversity trends across the five land-use types; 2 ) patterns of soil invertebrate rarity , phylogenetic diversity , and phylogenetic rarity across the five land-use types are more consistent among taxa than species richness or non-phylogenetic diversity; 3 ) rarity and phylogenetic rarity of soil invertebrates are more sensitive to land use than richness , diversity , or phylogenetic diversity .
We detected a total of 11 , 284 operational taxonomic units ( OTUs ) , of which 4549 ( 40 . 3% ) were identified as terrestrial invertebrates . The remainder were identified as protists ( 37 . 6% ) , fungi ( 14 . 9% ) , non-terrestrial metazoans ( 5% ) , bacteria ( 1 . 7% ) , and plants ( 0 . 5% ) . The terrestrial invertebrate OTUs mostly belonged to the phylum Arthropoda ( 2 , 626 OTUs , among which insects were most common ) , followed by Rotifera ( 772 OTUs ) , Nematoda ( 656 OTUs ) , Mollusca ( 219 OTUs ) , Annelida ( 204 OTUs ) , Platyhelminthes ( 44 OTUs ) , Tardigrada ( 22 OTUs ) , Gastrotricha ( four OTUs ) , and Onychophora ( two OTUs ) ( Appendix 1—figures 1 and 2 ) . Non-metric MDS ordinations showed clear differences between overall invertebrate community composition in samples from different land-use categories ( Figure 1 ) . Natural forest samples formed a distinct cluster with no overlap with any other land-use categories . Samples from the other four land-use categories overlapped , with planted forest communities most similar to those from low-producing grassland followed by high-producing grassland communities , and least similar to those from perennial cropland . Similar trends were observed when only Arthropoda , Mollusca , Nematoda , or Rotifera OTUs were included , whereas Annelida OTUs showed less distinction between land-use categories . PERMANOVA tests for composition differences among different land-use categories detected a significant difference based on the overall invertebrate community ( F4 , 61 = 1 . 804 , p≤0 . 001 ) , and based on each of the main phyla detected ( Annelida , Arthropoda , Mollusca , Nematoda and Rotifera; F4 , 44-61 = 1 . 447–2 . 288 , p≤0 . 001; Figure 1—source data 1A ) . To test for homogenisation effects of land use on soil invertebrate communities we compared multivariate heterogeneity/homogeneity of sample dispersions , mean pairwise beta diversity , and mean pairwise phylogenetic beta diversity , between land-use categories . For overall invertebrate communities , each of these measures differed significantly among land uses ( F4 , 61-442 = 3 . 59–14 . 99 , p≤0 . 011 ) , being highest in natural forest sites and lowest in grassland and/or cropland sites ( Figure 1—source data 1B; Figure 1—figure supplements 1–3 ) . Similar trends were observed for Arthropoda and Nematoda communities based on all three measures , and for Annelida and Mollusca communities based on phylogenetic beta diversity and multivariate heterogeneity of sample dispersions , whereas Rotifera communities showed different patterns . A heatmap based on the 1000 most relatively abundant terrestrial invertebrate OTUs detected suggested that low-producing grassland , high-producing grassland , and perennial cropland samples each had relatively consistent assemblages of abundant OTUs , both within and between each land-use category , whereas planted forest samples , and especially natural forest samples , each had more distinctive assemblages of abundant OTUs ( Figure 2 and Figure 2—figure supplement 1 ) . In particular , most of the natural forest samples had a subset of abundant OTUs that were not detected in any other sample . All biodiversity metrics ( except for mean pairwise distance ) showed a general trend of declining overall invertebrate biodiversity ( i . e . the biodiversity of the entire invertebrate community ) from forested and/or low-producing grassland sites to high-producing grassland and/or perennial cropland sites ( Figures 3 and 4 ) . Rarity and phylogenetic rarity metrics showed the largest and most consistent land-use-related biodiversity declines , with the highest mean values in natural forest sites followed by planted forest sites and low-producing grassland sites , and high-producing grassland sites , and lowest values in perennial cropland sites . Removing species found in only a single site did not substantially change these trends ( Appendix 1—figures 3–5 ) . Significant differences between mean biodiversity of overall invertebrate communities in different land-use categories were detected according to richness , rarity , phylogenetic diversity , phylogenetic rarity , and phylogenetic diversity and rarity SES metrics ( F4 , 64 = 3 . 56 to 17 . 986 , p = 0 . 012 to <0 . 001 ) , but not effective species numbers , mean pairwise distance , or mean pairwise distance SES metrics ( Figure 3—source data 1A; Figure 4—source data 1A ) . ANOVA tests of derived land-use rank trends provided similar results , with significant trends identified for all metrics except for mean pairwise distance and mean pairwise distance SES ( F1 , 67 = 4 . 66–31 . 94 , p = 0 . 034 to <0 . 001; Appendix 1—table 1 ) . The mean rarity of overall invertebrate communities was significantly lower in all four other land uses compared with natural forest ( t23-27 = −31 . 6 to −62 . 4 , P . adj = 0 . 03 to <0 . 001 ) . Similarly , the mean phylogenetic rarity of overall invertebrate communities was significantly lower in all four other land-use categories compared with natural forest ( t23-27 = −3 . 34 to −6 . 90 , P . adj = 0 . 043 to <0 . 001 ) , and in perennial cropland compared with planted forest ( t24 = −3 . 55 , P . adj = 0 . 046 ) . In contrast , the mean richness and phylogenetic diversity of overall invertebrate communities were similar in natural forest , planted forest , and low-producing grassland samples , and significantly lower in perennial cropland compared with natural forest ( t23 = −78 . 3 , P . adj = 0 . 023 , and t23 = −14 . 6 , P . adj = 0 . 008 , respectively ) and compared with low-producing grassland ( t23 = −84 . 2 , P . adj = 0 . 012 , and t23 = −13 . 3 , P . adj = 0 . 019 , respectively; Figure 3 ) . Mean phylogenetic diversity SES was significantly lower in low-producing grassland compared with natural forest ( t23-27 = −2 . 20 , P . adj = 0 . 048 ) , but did not otherwise differ between land-use categories , while phylogenetic rarity SES differences between land-use categories matched those based on non-SES phylogenetic rarity ( t23-27 = −3 . 68 to −8 . 61 , P . adj = 0 . 031 to <0 . 001; Figure 4 ) . A mixed-model ANOVA test for effects of derived land-use rank , land-use category , and taxonomic group effects showed that derived land-use rank and taxonomic group ( and interactions ) were the most consistently significant predictors of the diversity metrics ( F1-16 = 7 . 74 to 32 . 14 , p = 0 . 007 to <0 . 001; Appendix 1—table 2 ) . The further addition of land-use category to models already containing derived land-use rank did not explain additional variation for effective species , rarity , phylogenetic rarity and mean pairwise distance , but did for richness and phylogenetic diversity ( in the form of significant interactions between land-use category and taxonomic group; F48 = 1 . 41 and 1 . 82 , p = 0 . 037 and <0 . 001 ) . Most environmental variables showed clear land use-related trends of increasing or decreasing values in the order of natural forest , planted forest , low-producing grassland , high-producing grassland , and perennial cropland ( Appendix 1—figure 6 ) . An ANOVA test of spatial attributes ( latitude and altitude ) plus land-use category showed latitude had no effect on overall soil invertebrate biodiversity according to any metric , whereas altitude had significant effects on biodiversity of all metrics except for mean pairwise distance ( F1 = 9 . 41 to 22 . 33 , p = 0 . 003 to <0 . 001 ) . In addition to altitude , land-use category had a significant effect only on rarity and phylogenetic rarity metrics ( F1 = 4 . 40 and 4 . 60 , p = 0 . 003 and 002; Appendix 1—table 3 ) . The first three components of a PCA incorporating latitude , altitude , and soil chemistry variables explained 70 . 25% of variance . According to an ANOVA test of these three PCA components plus land-use category , the first component had significant effects on the rarity , phylogenetic diversity and phylogenetic rarity of the overall soil invertebrate biodiversity ( F1 = 4 . 79 to 15 . 25 , p = 0 . 032 to <0 . 001 ) , and the second component on the former three metrics plus richness ( F1 = 7 . 00 to 10 . 24 , p = 0 . 010 to 0 . 002 ) . The third component did not have a significant effect on any of the metrics . The addition of land-use category to these models explained further variation for richness , rarity , and phylogenetic rarity metrics only ( F4 = 2 . 71 to 4 . 72 , p = 0 . 038 to 0 . 006; Appendix 1—table 4 ) , indicating that there was some confounding between the environmental PCAs and land-use category . Biodiversity metrics for the main insect orders ( Coleoptera , Diptera , Hymenoptera , Lepidoptera , Hemiptera , and all other insects ) , other arthropod taxa ( Collembola , mites , non-mite Arachnida , Malacostraca , myriapods ) , and non-arthropod phyla ( Annelida , Mollusca , Nematoda , Platyhelminthes , Rotifera , and Tardigrada ) that were detected showed a general trend of declining biodiversity from forested to agricultural sites . Rarity , phylogenetic diversity , and phylogenetic rarity patterns were most consistent among different taxonomic groups ( Appendix 1—figures 7–12 ) , while land-use trends were most clear and consistent across taxonomic groups according to rarity and phylogenetic rarity ( Figure 3—figure supplements 1 and 2 ) . ANOVA tests detected significant differences among land-use categories for ten of the 17 taxonomic groups based on rarity ( all insect groups , non-mites , Annelida , Nematoda , and Platyhelminthes; F4 = 2 . 60 to 13 . 26 , p = 0 . 048 to <0 . 001 ) ; nine groups based both on phylogenetic rarity ( all insect groups except Hemiptera , mites and non-mites , Annelida , and Platyhelminthes; F4 = 2 . 74 to 11 . 07 , p = 0 . 036 to <0 . 001 ) and phylogenetic diversity ( all insect groups , Annelida , Mollusca , and Nematoda; F4 = 3 . 14 to 6 . 41 , p = 0 . 047 to <0 . 001 ) ; eight groups based on richness ( all insect groups , Nematoda , and Platyhelminthes; F4 = 2 . 55 to 6 . 32 , p = 0 . 048 to <0 . 001 ) ; five groups based on effective species numbers ( Diptera , Hymenoptera , Lepidoptera , mites , and Annelida; F4 = 2 . 73 to 4 . 36 , p = 0 . 037 to 0 . 004 ) ; and three groups based on mean pairwise distance differences ( Hymenoptera , mites , and Rotifera; F4 = 3 . 53 to 6 . 24 , p = 0 . 012 to <0 . 001; Figure 3—source data 1B ) . Tests of derived land-use rank trends for each metric and taxonomic group provided concordant results , with the same groups ( with few exceptions ) showing significant trends for each metric ( Appendix 1—table 5 ) . Post-hoc Tukey HSD tests showed that biodiversity was most commonly significantly higher in natural forest compared with perennial cropland ( Figure 3—figure supplements 1 and 2 ) . This was observed for nine taxonomic groups based on rarity ( t14-23 = 1 . 92 to 7 . 31 , P . adj = 0 . 040 to <0 . 001 ) , eight groups based on phylogenetic rarity ( t20-28 = 0 . 054 to 1 . 19 , P . adj = 0 . 024 to <0 . 001 ) , five groups based on phylogenetic diversity ( t20-23 = 1 . 16 to 2 . 63 , P . adj = 0 . 032 to <0 . 001 ) , four groups based on richness ( t14-23 = 3 . 69 to 9 . 47 , P . adj = 0 . 026 to <0 . 001 ) , three groups based on mean pairwise distance ( t22-23 = 0 . 03 to 0 . 35 , P . adj = 0 . 014 to 0 . 003 ) , and just one group based on effective species numbers ( t25 = 3 . 00 , P . adj = 0 . 012 ) . Biodiversity was also significantly higher in natural forest compared with high-producing grassland ( for two to six groups according to each of five metrics; t21-27 = 0 . 02 to 6 . 86 , P . adj = 0 . 029 to <0 . 001 ) , low-producing grassland ( one to five groups , four metrics; t20-26 = 0 . 04 to 4 . 71 , P . adj = 0 . 040 to <0 . 001 ) , and planted forest ( one to three groups , three metrics; t24-27 = 0 . 64 to 4 . 61 , P . adj = 0 . 041 to 0 . 007 ) ; in planted forest , low-producing grassland , or high-producing grassland compared with perennial cropland ( one to two groups , two to five metrics; t12-24 = 0 . 38 to 16 . 92 , P . adj = 0 . 045 to 0 . 001 ) ; and in planted forest or low-producing grassland compared with high-producing grassland ( one or two groups , two metrics; t23-30 = 2 . 14 to 3 . 33 , P . adj = 0 . 036 to 0 . 023 ) . All of the pairwise differences together implied a land-use category rank order of natural forest > planted forest > low producing grassland > high producing grassland > perennial cropland . Non-parametric bootstrapping of ANOVA sum of squares values for the ( non-SES ) biodiversity metrics and taxonomic groups for which significant land-use differences were detected showed that phylogenetic rarity followed by ( non-phylogenetic ) rarity explained the largest proportions of land-use category variance across the 17 taxonomic groups , while mean pairwise distance and richness explained the least variance ( Figure 5 ) . A Kruskal-Wallis test detected significant differences among the biodiversity metrics ( Chi square = 4782 . 6 , df = 5 , p<0 . 001 ) , with post-hoc tests indicating that the distributions of all metrics differed significantly from each other ( p<0 . 05 ) . Patterns of phylogenetic rarity SES values among land-use categories were more consistent across taxonomic groups , and with their corresponding non-SES metric patterns , than patterns of phylogenetic diversity SES and mean pairwise distance SES values ( Figure 4—figure supplements 1 and 2 ) . ANOVA tests detected significant differences among land-use categories for 11 of the 17 taxonomic groups based on phylogenetic rarity SES ( Collembola , Coleoptera , Diptera , Lepidoptera , other insects , mites and non-mites , Annelida , Mollusca , Nematoda , and Rotifera; F4 = 3 . 10 to 8 . 91 , p = 0 . 022 to <0 . 001 ) , six groups based on phylogenetic diversity SES ( Hymenoptera , Lepidoptera , mites , Malacostraca , Nematoda , and Rotifera; F4 = 2 . 76 to 7 . 39 , p = 0 . 035 to <0 . 001 ) ; and four groups based on mean pairwise distance SES ( Lepidoptera , mites , Malacostraca , and Rotifera; F4 = 4 . 40 to 11 . 28 , p = 0 . 016 to <0 . 001; Figure 4—source data 1B ) . All of the 11 taxonomic groups with significant phylogenetic rarity SES differences showed a consistent pattern of declining rarity from natural forest to planted forest to agricultural land-use categories . Post-hoc Tukey HSD tests detected significantly higher phylogenetic rarity SES values in natural forest ( for 11 groups ) and in planted forest ( for four groups ) compared with at least two of the agricultural land-use categories in each case ( t22-28 = −1 . 73 to −3 . 77 , P . adj = 0 . 047 to <0 . 001 ) . In contrast , only two groups ( mites and Rotifera ) showed this pattern based on either phylogenetic diversity SES ( t22-28 = −1 . 52 to −3 . 15 , P . adj = 0 . 031 to <0 . 001 ) or mean pairwise distance SES values ( t22-28 = −1 . 83 to −2 . 89 , P . adj = 0 . 047 to <0 . 001 ) . Otherwise , Lepidoptera phylogenetic diversity SES values were significantly lower in both planted forest and high-producing grassland compared with both natural forest and perennial cropland ( t21-27 = −1 . 07 to −1 . 44 , P . adj = 0 . 035 to 0 . 004 ) , whereas Hymenoptera , Malacostraca and Nematoda phylogenetic diversity SES values were higher in one or more of the anthropogenic land use categories compared with natural forest ( t3-28 = 1 . 46 to 2 . 87 , P . adj = 0 . 020 to 0 . 005 ) . Patterns of mean pairwise distance SES values across land use categories and taxonomic groups closely matched those observed for phylogenetic diversity SES values ( except significant differences among land-use categories were not detected for Hymenoptera or Nematoda ) .
The low beta diversity , heterogeneity , and rarity values detected in agricultural sites , and the overlap of samples from these sites in MDS ordinations , together strongly imply that these habitats tend to have relatively similar assemblages of species across locations . Agricultural practices have effects at a wide range of scales , from local-scale use of chemical fertilisers and pesticides to landscape-scale habitat simplification ( Tscharntke et al . , 2005 ) . Together these factors lead to homogenisation of communities and functions among sites , in which specialists in diverse natural communities are replaced by a smaller number of generalists that thrive in anthropogenic habitats ( Börschig et al . , 2013; Clavel et al . , 2011; Gámez-Virués et al . , 2015; McKinney and Lockwood , 1999; Smart et al . , 2006 ) . In contrast to the agricultural sites , the high diversity and rarity observed in natural forest sites indicates that these habitats tend to have richer and more unique assemblages of species . Forested sites tend to have greater physical habitat complexity and heterogeneity , providing more varied resources and niches for diverse communities including various specialists ( Jonsson et al . , 2009; Stein et al . , 2014 ) . Furthermore , natural forest habitats tend to be more disconnected , and located in more rugged and less accessible areas than agricultural sites , with more physical barriers to limit the dispersal of invertebrate fauna . Consequently , the distinct assemblages detected in natural forest sites are likely to reflect natural historical biogeographic distribution and evolutionary processes ( Buckley et al . , 2015; Trewick et al . , 2011 ) . Despite their varying sensitivity , most metrics of rarity and diversity ( not mean pairwise distance , phylogenetic diversity SES , or mean pairwise distance SES ) showed a consistent trend of lower biodiversity in agricultural land-use categories than in forested land-use categories . Further , while not all taxa showed significant evidence of declining biodiversity in relation to agricultural land use , no taxa responded positively . Many taxa not showing significant biodiversity declines had few species ( e . g . myriapods , Malacostraca and tardigrades ) , suggesting there was insufficient data to infer land-use differences . Among the most species-rich groups that did not show significant declines ( collembola , mites and rotifers ) , many of the diversity metrics were nonetheless lowest in grassland or perennial cropland sites , suggesting that while these groups may be more resilient to impacts of agricultural land use than others , the general trend was similar . These biodiversity declines are in contrast to previous research that suggested soil fauna are resilient to grassland intensification ( Gossner et al . , 2016 ) , likely because our study encompasses a broader range of land-use types . While it is likely that spatial and environmental factors associated with particular land uses contribute to these patterns , the fact that land use explained additional variation of richness and rarity metrics after these factors were statistically accounted for strongly indicates an independent role of land management practices . While rarity and phylogenetic rarity metrics showed the most consistent responses across land-use categories , the rank order of land-use categories implied by these ( and other ) metrics were not easily predicted prior to measurement . Planted forests , which were predominantly Pinus radiata plantations , are sometimes perceived as being biologically depauperate , while low-producing grasslands are frequently perceived as semi-natural in New Zealand ( Hobbs et al . , 2006 ) . Despite this , we found rarity and diversity in planted forest sites to be similar to those in low-producing grassland sites and higher than those in high-producing grassland or perennial cropland sites , consistent with suggestions that plantations can play an important role in insect biodiversity conservation ( Pawson et al . , 2009; Pawson et al . , 2010 ) . Similarly , high-productivity grasslands are often perceived as a more severe land use than perennial cropland due to high homogeneity of vegetation cover , low habitat complexity , and high fertiliser use . Nonetheless , our data suggest perennial cropland supports the lowest levels of invertebrate diversity and rarity of any of the measured land-use categories . This may reflect high chemical input in and intensive management of fruit production systems ( Manktelow et al . , 2005 ) . Overall , our results suggest pervasive impacts of agricultural land use upon soil invertebrate communities , with likely adverse consequences for ecosystem services . This adds to widespread evidence of declines in invertebrate biomass and diversity in response to anthropogenic land-use change and habitat loss ( Attwood et al . , 2008; Hallmann et al . , 2017; Hendrickx et al . , 2007; Powney et al . , 2019 ) , and suggests that efforts to conserve and restore soil invertebrate communities may be needed . Invertebrates tend to be neglected by conservation initiatives , due to the challenges of determining their identities , functions , and distributions ( Leandro et al . , 2017 ) . Indirect preservation of communities via flagship or umbrella species protection schemes tends to be ineffective ( Andelman and Fagan , 2000; Oberprieler et al . , 2019; Schuldt and Assmann , 2010 ) , and similarly , biomonitoring based on individual species is problematic . By allowing the efficient assessment of invertebrate community composition and distribution across large spatial scales , DNA metabarcoding methods may enable more informative biomonitoring and improved targeting of conservation initiatives based on multiple invertebrate taxa , if not entire invertebrate communities . While rarity and phylogenetic rarity were the most informative metrics of community change in this case , it is likely that consideration of these alongside richness and phylogenetic measures of diversity would provide the most comprehensive information for purposes such as biomonitoring and conservation planning ( Fleishman et al . , 2006 ) . Our results suggest that conserving a network of sites with high invertebrate diversity and rarity would preserve a diverse assemblage of species , communities , and functional traits , thus providing resilience of communities and ecosystem processes to environmental changes ( Balvanera et al . , 2006; Yachi and Loreau , 1999 ) . While diversity and rarity was typically highest in our natural forest sites ( of which many are protected ) , certain grassland and cropland sites with unusually high rarity values ( outliers on Figure 3 ) might be logical targets for further investigation and potential incorporation into conservation initiatives . In conclusion , our analysis of soil invertebrate biodiversity across land-use categories at a national scale shows that most soil invertebrate taxa have consistent rarity responses to land use , and that agricultural land use tends to cause the homogenisation and loss of soil invertebrate biodiversity . This research adds to evidence of widespread impacts of anthropogenic land use on invertebrate biodiversity , but also implies that these impacts may have been underestimated due to a widespread emphasis on species richness . DNA metabarcoding methods offer an efficient basis for measuring the diversity and rarity of invertebrate communities at large scales . Incorporating this information into conservation schemes would enable the protection of a broader range of biodiversity and enhance the preservation of terrestrial ecosystems .
Soil invertebrate communities were sampled from a total of 75 sites distributed across five different major land-use categories throughout New Zealand ( Figure 6 ) , during dry weather between November 2014 and March 2015 . The five land-use categories ( natural forest , planted forest , low-producing grassland , high-producing grassland , and perennial cropland ) represent differing states of anthropogenic modification ( Figure 6—source data 1 ) . The site locations were selected from a nationwide 8 km grid used for regular monitoring of native species and pests . For each land-use category , 15 replicate sites were randomly selected from the nationwide monitoring grid , excluding any that were >1000 m altitude and ensuring they were distributed across the length of New Zealand ( Makiola et al . , 2019 ) . At each site , a 20 m × 20 m plot was established according to a standardised protocol ( Hurst and Allen , 2007 ) . Twenty-four soil cores were collected within each plot on a regular grid ( min 3 . 54 m distance between cores ) to a depth of 15 cm using a sterile corer ( 5 . 08 cm diameter ) , following Wood et al . ( 2017 ) . Surface litter was removed prior to coring . The 24 soil cores were pooled together , homogenised , and stored at 4°C until laboratory processing . Invertebrates were extracted from a one-litre subsample of homogenised soil material from each site using Berlese-Tullgren funnels and stored in ethanol until DNA extraction . The altitude and latitude of plots were determined from topographic maps . Soil chemistry variables ( pH , C , N , C:N ratio , Olsen P , Total P , Ca , Mg , K , Na , cation exchange capacity , base saturation ) were determined for each plot according to Orwin et al . ( 2016 ) and Wood et al . ( 2017 ) . Bulk invertebrate concentrates were centrifuged for three minutes at 2 , 500 rpm ( 1258 rcf ) , after which ethanol was removed until <5 ml remained . The concentrates were then transferred into 5 ml tubes and homogenised with eight steel balls in a bead mill operated at 15 Hz for six intervals of 20 s each . A 1 . 5 ml aliquot of homogenised invertebrate concentrate from each sample was removed into a 1 . 5 ml microtube and centrifuged for one minute at 13 , 000 rpm ( 11 , 337 rcf ) , after which any ethanol was removed . The pelleted material was resuspended in purified water , re-centrifuged as before , then resuspended in 200 µl digestion buffer ( 10 mM Tris buffer , 10 mM NaCl , 5 mM CaCl2 , 2 . 5 mM EDTA , 2% SDS , 0 . 04 M dithiothreitol , and 0 . 1 M proteinase K ) with vortexing , and incubated overnight at 56 °C with shaking at 450 rpm ( Campos and Gilbert , 2012 ) . DNA was extracted from the digested samples using a Macherey-Nagel NucleoSpin Tissue kit ( MACHEREY-NAGEL GmbH and Co . KG , Düren , Germany ) , omitting sample lysis steps but otherwise according to the manufacturer’s directions , with a JANUS workstation laboratory robot ( PerkinElmer , Waltham , MA , USA ) . The DNA concentration was quantified in each extract using an Invitrogen Quant-iT PicoGreen dsDNA quantitation assay kit ( Thermo Fisher Scientific , Waltham , MA USA ) , and standardised across samples to 3 ng/µl . COI barcodes were amplified by PCR from each sample using metazoan-targeted primers mICOIintF ( 5'-GGWACWGGWTGAACWGTWTAYCCYCC-3' ) ( Leray et al . , 2013 ) and HCO2198 ( 5'-TAAACTTCAGGGTGACCAAAAAATCA-3' ) ( Folmer et al . , 1994 ) , which were respectively modified at their 5' ends with the linker sequences 5'-TCGTCGGCAGCGTC-3' and 5'-GTCTCGTGGGCTCGG-3' . PCRs were carried out in 20 µl volumes , containing 200 nM of the forward and reverse COI primers , 0 . 2 mM of each dNTP , 1 . 5 mM MgCl2 , 2 µg rabbit serum albumin , 0 . 5 U KAPA Plant 3G enzyme ( Kapa Biosystems , Wilmington , MA , USA ) , and 2 µl ( 6 ng ) DNA template . The PCR amplification protocol was 95 °C for 3 min; 35 cycles of 95 °C for 20 s , 52 °C for 15 s , and 72 °C for 30 s; and 1 min at 72 °C . Illumina sequencing adapters and sample-specific barcodes were added to the COI amplicons in a second round of PCR , carried out in 25 µl volumes containing the same reagents and concentrations as the first PCR , except for Illumina-tagged sequencing adaptors instead of COI primers , and 2 µl of the first PCR amplicon as template . The second-round PCR amplification protocol was 95 °C for 3 min; five cycles of 95 °C for 20 s , 54 °C for 15 s , and 72 °C for 30 s; and 1 min at 72 °C . The resulting libraries were purified and size-selected using a Pippin Prep system ( Sage Science , Beverly , MA , USA ) , to remove primer dimers and high molecular weight DNA , quantified , pooled , and sequenced on an Illumina MiSeq system with a 2 × 250 sequencing kit at the Australian Genome Research Facility Ltd . Demultiplexed forward and reverse DNA reads were merged and relabelled by sample using USEARCH ( Edgar , 2013 ) . Linker sequences and primers were trimmed from the merged sequences using cutadapt ( Martin , 2011 ) . The trimmed sequences were quality filtered to remove any with >1 maximum expected errors and dereplicated using VSEARCH ( Rognes et al . , 2016 ) . Non-singleton sequences ( i . e . those represented by at least two identical sequences ) were clustered into OTUs at a sequence identity threshold of 97% and simultaneously filtered for chimeras using the UPARSE algorithm in USEARCH ( Edgar , 2013 ) . OTU abundance was inferred by mapping the trimmed sequences back to the OTU centroid sequences at a sequence identity threshold of 97% . The OTUs were assigned a taxonomic identity using the RDP Naïve Bayesian classifier ( Wang et al . , 2007 ) in combination with an RDP-formatted animal mitochondrial COI sequence database ( Porter and Hajibabaei , 2018 ) , which includes bacterial , fungal , and protist COI sequences to enable the detection of non-metazoan OTUs . We excluded any OTUs that were not identified as belonging to an expected terrestrial invertebrate phylum . Data analyses were carried out using R version 3 . 5 . 1 ( R Development Core Team , 2016 ) and RStudio ( RStudio team , 2015 ) . Extraction blanks , negative and positive controls were examined for contamination . Tag jumping ( Schnell et al . , 2015 ) was accounted for by using a regression of contaminant abundances versus the maximum of total abundances in all other samples , after which the coefficient estimate for the 90th quantile regression was used to subtract that many sequences from the abundances of all OTUs ( Makiola et al . , 2019 ) . Comparisons of multivariate community composition and homogeneity between land-use categories were carried out for the overall terrestrial invertebrate dataset and the main terrestrial invertebrate phyla detected using the R package vegan v2 . 4–3 ( Oksanen et al . , 2017 ) . Non-metric MDS ordinations and PERMANOVA tests for community composition differences among land uses were based on the Jaccard distance metric and presence/absence data . Any samples with unusually low sequence abundance ( defined as less than 5% of the mean sequence abundance per sample for a given phylum ) were excluded from MDS ordinations . For the Mollusca-based MDS ordination , one further sample that resulted in an uninterpretable plot was excluded . To test for homogenisation effects of land use , multivariate homogeneity of sample dispersions was determined for each land-use category and compared between categories using the function betadisper in the R package vegan . Similarly , mean pairwise beta diversity and phylogenetic beta diversity ( UniFrac distances; Lozupone and Knight , 2005 ) were calculated for each land-use category , and compared between land-use categories using ANOVA and post-hoc Tukey HSD tests . Heatmaps of relative OTU abundance and distribution among sites were generated using phyloSeq ( McMurdie and Holmes , 2013 ) , for the 1000 terrestrial invertebrate OTUs with the highest proportional abundances across sites . Biodiversity estimates were calculated for each sample based on the overall terrestrial invertebrate communities , and for each of the main invertebrate groups detected , in such a way that all terrestrial invertebrate OTUs were represented: ( 1 ) the dominant insect orders detected ( Coleoptera , Diptera , Hymenoptera , Lepidoptera , and Hemiptera , each represented by >150 OTUs ) ; a further 18 insect orders represented by 1 to 36 OTUs were considered as a single pooled group ( ‘other insects’ ) ; ( 2 ) non-insect arthropod groups ( non-mite Arachnida , mites , Collembola , Malacostraca , myriapods ) ; and ( 3 ) non-arthropod phyla ( Annelida , Mollusca , Nematoda , Platyhelminthes , Rotifera , and Tardigrada ) . Because many OTUs were only found in a single site , biodiversity estimates were also calculated with these OTUs excluded , to check whether this affected the results . Species richness and effective species numbers ( exponential of Shannon entropy; Jost 2006 ) , were calculated for each invertebrate group using the R packages vegan v2 . 4–3 ( Oksanen et al . , 2017 ) and vegetarian v1 . 2 ( Charney , 2012 ) respectively . To calculate rarity , a weighting factor ( w ) was determined for each OTU as the reciprocal of its occurrence across all samples ( regardless of land use ) , so that w = 1 for OTUs that occur in only in a single sample , and w approaches zero for OTUs that occur in many samples . For each sample , values of w were then summed for all OTUs occurring in that sample . In other words , rarity represents the number of OTUs per sample adjusted for their occurrence across all samples ( Kier et al . , 2009; Kier and Barthlott , 2001 ) . To calculate phylogenetic diversity , phylogenetic rarity , and mean pairwise distance , OTU sequences were aligned using MAFFT v7 ( Katoh and Standley , 2013 ) , and phylogenetic trees constructed . Initially , phylogenetic trees were constructed separately for each phylum using both FastTree 2 ( Price et al . , 2010 ) and RAxML v8 ( Stamatakis , 2014 ) , and for the overall invertebrates using FastTree 2 ( construction of the overall invertebrates tree using RAxML failed ) . As phylum-level trees based on each method and the overall tree pruned to each phylum yielded similar results , the overall tree was used for estimation of phylogenetic biodiversity metrics per sample and taxonomic group . Phylogenetic diversity , in the form of total branch length per sample , and mean pairwise distance were calculated for each taxonomic group using the R package Picante ( Kembel et al . , 2010 ) . Phylogenetic rarity , in the form of the branch length unique to each sample ( based on occurrences across all samples ) , was calculated for each taxonomic group and sample according to Rosauer et al . ( 2009 ) using the R function phylo . endemism ( Niperess , 2010 ) . In addition , standardised effect size values were calculated for each of the phylogenetic metrics , by comparing observed values per site to a null distribution generated by 999 randomisations of the data using a regional null model ( Kembel et al . , 2010; Miller et al . , 2017 ) . ANOVA was used to test for significant differences among mean biodiversity values between land-uses , for overall invertebrate communities and for each of the taxonomic groups , based on each of the biodiversity metrics . We considered land use as an unordered categorical factor in these tests , because we had no a priori expectation about the relative intensity or impact of all five land uses . Any statistically significant ANOVA tests were followed with post hoc two-sided Tukey HSD tests to identify significant pairwise differences among land-use categories . Subsequently , based on our observed rank order of land uses , we derived a numeric rank of 1 to 5 in the order natural forest > planted forest > low producing grassland > high producing grassland > perennial cropland . We refer to this numeric rank as derived land-use rank ( DLUR in tables ) , to make clear that it is derived from our observed results , rather than on any a priori hypothesis as to which land uses might be considered more intense than others . We tested whether this provided the same conclusions as treating land use as a categorical factor for each metric and taxonomic group . We also included DLUR in a further ANOVA test for biodiversity and taxonomic group differences , to test whether different taxonomic groups showed the same patterns . We also investigated whether environmental covariates might explain biodiversity trends of overall soil invertebrate communities . To do so , we carried out ANOVA tests for effects of spatial variables ( latitude and altitude ) plus land-use category effects on overall biodiversity estimates for each metric . In addition , we generated a PCA based on spatial and soil chemistry variables . We then tested whether the most important PCA components , plus land-use category , had significant effects on overall biodiversity estimates for each metric . To investigate whether the biodiversity metrics differed in their sensitivity to land use , non-parametric bootstrapping stratified by taxonomic group with 999 replicates was used to estimate the proportion of variance attributable to land-use effects with 95% confidence intervals , across the set of taxonomic groups and metrics for which significant land-use differences were detected by ANOVA . These results were plotted as a histogram and compared between metrics using a non-parametric Kruskal-Wallis test . | Living within the Earth’s soil are millions of insects , worms and other invertebrates , which help keep the ground healthy and fertile . There is a growing concern that changing land-use habits , such as agriculture and urban development , are causing these populations of invertebrates to decline . However , to what extent different types of land use negatively impact soil invertebrates is not clear . Healthy habitats often have a greater variety of species . This biodiversity can be measured in a number of ways , ranging from counting the number of species , to more complex approaches that calculate a species’ role in an ecosystem or how close it is to extinction . Finding a way to sensitively measure the biodiversity of soil invertebrates could further researcher’s understanding of how different types of land use are affecting these communities . A new method known as DNA metabarcoding has made it easier to distinguish between different species and calculate the biodiversity of entire populations . Now , Dopheide et al . have used this technique to study invertebrate communities from 75 sites across New Zealand which have been impacted by different land-use habits . This revealed that the most reliable and consistent way to uncover how land use affects soil invertebrates was to measure the rarity of species ( i . e . the number of unique species present at each site ) . Dopheide et al . found that agriculture negatively affected soil invertebrates and that most types of invertebrates responded in a similar way . Horticulture – such as orchards and vineyards – had the most severe impact , with the lowest variety of species compared to grassland or forest . Other measurements of biodiversity , such as the number of different species , may underestimate the negative impact agriculture is having on invertebrate communities . The findings of Dopheide et al . highlight why developing strategies to preserve and restore these communities is so important . However , more work is needed to understand what specifically is causing biodiversity to decline and how this effect can be reversed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2020 | Rarity is a more reliable indicator of land-use impacts on soil invertebrate communities than other diversity metrics |
Metabolic network rewiring is the rerouting of metabolism through the use of alternate enzymes to adjust pathway flux and accomplish specific anabolic or catabolic objectives . Here , we report the first characterization of two parallel pathways for the breakdown of the short chain fatty acid propionate in Caenorhabditis elegans . Using genetic interaction mapping , gene co-expression analysis , pathway intermediate quantification and carbon tracing , we uncover a vitamin B12-independent propionate breakdown shunt that is transcriptionally activated on vitamin B12 deficient diets , or under genetic conditions mimicking the human diseases propionic- and methylmalonic acidemia , in which the canonical B12-dependent propionate breakdown pathway is blocked . Our study presents the first example of transcriptional vitamin-directed metabolic network rewiring to promote survival under vitamin deficiency . The ability to reroute propionate breakdown according to B12 availability may provide C . elegans with metabolic plasticity and thus a selective advantage on different diets in the wild .
Metabolic network rewiring to adjust metabolic flux in response to dietary or cellular cues can occur by transcriptional , post-transcriptional , or allosteric mechanisms ( Desvergne et al . , 2006 ) . For instance , genes encoding enzymes involved in the breakdown of galactose in the Leloir pathway are activated in yeast and other organisms upon a shift from glucose to galactose as a carbon source ( Fridovich-Keil , 2006 ) . As a second example , in both yeast and humans , glycolytic flux is temporarily re-routed through the pentose phosphate pathway to provide a first-line protection against oxidative stress ( Stincone et al . , 2014 ) . However , metabolic network rewiring to compensate for the absence of a vitamin or due to the toxic accumulation of a cellular metabolite has not yet been described . In both mammals and the nematode C . elegans , vitamin B12 is a critical cofactor in the canonical propionyl-CoA breakdown pathway ( Figure 1A and 1B ) . Propionyl-CoA is produced during the catabolism of odd chain fatty acids and branched chain amino acids , and is interconverted with the short chain fatty acid propionate derived from bacterial fermentation of dietary fibers in the intestine ( Kasubuchi et al . , 2015 ) . Many organisms , however , do not utilize vitamin B12 in the breakdown of propionate . For instance , Saccharomyces cerevisiae utilizes the methylcitrate cycle , whereas plants and Candida albicans use a β-oxidation-like pathway ( Halarnkar and Blomquist , 1989; Otzen et al . , 2014 ) ( diagrammed in Figure 1A ) . 10 . 7554/eLife . 17670 . 003Figure 1 . Propionate breakdown pathways in different organisms . ( A ) Vitamin B12-dependent species use a propionate carboxylation pathway to breakdown propionate . Other species use either the methylcitrate pathway or a β-oxidation-like pathway . ( B ) Diagram of canonical vitamin B12-dependent propionyl-CoA breakdown pathway indicating C . elegans and human enzymes and associated genetic diseases . MM – methylmalonyl , 3-HP – 3-hydroxypropionate , MSA – malonic semialdehyde , n . d . – not determined . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 003 Mutations in genes in the canonical vitamin B12-dependent propionate breakdown pathway cause propionic- and methylmalonic acidemias , diseases in which propionate and its derivatives accumulate to toxic levels ( La Marca et al . , 2007 ) . These diseases are diagnosed by elevated levels of specific metabolites such as 3-hydroxypropionate ( 3-HP ) , which is not normally detected at appreciable levels in healthy individuals ( Matsumoto and Kuhara , 1996 ) ( Figure 1B ) . Interestingly , 3-HP is an intermediate in the β-oxidation-like propionate breakdown pathway found in some vitamin B12-independent organisms ( Figure 1A ) . This observation suggests that propionic- and methylmalonic acidemia patients may break down propionate to some extent via an alternate oxidative route ( Ando et al . , 1972 ) . We previously identified numerous C . elegans metabolic genes that are transcriptionally repressed in response to vitamin B12 ( MacNeil et al . , 2013; Watson et al . , 2014 ) . This finding suggests that the C . elegans metabolic network is differentially wired under vitamin B12-deficient versus vitamin B12-replete nutritional conditions . However , the biological significance of the transcriptional rewiring by the vitamin B12/propionate axis remains unknown . Here , we find that C . elegans transcriptionally activates a β-oxidation-like propionate breakdown shunt under vitamin B12-deficient dietary conditions , or under genetic conditions mimicking propionic- or methylmalonic acidemia . This pathway is chemically similar to , but genetically distinct from the pathway found in Candida albicans . We detect elevated 3-HP in animals with a dysfunctional canonical propionate breakdown pathway , demonstrating that the C . elegans model faithfully recapitulates a metabolic phenotype of propionic- and methylmalonic acidemia . C . elegans is likely to encounter both vitamin B12-replete and B12-deficient diets in the wild because only a minority of bacterial species synthesize vitamin B12 ( Karasseva et al . , 1977; Sañudo-Wilhelmy et al . , 2014 ) . We find that activation of the C . elegans propionate shunt enables survival on vitamin B12-deficient diets . Altogether , our data suggest that metabolic network rewiring in response to vitamin B12 status enables the animal to thrive both when dietary vitamin B12 is low , and when this cofactor is in ample supply . This metabolic plasticity likely confers a selective advantage and evolutionary benefit .
Patients with propionic acidemia harbor loss of function mutations in both alleles of either PCCA or PCCB , which encode the two members of the propionyl-CoA carboxylase complex that catalyzes the first reaction in the canonical propionate breakdown pathway ( Deodato et al . , 2006 ) ( Figure 1A and B ) . These patients suffer from the toxic effects of propionate buildup , which manifest in several organ systems and lead to acute symptoms such as poor feeding , vomiting , hypotonia , lethargy , seizures , failure to thrive , intellectual disability , pancreatitis and cardiomyopathy ( Carrillo-Carrasco and Venditti , 2012 ) . Deletion of the C . elegans ortholog of PCCA , pcca-1 , slows development rate ( Watson et al . , 2014 ) and renders animal sensitive to propionate-induced toxicity: the LD50 of wild type animals is ~80mM propionate , while the LD50 of Δpcca-1 mutants is ~45 mM ( Figure 2A and B ) . As expected , vitamin B12 supplementation to wild type animals increases propionate tolerance on the low-B12 E . coli OP50 diet ( Watson et al . , 2014 ) , whereas it has no beneficial effect in Δpcca-1 animals ( Figure 2A and B ) . 10 . 7554/eLife . 17670 . 004Figure 2 . acdh-1 mutants are sensitive to propionate and synthetic lethal with pcca-1 mutants . ( A ) Dose-response curves showing that Δpcca-1 and Δacdh-1 mutants exhibit increased sensitivity to propionate compared to wild type animals . Three biological replicate experiments are shown , each with three technical replicates per data point with average and SEM plotted . ( B ) Average LD50 and standard deviation of data shown in ( A ) . Unpaired student’s T tests were used to calculate p-values . Black asterisks indicate significant difference compared to wild type , red asterisks indicate significant difference compared to wild type plus B12 . ( C ) Δacdh-1 mutants cannot survive on E . coli grown in vitamin B12 deficient media . ( D ) Δpcca-1 and Δacdh-1 are synthetically lethal because a cross between Δpcca-1 and Δacdh-1 mutants yielded no viable double homozygous mutants . pcca-1 +/+;acdh-1 +/- animals and pcca-1 -/-;acdh-1 +/- animals were grown on E . coli OP50 seeded plates containing 64nM vitamin B12 , and individual F1s were picked onto new plates , also containing 64nM vitamin B12 . The distribution of acdh-1 genotypes among the viable F1s picked from each P0 genotype is shown . ( E ) These genetic data support a role for acdh-1 parallel to the canonical propionate breakdown pathway . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 004 The C . elegans acyl-CoA dehydrogenase acdh-1 is differentially expressed depending on the vitamin B12/propionate axis: its transcript levels are very low when vitamin B12 is high , and increase several hundred fold in response to propionate accumulation ( Watson et al . , 2013; 2014 ) . A null mutation in acdh-1 also renders C . elegans sensitive to propionate: the LD50 in these animals is ~50 mM ( Figure 2A and B ) . However , in contrast to Δpcca-1 mutants , propionate sensitivity in Δacdh-1 mutants is completely rescued by vitamin B12 supplementation ( Figure 2A and B ) . Furthermore , Δacdh-1 mutants exhibit embryonic lethality on a very low-vitamin B12 diet ( E . coli OP50 grown on soy-peptone ) , and this phenotype can also be rescued by supplementing vitamin B12 ( Figure 2C ) . Acyl-CoA dehydrogenases catalyze the first step in β-oxidation of fatty acids ( Berg et al . , 2012 ) . Therefore , we hypothesized that acdh-1 may function in an alternate β-oxidation-like propionate breakdown pathway , hereafter referred to as the 'propionate shunt' , to enable survival of the animal on vitamin B12-deficient diets . Genes in parallel pathways often exhibit synthetic phenotypes ( Clark et al . , 1994; Costanzo et al . , 2010 ) . If acdh-1 does function in a propionate shunt , one would expect the propionate sensitivity to further increase when both acdh-1 and the canonical propionate-breakdown pathway are perturbed . To test this , we attempted to generate double null mutants that harbor deletions in both acdh-1 and in pcca-1 . However , a cross between Δpcca-1 and Δacdh-1 mutants yielded no viable double homozygous mutant offspring ( Figure 2D and Supplementary file 1 ) . This finding demonstrates that loss of function in both pcca-1 and acdh-1 results in synthetic lethality , and supports the hypothesis that acdh-1 functions in a parallel propionate breakdown pathway ( Figure 2E ) . To identify additional C . elegans genes that may function in a pathway with acdh-1 , we performed a synthetic genetic interaction screen using Δpcca-1 mutants and an RNAi library of 836 C . elegans metabolic genes ( Supplementary file 2 ) . RNAi of acdh-1 in the Δpcca-1 mutant resulted in complete lethality in the presence of 30 mM propionate ( Figure 3A ) . Therefore , we screened for knockdowns that led to non-viable offspring in the Δpcca-1 mutant supplemented with 30 mM propionate . Only three high-confidence hits were obtained from this screen: acdh-1 itself , ech-6 , an enoyl-CoA hydratase 6 , and F09F7 . 4 , which we named hach-1 , for hydroxyacyl-CoA hydrolase ( Figure 3B ) . Enoyl-CoA hydratases function in the second step of β-oxidation ( Berg et al . , 2012 ) and , therefore , ech-6 is an excellent candidate to catalyze the second reaction in the propionate shunt , directly downstream of acdh-1 . In the vitamin B12-independent yeast C . albicans , the Ehd3 enzyme converts 3-hydroxypropionyl-CoA into 3-HP and CoA in the third step of the β-oxidation-like propionate breakdown pathway ( Otzen et al . , 2014 ) . Ehd3 is the one-to-one ortholog of hach-1 , the third gene we identified , which we therefore placed downstream of ech-6 . Importantly , knockdown of either ech-6 or hach-1 resulted in similar phenotypes compared to loss of acdh-1: increased propionate sensitivity that was partially rescued by the addition of vitamin B12 ( Figure 3C ) . This observed phenocopying , along with the co-synthetic lethality with pcca-1 , supports the hypothesis that acdh-1 , ech-6 and hach-1 function together in a genetic pathway . 10 . 7554/eLife . 17670 . 005Figure 3 . A synthetic genetic interaction screen identifies candidate genes involved in the propionate shunt . ( A ) RNAi of acdh-1 is lethal in Δpcca-1 mutants supplemented with 30 mM propionate . ( B ) Synthetic genetic interaction screen of 836 metabolic genes in presence or absence of 30 mM added propionate , in wild type and Δpcca-1 mutant animals identifies three candidate genes , including acdh-1 . ( C ) Propionate toxicity dose response curve showing that the two candidate genes identified in the screen , ech-6 and hach-1 , phenocopy acdh-1 loss-of-function . ( D ) Genetic buffering of ech-6 and hach-1 RNAi phenotypes by loss of acdh-1 . Representative images of animals subjected to two generations of RNAi knockdown are shown . ( E ) Our data indicate that ech-6 and hach-1 function downstream of acdh-1 in the propionate breakdown shunt . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 005 The first reaction of the propionate shunt produces acrylyl-CoA , a highly toxic and reactive metabolite ( Hellwig et al . , 1993; Saillenfait et al . , 1999 ) . Since we predict that acrylyl-CoA is the substrate of ECH-6 , we hypothesized that ech-6 perturbation would result in a severe phenotype due to toxic buildup of acrylyl-CoA or its hydrolyzed derivative acrylate . Indeed , RNAi of ech-6 , and to a lesser extent hach-1 , strongly reduces C . elegans growth and viability ( Figure 3D ) . This phenotype was partially rescued by vitamin B12 supplementation , which facilitates propionate flux through the canonical pathway ( Figure 3D ) . This rescue depends on a functional canonical B12-dependent propionate breakdown pathway , as vitamin B12 supplementation had no beneficial effect when ech-6 or hach-1 was knocked down in Δpcca-1 mutants ( Figure 3D ) . We found that loss of acdh-1 largely suppressed the phenotypic effects of ech-6 and hach-1 knockdown , likely due to the lack of acrylyl-CoA production in the absence of acdh-1 ( Figure 3D ) . This genetic buffering supports the placement of ech-6 and hach-1 downstream of acdh-1 in the propionate shunt ( Figure 3E ) . The β-oxidation-like propionate breakdown pathway includes two additional reactions that convert the third metabolic intermediate 3-hydroxypropionate ( 3-HP ) to malonic semialdehyde ( MSA ) and finally to acetyl-CoA and CO2 ( Figure 1A ) . Importantly , the gene encoding the enzyme that converts 3-HP into MSA has not yet been identified in any metazoan . To identify enzymes that may catalyze the last two reactions in the C . elegans propionate shunt , we utilized WISP , a server for predicting tissue-specific functional networks based on the integration of a large compendium of diverse datasets ( http://wisp . princeton . edu , Yao et al . , in preparation; V . Yao , personal communication , June 2016 ) . The top predicted functional connections to acdh-1 , ech-6 and hach-1 included the metabolic genes Y38F1A . 6 and alh-8 ( Figure 4A ) , neither of which was tested in genetic interaction screen because they were not included in the ORFeome RNAi library ( Supplementary file 2 ) . These genes encode excellent candidate enzymes to catalyze the fourth and fifth reactions of the propionate shunt , respectively . Y38F1A . 6 is the ortholog of human ADHFE1 ( also known as HOT ) , a hydroxyacid-oxoacid transhydrogenase that has been found to metabolize β-hydroxybutyrate ( GHB ) , a structural analog of 3-HP ( Lyon et al . , 2009 ) . We will henceforth refer to Y38F1A . 6 as hphd-1 ( 3-hydroxypropionate-oxoacid transhydrogenase ) . ALH-8 is homologous to human ALDH6A1 , a decarboxylating dehydrogenase predicted to act on two structurally similar metabolites: methylmalonic semialdehyde from valine breakdown ( Marcadier et al . , 2013; Sass et al . , 2012 ) , and malonic semialdehyde ( Marcadier et al . , 2013 ) , the substrate in the fifth reaction of the propionate oxidation pathway ( Figure 1A ) . Additionally , hphd-1 and alh-8 are predicted to localize to the mitochondria along with acdh-1 , ech-6 , and hach-1 ( Yilmaz and Walhout , 2016; http://wormflux . umassmed . edu/ ) . 10 . 7554/eLife . 17670 . 006Figure 4 . Identifying additional putative propionate shunt genes . ( A ) hphd-1 and alh-8 ( blue ) are tightly connected to acdh-1 , ech-6 and hach-1 ( green ) in a C . elegans intestinal functional network and are candidates to catalyze the fourth and fifth reactions of the propionate shunt , respectively . ( B ) Structure of CRISPR/Cas9-generated alh-8 mutant . Diagram of the mutation generated by CRISPR/Cas9-mediated genome editing using an sgRNA ( red sequence ) targeting alh-8 . The alh-8 ( ww48 ) mutation consists of a 23 bp insertion and 399 bp deletion , and removes a part of the 5’UTR , the start codon , the first and second exons , and part of the third exon . Also shown is the Δhphd-1 ( ok3590 ) mutation . ( C ) Propionate toxicity dose response showing that Δhphd-1 and Δalh-8 mutants phenocopy acdh-1 , ech-6 and hach-1 perturbation . ( D ) Δhphd-1 and Δalh-8 mutants exhibit partial lethality on low-B12 conditions . Like the Δacdh-1 mutant phenotype , Δhphd-1 and Δalh-8 mutant phenotypes were rescued by 64nM B12 supplementation or by Comamonas aquatica DA1877 ( Coma . ) . The partial lethal phenotype of the Δhphd-1;Δpcca-1 double mutant was not rescued by B12 . ( E ) Combined deletion of hphd-1 and pcca-1 renders the animals more sensitive to propionate than mutation in either gene alone . Note that Δhphd-1 may not be a null allele . ( F ) The C . elegans propionate breakdown shunt pathway comprises five genes: acdh-1 , ech-6 , hach-1 , hphd-1 and alh-8 . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 00610 . 7554/eLife . 17670 . 007Figure 4—figure supplement 1 . Δhphd-1 and Δalh-8 mutants exhibit increased sensitivity to propionate compared to wild type animals . Three biological replicate corves are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 007 We obtained an hphd-1 deletion mutant from the C . elegans genetics center , and generated an alh-8 deletion mutant by CRISPR/Cas9-mediated genome editing ( Kim et al . , 2014 ) ( Figure 4B ) . Both mutants phenocopied acdh-1 , ech-6 and hach-1 loss of functions: they exhibited decreased propionate tolerance that was at least partially rescued by vitamin B12 supplementation ( Figure 4C , Figure 4—figure supplement 1 ) . Both Δhphd-1 and Δalh-8 mutants displayed partial lethality on low-B12 diets , which was rescued by vitamin B12 supplementation ( Figure 4D ) . This rescue was dependent on a functional canonical propionate breakdown pathway , as vitamin B12 failed to rescue the partial larval lethality exhibited by the double Δhphd-1;Δpcca-1 mutant ( Figure 4D ) . This result indicates that activation of the propionate shunt is required to sustain viability . Δhphd-1;Δpcca-1 mutants also exhibited increased propionate sensitivity compared to either single mutant , indicating a conditional genetic interaction between hphd-1 and pcca-1 , but not complete lethality like the Δacdh-1;Δpcca-1 double mutant ( Figure 4E ) . This may be due to Δhphd-1 not being completely null , or it is possible that an intact half-pathway is sufficient for at least partial survival . Altogether , these observations support the placement of hphd-1 and alh-8 in the same pathway as acdh-1 ( Figure 4F ) . We previously found reduced transcript levels of each of the five genes encoding propionate shunt enzymes in response to the vitamin B12-synthesizing bacteria Comamonas aquatica ( MacNeil et al . , 2013 ) . However , under genetic conditions mimicking propionic acidemia ( i . e . , when the animals cannot use vitamin B12 to breakdown propionate ) , vitamin B12 fails to reduce acdh-1 expression ( Watson et al . , 2014 ) . This observation led to the hypothesis that , rather than directly sensing vitamin B12 levels , the C . elegans gene regulatory network responds to elevated levels of propionate or propionyl-CoA ( or a derivative thereof ) to activate the shunt . Indeed , we found that reduced expression of all five shunt genes by vitamin B12 is reversed by supplementation of excess propionate ( Figure 5A ) . 10 . 7554/eLife . 17670 . 008Figure 5 . Transcriptional activation of the propionate shunt . ( A ) The expression of all five propionate shunt genes is repressed by vitamin B12 and activated by propionate . Condition matrices are shown for each shunt gene . Expression is normalized to levels in the control condition ( no vitamin B12 , no propionate ) . ( B ) Vitamin B12 reduces GFP levels in Pacdh-1::GFP transgenic animals , but not in those carrying a deletion in the canonical propionate pathway gene mce-1 or in the propionate shunt gene hphd-1 . ( C ) Quantification of GFP levels from part ( B ) DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 008 Activation of acdh-1 in response to propionate buildup occurs through its 1 . 5 kb promoter , indicating that it is governed by transcriptional mechanisms ( Watson et al . , 2014 ) . Not only is the acdh-1 promoter activated by propionate and by canonical pathway perturbations , it is also activated by perturbation of the propionate shunt genes ech-6 , hach-1 and acdh-1 itself ( Watson et al . , 2013 ) . To determine whether a deletion in hphd-1 also activates the acdh-1 promoter , we crossed the Δhphd-1 mutant to a transgenic strain expressing the green fluorescent protein ( GFP ) under the control of the acdh-1 promoter . Loss of hphd-1 did in fact lead to greater acdh-1 promoter activity , providing additional evidence that hphd-1 functions in propionate breakdown ( Figure 5B and C ) . 3-HP is a unique metabolic intermediate produced by the propionate oxidation pathway: to our knowledge neither KEGG , nor any other metabolic database , lists this metabolite being produced by any other metabolic pathway in metazoans , though it can be produced through several different pathways in microorganisms . The fourth reaction in the propionate shunt involves the conversion of 3-HP into MSA ( Figure 6A ) . Annotated with the enzyme commission number EC 1 . 1 . 1 . 59 , the gene encoding this enzyme has , to our knowledge , not yet been identified in any metazoan . Our co-expression network analysis and subsequent genetic investigation identified HPHD-1 as a candidate for this enzyme . If true , we would predict that 3-HP accumulates in the Δhphd-1 mutant . 10 . 7554/eLife . 17670 . 009Figure 6 . 3-Hydroxypropionate is a substrate for HPHD-1 . ( A ) Conversion of 3-hydroxypropionate ( 3-HP ) into malonicsemialdehyde ( MSA ) . ( B ) 3-HP mass spectrometry chromatogram for wild type , Δpcca-1 and Δhphd-1 animals . 3-HP was not detected in E . coli OP50 with or without supplemented propionate . ( C ) Propionyl-CoA chromatograms from E . coli and C . elegans samples . Propionyl-CoA quantifications are as follows: 1 . 86 , 0 . 20 , 0 . 14 , and 0 . 13 nmol/mg protein for E coli + PA , wild type C . elegans , Δpcca-1 and Δhphd-1 mutants , respectively . For E . coli -PA , propionyl-CoA was detectable but not quantifiable in our assay . ( D ) Average 3-HP quantities normalized to total protein levels from three biological replicates , +/- SEM . Animals were grown on E . coli OP50 . ( E ) 13C-labeled propionate fed to Δhphd-1 mutant animals for 2 hr yielded 13C-labeled 3-HP , demonstrating that C . elegans oxidizes propionate to 3-HP . Shown are SRM ( MS2 ) chromatograms specific for 3-HP . The peak corresponding to the natural 13C isotope distribution ( ~ 1 . 1% of 12C signal ) is illustrated for comparison in t = 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 009 Using liquid chromatography/selective reaction monitoring mass spectrometry ( LC-SRM ) we detected 3-HP in C . elegans , but not in its E . coli diet ( Figure 6B ) . We did detect ample propionyl-CoA in E . coli supplemented with propionate , so the lack of 3-HP in E . coli was not due to lack of pathway substrate ( Figure 6C ) . Therefore , we conclude that the 3-HP detected is derived from C . elegans and not from its bacterial diet . We observed a >4-fold increase in 3-HP levels in Δpcca-1 mutants , which mirrors elevated 3-HP levels observed in human patients with propionic acidemia caused by PCCA or PCCB mutations ( Carrillo-Carrasco and Venditti , 2012 ) , and confirms that the propionate shunt is active when the canonical pathway is perturbed ( Figure 6B and D ) . Importantly , we detected a >200-fold increase in 3-HP levels in Δhphd-1 mutants , which supports our prediction that HPHD-1 metabolizes 3-HP under low B12 conditions ( Figure 6B and D ) . To verify that 3-HP is indeed derived from propionate , we performed carbon tracing by feeding Δhphd-1 mutant animals E . coli OP50 supplemented with 13C-propionate . We detected the formation of 13C-3-HP after 2 hr , demonstrating that C . elegans indeed converts propionate to 3-HP ( Figure 6E ) . 3-HP is a specific diagnostic marker of propionic- and methylmalonic acidemias , as it is uniquely elevated in these diseases . This suggests that an alternative propionate breakdown pathway may be operational in humans as well , at least in patients with impaired canonical propionate breakdown . Interestingly , the closest human homologs of the C . elegans shunt enzymes are known to catalyze structurally similar reactions in other metabolic pathways , including the breakdown of the branched chain amino acids isoleucine and valine ( Figure 7A-Supplementary file 3 ) . Recent metabolomics data in patients with mutations in ECHS1 ( the homolog of C . elegans ech-6 ) and HIBCH ( the homolog of C . elegans hach-1 ) revealed elevated levels of acrylyl-CoA , a propionate shunt intermediate , in addition to the expected valine breakdown intermediates ( Peters et al . , 2014; 2015 ) ( Figure 7A ) . Further , patients with mutations in ALDH6A1 ( the homolog of C . elegans alh-8 ) exhibit elevated levels of 3-HP as well as elevated levels of valine breakdown intermediates ( Marcadier et al . , 2013 ) . Taken together , these observations suggest that the closest human homologs of several C . elegans propionate shunt genes may have conserved roles in propionate breakdown in humans in addition to their known roles in other pathways . 10 . 7554/eLife . 17670 . 010Figure 7 . Comparison between putative human and C . elegans propionate shunts . ( A ) Comparison between C . elegans propionate shunt genes ( red ) and candidate human shunt genes ( green , gray ) . Green text indicates higher confidence annotations based on patient mutations and metabolomics , or in the case of ADHFE1 one-to-one orthology of unique enzymes in both genomes . 3-HP is marked in magenta to indicate that it is a biomarker for impaired flux in the canonical , vitamin B12-dependent propionate breakdown pathway , such as occurs in patients with propionic or methylmalonic acidemia . ( B ) C . elegans propionate shunt genes and orthologs in mouse and humans are strongly co-expressed as a group compared to 10 , 000 random permutations of five genes from either the whole genome , a subset of only metabolic genes , or a subset of related metabolic genes from connected pathways , including BCAA breakdown and the TCA cycle . The expression data used for this analysis was compiled and weighted using the SEEK and modSEEK databases . Distributions of co-expression scores are shown for each set of randomizations , and vertical dashed lines indicate actual weighted co-expression score for propionate shunt genes and orthologs in human , mouse and C . elegans . ( C ) The expression of several candidate human genes is activated in response to propionate in HepG2 liver carcinoma cells . qRT-PCR experiment showing the average of four replicate experiments , each containing three technical replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 17670 . 010 We found that the closest human homologs of all five propionate shunt genes , ACADSB , ECHS1 , HIBCH , ADHFE1 and ALDH6A1 are significantly co-expressed in both mouse and human across a large compendium of transcriptomic data in the SEEK database ( Figure 7B ) ( Wang et al . , 2015 ) . This suggests that , like their C . elegans counterparts , these human genes may be co-regulated at the transcriptional level . Remarkably , the expression of ADHFE1 , HIBCH , and to a lesser extent ECHS1 and ALDH6A1 , is upregulated in response to propionate in HepG2 cells ( Figure 7C ) . These data suggest that the regulated response of these genes to propionate at the gene expression level is , at least to some extent , conserved between C . elegans and humans .
To our knowledge , our study presents the first example of transcriptional metabolic network rewiring in which the catabolic route of a cellular metabolite is dictated by the presence or absence of a dietary vitamin , in this case vitamin B12 . Other vitamins that have known roles in regulating gene expression include vitamins A and D , and this regulation is important for development , growth and homeostasis . However , vitamins A and D do not function directly in the metabolic network as cofactors of metabolic enzymes , and instead function more like hormones; in fact vitamin D can be synthesized endogenously . Vitamin D , via the vitamin D receptor ( VDR ) , regulates mineral uptake ( Carlberg and Seuter , 2009; Haussler et al . , 2008 ) , while vitamin A , via the retinoic acid receptor ( RAR ) , regulates developmental programs as well as the enzymes that interconvert the regulatory version of retinoic acid and the trans-retinal version required by rhodopsin for light-sensing ( D'Aniello and Waxman , 2015; di Masi et al . , 2015 ) . Less is known about potential regulatory roles of other vitamins , including those that function as true enzyme cofactors in the metabolic network . However , gene expression profiling in mammalian cells has revealed transcript-level responses to vitamins B1 ( thiamine ) ( Fraser et al . , 2012; Liu et al . , 2004 ) ( Tanaka et al . , 2007 ) , B2 ( riboflavin ) ( Nakano et al . , 2011 ) , B3 ( nicotinamide/niacin ) ( Choi et al . , 2011; Couturier et al . , 2014; Giammona et al . , 2006 ) , B6 ( pyridoxal 5′ phosphate , PLP ) ( Toya et al . , 2012; Zhang et al . , 2014 ) , B9 ( folic acid ) ( Barua et al . , 2014; Champier et al . , 2012; Lin et al . , 2011 ) , C ( ascorbic acid ) ( Canali et al . , 2014; Jun et al . , 2011; Takahashi et al . , 2014 ) , and E ( tocopherol/tocotrienols ) ( Landrier et al . , 2010; Makpol et al . , 2013; Mustacich et al . , 2009 ) . The mechanisms behind , and consequences of , these observed vitamin-induced gene expression changes have yet to be elucidated . Our study indicates that , in C . elegans , low vitamin B12 leads to accumulation of the short chain fatty acid propionate due to reduced flux through the B12-dependent propionate breakdown pathway . It remains to be determined whether vitamin B12 is directly sensed similar to vitamins A and D , or whether propionate or perhaps one of its derivatives is the sole proxy regulator ( see below ) . In mammals , propionate and other short chain fatty acids produced by the gut microbiota provide numerous benefits to the host , not only as nutrient sources that fuel colonocytes , but also potentially to inhibit cancer cell proliferation , induce cancer cell apoptosis ( Emenaker et al . , 2001; Hinnebusch et al . , 2002 ) , and reduce inflammation ( Louis et al . , 2014 ) . However , excess propionate accumulation , which occurs in patients with propionic- or methylmalonic acidemia , is toxic . It is possible that the metabolic network rewiring that we observe in C . elegans in response to the vitamin B12/propionate axis has evolved not only to optimize energy yield from propionate depending on the presence or absence of B12 , but also to prevent toxic propionate buildup . This represents a novel example of built-in metabolic network flexibility to mitigate the toxic accumulation of an endogenous metabolite . What is the biological function of metabolic network rewiring by the vitamin B12/propionate axis ? A loss of propionate breakdown capability , in double Δpcca-1/Δacdh-1 mutants , or in Δacdh-1 mutants on a very low vitamin B12 diet ( E . coli OP50 grown on soy peptone ) is not compatible with viability ( this study ) . Vitamin B12 is exclusively produced by bacteria , and studies of microbial communities have found that only 20–30% of community members synthesize vitamin B12 ( Sañudo-Wilhelmy et al . , 2014 ) . C . elegans is found all over the world in temperate climates and is likely to feed on a variety of bacterial species ( Félix and Duveau , 2012 ) . Our data suggest that the ability to catabolize propionate whether or not vitamin B12 is provided by the diet may provide the animal with the metabolic flexibility to survive in different dietary conditions , thus providing a selective advantage and evolutionary benefit . On vitamin B12-replete bacterial diets , such as Comamonas aquatica , expression of the propionate shunt is greatly reduced , indicating that the canonical vitamin B12-dependent propionate breakdown pathway is preferred . We speculate that this may be because the canonical pathway is more efficient at metabolizing propionyl-CoA than the propionate shunt due to the high redox potential between propionyl-CoA and acrylyl-CoA , the first shunt pathway intermediate ( Sato et al . , 1999 ) . Other advantages of the canonical pathway over the shunt include the use of fewer enzymes , and the lack of production of highly toxic intermediates ( e . g . , acrylyl-CoA ) . While all of the five genes identified in this study as propionate shunt members lead to similar phenotypes when mutated or knocked down ( sensitivity to propionate-induced toxicity and at least partial lethality on B12-deficient diets ) , the severity of these phenotypes differs depending on the gene disruption . This could potentially be explained by different levels of reactivity ( and therefore toxicity ) among the intermediates in the pathway , which may accumulate to different levels depending on which enzyme is disrupted . For instance , ech-6 knockdown results in a very severe phenotype , likely due to the accumulation of its substrate acrylyl-CoA , which is highly toxic . Simultaneously , the buildup of substrates containing CoA could lead to widespread metabolic impairment due to CoA sequestration ( Mitchell et al . , 2008 ) . Additionally , several enzymes that function in the shunt may also have roles in isoleucine and valine breakdown , and may therefore be pleiotropic . It should be mentioned that the Δhphd-1 and Δalh-8 deletion mutants used in this study may not be null since they are both partial locus deletions , and this may explain the less severe phenotypes observed for these mutants compared with the null Δacdh-1 mutant . It is also possible that C . elegans can ( somewhat ) tolerate B12 deficiency with only a partially intact shunt consisting of the first three reactions , or that there are other unidentified ( partially ) redundant enzymes that can compensate for loss of hphd-1 and alh-8 . How does C . elegans rewire its metabolic network in response to the vitamin B12/propionate axis ? Each of the five genes that encode enzymes of the propionate shunt is repressed by vitamin B12 and activated by propionate . By using a GFP reporter driven by 1 . 5 kb of acdh-1 promoter DNA we previously found that GFP levels are high when animals are fed bacterial diets low in vitamin B12 , whereas GFP is greatly reduced when the animals are fed bacteria that synthesize high levels of vitamin B12 , or upon direct supplementation of vitamin B12 ( MacNeil et al . , 2013; Watson et al . , 2013; Watson et al . , 2014 ) . This demonstrated that the response of acdh-1 occurs at the level of transcriptional regulation . Vitamin B12 is not sufficient to repress the acdh-1 promoter when enzymes within the B12-dependent propionate breakdown pathway are genetically perturbed , or when excess propionate is added to the media ( Watson et al . , 2014 ) . Therefore , we propose that the C . elegans gene regulatory network activates acdh-1 expression in response to the buildup of propionate , which occurs when this vitamin is in low supply . We have previously identified more than 50 C . elegans transcription factors that regulate acdh-1 ( MacNeil et al . , 2015; Watson et al . , 2013 ) , including the nuclear hormone receptor NHR-10 that directly binds its promoter ( Arda et al . , 2010 ) . Future studies will reveal which of these transcription factors mediate the response to propionate and/or vitamin B12 . Nuclear hormone receptors utilize binding to small molecule ligands to regulate gene expression . For instance , VDR directly interacts with vitamin D , and RAR binds biologically active forms of vitamin A ( Carlberg , 1999 ) . While humans have 48 nuclear hormone receptors , C . elegans has more than 270 ( Reece-Hoyes et al . , 2005 ) . It is tempting to speculate that one or more C . elegans nuclear hormone receptors directly respond to propionate , or its CoA derivative propionyl-CoA . Several lines of evidence indicate that humans also utilize a propionate detox shunt , at least to some extent . First , the detection of the unique shunt intermediate 3-HP is used as a diagnostic marker for propionic- and methylmalonic acidemia in newborns . Since 3-HP is not predicted as an intermediate in any other metazoan pathway , this finding suggests that a propionate shunt may also be functional in humans . Second , while the human homologs of several C . elegans shunt enzymes have well-established functions in the breakdown of branched chain amino acids , their genetic perturbation also results in the accumulation of propionate shunt intermediates . For instance , recent metabolomic analyses of patients with mutations ECHS1 and HIBCH revealed not only elevated upstream intermediates from valine catabolism , but also acrylyl-CoA , a unique intermediate from propionate oxidation ( Peters et al . , 2014 ) . Interestingly , global metabolomics has identified 3-HP in healthy individuals ( Bouatra et al . , 2013; Guneral and Bachmann , 1994 ) . This finding suggests that the propionate shunt may also be active to some degree when the canonical pathway is functional and thus may be part of central metabolism in humans . It is important to note that , in spite of evidence supporting alternative propionate breakdown mechanisms in humans , patients with an impaired canonical propionate breakdown pathway are very sick and must strictly adhere to diets low in the amino acids that are broken down to propionyl-CoA . This indicates that , in humans , alternative propionate catabolism routes are not sufficient to maintain propionate levels below the toxic threshold . Perhaps the most interesting gene we identified as a participant in the propionate shunt is hphd-1 , which is the one-to-one ortholog of human ADHFE1 . hphd-1 is the first metazoan gene to be associated with the reaction catalyzed by 3-hydroypropionate dehydrogenase ( EC 1 . 1 . 1 . 59 ) , which converts 3-HP to MSA . ADHFE1 is thought to metabolize a structural analog of 3-HP , GHB , which is commonly known as a recreational drug . However ADHFE1 is not assigned by KEGG or BRENDA enzyme databases to any endogenous metabolic pathway . ADHFE1 is unique in that , unlike most dehydrogenases that transfer electrons from their substrates to NAD or FAD , it transfers electrons to the TCA cycle intermediate α-ketoglutarate , thereby producing ( D ) -2-hydroxyglutarate ( Struys et al . , 2005 ) , a putative oncometabolite ( Dang et al . , 2009; Kranendijk et al . , 2010 ) . Interestingly , other than neomorphic isocitrate dehydrogenase ( IDH ) enzyme mutants found in many cancers , ADHFE1 is the only known enzyme to naturally produce ( D ) -2-hydroxyglutarate ( Struys et al . , 2005 ) . Currently , no patients have been identified with mutations in ADHFE1 so there is no metabolomics data available to determine which metabolites build up in humans lacking ADHFE1 enzyme function . However , our C . elegans mass spectrometry data in mutants lacking hphd-1 revealed greatly elevated 3-HP levels ( Figure 5 ) , and , since hphd-1 and ADHFE1 are clear one-to-one orthologs , ADHFE1 is a good candidate to function in propionate oxidation in humans directly downstream of 3-HP .
N2 ( Bristol ) was used as the wild type strain , and animals were maintained as described ( Brenner , 1974 ) . pcca-1 ( ok2282 ) , acdh-1 ( ok1489 ) , mce-1 ( ok243 ) and hphd-1 ( ok3580 ) strains were provided by the C . elegans Gene Knockout Consortium and were backcrossed 3 times against N2 prior to assays . The hphd-1 ( ok3580 ) allele removes only part of the C-terminus of the protein and may not be a complete loss-of-function mutation . For a diagram of deletion mutant loci , see Figure 4B and for a full list of genotyping primers refer to Supplementary file 4 . Approximately 100 synchronized L1s ( hatched overnight , 20 hr post-bleach ) were added to E . coli OP50-seeded 35 mM NGM ( bactopeptone ) agar plates containing various concentrations of pH-neutralized propionic acid . Each dose tested included four technical replicates . After 72 hr , un-arrested survivors ( animals that had developed past L1 stage ) were counted . Dose response curves were fit to the raw data using the following equation:Y=Bottom+ ( Top−Bottom ) / ( 1+10∧ ( ( LogLD50−X ) ∗HillSlope ) ) The dose required to kill 50% of the population ( LD50 ) was calculated according to the fitted dose response curves . Toxicity assays were performed in biological triplicate , and the average LD50’s are plotted +/- SEM . To obtain enough viable Δacdh-1 mutant animals for these assays , 64nM B12 was supplemented to animals two generations prior to assay . In the larval lethality quantifications , animals were fed for one generation on E . coli OP50 supplemented with 64nM B12 , and then grown for one generation on E . coli OP50 , E . coli OP50 +B12 or Comamonas aq . DA1877 . Offspring were harvested and live and dead L1s and embryos were quantified following a 24 hr arrest . A list of metabolic enzyme domain-containing genes was manually curated based on KEGG and WormBase databases , and available metabolic gene-targeting clones in the ORFeome RNAi library were re-arrayed in 96 well format . See Supplementary file 2 for the gene list . RNAi experiments were performed as follows: 24-well NGM ( bactopeptone ) agar plates containing 1 mM IPTG and 1 mM Ampicillin were seeded with one dsRNA-expressing E . coli HT115 clone per well the night before use . A separate set of plates containing 30 mM pH-neutralized propionate was also prepared and seeded with the same HT115 clones . The HT115 cultures were prepared by seeding 1 mL fresh LB + Ampicillin with 50 μL overnight culture , growing at 37oC for 6 hr , then centrifuged and resuspended in 150 μL LB + Ampicillin . 30 μL of this resuspended culture was placed in the center of NGM wells in the 24-well plates . Wild type and Δpcca-1 mutants were cultivated on E . coli OP50 , and eggs were harvested by bleaching , and hatched overnight in M9 media ( 20 hr ) , and synchronized L1s were added to prepared plates . Animals were observed after three days to observe effects in the 1st generation , and after six days to observe lethality in the 2nd generation . Animals were synchronized by L1 arrest and grown on plates containing bactopeptone and various doses of B12 and/or propionate , seeded with E . coli OP50 . Approximately 1500 adult animals were harvested for each condition , in biological duplicate . Animals were washed in M9 buffer , and total RNA was isolated using Trizol ( Invitrogen ) followed by DNAseI treatment and cleanup using Qiagen RNeasy columns . cDNA was prepared from RNA using oligo-dT and Mu-MLV enzyme ( NEB ) . Primer sequences for quantitative RT-PCR ( qRT-PCR ) were generated using the GETprime database ( Gubelmann et al . , 2011 ) and are listed in Supplementary file 4 . qPCR was performed in technical triplicate per gene per condition using the Applied Biosystems StepOnePlus Real-Time PCR system and Fast Sybr Green Master Mix ( Invitrogen ) . Relative transcript abundance was determined using the ΔΔCt method ( Schmittgen and Livak , 2008 ) , and normalized to averaged ama-1 and act-1 mRNA expression levels . Synchronized animals were cultivated on 15 cm NGM plates seeded with E . coli OP50 , and bleached after 3 days . Bleached eggs were washed three times in M9 , then allowed to hatch for 20 hr . 1 million synchronized L1s were added to 400 mL liquid NGM in a 2L Erlenmeyer flask , containing concentrated E . coli OP50 bacteria from 500 mL overnight LB culture , and total volume was adjusted to 450 mL with M9 . Some flasks contained 20 mM pH-neutralized propionic acid . Flasks were kept at 20°C , shaking gently at 100 rpm . Each day concentrated bacteria were added to the flasks to feed the worms . Adult animals were collected ( after 3 days of development for N2 , and four days for the mutants ) and washed 2 times in M9 in sterile Imhoff settling cones . The final pellet was flash-frozen and stored at −80°C until extraction . Cell extracts were obtained by re-suspending the frozen C . elegans or bacterial pellets in 4 mL of 5% trichloroacetic acid ( TCA ) . Cell suspensions were homogenized in a Polytron PT 1300 for 2 min at 20 , 000 rpm and neutralized with 1 mL of 2M of potassium monoacid phosphate . The samples were centrifuged at high speed for 10 min at 4°C and immediately injected for propionyl-CoA determination . The pellets were stored for protein quantification using the bicinchoninic acid method ( Thermo Scientific Pierce Protein BCA Kit ) . For 3-HP measurements , the cell extract was desiccated in a Speedvac and the resulting pellet was resuspended in the same volume of methanol . Then , the samples were centrifuged at high speed for 10 min at 4°C . For E . coli metabolite extraction , the cell extracts were obtained as mentioned above except that bacteria cells were disrupted by sonication for 2 min using intervals of 15 s of sonication followed by 15 s for cooling . The quantification of metabolites was performed using a LC-MS/MS system consisting of an ultra-high pressure LC system ( Agilent 1290 ) online coupled to a Triple Quadrupole mass spectrometer equipped with an electrospray ionization source ( Agilent 6460 ) . Propionyl-CoA was separated using a column Zorbax Eclipse Plus C18 Rapid Resolution HD 2 . 1 × 50 mm 1 . 8 Micron ( Agilent ) at 30°C . The mobile phase was composed of Buffer A: 10 mM tributylamine , 15 mM acetic acid and 5% methanol; and Buffer B: 100% methanol . The flow rate was 0 . 5 mL/min and the gradient method consisted of: 0–0 . 25 min , 2 . 5% B; 0 . 25–0 . 5 min , 2 . 5–30% B; 0 . 5–5 min , 30–70%B; 5–5 . 25 min , 70–100% B; 5 . 25–6 . 25 min , 100% B; 6 . 25–7 min , 2 . 5% B; 7–8 min 2 . 5% B . The 3-hydroxypropanoic acid was separated using a column Zorbax Eclipse Plus C18 Rapid Resolution HD 2 . 1 × 100 mm 1 . 8 Micron ( Agilent ) at 30 C . The mobile phase was composed of Buffer A: 0 . 1% formic acid in water; and Buffer B: 100% methanol . The flow rate was 0 . 35 mL/min and the gradient consisted of: 0–3 min , 5% B; 3–4 min , 5–70% B; 4–5 . 25 min , 70%B; 5 . 25–5 . 5 min , 70–100% B; 5 . 5–6 . 5 min , 100% B; 6 . 5–7 min , 5% B; 7–7 . 5 min 5% B . Q1/Q3 ( MRM ) transitions , ion source and collision energy settings were optimized according to the metabolites and were: 91->73 , 25 eV; 824 . 2->317 . 1 , 25 eV; and 92->74 , 25 eV ( in positive mode ) , for 3-HP , propionyl-CoA and 13C labeled 3-HP , respectively . Ion source settings were as follows: gas temperature , 300 C: gas flow , 8L/min; nebulizer 50 psi ( Nitrogen ) ; sheath gas temperature , 200 C: sheath gas flow , 11 L/min , capillary 3500 V and nozzle voltage , 500 V . We confirmed the peak identity of 3-HP by matching retention time , mass/charge ratio and MS/MS fragmentation spectra to a chemically synthesized 3-HP standard . The alh-8 mutant was generated by dual sgRNA directed-deletion ( Chen et al . , 2014 ) . We used a co-CRISPR strategy , which includes unc-22 as a CRISPR marker to enhance detection of genome-editing events ( Kim et al . , 2014 ) . The target sequences were manually derived to conform to the sequence N19NGG near the 5’ end of alh-8 . Two target sequences were chosen: CCGCCCATCTCTTGTGATTTTC and CTGTGCGACAGTTGTCGTATGG . We designed forward and reverse oligos containing the N19 sequence and ends of BsaI recognition sites . The forward and reverse oligos were annealed and ligated to BsaI-digested pRB1017 vector ( Arribere et al . , 2014 ) . The alh-8 sgRNA plasmids were prepared using a PureLink Quick Plasmid Miniprep Kit ( Invitrogen ) . The other co-injected DNA vectors were purified using a Qiagen midiprep kit . The DNA mixture used in microinjection contained Peft-3::Cas9 vector , pRF4::rol-6 ( su1006 ) , unc-22 sgRNA vector ( all gifts from the Mello lab ) and two alh-8 sgRNA vectors , each with a concentration of 40 ng/µl . Approximately 20 young adult hermaphrodite worms were injected . After recovering from injection , each worm was placed onto an individual E . coli OP50 plate . After 2–3 days , the F1 rollers ( dominant phenotype indicating presence of the pRF4::rol-6 ( su1006 ) construct ) were picked onto new plates . F1s with twitcher F2s were genotyped by PCR for mutations in alh-8 . The PCR primers are outside of the sgRNA-targeted region . Forward primer: TTCAATGTTCGCGTGTATTTTG; Reverse primer: TCAGCGAGCTTCTTCATGT . The amplicons with smaller size than wild type amplicons were reconfirmed by sequencing . Forward primer: ATTCGAAACGTGATCAGTAATG; Reverse primer: CTCTCTTGATCAAGGCTTGA . A mutant animal with a ~400 bp deletion ( 23 bp indel and 399 bp deletion ) was chosen for further analysis , and was outcrossed with N2 three times before use in phenotypic assays . The WISP tissue-specific functional networks were built using a semi-supervised regularized Bayesian approach that integrated 56 , 179 expression- and interaction-based measurements across 174 genome-level datasets [http://wisp . princeton . edu; Yao et al . , in preparation , V . Yao , personal communication , June 2016] . Using the intestine network , we found genes that were tightly connected to acdh-1 , ech-6 and hach-1 even after adjusting for average network connectivity , which improved the specificity of the retrieved genes to our seed genes . The SEEK ( seek . princeton . edu ) and modSEEK search engines have compiled thousands of publicly available expression datasets and given a gene set query , weights datasets by relevance ( using a cross-validation method ) and calculates a weighted coexpression score for every other gene in the genome to the gene set according to the dataset relevance weights . For every gene set , we can thus use the leave-one-out approach to calculate average weighted coexpression scores for the entire set . To construct null distributions for these 5-gene queries , we compared with average weighted coexpression scores based on random sets ( n = 10 , 000 ) from ( 1 ) all genes with sufficient data; ( 2 ) all genes that show enzymatic activity ( as indicated by known annotation to the catalytic activity GO term ) ; ( 3 ) all genes in similar gene families / related pathways ( members of acdh , ech , alh families , as well as those known to participate in branched chain amino acid breakdown , the canonical propionyl-CoA breakdown pathway , and the TCA cycle ) . HepG2 cells were seeded in 6 well plates at 0 . 6 × 106 cells/ml in 3 ml DMEM plus 1% FBS with or without 50 mM propionic acid . Cells were incubated for 48 hr at 37°C , 5% CO2 and 65% relative humidity . Cells were washed two times in PBS and before proceeding to Trizol lysis for RNA extraction . qRT-PCR was performed as described above , using actin and GAPDH to normalize expression levels . | Inborn errors of metabolism are human genetic diseases that cause developmental delays and are usually fatal . Propionic acidemia is an inborn error of metabolism where propionate , a byproduct created during the breakdown of fat and proteins , cannot be broken down efficiently . As a result , propionate builds up to toxic levels inside cells . Most animals , including humans , use a particular enzyme pathway to get rid of propionate . This pathway needs vitamin B12 in order to work , which is obtained from food . Newborns are screened for propionic acidemia using a test that measures the levels of a molecule called 3-hydroxypropionate ( 3-HP ) in the body . This molecule is not normally found in appreciable levels in healthy humans . However , it is not clear how 3-HP forms in individuals with propionic acidemia . In 2014 , researchers showed that in worms called Caenorhabditis elegans , propionate activates many genes when vitamin B12 levels are low . This suggests that the worms may have an alternate way to break down propionate when vitamin B12 is in short supply . Now , Watson et al . – including some of the researchers involved in the 2014 work – have used a combination of genetic , computational and biochemical techniques to identify five genes that the worms use to break down propionate when vitamin B12 is not available . Furthermore , the level of 3-HP rises in worms that cannot use B12 , just as is seen in humans with propionic acidemia . Thus , it appears that producing 3-HP may be an important step in an alternate pathway that does not require vitamin B12 to eliminate propionate . Having an alternate way of breaking down propionate may be essential for C . elegans worms living in the wild , which have to adapt to changing dietary conditions that may or may not provide them with vitamin B12 . Further studies are now needed to describe the metabolic effects of genes turned on by propionate and repressed by vitamin B12 , and to investigate how propionate alters the activity of these genes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"computational",
"and",
"systems",
"biology"
] | 2016 | Metabolic network rewiring of propionate flux compensates vitamin B12 deficiency in C. elegans |
Most perceptual decisions require comparisons between current input and an internal template . Classic studies propose that templates are encoded in sustained activity of sensory neurons . However , stimulus encoding is itself dynamic , tracing a complex trajectory through activity space . Which part of this trajectory is pre-activated to reflect the template ? Here we recorded magneto- and electroencephalography during a visual target-detection task , and used pattern analyses to decode template , stimulus , and decision-variable representation . Our findings ran counter to the dominant model of sustained pre-activation . Instead , template information emerged transiently around stimulus onset and quickly subsided . Cross-generalization between stimulus and template coding , indicating a shared neural representation , occurred only briefly . Our results are compatible with the proposal that template representation relies on a matched filter , transforming input into task-appropriate output . This proposal was consistent with a signed difference response at the perceptual decision stage , which can be explained by a simple neural model .
Human perception is flexible: the dimensions guiding perceptual decisions can be updated rapidly as a function of the current task . When decisions are based on perceptual analysis , task goals influence behaviour by creating an internal template: incoming sensory information is then matched against it . The representation of templates therefore plays a fundamental role in guiding perception and decision-making . Biased competition ( Desimone and Duncan , 1995 ) provides a broad framework for how the brain interprets new sensory information in light of the current search template . A central tenet is that attention tonically pre-activates visual cortical neurons with receptive fields for relevant , template-matching stimuli ( Reynolds and Chelazzi , 2004; Chelazzi et al . , 2011 ) . Single-cell neurophysiology ( Chelazzi et al . , 1993; Luck et al . , 1997; Chelazzi et al . , 1998 ) and human functional magnetic resonance imaging ( fMRI; Chawla et al . , 1999; Kastner and Ungerleider , 2000; Silver et al . , 2007; Kastner et al . , 2009; Reddy et al . , 2009 ) have demonstrated that template representation and stimulus processing can occur in overlapping neural populations in the visual cortex . Moreover , stimulus and template activity patterns cross-generalize ( when measured with fMRI , Stokes et al . , 2009 ) , implying that the two share a common neural code . In the simplest case , increasing baseline activity of a stimulus-specific representation could boost target processing ( Sylvester et al . , 2009 ) . This boost could facilitate target selection and reduce distractor competition for downstream processing resources ( Bundesen et al . , 2005; Maunsell and Treue , 2006 ) . However , recent findings complicate this simple model . Population-level analyses of time-resolved neural recordings show that stimulus decoding is highly time-specific ( King and Dehaene , 2014 ) , with discriminative activity patterns changing at the millisecond scale . Such dynamic coding has been observed at the level of population spiking patterns within individual brain areas ( Meyers et al . , 2008; Crowe et al . , 2010; Stokes et al . , 2013 ) , and at the level of distributed activation patterns across the cortex ( King et al . , 2013; Cichy et al . , 2014; Wolff et al . , 2015 ) , suggesting that this temporal dimension is an inherent aspect of neural coding ( Buonomano and Maass , 2009 ) . Importantly , neural populations in visual ( Meyers et al . , 2008; Sreenivasan et al . , 2014 ) and prefrontal cortex ( Hussar and Pasternak , 2012; 2013; Stokes et al . , 2013; Astrand et al . , 2015 ) appear to represent a memorized stimulus with an independent pattern from that used during initial encoding . As a consequence , it is necessary to distinguish between a neural pattern ( which may vary from moment to moment ) , and the representational content that is encoded in that pattern ( which may be stable even when the pattern changes over time , see Haxby et al . , 2014 ) . The highly dynamic trajectory that stimulus processing traces through activation state-space challenges classic models of template representation . These propose tonic activation of a static neural pattern , begging the question: which of the many points along the processing trajectory should be pre-activated ? An alternative scheme enables templates to guide perceptual decision-making even when stimulus processing is dynamic . If stimulus and template representations rely on different patterns of neural activity in the circuit , then a matched-filter process ( c . f . Sugase-Miyamoto et al . , 2008; Nikolic et al . , 2009; Stokes , 2015 ) could be envisaged in which the dynamic pattern of stimulus encoding would be automatically transformed into a pattern reflecting the degree of overlap to the template . This could be achieved if the pattern of activity elicited by the incoming stimulus is weighted by the neural pattern associated with the stored template information . While visual templates for target detection have been central to attention research , their role has been somewhat neglected in the study of perceptual decision-making . Perceptual decision-making tasks usually require the judgment of a visual stimulus feature against a fixed decision boundary or template ( Gold and Shadlen , 2007 ) . These tasks typically require judgments to be made at varying levels of perceptual difficulty ( Vogels and Orban , 1990; Ghose et al . , 2002; Purushothaman and Bradley , 2005; Summerfield and Koechlin , 2008; Scolari and Serences , 2010; Wyart et al , 2012 ) . The majority of perceptual decision-making studies have kept the decision boundary ( or template ) constant over the entire experiment , impeding a clear evaluation of the representation of templates as distinct from stimulus representation and from the sensory-to-template comparison . In the present study , we varied template and stimulus values independently , enabling us to examine the extent to which their coding and their temporal profiles overlap . We used pattern analysis of simultaneously recorded magneto- ( MEG ) and electroencephalography ( EEG ) to track visual template matching with high temporal resolution as human observers performed a parametric match-to-template orientation task ( Figure 1A ) . 10 . 7554/eLife . 09000 . 003Figure 1 . Task design and behavior . ( A ) Each block began with the presentation of a target orientation , which observers maintained for the duration of a task block . Template presentation was followed by a serial stream of randomly oriented stimuli . Observers were asked to respond with a button press whenever the stimulus matched the template orientation . ( B ) Response frequency was highest for target trials , and dropped off sharply for non-targets with increasing angular distance between template and stimulus orientation . Error bars indicate standard error of the mean across observers . The black line denotes a von Mises distribution fit to the responses . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 00310 . 7554/eLife . 09000 . 004Figure 1—figure supplement 1 . Reaction Time Distribution and Effects of Target Proximity . ( A ) Distribution of reaction time frequencies ( as a proportion of all responses ) , from stimulus onset ( collapsing over hits and false alarms ) . Beginning around 400 ms after stimulus onset , response frequency rises rapidly up to approximately 550 ms , after which responses slowly taper off . ( B ) Target proximity ( absolute angular distance between the stimulus and template angles ) does not affect reaction time ( F3 , 27 = 1 . 036 , p = 0 . 393 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 004 Neural responses rapidly traversed a cascade of discriminative patterns , transforming the initial task-invariant stimulus code , in conjunction with the template code , into a decision-relevant code . Template patterns and stimulus patterns cross-generalized only in a short time window during initial processing , suggesting some independence in the two neural codes . Despite these differences in the neural patterns , the content of the representation encoded in these patterns ( as measured by their representational similarity ) corresponded over a more sustained period . This might be expected if templates are encoded as a matched filter in the connections between stimulus-sensitive and decision-relevant populations . Interestingly , after the stimulus information was already reliably present and the response-relevant information had begun to emerge , neural signals also encoded the ( task-irrelevant ) signed difference between the current stimulus and the search template . This processing stage additionally suggests the presence of a matched filter that permits the flexible calculation of deviations from a search template . We argue , on the basis of a simple neural model , that this effect is consistent with the use of a population code for perceptual decision-making ( Ma et al . , 2006; Zemel et al . , 1998; Beck et al . , 2008 ) .
We recorded simultaneous MEG and EEG signals from 10 human observers as they performed a serial visual match-to-template task ( see Figure 1A and Materials and methods ) . At the beginning of each block , observers viewed a target orientation to be maintained in memory and used as a search template for the duration of the block . Each block consisted of a centrally presented stream of Gabor patches ( randomly drawn from a distribution of 16 orientations , uniformly spaced along the circle ) . Observers were instructed to respond with a button press whenever the target appeared . Over two sessions , each observer viewed a total of 7680 stimuli to maximize the statistical power of within-participant pattern analyses . On average , observers correctly detected approximately 70% of targets ( Figure 1B ) . They also made a large proportion of false alarms to near targets ( approximately 50% for offsets from the target angle of ± 11 . 25º ) , with false alarms rapidly dropping for more distant non-targets . Reaction times were distributed around 550 ms ( Figure 1—figure supplement 1A ) , with no strong effect of target proximity on reaction time ( p > 0 . 35 , Figure 1—figure supplement 1B ) . The stimulus information encoded in MEG/EEG signals was captured by calculating time-resolved population tuning curves ( Figure 2 , see Materials and methods ) . This approach transforms sensor-level responses into responses of virtual stimulus orientation channels: if a stimulus orientation is reflected in the MEG/EEG signal , virtual channel responses should peak at the corresponding orientation . In order to calculate the transformation of sensor data to tuning curves , the data were split into training and test sets . The training data were used to calculate each sensor’s sensitivity to each stimulus orientation , yielding a weight matrix . This weight matrix was then multiplied with the data in the independent test set and averaged over sensors . Single-trial virtual channel responses were then centered on the orientation presented on that trial and averaged over trials , providing an average population tuning curve . Tuning curves were calculated separately for each time point in the trial . Stimulus tuning curves showed that MEG/EEG signals ( EEG sensors were added to the analysis in combined MEG/EEG sessions ) reflected stimulus orientation shortly after its onset ( Figure 2B , Figure 3B , C , Figure 3A , 52–500 ms relative to stimulus onset , cluster-corrected p = 0 . 0019 ) . 10 . 7554/eLife . 09000 . 005Figure 2 . Stimulus-evoked population tuning curves . ( A ) Average population tuning curve , 50–300 ms after stimulus onset . ( B ) Time-resolved population tuning curve , showing a sharp increase in the tuning curve slope shortly after stimulus onset , tapering off within 500 ms . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 00510 . 7554/eLife . 09000 . 006Figure 3 . Task variable representation using population tuning curves ( see Figure 2 ) . ( A ) Stimulus orientation was represented in the early visual response . We fit weights ( using linear regression of stimulus orientation on the neural response ) using all trials in all training blocks and estimated virtual channel responses in the test block . Orientation-specific coding was estimated by calculating the linear slope of the tuning curve ( between 0° and 90° ) . Consistent positive slopes indicate orientation selectivity at a given time point . Shading indicates between-subject standard error of the mean . Black bars denote significant time points ( cluster-corrected ) . ( B ) Univariate sensitivity for stimulus orientation , calculated at each sensor and time point . Topography shows the shuffle-corrected orientation sensitivity ( z-scored against a distribution generated from permuting stimulus orientations 1000 times ) , averaged across sensor triplets ( two orthogonal planar gradiometers and one magnetometer ) and across the stimulus-decoding window . Color coding denotes the z-score , averaged across observers . ( C ) Tuning curve slope and topography ( D ) for template orientation sensitivity . E and F show the same analyses , sorting trials by the angular distance between template and stimulus ( i . e . , the decision value ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 006 Template orientation information was also present in the MEG/EEG signal during stimulus processing ( Figure 3C , –72 to 324 ms , cluster p = 0 . 0045 ) . A jack-knife analysis comparing onset latencies between template and stimulus coding showed that template coding began significantly earlier ( template: –72 ± 13 ms , stimulus: 52 ± 6 ms , t9 = –8 . 61 , p = 3*10–6 ) . To track the temporal evolution of task-relevant coding ( i . e . , the decision-value ) , we also decoded the distance of the current stimulus to the current template ( i . e . the signed angular distance between stimulus and template angles , from here on simply ‘angular distance’ ) . A strong effect of angular distance emerged around 200 ms ( Figure 3E , 164–596 ms , cluster p = 0 . 0024 , with an onset that was significantly later than for the stimulus orientation , t9 = 3 . 25 , p = 0 . 002 ) . This effect was also present when training only on trials without a response ( 172–596 ms , cluster p = 0 . 0014 ) , which discounts the possibility that this analysis simply reflected the difference in neural signals between responded trials ( which were most frequent for small angular distances ) and non-responded trials ( most frequent for large angular distances ) . While the three main task variables ( stimulus , template , and angular distance ) were all present in the signal , it is unclear whether different brain regions are involved . We calculated the sensor-level univariate sensitivity to each variable ( at MEG sensors only , averaging across the magnetometer and two gradiometers at each location ) to determine the topographies associated with different task variables . To a first approximation , all three variables were encoded in signals in visuo-parietal sensors ( Figure 3B , D , F ) . While sensitivity was again strongest to stimulus orientation , template and angular-distance responses nonetheless showed very similar topographies , indicating that all three variables might be computed in overlapping or nearby populations . To examine whether patterns of stimulus activity changed dynamically throughout the epoch , we tested for cross-temporal generalization of decoding ( as elaborated in King and Dehaene , 2014 ) . The population tuning curve approach was extended across time by calculating weights on one time point in the trial ( on a training data set ) and applying those weights to all time points in the trial ( on the left-out test data set ) . Stimulus orientation decoding was significant in one main cluster along the diagonal ( 64–544 ms in training time , 52–436 ms in test time , cluster p = 0 . 0024 , significant cluster extent indicated by color saturation in Figure 4A ) . More importantly , stimulus decoding was time-specific , with decoding dropping at off-diagonal train-test time points . To quantify the degree of dynamic coding statistically , we evaluated the off-diagonal results using a conjunction t-test: each off-diagonal combination of timepoints ( t1 , 2 ) was compared against both on-diagonal within-time pairs ( t1 , 1 and t2 , 2 ) . Evidence for dynamic decoding was inferred if decoding for t1 , 2 was significantly lower than both t1 , 1 and t2 , 2 . 10 . 7554/eLife . 09000 . 007Figure 4 . Cross-temporal generalization of orientation decoding . ( A ) Tuning curve amplitude for stimulus orientation , estimated by calculating weights at one time point and applying them to test data at all time points in a trial . While decoding is consistently high along the diagonal ( in the time window that contains significant stimulus information , between 52 and 544 ms , significant cluster indicated by color saturation/opacity ) , the slope drops sharply at off-diagonal train-test time coordinates . This indicates that the discriminative patterns are not consistent across time—rather they change rapidly , even while the stimulus can be readily decoded ( i . e . off-diagonal decoding is significantly lower than on-diagonal decoding , black outline ) . B and C show the same analyses as in A , but sorting all trials by the template angle and the decision-relevant angular distance , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 007 This drop-off is characteristic of dynamic coding ( King and Dehaene , 2014; Stokes , 2015 ) : despite significant decoding at two respective time points , the discriminative patterns do not generalize from one time point to the other . Off-diagonal generalization was significantly lower in a cluster ( black outline in Figure 4A ) stretching from 52–304 ms ( training time ) and from 88–436 ms ( generalization time , cluster p = 0 . 0031 , based on cluster extent , see Materials and methods ) . Since cross-generalization across time was significantly worse than within-time decoding , multiple stimulus-specific activity patterns seem to have been triggered in sequence . Importantly , Figure 4A shows that the cluster of significant decoding ( indicated by color saturation ) and the cluster of significant dynamic coding ( indicated by black outline ) partially overlap . In this overlapping region , training on timepoint t1 and testing on t2 still leads to significant decoding , but this generalization across time is nonetheless significantly lower than training and testing at either t1 or t2 alone . Such overlap can occur if decoding draws on a combination of dynamic and stationary patterns during the same epoch ( see also below ) . It is perhaps also interesting to note that we do not observe any evidence for periodic reactivation of orientation-specific patterns , which would be expected if the discriminating signal was oscillatory and phase-locked to the stimulus presentation ( King and Dehaene , 2014 ) . Template information ( Figure 4B ) was present in an early cluster ( training time: −140–340 ms , test time: −140–316 ms , relative to stimulus onset , cluster p = 0 . 0191 ) . In contrast to the stimulus decoding , template decoding showed no significant attenuation of decoding on the off-diagonal . Decision-relevant angular-distance decoding showed a dynamic pattern , although off-diagonal decoding appeared to be more pronounced ( Figure 4C ) compared with stimulus orientation decoding ( Figure 4A ) . Nevertheless , the strongest decoding was along the diagonal ( training time: 4–592 ms , test time: 64–592 ms , cluster p = 0 . 0009 ) , with significantly reduced off-diagonal generalization in this time window ( training time 316–424 ms , generalization time 472–592 ms , cluster p = 0 . 008 ) . Since off-diagonal decoding was nonetheless significant in a large time window , it is possible that the angular distance exhibits both a dynamic and a static element . This could happen for two reasons . First , it could occur if one population follows a dynamic processing cascade , while a separate population is tonically active in response to a given angular distance . Additionally , significant off-diagonal generalization could occur if there is temporal variability in the underlying processes , smoothing out the dynamic time-specificity across trials . Training the population tuning-curve weights on template orientations around the time of stimulus onset ( –150 to +300 ms ) showed a strong trend toward generalizing to stimulus decoding shortly after onset ( Figure 5A , 52–124 ms , corr . p = 0 . 063 ) . Using only the pre-stimulus time window ( –150 to 0 ms ) to train the template pattern , stimulus information could still be extracted in this immediate post-stimulus period ( Figure 5B , average over 50–150 ms after stimulus onset , t9 = 2 . 45 , p = 0 . 037 ) , indicating that template-specific neural patterns may be pre-activated immediately before stimulus onset . This result indicates that template activity patterns and stimulus activity patterns do cross-generalize ( e . g . , Stokes et al . , 2009 ) , but only transiently . The template- and the stimulus-discriminative patterns correspond only in the earliest encoding period , but not later ( even though stimulus decoding itself persisted up to 500 ms ) . 10 . 7554/eLife . 09000 . 008Figure 5 . Cross-generalization from template-discriminative patterns to stimulus-discriminative patterns . ( A ) Calculating tuning-curve weights relative to the template orientations in a training data set ( in window from –150 to +300 ms around stimulus onset ) , applying these weights on test data , and sorting them relative to the stimulus orientation , showed decoding early after stimulus onset that quickly returned to baseline . ( B ) Calculating population weights only on the pre-stimulus period ( with respect to the template orientations ) yielded a population tuning curve with a significant peak around the presented stimulus orientation ( e . g . a significant peak above the average response around 0º , and a significant positive tuning curve slope between ± 90º and 0º ) . Shading indicates the standard error of the mean . Black bars indicate significant time points or orientations ( p < 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 008 While the underlying patterns separating different stimulus orientations change dynamically after stimulus onset , the information content that is represented might be more stable . The basic decoding analysis already implies that dynamic neural patterns contain stable information: the same basis set is used for decoding throughout the epoch . Therefore , significant decoding along the time-specific diagonal axis in the cross-temporal analysis suggests that stimulus orientation ( Figure 4A ) , or angular difference ( Figure 4C ) , is represented throughout significant changes in the underlying neural patterns . However , a more formal test of the representational structure of multivariate activity is provided by representational similarity analysis ( RSA; Kriegeskorte et al . , 2008 ) . Rather than testing for discrimination per se , this approach focuses on the second-order pattern of condition-specific differences . This allows us to characterise the representational structure of the population code independently of the specific neural patterns associated with different stimulus orientations ( Kriegeskorte and Kievit , 2013 ) . As an example , a 40º-tilted stimulus might elicit a topography shortly after onset that is more similar to the topography from a 30º-stimulus than that of a 90º-stimulus . While the MEG patterns separating these stimuli might change throughout the trial , the relative difference between them could be preserved . This would indicate that the same kind of information about the stimuli is represented . This approach has been developed specifically to characterize implementation-independent representational geometry ( Kriegeskorte and Kievit , 2013 ) , and therefore is well-suited here to test whether dynamic neural patterns essentially code the same information . We tested this by repeating the cross-temporal analyses on the dissimilarity relationships between MEG responses evoked by different orientations . Dissimilarity was quantified by the Mahalanobis distance matrices between all pairs of stimulus orientations ( on one half of trials ) , separately for each time point in the trial . At each timepoint , this yielded a 16×16 distance matrix . We next calculated the same distance matrix for the remaining half of the data , and calculated the Pearson correlation coefficients between distance matrices from the two independent data sets , for each combination of time points . In contrast to the dynamically varying stimulus-discriminative patterns , the representational similarity remained much more stationary ( Figure 6A ) , with a stable plateau of high correlations ( Fisher-transformed Pearson’s rho ) from the earliest time of stimulus decoding . We found a significant ( on- and off-diagonal ) cluster early in the epoch ( 28–596 ms , cluster p = 0 . 005 ) . The temporal stability of the representation was , as above , tested by examining whether on-diagonal similarity was higher than off-diagonal similarity . While one short-lived dynamic cluster emerged ( training time 82–146 ms , generalization time 130–226 ms , cluster p=0 . 0075 , see black outline in Figure 6A ) , the majority of the epoch was dominated by time-stable correlations . 10 . 7554/eLife . 09000 . 009Figure 6 . Cross-temporal generalization of representational similarity . ( A ) Pearson correlations between stimulus-orientation-sorted distance matrices , calculated at different time points and on independent data sets . Color saturation shows significant cluster at the group level ( permutation test ) . The cluster extends off the diagonal in a square , indicating substantial cross-temporal generalization . In addition , there is a small dynamic cluster ( black outline ) , meaning that pairs of time points within the black outline showed significantly lower correlations than their corresponding time points on the diagonal ( even though they were still significantly greater than 0 ) . ( B ) shows the same analysis as in A , but sorting all trials by the decision-relevant angular distance . There were no significant dynamic clusters . RSA , representational similarity analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 009 Similarly , the representational similarity of angular distance was stable throughout the trial ( Figure 6B ) . Dissimilarity matrices correlated significantly in a later window in the trial ( 172–588 ms , cluster p = 0 . 0026 ) , with no time points where within-time correlations were significantly higher than between-time correlations ( all p > 0 . 20 ) . These complementary analyses highlight the cardinal feature of dynamic coding: discriminative dimensions vary with time even though the information content remains constant ( Laurent , 2002; Stokes , 2011 ) . As the previous section indicates , the representational similarity of different stimulus orientations is more temporally stable than the underlying discriminative pattern . If RSA can reveal stable representations over time , it could also uncover representational similarity between the template and the stimulus . In other words , even though the MEG patterns did not persistently cross-generalize between stimulus and template decoding , the dissimilarity matrices calculated for template orientations and for stimulus orientations might reveal a more stable match . This would indicate that similar content is being stored about stimuli and templates , even though the precise neural implementation might differ . To quantify this relationship , we again tested for cross-temporal generalization of the dissimilarity matrix . However , here we correlated the dissimilarity matrix calculated between template orientations with the matrix calculated between stimulus orientations . Specifically , we correlated the Mahalanobis distance matrix between all eight template orientations , calculated at each time point , with the distance matrix between stimulus orientations ( limiting our analyses to the same eight stimulus orientations that served as targets in the experimental session ) . The template-sorted dissimilarity matrix correlated significantly with the stimulus-sorted dissimilarity around the time of visual processing ( cluster in Figure 7A , template structure from –48 to 196 ms , stimulus structure from 88–208 ms , cluster p = 0 . 038 ) . The within-time comparison between templates and stimuli ( i . e . , the values along the diagonal ) also showed a significant correlation ( Figure 7B , 104–176 ms , p = 0 . 036 , with trends toward significance between 412–464 ms , p = 0 . 079 , and 552–596 ms , p = 0 . 086 ) . 10 . 7554/eLife . 09000 . 010Figure 7 . Geometry of stimulus and template coding . ( A ) The representational similarity structures between template- and stimulus-ordered responses were significantly correlated in the early stimulus-processing window ( saturated colors indicate significant cluster ) . ( B ) The within-time comparison also showed a significant correlation in the representational similarity structure from 104 to 176 ms . Values correspond to the mean regression coefficient across all observers . Shading is between-subjects standard error of the mean . ( C ) Multi-dimensional scaling of the distances between stimulus orientations was not visible before stimulus onset . ( D ) Shortly after stimulus onset , the circular structure indicated that responses used a circular geometry . ( E To quantify the representational structure over time , we fit ( using regression ) to the neural distance matrix between all angles ( 16 different angles , split randomly into two sets of trials , resulting in a 32×32 distance matrix of Mahalanobis distances ) the distance matrix of a 16-point circular simplex , shown in ( F ) . ( G ) Similarly , relationships between the eight template orientations fit a circular structure , particularly around stimulus onset time . ( H ) An example of a simplex from one session , with the eight chosen template angles highlighted in color , and the eight stimulus orientations which were never targets shown in gray . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 010 What is the basis of this representational similarity between templates and stimuli ? Given the simplicity of the stimulus set , a straightforward representational structure comes to mind: more similar stimulus ( or template ) orientations evoke more similar MEG topographies . However , this cannot be deduced from the population tuning-curve analysis alone . To evaluate the possibility , we calculated neural dissimilarity matrices between the mean responses evoked by each of the 16 stimulus orientations , and projected this 16×16 neural dissimilarity matrix into two dimensions for visualization ( using multi-dimensional scaling , Figure 7D ) . During the stimulus-encoding period ( 50–250 ms after stimulus onset ) , conditions fell onto a well-ordered circle: topographies were more similar ( had a smaller Mahalanobis distance ) if they were evoked by more similar stimulus orientations . This geometry was not present in the data before the onset of stimulus processing ( –50 to +50 ms relative to stimulus onset , Figure 7C ) . We tested for the temporal stability of this representational structure by correlating the data-derived ( Mahalanobis ) distance matrix at each time point with an idealized distance matrix , derived from the angular distances of the respective stimulus orientations ( i . e . a 16-point simplex , corresponding to the pairwise angular distances between all 16 presented orientations , Figure 7F ) . We used linear regression to fit the idealized distance matrix of the stimuli to the neural distance matrix at each time point . The stimulus similarity matrix significantly fit the neural data ( Figure 7E , 44–432 ms , cluster p = 0 . 002 ) . Likewise , the neural dissimilarity matrix between different template orientations was well described by the same simplex structure ( Figure 7G , H , 48–300 ms , cluster p = 0 . 0012 , with a second cluster around the time of the next stimulus ) . Therefore , while the discriminative patterns for stimuli and templates cross-generalized only briefly , the content of their representation appears to be both stable over time and similar between task variables . We tested whether the decision-relevant angular distances ( between the current stimulus and the template ) , which were necessary for guiding behavior , showed a comparable similarity structure . The angular distance could be decomposed into two components: its magnitude and its sign . The magnitude of the angular distance ( i . e . the absolute difference between stimulus and template orientation ) solely determined the task-relevance of a stimulus: the closer the magnitude is to 0º , the more likely the stimulus led to a target response . By contrast , the sign of the angular distance ( i . e . , whether a stimulus was oriented clockwise or counter-clockwise with respect to the template ) had no relevance to the task , because it did not influence how close that stimulus was to the current template . In the next analysis , we therefore attempted to isolate the effects of magnitude and sign on the neural response . We projected ( using independently calculated weights , see Materials and methods ) neural responses ( from all 16 possible angular distances , and from all MEG/EEG sensors ) onto two axes measuring separately the influence of magnitude and sign of the angular distance at each point in the trial ( Mante et al . , 2013 ) . This allowed an analysis of the MEG signal’s sensitivity to the decision-relevant magnitude ( measured by the amplitude of the response along the magnitude axis ) , independently of its sensitivity to the decision-irrelevant sign of the angular distance ( measured along the sign axis ) . The mean responses to the 16 angular distances fell roughly onto a circle that stretched out along the magnitude axis , with targets and near-targets clearly separable from the definite non-targets ( Figure 8A ) . Interestingly , near non-targets that were either clockwise or counter-clockwise to the target also separated along the decision-irrelevant sign axis , indicating an unexpected result: angular distances with equal magnitude but different sign ( i . e . , stimuli at an identical distance to the template orientation , such as –23º and +23° ) evoked distinct and separable neural responses . Mean projections along the task-irrelevant axis for conditions with equal magnitude but different sign diverged around the time of decision formation ( 348–588 ms , corrected p = 0 . 004 , Figure 8B ) . We verified that this was the case even without relying on the task-projection approach by calculating Mahalanobis distances between pairs of angular distance trials that had equal magnitude ( i . e . , –11° vs . +11° , –22° vs . +22° , –34° vs . +34° ) , and found similar results ( see Figure 8—figure supplement 1 ) . In addition , we confirmed in a control analysis that different angular distances were not separable merely because of possible differences in response bias ( Figure 8—figure supplement 2 ) . 10 . 7554/eLife . 09000 . 011Figure 8 . Geometry of response-related coding . ( A ) Dissimilarity structure of angular distances . Data dimensionality was reduced using PCA , and weights calculated between sensor activity and different task variables using independent training data . Mean responses for each angular distance , calculated using the left-out data , were then projected via the calculated weights onto the task axes ( the magnitude and sign of the angular distance ) . Since the task relevance of a particular angular distance was defined solely by its magnitude , projections onto the sign axis measured sensitivity to task-irrelevant signed differences between conditions . Prior to decision onset ( 250–500 ms after stimulus onset ) , the neural geometry is elliptical: in addition to conditions separating along the target-relevant magnitude axis ( horizontal ) , near non-targets separate along the task-irrelevant sign axis ( vertical ) . ( B ) Task-irrelevant coding emerges approximately 350 ms after stimulus onset . Time courses for the three nearest non-targets ( 11º , 22º , 34º offset from target angle ) separate along the task-irrelevant axis , depending on whether they are clockwise or counterclockwise to the target . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 01110 . 7554/eLife . 09000 . 012Figure 8—figure supplement 1 . Multidimensional Scaling and Pairwise Mahalanobis distances between Angular Distances . ( A ) Dissimilarity structure of angular distances . We used MDS , which maps the multidimensional ( 32× 32 ) Mahalanobis distance matrix between target-relative angles into two dimensions . During relatively early stimulus processing ( 250–400 ms after stimulus onset ) , geometry is elliptical—that is , in addition to conditions separating along the target-relative axis ( horizontal ) , conditions separate along a task-irrelevant axis ( vertical ) . During later processing stages ( B: 450–900 ms ) , the task-related axis accounts for most of the condition differences . Since MDS is rotation-invariant , the solution in B happens to have flipped axis 2 , without affecting the geometrical relationship between points . ( C ) Mahalanobis distances ( shuffle-corrected ) between trials with equal target proximity , but different direction ( i . e . , clockwise vs . counter-clockwise deviations of the stimulus angle , with respect to the template angle ) . The figure shows the mean z-score ( with respect to 250 random permutations of the trial labels ) of pairwise distances between equal target proximities , averaged over the pairs ± 11 . 25º , ± 22 . 5º , and ± 33 . 75º . Shading indicates standard error of the mean . The black bar denotes significant time points ( p<0 . 05 , cluster-corrected ) . MDS , multi-dimensional scalingDOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 01210 . 7554/eLife . 09000 . 013Figure 8—figure supplement 2 . Figure is identical to Panel 8c , but includes in the graph the fit to the distance matrix provided by the linear decision value ( i . e . , the unsigned target proximity , stimulus—target ) . This variable was also included in the analysis described in the main text , but omitted from Figure 8C for clarity ( since it a nuisance variable ) . Shading indicates standard error of the mean and colored bars at the bottom denote significant time points for each regressor ( p<0 . 05 , cluster-corrected ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 01310 . 7554/eLife . 09000 . 014Figure 8—figure supplement 3 . Neural Population Model . ( A ) Probabilistic population code model architecture . ( B ) Dissimilarity structure of responses in the stimulus layer ( left panel ) and the decision layer ( right panel ) . ( C ) Accumulator model architecture . In contrast to the population code , decision value here is represented only in a single node ( red unit in the decision layer ) . Otherwise , the architectures are identical . ( D ) Dissimilarity structure of responses in the accumulator model . While responses in the stimulus layer are identical in both cases , the decision layer differs from the population code model , in that the magnitude , but not the direction , of angular differences between stimulus and template , is represented . ( E ) Model response on an exact template match trial . ( F–H ) Model responses on mismatch trials . DOI: http://dx . doi . org/10 . 7554/eLife . 09000 . 014 The neural encoding and sustained representation of the signed difference between the current stimulus and the template was unexpected because representing the sign of the angular distance was not necessary for solving the task ( since it would be sufficient to calculate only the magnitude ) . However , the result yields some insight into the particular neural implementation of the decision process in this task . Specifically , a representation of both magnitude and sign is consistent with the use of a probabilistic population code ( Ma et al . , 2006; Beck et al . , 2008 ) . Probabilistic population codes assume that the brain uses the activation pattern across a population of neurons , each tuned to a different stimulus value ( angular distance , in our case ) , to encode a probability distribution across the entire stimulus space ( Ma et al . , 2006; Zemel et al . , 1998 ) . The peak of activity of this distribution lies in or near neurons tuned to the presented angular distance . Therefore , angular distances with equal magnitude but different sign can be naturally separated in the population response , even if that dimension of the neural pattern is task-irrelevant . Importantly , this would not be the case if the entire population simply encoded the magnitude of the absolute distance ( i . e . , the overall match between stimulus and template , as is the case with some accumulator models of decision-making ) . Therefore , the presence of signed difference signals in the MEG response suggests that the brain uses a ( probabilistic ) population code to represent the decision variable in this task . For illustration , we created a simple neural architecture to elaborate this argument for a population code and to link it to the neural data . The model consisted of three interconnected layers ( see Materials and methods and Figure 8—figure supplement 3 for details of model behavior ) . Each layer encoded information about one of the three task variables ( stimulus , template , and angular distance ) . Each unit in a layer was tuned to a different orientation . Tuning in the decision layer represented decision-relevant angular distances , meaning that angles closer to 0° represented stimuli closer to the current template . We created the same dissimilarity matrices used in the MEG/EEG analyses from synthetic responses generated by each layer in the model . Identical to Figure 8A , conditions with equal magnitude but different sign led to separable population responses in the template and angular distance layers . In contrast , a simpler model with only a single accumulator unit in the decision layer showed only a differentiation of conditions if they differed in magnitude , reflecting that here the signed angular difference was not encoded . The population coding model used here is almost identical in architecture and behavior to a more elaborate biophysical model that was recently developed to predict the learning of new categories in a population of lateral intraparietal neurons in monkeys performing an orientation discrimination task ( Engel et al . , 2015; Freedman and Assad , 2006 ) .
In a visual match-to-template orientation task , we found distinct , dynamically evolving neural responses that reflected the orientation of the stimulus and the template , as well as the angular distance between the two ( i . e . , the task-relevant variable ) . Contrary to standard models of top-down attention , we did not find a tonic activation of the template neural pattern . Instead , the template pattern emerged transiently around the time of the stimulus onset and then quickly returned to baseline . While pattern analysis is a well-established methodology for intracranial multi-unit recordings and for fMRI , it is becoming clear that it can provide a useful approach to MEG and EEG as well . Although MEG/EEG measures neural activity at a larger scale relative to micro-electrode recordings , recent modelling demonstrates that the electromagnetic signal contains rich spatiotemporal information suitable for multivariate decoding ( Cichy et al . , 2015 ) . Even subtle differences in dipole position or angle elicit statistically separable patterns at the scalp surface . For orientation decoding , these differences presumably depend on idiosyncrasies in the distribution of orientation columns along the cortical surface ( Cichy et al . , 2015 ) . Such subtleties average out at the group level ( or are lost during source localization due to inherent ambiguities with the inverse solution ) , but can be characterized within individual participants using pattern analysis ( see Stokes et al . , 2015 ) . This logic extends in time: small differences in the spatial distribution of activity patterns at different time points would result in idiosyncratic changes in the dipole , resulting in a time-varying signal at the scalp surface . Indeed , the cross-temporal analyses suggest that orientation-specific patterns are also time-specific ( see also Wolff et al . , 2015 ) . In animal models , similar spatiotemporal patterns have been attributed to a cascade of neural engagement within the same brain area ( Harvey et al . , 2012 ) or time-specific changes in cell preferences ( Sigala et al . , 2008; Stokes et al . , 2013 ) . It is important to appreciate that decodability does not necessarily imply that the brain is making use of the decodable information ( Tong and Pratte , 2012 ) . Nonetheless , neural circuits with complex spatiotemporal dynamics could plausibly provide a rich source of information for guiding flexible ( Miller and Fusi , 2013 ) and context-dependent behavior ( Buonomano and Maass , 2009 ) . The rapidly changing patterns encoding the stimulus orientation raised the question of when the stimulus is compared to the template . Even though template decoding lasted up to 300 ms after stimulus onset , the template-specific neural patterns cross-generalized to stimulus-specific patterns only in the earliest encoding phase . The transient , rather than sustained , activation of template-specific patterns may reflect the reactivation of a latent code ( Mongillo et al . , 2008; Buonomano and Maass , 2009 ) that was laid down in altered synaptic weights , but which is reactivated via top-down or stimulus-driven input . Template decoding began shortly before stimulus onset , suggesting that the semi-regular timing of events may have allowed for top-down re-activation of the template ( as in ‘rhythmic sampling’ , c . f . Schroeder and Lakatos , 2009; Lakatos et al . , 2013 ) . Since template decoding peaked during the stimulus presentation period , bottom-up stimulus drive may have additionally activated the template pattern . The rapid dynamics of stimulus decoding further raise the question of how the brain compares dynamically evolving population codes ( Laurent , 2002; Meyers et al . , 2008; Stokes et al . , 2013; King and Dehaene , 2014 ) . Representational similarity analysis ( Kriegeskorte and Kievit , 2013; Haxby et al . , 2014; Nili et al . , 2014 ) permits higher-order comparisons between different task variables , even if their underlying neural patterns are different . We speculate that matched filters provide a natural solution to the problem of comparing a dynamically evolving stimulus-encoding pattern to a template-encoding pattern: the stimulus pattern is filtered by a population that is matched to the visual characteristics of the template , leading to output that quantifies their overlap . Unexpectedly , the task-irrelevant sign of the angular distance was encoded in the MEG/EEG response pattern . This finding could provide an interesting insight into the potential mechanism underlying perceptual decision-making in our task . Specifically , probabilistic population codes may underlie the representation of the angular distance , and encode the sign of the angular distance as a by-product of decision-making . This could be a simple result of the use of a matched filter at an earlier stage . There is evidence that stimulus orientation is represented via population codes in early visual cortex ( Graf et al . , 2011; Berens et al . , 2012 ) , with the activation profile across neurons tuned to many different orientations reflecting a probability distribution peaking at the orientation that is most likely present in the environment . If this population activity pattern is passed through a filter tuned to the orientation of the template , the resulting output population pattern could again reflect a probability distribution peaking at the most likely relative orientation of the stimulus with respect to the template . Because of the orientation symmetry of the filter mechanism , the output would also reflect the direction of the angular distance ( clockwise or counterclockwise ) . Recent computational and empirical work on the maintenance of items in working memory has argued that mnemonic information ( such as a visual template ) can be stored through a reconfiguration of synaptic weights ( Mongillo et al . , 2008; Lewis-Peacock et al . , 2012; Stokes et al . , 2013 ) , without requiring strong persistent activity ( Watanabe and Funahashi , 2014 ) . The encoding of decision-relevant mnemonic templates in the weights of a network has the crucial advantage that processing of any new information can immediately pass through the modified weights and produce a matched filter response ( Mongillo et al . , 2008 ) . Conceived in this way , contents in working memory are decision rules that enforce the current stimulus-response mapping ( Stokes , 2015 ) . Read-out at the time of the probe then consists in a perceptual decision ( Martínez-García et al . , 2011; Pearson et al . , 2014 ) . These hidden states , reflecting the current task context , could be in operation in our task , and would map onto the representation of the target orientation in the template layer of our toy model . The transient reactivation of the template shortly before stimulus onset could also reflect the nature of our task , where the majority of stimuli were non-targets that may have discouraged the use of tonic template activation . For instance , single-cell studies have shown that visual distractors presented in a memory delay can disrupt tonic activity of cells coding the remembered item ( in IT , Miller et al . , 1993 , and transiently in PFC , Miller et al . , 1996 ) . By contrast , tasks without intervening distractors may be more conducive to the use of tonic activation ( Chelazzi et al . , 1993 , 1998 ) .
Ten healthy , right-handed volunteers ( age range: 21–27 years , 6 females ) took part in the study , and were paid £10/hr for their time . Visual acuity was normal or corrected to normal . Ethical approval for methods and procedures was obtained from the Central University Research Ethics Committee of the University of Oxford . Each participant completed two experimental sessions , approximately 1 week apart , with each lasting approximately 2 hr ( of which approximately 1 hr was spent on performing the task ) . Each participant completed a large number of trials ( 7680 across two sessions ) , providing robust within-participant statistical power for within-participant decoding . Participants completed the MEG/EEG scan inside a sound-attenuated , dimly lit , and magnetically shielded room . Stimuli were displayed onto a rear-projection screen ( placed at a viewing distance of 90 cm ) via a projector ( Panasonic DLP Projector , PT-D7700E ) with a spatial resolution of 1024 × 768 pixels and a refresh rate of 60 Hz . Stimuli were presented using Psychophysics Toolbox ( Brainard , 1997 ) , running on MATLAB ( Mathworks , Natick , WA ) . Participants responded using an optic-fibre response box by lifting their right index finger to indicate whenever they had seen a target . Participants were instructed to respond as quickly and accurately as possible . The task required the detection of visual targets within a stream of sequentially presented stimuli . The stream consisted of oriented Gabor patches ( diameter: 4° visual angle , spatial frequency: 2 cycles/° ) , presented foveally for 100 ms , at an average rate of 650 ms ( inter-stimulus interval , ranging from 516 to 783 ms ) . Orientations were drawn without replacement from a set of 16 possible angles . Stimuli were equally spaced from 5 . 625° to 174 . 375° , in steps of 11 . 25° . The task consisted of eight brief ( approximately 6 min ) blocks , in which 480 stimuli were presented ( resulting in a total of 3840 stimulus presentations per session ) . Each block began with the presentation of a target orientation ( drawn at random , without replacement , from the 16 stimulus orientations ) , displayed centrally as a green line ( 4° length ) . Thus , each session contained eight randomly drawn target orientations ( they did not need to repeat across experimental sessions ) . The participants were instructed to respond whenever a Gabor patch with a matching orientation appeared . Since stimuli were drawn equiprobably from the 16 possible orientations , 1/16 of all stimuli were targets . Each block was cut into three shorter segments , giving participants brief rest periods . During the rest periods , the target orientation was presented again as a reminder . In accordance with the principles of open evaluation in science ( Walther and van den Bosch , 2012 ) , all data and fully annotated analysis scripts from this study are publicly available at http://datasharedrive . blogspot . co . uk/2015/08/testing-sensory-evidence-against . html ( see also Myers et al . , 2015 ) . We also hope these will provide a valuable resource for future re-use by other researchers . In line with the Organisation for Economic Cooperation and Development ( OECD ) Principles and Guidelines for Access to Research Data from Public Funding ( Pilat and Fukasaku , 2007 ) , we have made every effort to provide all necessary task/condition information within a self-contained format to maximise the re-use potential of our data . We also provide fully annotated analysis scripts that were used in this paper . Any further queries can be addressed to the corresponding author . Because of the rapid succession of stimuli , it is difficult to attribute unequivocally each response to a single stimulus . Therefore , a stimulus-response assignment procedure was designed in order to attribute , in a probabilistic fashion , each response to a single stimulus . First , response-time ( RT ) distributions to stimuli were computed on the basis of their absolute angular distance ( tilt ) from the target orientation ( from 0 to ± 90º ) . When RTs were averaged relative to the orientation of the stimuli , it was clear that the responses fell within a certain time window ( from approximately 200 to 1000 ms ) , consistent with the approximately periodic presentation of stimuli . Tilt-dependent RT distributions were used to estimate the tuning of responses to the target . At each RT , the response tuning profile—the probability of a response given the tilt of the stimulus , from 0 to 90º—was fitted with a von Mises distribution having two free parameters: the peak of the distribution PMAX , and the concentration parameter kappa κ . The von Mises distribution was constrained to be centred at the target orientation ( tilt = 0 ) , and the definition of κ was modified such that κ = 0 indicates no tuning , κ > 0 indicates a preferred tuning for the target orientation , and κ < 0 indicates a preferred tuning for the orientation perpendicular/opposite to the target . For each subject , the tuning concentration showed a clear positive response following stimulus onset ( approximately 200 to 1000 ms post-stimulus ) . This tuning information was then used to assign probabilistically each response to an individual stimulus . First , for each response , all stimuli that fell into the time window during which the tuning concentration was positive were preselected . Next , among these candidate stimuli ( which had different tilts with respect to the target ) , the stimulus that maximised the probability of a response at the observed RT was selected . The resultant RT distributions truncated the low and high RT values leaving the central part of the original RT distributions Each participant completed two sessions: one MEG-only session , and one session in which EEG data were recorded concurrently . Participants were seated in the MEG scanner in a magnetically shielded room . Their legs were placed on leg rests and arms on their lap to avoid movements . Both experimental sessions lasted approximately one hour . Participants were instructed to maintain fixation on the centre of the screen during the stimulus blocks and minimize blinking . Neuromagnetic data were acquired using a whole-head VectorView system ( 204 planar gradiometers , 102 magnetometers , Elekta Neuromag Oy , Helsinki , Finland ) . Magnetoencephalographic signals were sampled at a rate of 1 , 000 Hz and on-line band-pass filtered between 0 . 03 and 300 Hz . The participant’s head position inside the scanner was localised and tracked continuously using head-position index coils placed at four distributed points on the head . Electrodes were placed above and below the right eye for the vertical electro-oculogram ( EOG ) and to the side of each eye for the horizontal EOG . In addition , eye movements were monitored using a remote infrared eye-tracker ( SR research , EyeLink 1000 , sampling one eye at 1000 Hz , controlled via Psychophysics Toolbox , Cornelissen et al . , 2002 ) . EEG data were collected in half of the sessions ( for each participant ) , using 60 channels distributed across the scalp via the international 10–10 positioning system ( AEEGS , 1991 ) . Filtering , downsampling , epoching , and rejection of artefactual trials were performed on EEG data in the same way as on the MEG data . EEG data were added to all decoding analyses for the MEG+EEG sessions ( except for the topographies in Figure 3 ) . We found no substantial differences in decoding between MEG-only and MEG+EEG sessions , apart from a small increase in decoding sensitivity in the latter . Therefore , all within-session analyses were averaged to arrive at participant-level results . Data were preprocessed using the in-house OHBA software library ( OSL ) , drawing on SPM8 ( http://www . fil . ion . ucl . ac . uk/spm ) , Fieldtrip ( Oostenveld et al . , 2011 ) , and Elekta software . The raw MEG data were visually inspected to remove and interpolate any channels with excessive noise , and were further de-noised and motion-corrected using Maxfilter Signal Space Separation ( Taulu et al . , 2004 ) . Next , data were downsampled to 500 Hz . Remaining epochs with unsystematic noise corruption were then excluded via visual inspection . Systematic artefacts , arising from eye blinks and heart beats , were identified via independent component analysis , and regressed out of the raw data . The cleaned data were then epoched with respect to each stimulus onset ( from –1 to + 1 s ) . In a final step , data were imported into Fieldtrip and inspected using the semi-automatic rejection tool to eliminate any remaining trials with excessive variance . All data were then baseline-corrected by subtracting the mean signal between –150 and –50 ms relative to stimulus onset ( for analyses relating to the template , we used an earlier baseline , from –200 to –150 ms relative to stimulus onset , to explore the possibility that template information might be ‘pre-activated’ around the expected onset time . Using the standard baseline from –150 to –50 ms , however , did not change the results presented here ) . In addition , the data were smoothed with a 32-ms Gaussian kernel for template-based analyses to reduce noise . We used a population tuning curve model to recover information about the stimulus orientation from the full M/EEG signal . Instead of looking to relate imaging data to different stimulus orientations directly , each stimulus orientation is represented using weights from a linear basis set of population tuning curves . Tuning curve models are well suited to recovering information about parametric features like orientations ( Saproo and Serences , 2010; Brouwer and Heeger , 2011; Serences and Saproo , 2012; Garcia et al . , 2013 ) or colors ( Brouwer and Heeger , 2009 ) . To recover stimulus orientations , data were separated into a training set ( all trials from 7 of 8 blocks ) and a test set ( the left-out block ) . For all trials in the training set , we then created a matrix of 16 regressors , with the height of the each regressor on any trial reflecting that trial’s stimulus orientation ( i . e . a regressor was set to 1 when the corresponding orientation was presented on that trial , and to 0 otherwise ) . The regressor matrix was then convolved with a half-cosine basis set ( raised to the 15th power , see Brouwer and Heeger , 2009 ) , in order to pool information across similar orientations . Orientation sensitivity at each MEG/EEG sensor was then calculated by regressing the design matrix against the signal ( across all 306 sensors or all 366 sensors in MEG+EEG sessions ) , separately for all time points in the epoch ( in 4 ms steps , using a sliding window of 20 ms ) . We solved the linear regression equation: ( 1 ) B1 = WC1; where C1 is the design matrix ( 16 regressors × no . of training trials ) , B1 is the training data set ( 306/366 sensors × no . of training trials ) , and W is the weight matrix ( 306/366 sensors × 16 orientation values ) that we want to estimate . This was done using ordinary least squares: ( 2 ) W = B1C1T ( C1C1T ) -1; Overall differences in signal magnitude between sensors were modeled out using a constant regressor in C1 . We used W to estimate the population orientation response ( or tuning curve ) in the test set , B2 ( 306/366 sensors × no . of test trials ) : ( 3 ) C2 = ( WTW ) -1WTB2; where C2 is the tuning curve , W is the weight matrix , WT is its transpose , and W–1 is its pseudo-inverse . Since both the design matrix and the estimated weight matrix were of full rank , this approach was equivalent to using the pseudoinverse for estimation . For each trial , this curve was then zero-centered relative to the presented orientation . This procedure was repeated for each time point in the epoch before moving to the next iteration in the leave-one-out procedure . Zero-centered orientation curves were then averaged across trials . The time course of the tuning curve was then converted into a stimulus information time course by calculating the linear slope of the tuning curve from –90° to 0° . We first averaged stimulus channels that were equidistant from 0° ( i . e . +11 . 25° and –11 . 25° , +33 . 75° and –33 . 75° , etc . ) and smoothed each resulting ( sign-invariant ) orientation channel time course ( with a 16-ms Gaussian kernel ) . We then fit a linear slope across the orientation channels ( from –90° to 0° ) , separately for each time point , session , and participant . Decoding accuracy was then evaluated using one-sample t-tests ( against 0 ) , under the assumption that slopes are randomly distributed around 0 if there is no stimulus information in the signal . Multiple comparisons across time were corrected for using cluster-based permutation testing ( 10 , 000 permutations , Maris and Oostenveld , 2007 ) . We used a similar approach to test for encoding of the current template orientation , with the exception that here we used a 32-ms sliding window to increase sensitivity to a more slowly evolving effect . Since there were only eight template orientations per session , and these were randomly selected from the 16 possible stimulus orientations , they were not always equally spaced across the circle . We estimated orientation tuning curves across the eight irregularly spaced angles ( using eight equally spaced regressors ) , and then linearly interpolated the estimated tuning values at the eight intermediate values . After interpolation , the template orientation tuning curves were treated as above to derive decoding time courses . Finally , we also applied this approach to calculating tuning profiles for information about the angular distance between the orientation of the current stimulus and the template ( ranging from 0° for template matches , in steps of 11 . 25° , to 90° , for stimuli that were orthogonal to the template ) . Onset latencies between stimulus , template , and angular distance were compared using a jack-knife approach ( Miller , Patterson and Ulrich , 1998 ) . We compared the onset times of significant coding ( p<0 . 05 , corrected ) using t-tests . To estimate the variance of each onset time , we used an iterative procedure that left out one participant in turn and calculated the onset time of significant coding across all remaining participants . The standard error of the latency difference was calculated using a revised measure that takes into account the reduced variability caused by the jack-knife procedure ( Miller et al . , 1998 ) . The latency difference calculated across the entire set of participants was divided by this standard error estimate to provide t-statistics that were then evaluated using the conventional t-distribution . In addition to the pattern analyses that averaged signals over all sensors , we tested the orientation sensitivity of individual MEG/EEG sensors , to generate a topographical distribution of the sensitivity to the three task variables ( stimulus orientation , template orientation , and decision-relevant angular distances ) . The baseline-corrected signal at each sensor and time point in the epoch was fit ( across all trials ) using a general linear model ( GLM ) consisting of pairs of regressors containing the sine and cosine of the three task orientations , along with a constant regressor . From the pair of regression coefficients for the sine ( βSIN ) and cosine ( βCOS ) of an orientation , we calculated orientation sensitivity A: ( 4 ) A = √ ( βCOS2 + βSIN2 ) ; We calculated the amplitude A expected by chance alone by permuting the design matrix and repeating the amplitude analysis 1000 times . The observed ( unpermuted ) amplitude was ranked within the permutation distribution of amplitudes to calculate a p-value , which was transformed into a z score using the inverse of the cumulative Gaussian distribution ( with center 0 and standard deviation 1 ) . Sensitivity at the group level was then estimated by averaging z-scored amplitudes across session , participant , and the magnetometer and two gradiometers at each sensor location . These values were then plotted as topographies to illustrate the distribution of orientation sensitivity for the three task variables . To assess the temporal stability of stimulus-specific topographies , we trained the population tuning-curve model on one time point in the epoch , and applied the estimated weights to all time points in the test data ( using a sliding window of width 20 ms , applied every 12 ms ) . This was then repeated for all time points , creating a two-dimensional matrix of cross-temporal tuning-curve slopes ( with no additional smoothing ) . Dynamic coding can be inferred by comparing the decoding slopes on within-time training ( i . e . , training and testing on time t1 , or time t2 ) with the decoding slopes on between-time training ( i . e . , training on t1 and testing on t2 ) . Our criterion for a dynamic epoch was: for each pair of time points ti , j , coding is dynamic if the tuning curve slope is significantly higher ( as measured by a paired t-test across 10 participants ) within time than across time ( ti , i > ti , j AND tj , j > ti , j ) . Time windows of significant decoding ( ti , j > 0 ) and windows of significant dynamic coding were identified using 2-dimensional cluster-based permutation testing ( i . e . , across both time axes ) . To test whether stimulus-specific patterns cross-generalize to template-specific patterns , we repeated the cross-temporal tuning-curve analysis , but calculated weights based on the presented template orientations in the training set , and then zero-centered the tuning curves of the test set with respect to the stimulus orientations . Here , a significantly positive tuning curve slope at time pair ti , j indicates that stimulus coding around time point i shares orientation-specific topographic patterns with template coding around time point j . For consistency with the other analyses , we treated the training data as in the analyses evaluating template coding , and treated the training data as in the stimulus decoding analyses . Therefore we used a baseline of –200 to –150 ms for the training data , and smoothed with a Gaussian kernel of width 32 ms . For the test data , we used a baseline of –150 to –50 ms and did not smooth . For calculating weights , we used a sliding window of 32 ms , moving in 12-ms steps . The results were smoothed with a 20-ms Gaussian kernel . Again , we used permutation testing to correct for multiple comparisons . In light of the rapid dynamics of the population tuning curve data , we reasoned that , while the exact neural pattern might differ between time points and task variables , the information represented ( as measured by their representational geometry ) might be more constant over time . We tested for this possibility with RSA . Specifically , our approach involved calculating neural dissimilarity matrices between the MEG/EEG topographies evoked by different stimulus orientations . For each session , we sorted all trials by the presented stimulus orientation ( into 16 bins ) , and then split each of these in half ( separating odd and even trials ) . The odd–even split allowed us to compare dissimilarity structures in two independent data sets , and to verify the reliability of the RSA . For each of the 32 bins , we calculated the baseline-corrected average evoked response ( across trials ) at all sensors and time points . Next , for each time point ( moving in 4-ms steps ) , we calculated the neural dissimilarity matrix by computing all pairwise Mahalanobis distances between orientations ( using the within-condition covariance , pooled over all conditions ) . We interpret these dissimilarity matrices as reflections of the representational structure at each time point in the epoch . In the first instance , we were simply interested in whether the neural dissimilarity structure was more stable over time than the underlying neural patterns ( that were calculated in the tuning curve analyses ) . To test for this , we correlated ( with Pearson correlations ) the dissimilarity matrix from one-half of trials at one time point with the dissimilarity matrix from the other half of trials at all time points , generating a cross-temporal matrix of correlations between dissimilarity structures . If the dissimilarity structure is stable over time , this should result in significant correlations ( measured via one-sample t-tests at the group level on the Fisher-transformed correlation coefficients ) between time points ( e . g . , off-diagonal coding ) . We repeated this analysis for the decision-relevant angular distance . For the analyses comparing the neural dissimilarity structures for template coding and stimulus coding , we used one-half of trials to calculate the 8×8 template-based dissimilarity matrix ( on data baselined at –200 to –150 ms , as above ) , and the other half of trials ( on data baselined at –150 to –50 ms , as above ) to calculate the 8×8 stimulus-based dissimilarity matrix ( using the eight stimulus orientations that also served as target orientations in that session ) . As above , the resulting within-time correlations were then smoothed with a 20-ms Gaussian kernel . Next , we asked whether the neural dissimilarity structure , or geometry , was related to the parametric dissimilarity structure of the stimuli: since a 45° angle is more similar to a 60° angle than a 90° angle , the corresponding MEG/EEG topographies might be more similar as well . The stimulus dissimilarity matrix based on the pairwise angular distances between all presented orientations was regressed against the MEG/EEG dissimilarity matrix using a general linear model , fitting the model separately for each time point , session , and participant . Significant fits were assessed via one-sample t-tests . As an illustration of the presence of circular structure in the representational geometry , we projected the 32×32 dissimilarity matrix into two dimensions using multi-dimensional scaling ( MDS ) . We repeated geometric analyses on the dissimilarity structure with respect to the template orientation , and the decision-relevant angular distance . For the latter ( angular distances ) , the MDS results indicated that the circular geometry of stimulus relationships was distorted by the decision likelihood . Therefore we used multiple regression to account for its possible influence ( pattern component modeling , Diedrichsen et al . , 2011 ) . A first nuisance regressor captured the differences in the absolute decision value , i . e . the distance between the unsigned angular distances ( i . e . the distance between 0° and –22 . 5° was 22 . 5° , but the distance between –22 . 5° and +22 . 5° was 0° , rather than 45° ) . The second nuisance regressor was based on the similarity in response likelihood , which we estimated by calculating the participant-wise differences in response frequency between all decision-relevant angular distances . This regressor reflected differences in response likelihood between orientations , accounting for any effect of motor preparation . This regressor accounted for the effect of linear decision value on the MEG pattern . This pattern was regressed out because it could reflect two possible decision mechanisms: population-based coding , as proposed here , or linear evidence accumulation ( Gold and Shadlen , 2007 ) . While the latter may still be at play in this task , we were specifically interested in dissociating the two coding mechanisms . A final analysis examined how the entire population of MEG/EEG sensors dynamically encodes different task variables relating to angular distance representation . This was done by representing population responses as trajectories in neural state space ( with each dimension representing a unique task variable ) . One approach , emulated here , has recently been described for populations of neural spike trains ( Mante et al . , 2013 ) . First , in order to de-noise the data , we smoothed data with a 20-ms Gaussian kernel and reduced the dimensionality of the MEG signal from 306 sensors ( or 306+60 sensors for MEG+EEG sessions ) to 30 principal components ( PCs ) by calculating coefficients over the average time series at each sensor . We then fit the task variables to the reduced-dimensionality data using a GLM . The regressors were derived from the three main task variables: stimulus orientation , template orientation , and angular distance . Since all three are circular variables , we used pairs of regressors , consisting of the sine and cosine of each task angle , yielding a design matrix consisting of six regressors in total . The fitting was done in a leave-one-block-out procedure: in turn , we held out all trials from one task block as a test set , and fit the GLM on the trials in the remaining seven blocks ( the training set ) . The GLM was solved on normalized data ( by subtracting the mean and dividing by the variance across all trials in the training set ) . This yielded a set of six regression coefficients ( ‘betas’ ) for each time point in the trial and for each of the 30 PCs , which were then symmetrically orthogonalized ( Colclough et al . , 2015; following Mante et al . , 2013 ) . After normalizing the data from the test set ( using the mean and variance from the training set ) , we calculated mean responses for all 16 angular distances in the test set ( yielding a 16 angles × 30 PCs matrix ) . The means ( 16×30 ) were then projected onto the task axes by multiplying them , time point by time point , with the betas ( 30 PCs × 6 regressors ) from the training set , creating a 16×6 matrix at each timepoint for each left-out block . We then averaged projections across the eight cross-validation folds . The resulting projections estimate the sensitivity of each condition ( i . e . , the 16 angular distances ) to each task variable ( i . e . , the six regressors ) , separately for each time point in the trial . In line with Mante and colleagues , we interpreted consistent deviations from 0 ( as measured by one-sample t-tests across observers ) , in either direction along an axis , as task variable sensitivity . In particular , the two regressors for the angular distance partialled out task-relevant and task-irrelevant aspects of the angular distance between stimulus and template: the cosine regressor , with a maximum of 1 at 0º ( targets ) , a minimum of –1 at the farthest non-targets ( ± 90º , in our 180º orientation space ) , and equal magnitudes for equivalent non-targets ( e . g . , 0 . 92 for both +11º and –11º ) , measured only the task-relevant aspect of the angle ( i . e . , the decision value , as shown in Figure 8—figure supplement 2 ) . By contrast , the sine regressor is insensitive to decision value ( since , in 180º orientation space , sin ( 0º ) = sin ( ± 90º ) = 0 ) , but distinguished between signed differences between non-targets ( e . g . , sin ( +11º ) = 0 . 38 = −sin ( –11º ) ) . To summarize our MEG results , and to illustrate how they could arise from a very simple decision circuit , we created a population-based neural coding model capable of performing the template-matching task used in our experiment . The model consisted of a three-layer architecture , with each layer consisting of neurons coding for different task variables ( a stimulus layer , a template layer , and a decision layer ) . The stimulus layer consisted of a set of 100 units , each tuned to a different veridical stimulus orientation , with tuning determined by a von Mises distribution: ( 5 ) Ri ( θ ) = exp ( κ*cos ( θ-θi ) ) /A; Ri ( θ ) indicates the response R of model unit i ( tuned to θi ) to orientation θ , with concentration parameter κ determining the tuning width of the response , and A reflecting a normalizing constant . Activation in the layer was then normalized to a range between 0 and 1 . In the stimulus layer , the concentration parameter was set to 5 ( see Beck et al . , 2008 , for the same parameter choice ) . The template layer was identical to the stimulus layer , with the exception that template tuning was broader ( κ = 2 ) , under the assumption that remembered stimuli would be encoded with lower precision than currently visible stimuli . Finally , the decision layer was identical to the template layer , with the conceptual difference that here , units were not tuned to veridical stimulus orientations , but to decision-relative orientations . In other words , activation of units tuned to near 0° in the decision layer reflected choice-relevant signals , irrespective of the current template orientation . The stimulus and template layers were connected via a one-to-one mapping between identically tuned units ( weight matrix WST , with all connections between non-identical orientations set to 0 ) . On each trial , the stimulus layer was initialized by setting the population response vector RS in accordance with the stimulus orientation , and the template layer response vector RT in accordance with the template orientation . In a second step , corresponding to a later processing stage , activation in the template layer was updated as a function of activation in the stimulus layer , by computing the element-wise product between RS and RT . This step is similar to a Bayesian update , in which the prior distribution ( the template layer response ) is multiplied with the current evidence ( the stimulus layer response ) to produce a posterior distribution . The crucial mapping for the task was between the template and decision layers , which consisted of an all-to-all reciprocal weight matrix WTD . The template layer unit tuned to the current target orientation had the strongest connection with the 0° unit in the decision layer ( and neighboring template layer units were connected to correspondingly shifted units in the decision layer ) . All other connection weights fell off according to a von Mises distribution with κ = 5 ( although the exact tuning width did not substantially alter model behavior ) . This weight matrix shifted the response profile RT in the template layer ( which was still in veridical orientation space ) to a response RD in decision space . The decision layer response therefore permitted a direct mapping to decision- or motor-related output regions ( which are omitted here ) . Importantly , only the weight matrix WTD needs to change in response to a change in the current template orientation . Since we were mainly interested in the effects of reading out population activity in the decision layer , this model contained the simplification that codes in the three layers did not change over time . However , the dynamics of coding were not of interest for the question of whether population activity , in principle , could account for our neural results . The model behavior ( Figure 8—figure supplement 3E–H ) followed a simple trajectory over the course of a hypothetical trial . At the beginning of the trial , before current stimulus input has been processed , the template layer encodes the current template layer , via a bump in activation in template-tuned neurons . This input can be instantiated in the template layer via top-down input from the 0° unit in the decision layer ( although other mechanisms for activating the template are also conceivable , such as periodic reactivation , e . g . Buzsáki and Moser , 2013; Eichenbaum , 2013; Johnson and Redish , 2007; Lisman and Jensen , 2013; Schroeder and Lakatos , 2009 ) . This activation of the template layer around the time of stimulus onset might correspond to the decoding profile for template information in the MEG data ( Figure 3C ) . Next , stimulus input is represented in the stimulus layer , again via a population activity profile peaking at neurons tuned to the currently presented stimulus ( again corresponding to the decoding profile , Figure 3A ) . Stimulus layer activation is then fed forward into the template layer , where activation is multiplied point-by-point with the existing activation state . This could happen , for example , if neurons in the template layer change their gain to all input depending on their proximity to the current target orientation ( Carandini and Heeger , 2012; McAdams and Maunsell , 1999; Reynolds and Heeger , 2009; Silver et al . , 2007; Treue and Trujillo , 1999 ) . The resulting activation profile now represents the stimulus orientation , scaled by its similarity to the template—while neurons tuned to the current orientation again have high activation , the height of that activation depends on stimulus-template similarity ( and the peak is shifted towards the template orientation , with the magnitude of the shift depending on the ratio of stimulus and template tuning widths ) . This representation of the template and the stimulus in the same population ( at slightly different but overlapping timepoints ) might reflect why stimulus and template geometries cross-generalize ( Figure 7G , H ) . By passing the template layer profile on to the decision layer , it is shifted into a decision-relative ( i . e . stimulus-invariant ) space . Here , the response exhibits two decision-relevant features . First , the closer the current stimulus is to the template , the closer the decision layer peak is to the 0° neuron . Second , as in the template layer , the height of the decision layer profile also depends on the proximity between stimulus and template , with highest activation for targets ( Figure 8—figure supplement 3B ) , and the peak dropping off for increasingly distant non-targets ( Figure 8—figure supplement 3C–E ) . This decreasing amplitude may explain why near-targets were more separable than definite non-targets later in the trial epoch ( Figure 7B ) . Downstream read-out units could use either the population’s peak location ( i . e . how close the maximum response is to the 0° unit , as in Beck et al . , 2008 ) or its peak activation ( as in more classical decision models ) to determine whether a target is present or absent . | Imagine searching for your house keys on a cluttered desk . Your eyes scan different items until they eventually find the keys you are looking for . How the brain represents an internal template of the target of your search ( the keys , in this example ) has been a much-debated topic in neuroscience for the past 30 years . Previous research has indicated that neurons specialized for detecting the sought-after object when it is in view are also pre-activated when we are seeking it . This would mean that these ‘template’ neurons are active the entire time that we are searching . Myers et al . recorded brain activity from human volunteers using a non-invasive technique called magnetoencephalography ( MEG ) as they tried to detect when a particular shape appeared on a computer screen . The patterns of brain activity could be analyzed to identify the template that observers had in mind , and to trace when it became active . This revealed that the template was only activated around the time when a target was likely to appear , after which the activation pattern quickly subsided again . Myers et al . also found that holding a template in mind largely activated different groups of neurons to those activated when seeing the same shape appear on a computer screen . This is contrary to the idea that the same cells are responsible both for maintaining a template and for perceiving its presence in our surroundings . The brief activation of the template suggests that templates may come online mainly to filter new sensory evidence to detect targets . This mechanism could be advantageous because it lowers the amount of neural activity ( and hence energy ) needed for the task . Although this points to a more efficient way in which the brain searches for targets , these findings need to be replicated using other methods and task settings to confirm whether the brain generally uses templates in this way . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Testing sensory evidence against mnemonic templates |
Electroencephalogram ( EEG ) approaches may provide important information about developmental changes in brain-state dynamics during general anesthesia . We used multi-electrode EEG , analyzed with multitaper spectral methods and video recording of body movement to characterize the spatio-temporal dynamics of brain activity in 36 infants 0–6 months old when awake , and during maintenance of and emergence from sevoflurane general anesthesia . During maintenance: ( 1 ) slow-delta oscillations were present in all ages; ( 2 ) theta and alpha oscillations emerged around 4 months; ( 3 ) unlike adults , all infants lacked frontal alpha predominance and coherence . Alpha power was greatest during maintenance , compared to awake and emergence in infants at 4–6 months . During emergence , theta and alpha power decreased with decreasing sevoflurane concentration in infants at 4–6 months . These EEG dynamic differences are likely due to developmental factors including regional differences in synaptogenesis , glucose metabolism , and myelination across the cortex . We demonstrate the need to apply age-adjusted analytic approaches to develop neurophysiologic-based strategies for pediatric anesthetic state monitoring .
In the United States , 200 , 000 children a year receive general anesthesia during the first year of life ( Rabbitts et al . , 2010 ) . Anesthetics have profound effects on all physiological systems . As a consequence , since the mid-1980s , anesthesia caregivers have been required to monitor blood pressure , heart rate , body temperature , and oxygen saturation along with anesthetic gas and oxygen delivery for all patients receiving general anesthesia . The states of general anesthesia and sedation are produced by the anesthetics acting in the brain and spinal cord ( Brown et al . , 2010 , 2011; Ching and Brown , 2014 ) . However , brain monitoring using electroencephalogram ( EEG ) -derived indices is used to only a limited extent in adults , rarely in children , and essentially not at all in infants . These EEG-derived indices , which been developed in adults can give inaccurate indications of anesthetic states in infants and younger children ( Davidson , 2007; Constant and Sabourdin , 2012 ) . The lack of principled strategies for monitoring the brains of infants and children receiving anesthesia care is especially troubling in view of growing concern about anesthetic toxicity to the developing brain ( McCann and Soriano , 2012; Jevtovic-Todorovic et al . , 2013; Lin et al . , 2014 ) . A plausible inference to be drawn from the few available EEG studies of children under general anesthesia is that the inaccuracy of EEG-derived indices in pediatric practice is likely due to differences between children and adults in their brain responses to the anesthetics . A consensus has not been achieved on what these differences are because the available pediatric investigations have studied a limited set of anesthetics and frequently used EEG montages with few electrodes; different analysis methods have been used in different studies; multi-electrode recordings available in children have not been analyzed in function of age; and dose-titration experiments commonly conducted in adults cannot for ethical reasons be conducted in children ( Davidson et al . , 2008; Lo et al . , 2009; Hayashi et al . , 2012; McKeever et al . , 2012; Sury et al . , 2014 ) . For adults , characterizing the anesthetic response of the brain under general anesthesia and sedation is an active research field in which EEG , intracranial recordings , functional magnetic resonance imaging , and positron emission tomography are being used to study spatio-temporal changes in brain activity for different anesthetics and for different doses of the same anesthetic ( Feshchenko et al . , 2004; Boveroux et al . , 2010; Breshears et al . , 2010; Mhuircheartaigh et al . , 2010; Cimenser et al . , 2011; Martuzzi et al . , 2011; Murphy et al . , 2011; Boly et al . , 2012; Lewis et al . , 2012; Casali et al . , 2013; Lee et al . , 2013; Liu et al . , 2013; Ní Mhuircheartaigh et al . , 2013; Purdon et al . , 2013; Akeju et al . , 2014a; Vizuete et al . , 2014 ) . For example , a significant EEG pattern observed in adults under propofol general anesthesia and anesthesia maintained by an ether-derived anesthetic is incoherent slow oscillations across the entire head and strongly coherent alpha oscillations across only the front of the head ( Purdon et al . , 2013 ) . Modeling studies suggest that the alpha oscillations are thalamo-cortical in origin ( Vijayan et al . , 2013 ) . In addition , it is now appreciated that different anesthetics have different EEG signatures that are readily visible in the unprocessed EEG and its spectrogram ( Purdon et al . , 2013; Akeju et al . , 2014b , 2014c ) . These EEG signatures can be related to the altered arousal states the anesthetics produce and to the mechanisms through which the drugs are believed to act at specific receptors in specific brain circuits . These neurophysiological studies in adults have led to the idea of using the unprocessed EEG and its spectrogram as an alternative to using EEG-derived indices to monitor the brain states of adults receiving general anesthesia or sedation ( Brown et al . , 2015 ) . Use of the unprocessed EEG and its spectrogram to develop a neurophysiological-based paradigm tailored to pediatric practices for monitoring anesthetic state requires studying in children the spatio-temporal dynamics of brain responses to anesthetics . To begin to address this key knowledge gap , we used multi-electrode EEG recordings to study the spatial and temporal dynamics of brain activity in infants 0–6 months of age when awake , and during maintenance of and emergence from sevoflurane general anesthesia administered for routine surgical care .
We evaluated the EEG features during the awake state ( prior to anesthesia exposure ) in infants whom awake EEG could be recorded . This was 7 infants in the 0–3 month age group and 12 infants in the 4–6 month age group . Body movement and/or eye opening prior to exposure of sevoflurane general anesthesia were used to confirm the presence of the awake state . Frontal group-median spectrograms at F7 showed dominant slow and delta power , with low theta and alpha power in all ages , were the predominant frequencies during the awake state at all ages ( Figure 10—figure supplement 1A–C ) . Power across all frequencies was significantly lower in infants at 0–3 months compared to infants at 4–6 months of age ( 95% CI , bootstrap analysis; Figure 10—figure supplement 1D–G ) .
Age-related differences in EEG properties have been shown in previous studies of infants and children receiving general anesthesia . Davidson and colleagues monitored 8 leads of EEG in 17 infants , 0–6 months of age , who received either sevoflurane or propofol general anesthesia ( Davidson et al . , 2008 ) . These infants were part of a study cohort that ranged in age from 9 days to 12 years . While total spectral power—limited to 2–20 Hz—did not differ between the anesthetized state and emergence , the EEG traces of the infants during emergence showed bursts of electrical activity interspersed with periods of low-amplitude activity . In contrast , the 21 children aged 2–10 years showed a significant increase in SEF90 and a significant decrease in total frontal EEG power during emergence . Lo and colleagues studied emergence by monitoring 128 leads of EEG in children 22 days to 3 . 6 years who received either sevoflurane or isoflurane anesthesia ( Lo et al . , 2009 ) . These investigators reported differences between the spatial EEG patterns of the two agents . However , none of the analyses was stratified by age . Hayashi and colleagues conducted a retrospective study on EEG data collected using a 2-electrode montage in 62 neonates and infants ranging in age from 1 day to 2 years who received sevoflurane general anesthesia ( Hayashi et al . , 2012 ) . They analyzed changes in 90% spectral edge frequency ( SEF90 ) , burst suppression ratio ( BSR ) , relative beta ratio ( RBR ) , and approximate entropy ( ApEn ) to assess changes in brain activity during emergence . They found that infants less than 6 months had little to no change in EEG parameters during emergence , whereas the infants 6 months and older showed the greatest changes in the EEG parameters , which were an increase in the SEF90 , decrease in the BSR , increase in the RBR , and decrease in ApEn . The infants in the range of 3–5 month old marked the transition between no EEG changes on emergence to highly visible changes on emergence . Sury and colleagues studied 20 infants at 1 week to 10 months of age during emergence from sevoflurane general anesthesia using a 4-lead EEG montage to analyze spectral power between 1 and 28 Hz ( Sury et al . , 2014 ) . They found that infants 3 months and older had greater power in the 5–20 Hz range during maintenance than infants less than 3 months and that the power in the older infants decreased on emergence . Our results refine the results in these reports by showing that there are readily discernible differences in the anesthetic responses of infant 0–3 months of age and infants 4–6 months of age . We report for the first time detailed spatio-temporal analyses of infant 33-lead EEGs with Laplacian referencing using spectra ( analyzed from 0 . 1–30 Hz ) , spectrograms , coherograms , and global coherence computed by multitaper spectral analysis methods . These methods , which are known to have several important optimality properties , have been used successfully to characterize the EEG dynamics of adults receiving general anesthesia and sedation ( Cimenser et al . , 2011; Ní Mhuircheartaigh et al . , 2013; Purdon et al . , 2013; Akeju et al . , 2014b , 2014c ) . Moreover , the spectrograms are particularly important because they are the frequency domain representations of the EEG displayed on commercially available EEG monitors . It is now appreciated that in adults at surgical levels of general anesthesia maintained by GABAergic anesthetics ( ether-based agents , barbiturates , propofol ) , the EEG shows slow oscillations and alpha oscillations , in addition to anteriorization , defined as the movement predominantly of the EEG power in the alpha band from the occipital to the frontal area of the head ( Feshchenko et al . , 2004; Cimenser et al . , 2011; Murphy et al . , 2011; Ní Mhuircheartaigh et al . , 2013; Purdon et al . , 2013; Akeju et al . , 2014b , 2014c ) . Studies of propofol have shown that these alpha oscillations are strongly coherent across the front of the head during unconsciousness , whereas the slow oscillations are not coherent anywhere across the head ( Lewis et al . , 2012; Purdon et al . , 2013 ) . With return of consciousness , the frontal alpha power disappears , and posterior alpha reappears ( Purdon et al . , 2013 ) . Because we did not record EEG in the infants continuously and then administer sevoflurane in a gradual escalating and a gradual deescalating manner , our study could not assess anteriorization . During MOSSA , we see a small , but significant , increase in frontal alpha power when compared to occipital power in infants 4–6 months , which suggests this feature of anesthetic-induced consciousness emerges in early life . Nevertheless , our findings demonstrate that the spatio-temporal dynamics of the anesthetized states of 0- to 3-month-old and 4- to 6-month-old infants differ appreciably from each other and from those of adults . Recent modeling studies of propofol offer insights into the possible origins of these differences . These studies suggest that the frontal EEG alpha oscillations likely represent oscillatory activity between the cortex and thalamus ( Ching et al . , 2010 ) . Other modeling work suggests that anteriorization is due to different effects of propofol on thalamic nuclei that project to occipital areas compared to thalamic nuclei that project to areas of the pre-frontal cortex ( Vijayan et al . , 2013 ) . Propofol disrupts the normal , depolarized alpha oscillations in posterior-projecting thalamic nuclei whereas it induces , hyperpolarized alpha oscillations in frontal-projecting thalamic nuclei . The differential effect appears to be due primarily to propofol's inhibition of subset of posterior-projecting neurons with hyperpolarizing activating Ih currents that typically produce alpha oscillations at resting membrane voltages above 60 millivolts ( Vijayan et al . , 2013 ) . The slow and delta oscillations during general anesthesia are consistent with anesthetic-induced decreases in major excitatory brainstem inputs to the cortex . GABAergic anesthetic likely produce these effects by acting at the GABAergic synapses arising from inhibitory neurons in the pre-optic area of the hypothalamus that project onto each of the major arousal nuclei in the midbrain and upper pons ( Brown et al . , 2011 ) . Both the alpha , slow and delta oscillations likely reflect disruption of thalamo-cortical and cortico-cortical processing required to maintain consciousness ( Ching et al . , 2010; Purdon et al . , 2013 ) . The model predictions for the frontal predominance of alpha oscillations await experimental verification . However , the neuroanatomy and neurophysiology underlying the putative mechanism for alpha oscillations suggest that the thalamocortical connections required to produce the alpha oscillations are absent in infants 0–3 months old yet , present in infants 4–6 months old . The putative mechanism for frontal predominance of alpha power suggests that the differential thalamic connectivity required to produce this phenomenon is not present in infants 6 months of age or younger . Developmental changes in the brain are the obvious mechanisms to explain the changes in EEG dynamics that occur with age . Gross brain development occurs in a caudal to rostral direction , with myelination of the medulla , pons , and thalamus starting within the first few postnatal weeks and frontal cortex myelination starting around postnatal months 3–4 ( Brody et al . , 1987; Kinney et al . , 1988 ) . In addition , there is significant synaptogenesis , neuronal differentiation , and pruning , along with changes in GABAergic neurotransmission ( Huttenlocher and Dabholkar , 1997; Hensch , 2004; Tau and Peterson , 2010; Dehorter et al . , 2012; Catts et al . , 2013; Semple et al . , 2013 ) . Distinct regional differences in the rate of synaptogenesis , glucose metabolism , and myelination across the cortex occur between subcortical and cortical regions , and between different regions of the cortex during the first 12 postnatal months in human infants . For example , in the visual cortex , there is a rapid burst of synapse formation between 3 and 4 months , and the maximum density is reached between 8 and 12 months ( Huttenlocher et al . , 1982 ) . Synaptogenesis starts at the same time in the pre-frontal cortex , but the density of synapses increases much more slowly and does not peak until after the first 12 months ( Johnson , 2001 ) . Additionally , Positron Emission Tomography ( PET ) studies show that between 0 and 3 months postnatal age , glucose uptake is highest in the occipital parietal , and temporal cortices; and by 6–8 months , glucose uptake extends to the frontal cortex appearing with higher cortical function ( Chugani and Phelps , 1986; Chugani et al . , 1987; Kinnala et al . , 1996 ) . A key role in brain development is played by the subplate neurons , the first neurons generated in the cerebral cortex , which guide formation of thalamocortical connections ( Kanold and Luhmann , 2010; Kostović and Judas , 2010 ) . The subplate cells form the first functional connections and are a must for relaying early oscillatory activity in the developing brain ( Kanold and Luhmann , 2010 ) . To the extent that the alpha oscillations in the anesthetized brain are postulated to be produced by thalamocortical circuits , the appearance of the alpha oscillations at 4 months of age may suggest that an important developmental milestone has been reached in the processes guided by the subplate neurons . The conduct of research in children carries with it all the ethical obligations of adult research along with the additional obligation of not exposing children to risks beyond those associated with their routine medical care . As a consequence , observational studies of children receiving general anesthesia or sedation as part of routine diagnostic or therapeutic care are and will continue to be the principal approach to studying the neurophysiology of general anesthesia in children . It is crucial to plan carefully these studies and their subsequent analyses in order to maximize the information learned on this important topic . Anesthetic management was only standardized in our study to the extent that the staff at Boston Children's Hospital uses similar practices . The specific anesthetic management in each case was carried out by the attending anesthesiologist . Given the strength of the findings we report , a more systematic protocol would likely provide more evidence for our findings . Because administration of general anesthesia is a high-risk human study , dose-titration protocols that are common in adults for studying controlled induction and emergence cannot be conducted in children . Therefore , we conducted our dose–response analysis ( Figure 11 ) , like those reported in previous studies , by observing changes in end-tidal sevoflurane concentration during emergence instead of by recording EEG while systematically changing the anesthetic dose during induction . In place of formal behavioral assessments of consciousness , which would not be feasible in children , we used movement recorded on video as an approximate behavioral marker of emergence from general anesthesia . The advantage of using body movement as a measure of emergence is that it is well-defined in children , and video recordings of body movement can be time-locked to the EEG recording for quantitative analysis . Relating onset of emergence to changes in respiration and heart rate , as well as blood pressure , also reflect effects of anesthesia but in a more complex manner being dependent on drug administration and clinical state of the patient . Similar to what we observed in adults , we found concentration- and behavior-dependent reductions in alpha and theta power during emergence in 4- to 6-month-old infants . Unlike in adults , we did not observe any relationship between slow/delta power and anesthetic concentration or behavior . The differences in slow/delta oscillation dynamics between these infants and adults may be attributable to a number of factors , including mechanistic differences the two groups , low signal-to-noise in the slow/delta band due to patient movement , and the possibility that the infants remained sedated or unconscious despite recovering movement immediately following surgery . Our comparisons between the awake state , MOSSA and emergence indicate that state-specific EEG spectral properties begin to emerge at 4–6 months of age . They demonstrate the challenges faced when using EEG measures to evaluate anesthetic depth in younger infants ( 0–3 months ) . Our findings can be strengthened by studying sevoflurane general anesthesia in children across the entire pediatric age spectrum , and by conducting similar studies of other the anesthetics , for example , propofol , dexmedetomidine , and isoflurane , commonly used in children . In summary , we have shown that infants 0–6 months of age have markedly different EEG patterns from each other and from adults under general anesthesia . These differences are likely due to differences in structural and functional aspects of cortico-cortical and thalamocortical connectivity , and help explain why EEG-based indices provide inaccurate measures of anesthetic states in children , especially during the first three months of life . We introduced the use of multitaper spectral methods in the analysis of pediatric EEG recordings to facilitate comparisons with adult analyses . We provide spectrograms to show how these brain dynamics appear on available EEG monitors . The design of strategies to track the brain states of children receiving general anesthesia and sedation has not received the attention that this topic has received in adults . Systems neuroscience research that takes account of brain development will be required to accurately define anesthetic states for the entire pediatric age spectrum and to devise principled neurophysiological-based strategies for anesthetic dosing in older infants ( >3 months ) and children . Moreover , until the question of whether anesthetics are toxic to the developing brain is answered , design of neurophysiological-based definitions of anesthetic states and design of neurophysiological-based brain monitoring strategies offer the most prudent approaches to mitigating anesthetic risk in this vulnerable population .
The objective of this observational study was to evaluate the effect of postnatal age on electroencephalographic ( EEG ) activity during sevoflurane general anesthesia in infants 0–6 months old . We recorded multichannel EEGs during administration of sevoflurane general anesthesia for elective surgery , per clinical protocol . End-tidal anesthetic gas volume and video recordings of behavioral activity were time-locked to the EEG recording . The spatial and temporal properties of the infant EEG were evaluated during the awake state , and at two distinct periods during administration of sevoflurane general anesthesia: ( 1 ) MOSSA and ( 2 ) emergence from sevoflurane general anesthesia ( Figure 1 ) . The age-dependent effects of sevoflurane general anesthesia on the EEG were compared in two groups of infants who were ( i ) 0–3 months and ( ii ) 4–6 months of age . All infants were in the supine position throughout the study . Each infant was studied once . Behavioral data during ( 1 ) the awake state and ( 2 ) emergence from general anesthesia were analyzed post hoc . ( 1 ) For awake analysis: videos were reviewed frame-by-frame prior to anesthesia to identify the sleep state of each infant . The sleep state of the infant was determined using behavioral and vocal markers . The awake state was defined as from one of three features ( a ) eyes open , ( b ) body movement , and/or ( c ) crying . ( 2 ) For emergence analysis: videos were reviewed frame-by-frame to identify the time point ( in seconds ) where gross body movement first occurred . The corresponding end-tidal sevoflurane concentration was extracted from the ixTrend data recording . The percentage of infants who displayed gross body movement in each end-tidal sevoflurane concentration range was evaluated for each group . Data are shown as median ( 95% CI of median ) unless otherwise stated . Statistical analysis performed using SPSS Statistics v . 21 ( IBM , Armonk , NY ) and custom-written MATLAB code ( MathWorks Inc . ; Source code 2 ) . To assess statistical significance for the difference in power at each frequency , we computed the 95% CI by using a frequency domain-based bootstrapping algorithm ( Kirch and Politis , 2011 ) . We drew Fourier coefficients from normal Gaussian distribution with variance of its spectral power for each subject . From the Fourier coefficients , we computed replicates of spectral power for each subject and took the median value ( i . e . , power , or coherence , where relevant ) across infants within each postnatal age group and computed group differences ( using paired comparisons , where relevant ) , following this a new set was randomly selected in each postnatal age group and the analysis repeated . We repeated this 2000 times and calculated the 95% CI using the median difference at each frequency . Custom-written MATLAB code ( with simulated data ) for computing global coherence ( Source code 1 ) , and multitaper spectra and bootstrap CIs ( Source code 2 ) is given in the Supplemental Materials . | Every year about 200 , 000 infants in the United States are given general anesthesia during their first year of life . Though anesthesia is essential to control pain during surgery and other medical procedures on infants , it involves some risks . There are some controversial studies suggesting that repeated anesthetics early in life may impact how the brain develops , but other studies have been reassuring and found no such effects . To reduce the risks , doctors carefully monitor infants' blood pressure , heart rate , body temperature , and oxygen levels while they are receiving anesthesia . Electroencephalograms ( EEGs ) have proven to be a useful tool for monitoring the brain activity of adults undergoing anesthesia , but studies have found EEG-based monitoring to be unreliable in infants under anesthesia . A more reliable method of monitoring the brains of infants during anesthesia is needed . Anesthesiologists nevertheless need to better understand how the infant's brain works under general anesthesia , and novel EEG techniques hold promise for monitoring brain well-being and for adjusting anesthetic dosing in infants of different ages . Differences in the way infants' brains respond to anesthesia may explain why current EEG-based monitoring methods developed for adults don't work as well in infants as in adults . Now , Cornelissen , Kim et al . have used a new EEG-based approach to demonstrate that as infants' brains develop , their responses to anesthesia change . The experiments involved 36 infants aged up to six months old who were going through routine surgical procedures . The brain activity of the infants was recorded using EEG—via electrodes placed on their scalps—when they were awake , during anesthesia , and as they recovered afterwards . The infants were videoed at the same time . Comparing the video with the EEG recordings allowed the brain activity of the infants to be matched up with their state of consciousness . Cornelissen , Kim et al . detected slow waves of brain activity across the entire scalp of infants who are under six months old and under anesthesia . Infants who are older than about four months old also display some faster brain waves , which decreased in power as the infants emerged from anesthesia . However , none of the infants has the same pattern seen in adults—where faster waves appear near the front of the brain . These findings may help scientists develop more reliable ways to monitor infants' brains during anesthesia . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"neuroscience"
] | 2015 | Age-dependent electroencephalogram (EEG) patterns during sevoflurane general anesthesia in infants |
Despite the critical role of endothelial Wnt/β-catenin signaling during central nervous system ( CNS ) vascularization , how endothelial cells sense and respond to specific Wnt ligands and what aspects of the multistep process of intra-cerebral blood vessel morphogenesis are controlled by these angiogenic signals remain poorly understood . We addressed these questions at single-cell resolution in zebrafish embryos . We identify the GPI-anchored MMP inhibitor Reck and the adhesion GPCR Gpr124 as integral components of a Wnt7a/Wnt7b-specific signaling complex required for brain angiogenesis and dorsal root ganglia neurogenesis . We further show that this atypical Wnt/β-catenin signaling pathway selectively controls endothelial tip cell function and hence , that mosaic restoration of single wild-type tip cells in Wnt/β-catenin-deficient perineural vessels is sufficient to initiate the formation of CNS vessels . Our results identify molecular determinants of ligand specificity of Wnt/β-catenin signaling and provide evidence for organ-specific control of vascular invasion through tight modulation of tip cell function .
Endothelial cells ( ECs ) acquire organ-specific characteristics to adapt to the requirements of their host tissues . The central nervous system ( CNS ) vascular microenvironment serves as a paradigm for blood vessel specialization because CNS ECs develop a set of junctional , cellular trafficking , and metabolic properties , collectively called the blood–brain barrier ( BBB ) , that protect the CNS from blood-borne toxins and pathogens . Brain angiogenesis and barriergenesis are temporally coupled through a distinct , and tissue-specific , developmental program ( Obermeier et al . , 2013; Engelhardt and Liebner , 2014; Vallon et al . , 2014 ) . The best characterized class of angiogenic and BBB-inductive signals operates through the Wnt/β-catenin pathway ( canonical Wnt signaling ) ( Xu et al . , 2004; Liebner et al . , 2008; Stenman et al . , 2008; Daneman et al . , 2009; Ye et al . , 2009 ) , with distinct sets of ligands , receptors , and co-receptors controlling vascular development in different CNS locations . For example , in the embryonic forebrain and ventral spinal cord , neural progenitor-derived Wnt7a and Wnt7b activate Wnt/β-catenin signaling in ECs to control both angiogenesis and BBB formation ( Liebner et al . , 2008; Stenman et al . , 2008; Daneman et al . , 2009 ) . In the retina , the Muller glia-derived ligand Norrin , in conjunction with the receptor Frizzled4 ( Fz4 ) , co-receptor Lrp5 , and co-activator Tspan12 , mediate Wnt/β-catenin signaling to control angiogenesis and blood-retina barrier ( BRB ) formation and maintenance ( Xu et al . , 2004; Junge et al . , 2009; Ye et al . , 2009 ) . In addition to classical components of the Wnt/β-catenin pathway , such as Frizzled receptors and Lrp5/Lrp6 co-receptors ( Zhou et al . , 2014 ) , recent evidence indicates that a unique signal transduction complex containing Gpr124 , an orphan receptor of the adhesion GPCR family , operates specifically in CNS ECs to promote Wnt7a and Wnt7b angiogenic signaling ( Kuhnert et al . , 2010; Anderson et al . , 2011; Cullen et al . , 2011; Zhou and Nathans , 2014; Posokhova et al . , 2015 ) . In mouse embryos , eliminating neuroepithelial Wnt7a and Wnt7b , or endothelial Gpr124 or β-catenin , leads to reduced CNS angiogenesis with production of abnormal vascular structures , termed glomeruloids , that fail to acquire BBB characteristics ( Liebner et al . , 2008; Stenman et al . , 2008; Daneman et al . , 2009; Kuhnert et al . , 2010; Anderson et al . , 2011; Cullen et al . , 2011 ) . Whether these vascular malformations are exclusively the result of defective endothelial Wnt/β-catenin signaling or are also influenced by signals from hypoxic tissues has not been determined ( Sundberg et al . , 2001; Cullen et al . , 2011 ) . At present , the cellular mechanisms by which Wnt/β-catenin signaling controls the complex and multistep process of CNS blood vessel formation remain largely unexplored . Here , we leverage the zebrafish model to study the molecular machinery governing Wnt-dependent brain angiogenesis using a combination of targeted mutagenesis , morpholino knock-downs , RNA injections , genetic mosaics , and single-cell resolution real-time imaging . We find that EC-specific and Gpr124-dependent Wnt/β-catenin signaling is required for angiogenic sprouting throughout the zebrafish brain . In addition , we identify Reck ( reversion-inducing-cysteine-rich protein with Kazal motifs ) , a GPI-anchored MMP inhibitor and angiogenic modulator ( Oh et al . , 2001 ) , as a novel and essential activator of Wnt/β-catenin signaling during CNS angiogenesis , and we show that Gpr124 and Reck physically interact and strongly synergize in mammalian cells to promote Wnt/β-catenin signaling exclusively via Wnt7a and Wnt7b . Finally , by using live imaging of genetically mosaic animals , we have discovered a tip cell-autonomous requirement for Gpr124- and Reck-dependent Wnt/β-catenin signaling during sprouting angiogenesis in the CNS . These experiments demonstrate that Wnt/β-catenin signaling specifically regulates tip cell function and reveal that coordination of tip and stalk cell behaviors within nascent vessels and organ-specific specialization , generally viewed as distinct aspects of vascular development , can in fact be tightly coupled .
To examine the function of Gpr124 during zebrafish development , we generated two gpr124 mutant alleles using TAL effector nucleases ( Cermak et al . , 2011; Dahlem et al . , 2012 ) . The TALEN pairs were directed towards sequences within exons 7 and 16 , corresponding to the Ig-like domain and second transmembrane helix , respectively ( Figure 1A ) . We identified frame-shift mutant alleles , gpr124s984 and gpr124s985 , which lead to premature stop codons after 10 and 39 amino acid-long missense segments following the lesion site ( Figure 1—figure supplement 1 ) . Heterozygous carriers of either mutant allele display no obvious anatomical or behavioral phenotype . Homozygous gpr124s984 and gpr124s985 mutants , although morphologically indistinguishable from wild-type siblings ( Figure 1B ) , exhibit specific and highly penetrant brain vascular defects ( Figure 1C , D ) . The initial assembly of the perineural vessels is unaffected in the absence of Gpr124 . Between 28 and 32 hpf ( hours post fertilization ) , the paired ventro-lateral primordial hindbrain channels ( PHBC ) and primordial midbrain channels ( PMBC ) that extend along the rostro–caudal axis , establish wild-type-like connections with the medial basilar artery ( BA ) and the more rostral V-shaped posterior communicating segments ( PCS ) . 10 . 7554/eLife . 06489 . 003Figure 1 . CNS vascular defects in gpr124 mutants . ( A ) Schematic representation of Gpr124 structure and TALEN target site locations corresponding to gpr124s984 and gpr124s985 alleles . LRR: leucine-rich repeats; LRRCT: leucine-rich repeat C-terminal domain; Ig: Ig-like domain; HBD: hormone binding domain; GAIN: GPCR-autoproteolysis inducing domain; GPS: GPCR proteolysis site; PBD: PDZ binding domain . ( B ) Lateral views of wild-type , gpr124s984/s984 and gpr124s985/s985 larvae at 5 dpf . ( C ) Lateral views of wild-type , gpr124s984/s984 and gpr124s985/s985 Tg ( kdrl:GFP ) embryos at 36 hpf ( hindbrain region , upper panels ) and 24 hpf ( trunk region , bottom panels ) . MCeV: middle cerebral vein . Scale bar , 50 µm . ( D ) Maximal intensity projection of a confocal z-stack of the cranial vasculature of Tg ( kdrl:GFP ) wild-type and gpr124s984/s984 embryos at 60 hpf in dorsal views ( anterior to the left ) and wire diagram of the brain vasculature in lateral ( middle panels ) and dorso-lateral ( bottom panels ) views . Red vessels in the 3D renderings represent the intra-cerebral central arteries ( CtAs ) , blue vessels represent the extra-cerebral connections between the PHBC and BA lining the hindbrain ventrally , and gray vessels represent the perineural vessels ( PHBC , PMBC , BA , and PCS ) to which the central arteries connect in wild-type embryos . Scale bar , 100 μm . ( E ) Quantification of hindbrain CtAs upon Gpr124 depletion in 60 hpf embryos . ( F ) Quantification of hindbrain CtAs in control and gpr124 morphants at 60 hpf after injection at the one-cell stage of 100 pg RNA encoding the depicted receptors or Gpr124/Gpr125 hybrid receptors . ( G ) Vasculature of wild-type and gpr124 mutant adults . Single plane confocal image of the vascular network ( upper panels: scale bar , 100 µm ) and immunostaining for Slc2a1 and Pgp in sections through the optic tectum ( middle panels: scale bars , 20 µm ) . Evaluation of the optic tectum and liver vessel permeability by fluorescent streptavidin labelling ( red signal ) 60 min after intracardial injection of sulfo-NHS-biotin in live animals ( bottom panels; scale bar , 20 µm ) . In all panels , values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; Kruskal–Wallis test ) . Morpholino and RNA injections were performed as described in ‘Methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 00310 . 7554/eLife . 06489 . 004Figure 1—figure supplement 1 . Generation of gpr124 mutant zebrafish . ( A ) A TALEN was designed to target sequences within exon 7 ( upper panel ) corresponding to the Ig-like domain ( right panel ) . The left and right TAL repeats were linked to FokI DD and FokI RR , respectively . The spacer sequences are highlighted in yellow . The bottom panel shows the sequence alignment of the wild-type allele and the TALEN-generated gpr124s984 allele . The gpr124s984 allele contains an indel mutation ( −5 +15 [red] ) leading to a premature stop codon after a 10 amino acid-long missense segment ( in bold ) . ( B ) A TALEN was designed to target sequences within exon 16 ( upper panel ) corresponding to the second transmembrane helix ( right panel ) . The left and right TAL repeats were linked to FokI DD and FokI RR , respectively . The spacer sequences are highlighted in yellow . The bottom panel shows the sequence alignment of the wild-type allele and the TALEN-generated gpr124s985 allele . The gpr124s985 allele contains a two-nucleotide deletion ( red ) leading to a premature stop codon after a 39 amino acid-long missense segment ( in bold ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 00410 . 7554/eLife . 06489 . 005Figure 1—figure supplement 2 . gpr124 mutant survival curve and vascular phenotypes at 5 and 14 dpf . ( A ) Maximal intensity projection of a confocal z-stack of Tg ( kdrl:GFP ) wild-type and gpr124s984/s984 cranial vasculature at 5 dpf in dorsal views color-coded by depth . Dorsal most vessels are blue , ventral most vessels are red . Scale bar , 100 µm . ( B ) DIC images ( left panel ) and maximal intensity projection of a confocal z-stack of Tg ( kdrl:GFP ) wild-type and gpr124s984/s984 cranial vasculature at 14 dpf in dorsal views color-coded by depth ( right panels ) . Dorsal most vessels are blue , ventral most vessels are red . Scale bar , 100 µm . ( C ) Survival curves of wild-type and gpr124s984/s984 mutants . The survival of respectively 34 and 26 animals was recorded . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 005 From 32 hpf in wild-type embryos , the intracerebral central arteries ( CtAs ) begin to form by angiogenic sprouting from the dorsal wall of the PHBCs . These sprouts progress dorso-medially into the neural tissues to connect with the basilar artery and by 36 hpf , an average of four sprouts per hindbrain hemisphere have formed ( Bussmann et al . , 2011; Fujita et al . , 2011; Ulrich et al . , 2011 ) . In contrast , gpr124 mutants completely lack these forming vessels ( Figure 1C ) . Defects in CNS vascularization are fully penetrant in gpr124 mutants and at 60 hpf the entire brain remains avascular , while intersegmental vessel ( ISV ) sprouting from the dorsal aorta appears to be unaffected ( Figure 1C , D , E , Videos 1 and 2 ) . Injection of an anti-sense morpholino targeting the splice donor site of gpr124 exon 6 dose-dependently mimicked the mutant vascular phenotype ( Figure 1E ) and both the mutant and morphant phenotypes could be partially rescued by the injection of RNA encoding the full-length receptor ( Figure 1F ) . Asymmetric or unilateral rescues are sometimes observed ( Figure 2G ) , possibly as a result of uneven distribution of the injected RNA within the yolk cell . Gpr125 , a closely related adhesion GPCR , did not demonstrate angiogenic potential in a similar rescue assay . We mapped the functional differences between the receptors to components of the extracellular domain ( Figure 1F ) . 10 . 7554/eLife . 06489 . 006Video 1 . Wild-type cerebral vasculature at 60 hpf . Three-dimensional rotation of a wire diagram representation ( Imaris; Bitplane ) of the wild-type zebrafish cerebral vasculature at 60 hpf . Red vessels in the 3D renderings represent the intra-cerebral central arteries ( CtAs ) , and gray vessels represent the perineural vessels ( PHBC , PMBC , BA , and PCS ) to which the central arteries connect . Scale bar , 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 00610 . 7554/eLife . 06489 . 007Video 2 . gpr124s984/s984 cerebral vasculature at 60 hpf . Three-dimensional rotation of a wire diagram representation ( Imaris; Bitplane ) of the gpr124s984/s984 cerebral vasculature at 60 hpf . Blue vessels represent the extra-cerebral connections between the PHBC and BA lining the hindbrain ventrally , and gray vessels represent the perineural vessels ( PHBC , PMBC , BA , and PCS ) . Scale bar , 100 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 00710 . 7554/eLife . 06489 . 008Figure 2 . Essential regulation of CNS angiogenesis by Reck . ( A ) Lateral views of skin pigmentation patterns of wild-type and gpr124s984/s984 adults . Bottom panels are high magnification images of the upper panel boxed areas . ( B ) Dorsal views of DRGs in the trunk region of wild-type and gpr124s984/s984 Tg ( ngn1:GFP ) ; Tg ( sox10:mRPF ) larvae at 72 hpf; arrows point to DRGs ( anterior to the left ) . Scale bar , 50 µm . ( C ) Quantification of ngn1:GFP+ DRGs in 72 hpf wild-type , gpr124s984/+ and gpr124s984/s984 larvae . ngn1:GFP+ DRGs were counted on one side of the larvae . ( D ) Lateral views of DRG in Tg ( ngn1:GFP ) ;Tg ( sox10:mRPF ) wild-type and gpr124s984/s984 animals . Scale bar , 20 µm . ( E ) Lateral views ( anterior to the left ) of Tg ( kdrl:GFP ) control and morphant embryos at 36 hpf ( hindbrain region , upper panels ) and 24 hpf ( trunk region , bottom panels ) . Arrows point to the forming CtAs ( upper panels ) and ISVs ( lower panels ) . Scale bar , 50 µm . ( F ) Maximal intensity projection of a confocal z-stack of Tg ( kdrl:GFP ) wild-type and reck morphant cranial vasculature at 60 hpf in dorsal views ( anterior to the left; scale bar , 100 µm ) and quantification of hindbrain CtAs after Reck downregulation by anti-sense morpholino injections at various doses . ( G ) Quantification of hindbrain CtAs in control and gpr124 or reck morphants at 60 hpf after injection at the one-cell stage of 100 pg RNA encoding Gpr124 or Reck . The insert shows a typical rescue . ( H ) Quantification of ngn1:GFP+ DRGs in control and gpr124 or reck morphants at 72 hpf after injection at the one-cell stage of 200 pg RNA encoding Gpr124 or Reck . ngn1:GFP+ DRGs were counted on one side of the larvae . ( I ) Lateral view of wild-type and reck morphants at 5 dpf . In all panels , values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; Kruskal–Wallis test ) . Morpholino and RNA injections were performed as described in ‘Methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 008 Remarkably , gpr124 mutant zebrafish can proceed through organogenesis in the complete absence of intracerebral blood vessels and approximately half of them reach adulthood ( Figure 1—figure supplement 2 ) , some becoming fertile . Adult gpr124 mutants exhibit a CNS vascular network that appears of equal density to that of wild-type animals ( Figure 1G , upper panels ) , indicating that CNS vascularization can ultimately occur in the absence of Gpr124 . This late-onset brain vascularization program starts after 5 dpf ( days post fertilization ) with a varying degree of expressivity ( Figure 1—figure supplement 2 ) . The adult gpr124-deficient CNS vessels can acquire all tested BBB characteristics as revealed by immunostaining for Slc2a1 ( Glut1 ) , Pgp and by permeability assays ( Figure 1G ) . Non-transgenic homozygous gpr124 mutants are first distinguishable from wild-type siblings by their reduced growth rate and eventually by the disrupted pigmentation patterns of their skin after metamorphosis ( after 14 dpf ) . In contrast to the regular pattern of alternating longitudinal melanophore-rich dark stripes and light interstripes found in wild-types , gpr124 mutants exhibit discontinuous stripes formed by clusters of melanophores bordered by xanthophores and iridophores , most prominently in the dorsal aspect of the trunk ( Figure 2A ) . Metamorphic melanophores derive from a postembryonic stem cell population residing in close association with the segmentally arranged dorsal root ganglia ( DRG ) ( Dooley et al . , 2013 ) , and adult pigmentation defects have been previously correlated with abnormal DRG formation ( Budi et al . , 2008; Honjo et al . , 2008; Malmquist et al . , 2013 ) . Accordingly , gpr124 mutants fail to form DRGs with both sox10:mRFP+ satellite glial cells and ngn1:EGFP+ neurons missing in most segments at 72 hpf ( Figure 2B , C ) . On rare occasions , ganglia could be identified in the anterior-most segments of the trunk . The DRG defects do not result from the initial failure to specify the neuroglial lineage , as sox10:mRFP+ cells formed transient aggregates in all segments of gpr124 mutants at earlier stages . The ganglia however never contained ngn1:EGFP+ neurons as seen in wild-type siblings and were not retained at later stages ( Figure 2D ) . Other neural crest derivatives , like the embryonic pigment cells and lateral line glia , appear to develop normally , ruling out a general requirement for Gpr124 in neural crest-derived tissues . In order to identify the components of the molecular pathway through which Gpr124 operates , we tested whether previously described regulators of DRG formation ( Budi et al . , 2008; Honjo et al . , 2008; Prendergast et al . , 2012; Malmquist et al . , 2013 ) could , like Gpr124 , additionally control CNS angiogenesis . Using a morpholino knock-down approach , we identified Reck , a GPI-anchored MMP inhibitor and angiogenic modulator ( Oh et al . , 2001 ) as a novel essential regulator of brain vascularization ( Figure 2E ) . Reck and Gpr124 knockdown embryos exhibit identical CNS-specific vascular defects without detectable ISV or gross morphological phenotypes ( Figure 2F , I ) . Other DRG regulators like Sorbs3 ( Malmquist et al . , 2013 ) and Erbb3b ( Honjo et al . , 2008 ) do not appear to modulate CNS angiogenesis specifically ( Figure 2E ) . Reck knock-down dose-dependently blocked brain vascularization ( Figure 2F ) , and at doses of morpholino above 1 ng , brains remained avascular . As previously reported , reck morphants completely lack DRGs at 72 hpf ( Prendergast et al . , 2012 ) despite normal initial specification of sox10:mRFP+ neuroglial cells , much like gpr124 mutants ( Figures 2D and 3C ) . 10 . 7554/eLife . 06489 . 009Figure 3 . Wnt/β-catenin signaling is controlled by Gpr124 and Reck . ( A ) Maximal intensity projection of a confocal z-stack of Tg ( 7xTCF-Xia . Siam:GFP ) Wnt/β-catenin reporter expression during brain vascular development in wild-type , gpr124 or reck morphant Tg ( kdrl:ras-mCherry ) embryos . ( B ) Activity of the endothelial-specific Wnt/β-catenin reporter in the PHBC and BA in wild-type , gpr124 or reck morphant Tg ( fli1:Myr-mcherry ) embryos at 40 hpf . ( C ) Lateral view of Tg ( 7xTCF-Xia . Siam:GFP ) Wnt/β-catenin reporter expression at 54 hpf in the trunk region of wild-type , gpr124 or reck morphant Tg ( sox10:mRPF ) embryos ( anterior to the left ) . ( D ) Quantification of hindbrain CtAs at 60 hpf after genetic and pharmacological inhibition of Wnt/β-catenin signaling . Heat-shock and pharmacological inhibition were performed as described in ‘Methods’ . ( E ) Quantification of ngn1:GFP+ DRGs at 72 hpf after genetic inhibition of Wnt/β-catenin signaling . ngn1:GFP+ DRGs were counted on one side of the larvae . ( F ) Maximal intensity projection of a confocal z-stack of the cranial vasculature and quantification of hindbrain CtAs in control or gpr124 morphant Tg ( kdrl:EGFP ) embryos after exposure to the indicated GSK-3β inhibitors from the 16-somite stage onwards . Red arrows point to CtAs . Pharmacological inhibitions were performed as described in ‘Methods’ . ( G ) Dorsal views and quantification of DRGs in control or gpr124 morphant Tg ( ngn1:GFP ) ;Tg ( sox10:mRPF ) embryos after exposure to the indicated GSK-3β inhibitors from the 16-somite stage onwards . ngn1:GFP+ DRGs were counted on one side of the larvae . Pharmacological inhibitions were performed as described in ‘Methods’ . Scale bars , 50 µm . In all panels , values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; Kruskal–Wallis test: ( D ) , ( F ) , and ( G ) ; Mann–Whitney test: ( E ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 009 The remarkable phenotypic similarities observed after gpr124 and reck knock-downs in two settings of distinct embryological origin led us to probe their functional relationship . In both the CNS vascular and peripheral neurogenic settings , ectopic expression of Gpr124 or Reck could compensate for their respective loss-of-function but no functional epistatic relationship could be detected , that is , the absence of one protein could not be rescued by overexpression of the other ( Figure 2G , H ) . This result is compatible with one of the following two scenarios: Gpr124 and Reck act in independent parallel pathways , or they act in concert to control a common signaling pathway during both CNS angiogenesis and DRG neurogenesis . Wnt/β-catenin signaling controls CNS vascular formation in mouse ( Liebner et al . , 2008; Stenman et al . , 2008; Daneman et al . , 2009 ) and instructs sensory neural cell fates during DRG development ( Hari et al . , 2002; Lee et al . , 2004 ) . We tested whether Gpr124 or Reck participate in Wnt/β-catenin signaling during these processes , by evaluating the expression of Wnt/β-catenin reporter transgenes . We could detect Tg ( 7xTCF-Xla . Siam:GFP ) expression in a subset of wild-type ECs of the PHBCs starting at 26 hpf . Gradually , both the intensity and the number of GFP+ cells increased and these cells contributed to the PHBC , BA , and CtAs of 48 hpf wild-type embryos . In contrast , no GFP+ ECs could be detected in the PHBC or BA after gpr124 or reck knock-down ( Figure 3A ) , while the expression of the 7xTCF-Xla . Siam:GFP transgene appeared unaffected in the surrounding neural tissues in the absence of Gpr124 or Reck . Examination of an endothelial-specific Wnt/β-catenin reporter transgene , in which the fusion of the DNA binding domain of Gal4 to the β-catenin binding domain of TCF4 acts as a sensor for nuclear β-catenin ( Kashiwada et al . , 2015 ) , confirmed that Gpr124 and Reck are both required for Wnt/β-catenin signaling in the PHBCs ( Figure 3B ) . Similarly , both proteins are required to establish Tg ( 7xTCF-Xla . Siam:GFP ) expression in the forming DRGs at 54 hpf ( Figure 3C ) . We next tested whether Wnt/β-catenin signaling inhibition could mimic the phenotypes induced by Gpr124 or Reck depletion . Global heat-shock induced expression of Dickkopf1 ( Stoick-Cooper et al . , 2007 ) , the secreted inhibitor of Wnt/β-catenin signaling or Axin1 ( Kagermeier-Schenk et al . , 2011 ) , a component of the GSK-3/Axin/APC β-catenin destruction complex , induced respectively a partial and near-complete inhibition of brain vascular development . Similarly , pharmacological inhibition of Wnt/β-catenin signaling by the Axin-stabilizing compound IWR-I dose dependently blocked brain vessel formation ( Figure 3D ) . DRG neurogenesis was similarly sensitive to Axin1 overexpression ( Figure 3E ) . These observations suggest that the downregulation of Wnt/β-catenin signaling in Gpr124 or Reck deficient embryos could explain their angiogenic and neurogenic defects . We tested this hypothesis by artificially restoring β-catenin levels through GSK-3β inhibition using structurally distinct compounds . These drugs could partially restore brain angiogenesis ( Figure 3F ) and completely restore DRG neurogenesis ( Figure 3G ) in gpr124 morphants . In a previous study , we found that mouse Gpr124 co-activates Wnt/β-catenin signaling via Wnt7a/Wnt7b in a reporter cell line ( Super Top Flash; STF [Xu et al . , 2004] ) and that signaling was further enhanced by co-transfection with Fz4 and Lrp5 ( Zhou and Nathans , 2014 ) . ( We note that RNAseq analysis of STF cells showed low level expression of many Wnt signaling components , including Frizzled receptors , Lrp co-receptors , Gpr124 , and Reck [Zhou and Nathans , 2014] , which likely accounts for the signals observed when individual components are omitted from the transfection . ) To determine whether Reck influences the Gpr124 dependence of Wnt7a- and Wnt7b-induced signaling , we co-transfected various combinations of Wnt7a or Wnt7b , Fz4 , Lrp5 or Lrp6 , Gpr124 , and Reck into STF cells ( Figure 4A ) . These experiments showed that Reck dramatically synergizes with Gpr124 in activating Wnt/β-catenin signaling in response to Wnt7a and Wnt7b , and that signaling is further increased by co-transfection with Fz4 together with Lrp5 or Lrp6 . Both Gpr124 and Reck show well behaved dose–response curves , with synergistic activity over a wide range of DNA concentrations ( Figure 4B ) . 10 . 7554/eLife . 06489 . 010Figure 4 . Synergy between Gpr124 and Reck in co-activating Wnt7a- and Wnt7b-dependent Wnt/β-catenin signaling in cell culture . ( A ) Plasmids encoding Wnt7a ( left ) or Wnt7b ( right ) , together with Lrp5 ( top row ) , Lrp6 ( center row ) , or no Lrp ( bottom row ) were transfected with Gpr124 , Fz4 , and/or Reck plasmids as indicated . ( B ) Titration of Gpr124 with or without Reck ( left ) , and titration of Reck with or without Gpr124 ( right ) . ( C ) Comparison of the ten mouse Frizzleds transfected with Lrp5 and Wnt7a , with or without Gpr124 and/or Reck . ( D ) Comparison of the nineteen mouse Wnts and Norrin , transfected with or without Gpr124 and/or Reck . Arrows point to Wnt7a and Wnt7b responses . Luciferase assays in transiently transfected STF cells were performed as described in ‘Methods’ . Values represent means ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01010 . 7554/eLife . 06489 . 011Figure 4—figure supplement 1 . Synergy between Gpr124 and Reck in co-activating Wnt7a- and Wnt7b-dependent Wnt/β-catenin signaling in cell culture . ( A ) Comparison of the ten mouse Frizzleds with Lrp6 and Wnt7a , with or without Gpr124 and/or Reck . Bars show mean ± SD from triplicate determinations . ( B ) Mean ratio of relative luciferase activity in wells with Gpr124 and/or Reck divided by the activity in wells with vector . For each set of transfections ( Wnt , Norrin , or no ligand ) , the plot shows the mean ratio of relative luciferase activity in wells with Gpr124 and/or Reck divided by the activity in wells with vector . The largest ratio increases are observed for Wnt7a and Wnt7b . The relative luciferase activities used for this plot were obtained from Figure 4D . ( C ) Mean ratio of relative luciferase activity in wells with Gpr124 and Reck divided by the activity in wells with Gpr124 or Reck . For each set of transfections ( Wnt , Norrin , or no ligand ) , the plot shows the mean ratio of relative luciferase activity in wells with Gpr124 and Reck divided by the activity in wells with Gpr124 or Reck . Large ratio elevations are observed for Wnt7a and Wnt7b . Modestly elevated ratios are also observed for Wnt5a , Wnt8a , and Wnt10a , but the relative luciferase activities for these Wnts were extremely low ( Figure 4D ) and the elevated ratios may simply reflect noise in the measurements . The relative luciferase activities used for this plot were obtained from Figure 4D . Luciferase assays in transiently transfected STF cells were performed as described in ‘Methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 011 To test whether the combination of Reck and Gpr124 can co-activate signaling by Wnts other than Wnt7a and Wnt7b , we screened all 19 Wnts and Norrin for co-activation by Gpr124 alone , Reck alone , or Gpr124 plus Reck ( Figure 4D , Figure 4—figure supplement 1B , C ) . This experiment shows that the highest co-activation with Gpr124 plus Reck occurs with Wnt7a and Wnt7b , with other Wnts and Norrin showing little or no response . An analogous comparison among the ten Frizzled receptors showed that Gpr124/Reck stimulates Wnt7a- and Wnt7b-mediated signaling by multiple Frizzleds ( Figure 4C , Figure 4—figure supplement 1A ) . The data predict that cells expressing Gpr124 , Reck , Lrp5 and/or Lrp6 , and any of multiple Frizzleds will be responsive to Wnt7a or Wnt7b . In earlier experiments , we observed that Gpr124 co-activation of Wnt7a- and Wnt7b-dependent signaling in STF cells could be stimulated by Lrp5 but not by Lrp6 , and we postulated that one or more additional proteins might be required to enhance signaling in the presence of Lrp6 ( Zhou and Nathans , 2014 ) . The present experiments identify Reck as the missing protein since signaling in the presence of Reck can be stimulated by both Lrp5 and Lrp6 ( Figure 4A ) , and they provide functional evidence that Gpr124 and Reck are major determinants of both the amplitude and ligand specificity of Wnt/β-catenin signaling . To examine whether the functional interactions between Gpr124 , Reck and the Fz/Lrp5/6 complex reflect a multi-component membrane receptor complex involved in Wnt7 binding , their subcellular localization was investigated in 293 cells . We generated amino-terminal epitope-tagged version of Gpr124 ( FLAG-Gpr124 ) and Reck ( HA-Reck ) and validated the biological activity of the fusion proteins in STF assays ( data not shown ) and brain vascular rescue experiments in zebrafish embryos ( Figure 5A ) . When expressed in 293 cells , FLAG-Gpr124 and HA-Reck colocalized at the plasma membrane as revealed by indirect immunofluorescence assays ( Figure 5B ) . Both Gpr124 and Reck reached and resided at the plasma membrane irrespective of the presence of the other ( Figure 5B ) , and within this compartment they co-localized with GFP-tagged Fz4 ( Figure 5—figure supplement 1 ) . We next evaluated the proximity of Gpr124 and Reck through a highly sensitive in situ proximity ligation assay ( PLA ) which allows the localized detection of protein interactions ( Söderberg et al . , 2006 ) . While PLA assays on 293 cells expressing either FLAG-Gpr124 or HA-Reck individually remained negative , simultaneous expression of the fusion proteins yielded extensive fluorescence hybridization signals at the cell surface ( Figure 5C , D ) . As an additional control , the FLAG-Dvl2/HA-Reck protein pair , generated weaker and less frequent PLA signals , despite increased anti-FLAG immunoreactivity at the plasma membrane ( Figure 5C ) . These observations , coupled with the synergistic capacity of Gpr124 and Reck to stimulate and confer ligand specificity to Wnt/β-catenin signaling in a Frizzled and Lrp5/6-dependent manner , are compatible with the assembly of a multi-component Wnt7 membrane receptor complex , whose precise stoichiometry and composition remains to be determined ( Figure 5E , see also ‘Discussion’ ) . 10 . 7554/eLife . 06489 . 012Figure 5 . Interaction between Gpr124 and Reck in cultured cells . ( A ) Quantification of hindbrain CtAs in control and gpr124 morphants at 60 hpf after injection at the one-cell stage of 100 pg RNA encoding Gpr124 , Reck or their epitope-tagged versions , FLAG-Gpr124 and HA-Reck . ( B ) and ( C ) Plasmids encoding FLAG-Gpr124 , HA-Reck , Lrp5 , Fz4 , and Wnt7a were transfected in 293 cells as indicated . 48 hours after transfection , epitope-tagged fusion proteins were detected by indirect immunofluorescence ( B ) or by proximity ligation assays as described in ‘Methods’ . ( D ) High magnification views of PLA signals between FLAG-Gpr124 and HA-reck . ( E ) Illustration of the plasma membrane protein complexes that mediate Wnt7a or Wnt7b signaling during CNS angiogenesis . Values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; Kruskal–Wallis test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01210 . 7554/eLife . 06489 . 013Figure 5—figure supplement 1 . Gpr124 , Reck , and Fz4 colocalize at the plasma membrane in cultured cells . Plasmids encoding FLAG-Gpr124 , HA-Reck , Lrp5 , Fz4-GFP , and Wnt7a were transfected in 293 cells as indicated . 48 hours after transfection , FLAG-Gpr124 ( A ) and HA-Reck ( B ) fusion proteins were detected by indirect immunofluorescence and superposed with the endogenous GFP fluorescence of the Fz4-GFP fusion protein . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 013 In contrast to mouse , the size of zebrafish embryos allows oxygen to reach tissues independently of cardiovascular-based convection . We performed microarray gene expression analysis and found , in line with normoxic conditions , that the transcript levels of various vegf genes were not increased in the absence of Gpr124 at 48 hpf ( Figure 6A ) . We took advantage of the unique normoxic nature of zebrafish embryos , combined with their optical clarity to investigate the requirements for Wnt/β-catenin during CNS angiogenesis at single-cell resolution . 10 . 7554/eLife . 06489 . 014Figure 6 . Tip cell defects in Wnt/β-catenin-deficient PHBCs . ( A ) Scatterplot of a microarray comparison of wild-type and gpr124s984/s984 embryos at 48 hpf . Each dot refers to the relative signal intensity of a given probe in mutant vs wild-type embryos . Circles identify independent probes for vegfaa , vegfab , and vegfc . ( B ) Endothelial cell number in the PHBCs of wild-type and gpr124s984/s984 embryos at 30 hpf . kdrl:NLS-GFP+ nuclei were counted on one side of the embryos . ( C ) Dorsal views of Tg ( pdgfrb:mCitrine ) ;Tg ( kdrl:ras-mCherry ) wild-type and gpr124s984/s984 embryos at 40 hpf . ( D ) Stills from Video 3 recording sprouting angiogenesis in the hindbrain of Tg ( kdrl:GFP ) wild-type and gpr124s984/s984 embryos . Below are kymograph plots showing a time course of pixel intensity across the yellow virtual lines running 15 µm dorsal to the PHBC ( from the anterior MCeV to the posterior PCeV [posterior cerebral vein] ) and depicted in the upper left panel . ( E ) Lateral views of Tg ( kdrl:lifeact-GFP ) ;Tg ( kdrl:NLS-mCherry ) wild-type and gpr124s984/s984 hindbrains at 30 hpf . Arrows point to actin-dense structures forming in the PHBC vessel wall . Scale bars , 50 µm . Values represent means ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01410 . 7554/eLife . 06489 . 015Figure 6—figure supplement 1 . Endothelial cell number in wild-type and gpr124s984/s984 brain vessels at 54 hpf . Maximal intensity projection of a confocal z-stack of the Tg ( kdrl:NLS-GFP ) wild-type and gpr124s984/s984 cranial vasculature at 54 hpf ( left panels ) . kdrl:NLS-GFP+ EC nuclei were counted in the PHBCs , ventral PHBC-BA connections , BA and CtAs ( right panels ) . In all panels , values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; Mann–Whitney test ) of a minimal of five embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 015 We first asked whether defective brain vascularization might result from insufficient EC abundance in the parental vessels . Using the endothelial-specific nuclear reporter Tg ( kdrl:NLS-GFP ) ( Blum et al . , 2008 ) , we determined EC number in the PHBC at 30 hpf and observed no difference between gpr124 mutants and their wild-type siblings ( Figure 6B ) . As expected , marked differences were detected after the onset of brain invasion; while wild-type siblings accumulated intra-cerebral ECs contributing to the CtAs , gpr124 mutants maintained marginally higher numbers of cells in the PHBC , BA and most prominently in ventral connections between the PHBCs and the BA ( Figure 6—figure supplement 1 , Figure 1D , Videos 1 and 2 ) . Mammalian CNS vessels show extensive pericyte coverage and these support cells convey essential properties to the BBB ( Armulik et al . , 2010; Bell et al . , 2010; Daneman et al . , 2010 ) . To determine whether EC-pericyte interactions were compromised in gpr124 mutants , we generated a Pdgfrb transgenic reporter line TgBAC ( pdgfrb:mCitrine ) that labels perivascular cells presenting all characteristics of bona fide pericytes . We found extensive vascular wall coverage in both the presence and absence of Gpr124 function ( Figure 6C ) . We then examined EC behaviors using time-lapse confocal microscopy in wild-type and Wnt/β-catenin-deficient vessels during CNS angiogenesis . Dynamic filopodial extensions were observed from the PHBC of both wild-type and gpr124 mutants embryos ( Figure 6D , Videos 3–5 ) , but only in wild-type siblings did those extensions progress dorsally to allow the cell body of a stereotypic tip cell ( Gerhardt et al . , 2003 ) to emerge from extra-cerebral vessels and invade the brain tissue . Intra-cerebral EC nuclei were absent in gpr124 mutants ( Videos 6 and 7 ) . Similar results were seen upon Reck knockdown as well as Wnt/β-catenin inhibition by heat-shock induced Axin1 overexpression . 10 . 7554/eLife . 06489 . 016Video 3 . Brain angiogenesis in wild-type and gpr124s984/s984 embryos . Time-lapse confocal video of Tg ( kdrl:GFP ) wild-type ( upper right ) and gpr124s984/s984 ( bottom left ) embryos , starting at 32 hpf and ending at approximately 42 hpf . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01610 . 7554/eLife . 06489 . 017Video 4 . Hindbrain angiogenesis in wild-type embryos . High magnification view of the time-lapse confocal video presented in Video 3 , focusing on the wild-type hindbrain ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01710 . 7554/eLife . 06489 . 018Video 5 . Hindbrain angiogenesis in gpr124s984/s984 embryos . High magnification view of the time-lapse confocal video presented in Video 3 , focusing on the gpr124s984/s984 hindbrain ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01810 . 7554/eLife . 06489 . 019Video 6 . Endothelial cell invasion into the wild-type hindbrain . Time-lapse confocal video generated from maximum intensity confocal projections through one hindbrain hemisphere of a Tg ( kdrl:NLS-GFP ) ;Tg ( kdrl:ras-mCherry ) wild-type embryo . Endothelial cell nuclei are green ( GFP ) ; endothelial cells are red ( mCherry ) . Video starts at 32 hpf and ends at approximately 42 hpf ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 01910 . 7554/eLife . 06489 . 020Video 7 . Lack of endothelial cell invasion into the gpr124 mutant hindbrain . Time-lapse confocal video generated from maximum intensity confocal projections through one hindbrain hemisphere of a Tg ( kdrl:NLS-GFP ) ;Tg ( kdrl:ras-mCherry ) gpr124s984/s984 embryo . Endothelial cell nuclei are green ( GFP ) ; Endothelial cells are red ( mCherry ) . Video starts at 32 hpf and ends at approximately 42 hpf ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 020 Polar filopodial extensions and abluminal emergence of ECs from pre-existing vessels are easily detectable angiogenic events occurring after the VEGF/Notch-controlled ( Hellström et al . , 2007; Leslie et al . , 2007; Lobov et al . , 2007; Siekmann and Lawson , 2007; Suchting et al . , 2007; Phng and Gerhardt , 2009; Eilken and Adams , 2010 ) specification of presumptive tip and stalk cells within the parental vessel . We sought to distinguish between a defect in filopodial extension and an earlier event occurring within the parental vessel , by contrasting F-actin structures within the pre-angiogenic PHBCs of wild-type and gpr124 mutant embryos using a novel Tg ( kdrl:lifeact-EGFP ) line that labels filamentous actin ( Riedl et al . , 2008; Phng et al . , 2013 ) in ECs . While the use of this transgene confirmed the presence of filopodial extensions independently of Gpr124 function , wild-type siblings in addition displayed actin-rich structures at the dorsal wall of the PHBCs at sites where angiogenic tip cells later emerged . These structures were not observed in gpr124 mutants , indicating that Gpr124 regulates actin cytoskeletal rearrangement within extra-cerebral ECs before the onset of brain invasion ( Figure 6E ) . These observations are consistent with a defect in tip cell specification or behavior within the PHBCs of gpr124 mutants . To test this hypothesis , we generated genetically mosaic PHBCs by cell transplantation at mid-blastula stages and examined the behavior of wild-type Tg ( kdrl:GFP ) ECs in the context of Tg ( kdrl:ras-mCherry ) gpr124 , reck or Wnt/β-catenin deficient endothelial neighbors ( Figure 7A , B ) . Time-lapse imaging ( Video 8 and stills presented in Figure 7A ) revealed that mosaic PHBCs were competent for brain vascular invasion and that in all sprouts examined , tip cells were wild-type ( green ) . Notably , trailing stalk cells could be wild-type ( green ) or Gpr124/Reck/Wnt/β-catenin-deficient ( red ) ( red arrows in Figure 7A ) revealing a tip cell-autonomous role for Gpr124/Reck/Wnt/β-catenin signaling during brain angiogenesis . A single wild-type cell was sufficient to instruct Wnt/β-catenin-deficient ECs to assemble mosaic vessels ( Video 9 ) . These mosaic sprouts invariably led by wild-type tip cells ( Figure 7C ) invaded the hindbrain in a wild-type manner and lumenized after connecting medially to the BA ( Figure 7B , D ) . In contrast , when wild-type Tg ( kdrl:GFP ) cells were transplanted into Tg ( kdrl:ras-mCherry ) wild-type hosts , tip cell genotypes were randomized ( Figure 7C ) . In mosaic ISV sprouts , gpr124 or reck loss of function did not impact tip cell genotype ( Figure 7E , F ) . The mosaic cerebral vessels were maintained for several days with no indication of vascular instability during embryonic or early larval development ( Figure 7G ) . 10 . 7554/eLife . 06489 . 021Figure 7 . Tip cell-specific requirement for Gpr124 and Reck-controlled Wnt/β-catenin signaling . ( A ) Stills from Video 8 recording sprouting angiogenesis from a mosaic PHBC obtained by blastula-stage transplantation . Green kdrl:GFP+ endothelial cells derive from a wild-type donor embryo , red kdrl:ras-mCherry+ endothelial cells are from the gpr124s984/s984 host . ( B ) Dorsal views of 48 hpf mosaic cranial vasculatures of the indicated genotypes . Right panels are high magnification views of a confocal z-stack of relevant depth illustrating intra-cerebral mosaic CtAs corresponding to the boxed areas in the left panels . ( C ) Contribution of cells of defined genotype to the tip cell position of mosaic CtAs after transplantation of wild-type Tg ( kdrl:GFP ) donor cells ( green ) into Tg ( kdrl:ras-mCherry ) host blastulae of the indicated genotype ( red ) . Number of mosaic vessels analyzed is indicated above each bar . ( D ) 3D Imaris ( Bitplane ) reconstruction of representative hindbrain mosaic vascular networks . Arrows point to the cells connecting with the BA , after leading the CtA as tip cells . ( E ) Lateral views of the trunk region after transplantation of Tg ( kdrl:GFP ) wild-type donor cells into Tg ( kdrl:ras-mCherry ) host blastulae of the indicated genotype . ( F ) Contribution of cells of defined genotype to the tip cell position of mosaic ISVs after transplantation of wild-type Tg ( kdrl:GFP ) donor cells ( green ) into Tg ( kdrl:ras-mCherry ) host blastulae of the indicated genotype ( red ) . Number of mosaic vessels analyzed is indicated above each bar . ( G ) Dorsal view of the cranial vasculature and 3D Imaris ( Bitplane ) reconstruction of the brain vessels of a 96 hpf mosaic larva after transplantation of wild-type Tg ( kdrl:GFP ) donor cells into reck morphant Tg ( kdrl:ras-mCherry ) host blastula . ( H ) Cellular requirement for Wnt/β-catenin , Gpr124 , and Reck during sprouting angiogenesis in the zebrafish CNS ( *p < 0 . 05; **p < 0 . 01; exact Fisher test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 02110 . 7554/eLife . 06489 . 022Figure 7—figure supplement 1 . Transgenic endothelial dll4 expression is not sufficient to rescue the gpr124 mutant vascular defects . ( A ) dll4 whole-mount in situ hybridization in wild-type and gpr124 mutant embryos at 36 hpf . ( B ) Quantification of the proportion of gpr124 morphant embryos with one or more central arteries at 60 hpf . Embryos were co-injected with gpr124 MO , tol2 transposase mRNA and pTol2-fli1:egfp , pTol2-fli1:gpr124-2A-EGFP , pTol2-fli1:EGFP-2A-gpr124 , pTol2-fli1:dll4-2A-EGFP or pTol2-fli1:EGFP-2A-dll4 ( C ) Maximal intensity projection of a confocal z-stack of the cranial vasculature of Tg ( kdrl:ras-mCherry ) gpr124 morphant embryos at 60 hpf in dorsal views ( anterior to the left ) after transgenic endothelial overexpression ( green fluorescence ) of Gpr124-2A-EGFP ( upper panel ) or Dll4-2A-EGFP ( bottom panel ) . Values represent means ± SD ( *p < 0 . 05; **p < 0 . 01; one-way ANOVA test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 02210 . 7554/eLife . 06489 . 023Video 8 . Single-cell analysis of CtA formation in gpr124 mosaic animals . CtAs are led by wild-type tip cells and gpr124 mutant cells can incorporate into the sprouts as stalk cells . Time-lapse confocal video generated from maximum intensity confocal projections through one hindbrain hemisphere of a mosaic embryo containing wild-type kdrl:GFP+ ECs ( green ) and gpr124s984/s984 kdrl:ras-mCherry+ ECs ( red ) . Video starts at 30 hpf and ends at approximately 40 hpf ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 02310 . 7554/eLife . 06489 . 024Video 9 . Single-cell analysis of CtA formation in gpr124 mosaic animals . A single wild-type tip cell is sufficient to initiate the formation of an intra-cerebral CtA . Time-lapse confocal video generated from maximum intensity confocal projections through one hindbrain hemisphere of a mosaic embryo containing wild-type kdrl:GFP+ ECs ( green ) and gpr124s984/s984 kdrl:ras-mCherry+ ECs ( red ) . Video starts at 30 hpf and ends at approximately 40 hpf ( anterior to the left ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06489 . 024 The tip cell-restricted requirement for Gpr124 and Reck mediated Wnt/β-catenin signaling could conceptually be linked to a role in initiating a tip cell-permissive arterial-biased transcriptional program ( Corada et al . , 2010 ) that is known to operate in the PHBCs during brain angiogenesis ( Bussmann et al . , 2011 ) and that results notably in dll4 expression . At 36 hpf however , dll4 transcripts could be detected in the PHBCs irrespective of Gpr124 function ( Figure 7—figure supplement 1 ) and accordingly , mosaic endothelial-specific overexpression of Dll4 was not sufficient to restore brain angiogenic competence to Gpr124-deficient PHBCs , while in a similar setting Gpr124-positive tip cells led new brain vascular sprouts ( Figure 7—figure supplement 1 ) .
During retinal vascular development in mouse , wild-type ECs can instruct Fz4−/− ECs to assemble into mosaic vessels , suggesting that Wnt/β-catenin signaling ( Wang et al . , 2012 ) is not a uniform requirement for all ECs during vascular network formation . The cellular basis of this phenomenon has , until now , not been investigated . Using genetic mosaics and live imaging , we present evidence that the control of brain vascular invasion by Wnt/β-catenin signaling operates at the level of the tip cells , and that stalk cells that are deficient in Wnt/β-catenin signaling can follow wild-type tip cells and contribute to the developing vasculature ( Figure 7H ) . These observations reveal a heterogeneous requirement for Wnt/β-catenin signaling among ECs during brain invasion in zebrafish , and they predict a similar differential requirement between tip and stalk cells in the mammalian brain and retina . In recent years , through the investigation of a limited number of stereotypical in vivo settings , most notably the postnatal mouse retina and the zebrafish ISVs , a coherent model of sprouting angiogenesis integrating controlled behaviors of VEGF-selected tip cells and Notch-induced stalk cells within nascent vessels has emerged . A question of pressing interest is whether tissue and organ-specific angiogenic programs might refine our understanding of these processes . The Wnt/β-catenin-dependency of brain angiogenesis in zebrafish ( Figure 3 ) coupled to their unique optical attributes has permitted us , through live imaging of mosaic animals ( Figure 7 ) , to collect evidence that local angiogenic programs can impact directly on the basic cellular behaviors of nascent angiogenic sprouts . An important focus for future studies will be to delineate the molecular mechanisms by which Wnt/β-catenin signaling affects tip cell specification and/or behavior . According to the current model , when pro-angiogenic signals stimulate a quiescent vessel , ECs that experience the highest level of VEGFA-VEGFR2 signaling become tip cells . As these nascent tip cells egress from the parental vessel , they accumulate tip cell-enriched transcripts , including dll4 and vegfr3 , while neighboring cells become stalk cells through Notch-mediated lateral inhibition ( Hellström et al . , 2007; Leslie et al . , 2007; Lobov et al . , 2007; Siekmann and Lawson , 2007; Suchting et al . , 2007 ) . The multiple interactions between the Wnt/β-catenin , Notch , and VEGFR signaling systems ( Phng et al . , 2009; Corada et al . , 2010; Gore et al . , 2011 ) , as seen , for example , by the strong reduction of hindbrain CtAs after loss of VEGFR2 signaling or Notch activation ( Bussmann et al . , 2011 ) , imply that regulation of brain vascularization is highly integrated . Despite current evidence for the tight integration of multiple signaling systems in EC development , our data argue that the tip cell requirement for Wnt/β-catenin signaling operates through a mechanism that does not simply reflect a global modulation of VEGF and/or Notch signals . In gpr124 mutants and reck morphants , a causal role for reduced VEGFR2 signaling in the PHBC appears unlikely because the formation of the BA from the PHBCs , which is known to be sensitive to VEGF inhibition ( Bussmann et al . , 2011 ) , appears unaffected . We have also shown that at 36 hpf dll4 transcripts could be detected in gpr124 mutant PHBCs as in wild-type , and that clonal endothelial dll4 overexpression was not sufficient to restore the angiogenic behavior . With respect to Notch signaling , although gpr124 and reck-depleted cerebral vasculatures exhibit defects resembling those generated by pan-endothelial expression of a constitutively active Notch intracellular domain ( NICD ) ( Bussmann et al . , 2011 ) , gpr124 and reck morphant vascular defects were not rescued by dll4 knock-down or gamma-secretase inhibitor ( DAPT ) treatment . In our experiments , the Wnt reporter transgene labels both tip and stalk cells in nascent sprouts , as well as ECs in adult CNS vessels ( Moro et al . , 2012 ) , reflecting the role of Wnt/β-catenin signaling in angiogenesis , BBB acquisition , and BBB maintenance ( Reis et al . , 2012; Wang et al . , 2012 ) . The multiple roles of Wnt/β-catenin signaling in endothelial development complicate the interpretation of experiments designed to discriminate between a permissive role for Wnt/β-catenin signaling during CNS angiogenesis vs a selective role in tip cell differentiation . For example , Wnt/β-catenin signaling could regulate a transcriptional program that is required for tip cell function but plays no role in tip cell selection or , alternatively , tip cells could be selected through a specific transcriptional program that is only initiated by a certain threshold level of Wnt/β-catenin activity . Although the present data do not distinguish between these two scenarios , we note that Wnt/β-catenin reporter expression does not begin uniformly within PHBCs at 26–28 hpf , but instead labels a small number of cells , and even at later stages ( 32 hpf ) , different levels of Wnt/β-catenin reporter expression are observed among PHBC ECs ( Figure 3A ) . It is also notable that sprouting tip cells display high levels of Wnt/β-catenin reporter expression . Our discovery of Reck's central role in Wnt/β-catenin signaling is surprising because earlier work had implicated Reck in tissue remodeling via inhibition of MMP function at the transcriptional level ( Takagi et al . , 2009 ) , and at the level of protein–protein interactions ( Takahashi et al . , 1998 ) . Down-regulation of RECK has been shown to correlate with enhanced tumor invasion , angiogenesis , and metastasis ( Noda and Takahashi , 2007 ) , and Reck−/− mouse embryos die in utero at around E10 . 5 with prominent vascular defects including vessel dilation , hemorrhaging , and arrested development of the primary vascular plexus ( Oh et al . , 2001; Chandana et al . , 2010 ) . The mechanisms underlying these vascular defects have been unclear , but in light of the present work , a contribution of defective Wnt/β-catenin signaling seems possible . A question of central importance is the mechanism by which Gpr124 and Reck promote ligand-specific Wnt/β-catenin signal transduction . It is possible that Reck's role as an MMP inhibitor may contribute to its activity in the context of Wnt/β-catenin signaling . For example , Reck might protect ECM- or membrane-bound Wnt ligands and/or receptor complexes from proteolytic degradation , a possibility that would be consistent with its capacity to modulate Notch ligand ectodomain shedding through ADAM10 antagonism during cortical neurogenesis ( Muraguchi et al . , 2007 ) . In the context of this model , Gpr124 itself would be an attractive candidate for Reck-dependent regulation given previous evidence that MMP-processing of Gpr124 might affect its function ( Vallon and Essler , 2006 ) . Arguing against this model is our failure to rescue reck morphant vascular or DRG phenotypes by treating embryos with various MMP or ADAM inhibitors ( data not shown ) , in agreement with results from a previous study that examined Reck's role in DRG development ( Prendergast et al . , 2012 ) . We note that Reck has additionally been implicated in endocytic trafficking ( Miki et al . , 2007 ) , focal adhesion formation and cell polarity ( Morioka et al . , 2009 ) processes that may affect Wnt/β-catenin signal transduction ( Fonar and Frank , 2011; Feng and Gao , 2014 ) . The striking similarities between Gpr124 and Reck loss-of-function phenotypes , together with their synergistic roles in Wnt7a/Wnt7b-specific activation of Wnt/β-catenin signaling , suggest a more focused view of Reck function . As an extension of a previous model for Gpr124 action ( Zhou and Nathans , 2014 ) , we propose that Reck and Gpr124 cooperate to activate Wnt/β-catenin signaling by selectively promoting Wnt7a/Wnt7b binding to cell surface receptor complexes , thereby ( 1 ) conferring specificity among Wnt ligands that would not otherwise be discriminated by Frizzled-binding ( Janda et al . , 2012 ) , and ( 2 ) greatly enhancing the amplitude of Wnt7a and Wnt7b signals . Moreover , the observation that maximal Wnt/β-catenin signaling activity in Wnt7a/Wnt7b-stimulated cells requires the simultaneous presence of Gpr124 , Reck , Fz , and Lrp5/Lrp6 ( Figure 4 ) , suggests that the Wnt7a/Wnt7b receptor complex might be composed of all four proteins ( Figure 5E ) . An alternative model is suggested by the intriguing observation that Reck contains a cysteine-rich domain that is homologous to the Wnt-binding cysteine-rich domain of Frizzled receptors ( Pei and Grishin , 2012 ) and by our demonstration that the intracellular domain of Gpr124 is functionally interchangeable with the homologous domain of Gpr125 which has been shown to bind the cytoplasmic Wnt signaling adaptor Dishevelled ( Li et al . , 2013; Figure 1F ) . Taken together , these two observations suggest that Gpr124 and Reck might be capable of signaling as an autonomous Wnt-binding receptor complex ( Figure 5E ) . Vascular morphogenesis is a complex process that requires coordinated control of EC behaviors , including a separation of ECs into leading tip cells and trailing stalk cells . In addition , ECs must respond to organ-specific signals to meet the specialized needs of the vascularized tissue . By observing CNS angiogenesis in genetic mosaics at single-cell resolution , we have identified a regulatory system whereby endothelial Wnt/β-catenin signaling controls CNS angiogenesis through selective modulation of tip cell function . This Wnt/β-catenin signaling pathway requires the membrane proteins Gpr124 and Reck , further expanding the complexity of Wnt signal transduction mechanisms and holding the promise of new insights into the etiology and treatment of cerebral vascular defects , a major cause of morbidity and mortality .
Zebrafish were maintained and embryos were obtained and raised under standard conditions . Establishment and characterization of the Tg ( kdrl:GFP ) s843 , Tg ( kdrl:ras-mCherry ) s896 , Tg ( kdrl:NLS-GFP ) zf109 , Tg ( hsp70l:dkk1-GFP ) w32 , Tg ( hsp70l:Mmu . Axin1-YFP ) w35 , Tg ( 7xTCF-Xla . Siam:GFP ) ia4 , Tg ( ngn1:gfp ) , Tg ( UAS:EGFP ) , Tg ( sox10:mRFP ) vu234 , Tg ( kdrl:NLS-mCherry ) is4 transgenic lines have been described elsewhere ( Blader et al . , 2003; Jin et al . , 2005; Kirby et al . , 2006; Stoick-Cooper et al . , 2007; Asakawa and Kawakami , 2008; Blum et al . , 2008; Chi et al . , 2008; Wang et al . , 2010; Kagermeier-Schenk et al . , 2011; Moro et al . , 2012 ) . The Tg ( fli1:Gal4db-TΔC-2A-mC ) ( Kashiwada et al . , 2015 ) was generated by fusing a cDNA fragment encoding the β-catenin binding domain of human TCF4E ( amino acid 1–314 ) to the DNA binding domain of Gal4 followed by a 2A peptide and mCherry . It was subcloned into the pTol2-fli1 vector to generate the pTol2-fli1:Gal4db-TΔC-2A-mC plasmid . The Tg ( fli1:Myr-mCherry ) was generated by fusing an oligonucleotide encoding the myristoylation ( Myr ) signal derived from Lyn to mCherry in the pTol2-fli1 vector , yielding the pTol2-fli1:Myr-mC plasmid . The Tg ( kdrl:Lifeact-EGFP ) s982 line was generated following cloning of the F-actin binding Lifeact-EGFP downstream of a 6 . 8 kb kdrl promoter in pTol2 . The TgBAC ( pdgfrb:citrine ) s1010 line was generated using a BAC clone containing the start codon of pdgfrb ( CH73-289D6 ) using a pRedET based protocol ( Bussmann and Schulte-Merker , 2011 ) . Tol2 LTRs were inserted into the vector backbone . Recombination was performed by amplifying a mCitrine-Kan ( R ) cassette using primers that contained 50bp of homology flanking the start codon: pdgfrb_HA1_mCitrine , TTTGGCTTTGAGGCGAATCAGTCATGTTGTTTTCTCTCCGTCTGCAGTGTACCATGGTGAGCAAGGGCGAGGAG and pdgfrb_HA2_Kan ( R ) , TGGATGCGGCTGATGGTCGAACTCTTCATGCTTTCTTCTAGAGCAGGACATCAGAAGAACTCGTCAAGAAGGCG . Zebrafish transgenic lines were generated according to previously described protocols ( Kawakami , 2004 ) . All images were acquired using a Leica ( Wetzlar , Germany ) M165 stereomicroscope or a Zeiss ( Oberkochen , Germany ) LSM710 confocal microscope after embryo anesthesia with a low dose of tricaine and immobilization in 1% low-melting agarose in glass-bottom Petri dishes ( MatTek Corporation , Ashland , MA ) . Time-lapse images were recorded every 10 min for a total period of 16 hr . The microscope stage was enclosed in a temperature-controlled chamber , and samples were kept at 28 . 5°C . The Sulfo-NHS-Biotin leakage assay was performed by intra-ventricular injection of 5 nL of 0 . 5 mg/mL EZ-link Sulfo-NHS-Biotin ( Pierce; 0 . 5 kDa ) in D-PBS , with the help of a micromanipulator , in animals treated with a low dose of tricaine . After 60 min , animals were anaesthetized with an overdose of tricaine , and the brains and livers were dissected and fixed in 4% PFA overnight at 4°C before embedding in 4% agarose . Vibratome sections of 150 μm thickness were permeabilized in 0 . 4% PBT ( Triton X-100 ) and blocked in 4% PBTB ( BSA ) . Primary and secondary antibody staining was performed at 4°C overnight . The following primary antibodies were used: anti-GLUT1 ( NB300-666; Novus Biologicals , Littleton , CO ) , anti-P Glycoprotein [C219] ( ab3364; Abcam , Cambridge , UK ) . Biotin was detected with Alexa Fluor 647 streptavidin . In all confocal images and time points , brightness and contrast were adjusted linearly and equally . Three-dimensional reconstructions were done with the Imaris FilamentTracer software ( Bitplane , Zurich , Switzerland ) in automatic detection mode before manual false-coloring and editing to highlight extra- and intra-cerebral vessels . Two TALENs targeting exons 7 and 16 of gpr124 were designed and cloned using Golden Gate assembly ( Cermak et al . , 2011 ) into the pCS2TAL3RR or pCS2TAL3DD expression vectors ( Dahlem et al . , 2012 ) . The TALEN targeting exon 7 was composed of the following TAL effector domains RVDs: NN HD NI HD NI HD NI NG NG NN NG HD NN NG NG NI NG and HD HD HD NG NG NI NI NI NI NI HD NI NI HD HD NG . The TALEN targeting exon 16 was composed of the following TAL effector domains RVDs: HD HD NI NG NN NG NG NN HD NI NI NG NN NI HD NN NG and NI NG HD HD NG NN NG HD NI NN NI NN NG NN NI NG NN HD HD NG . One-cell stage embryos were injected with 50 pg total TALEN capped messenger RNA synthesized using the mMESSAGE mMACHINE kit ( Ambion , Carlsbad , CA ) . Mutant alleles were identified by high-resolution melt analysis of PCR products generated with the following primers: gpr124-exon 7-Fo: AGGGTCCACTGGAACTGC; gpr124-exon 7-re: CAATGGAAAGGCAGCCTG; gpr124-exon 16-Fo: GCACACTCTCCTGAACACTAG; gpr124-exon 16-Re: TGCCTGGCATACAATTGGG . Capped messenger RNA was synthesized using the mMESSAGE mMACHINE kit ( Ambion , Carlsbad , CA ) . The following expression plasmids were generated and used in this study: full-length zebrafish gpr124 and gpr125 were amplified from 48 hpf cDNA and recombined into BamHI-XhoI digested pCS2 using In-fusion cloning ( Takara , Mountain View , CA ) with the following primers: gpr124-Fo: CTTGTTCTTTTTGCAGGATCCGCCACCATGTCGGGACCCTGCGTC; gpr124-Re: CTATAGTTCTAGAGGCTCGAGTTATACAGTCGTCTCGCTCTTCC; gpr125-Fo: CTTGTTCTTTTTGCAGGATCCGCCACCATGTCGGTGCTTTGCGTCC; gpr125-Re: CTATAGTTCTAGAGGCTCGAGCTACACAGTAGTTTCATGCTTCCAC . Full-length zebrafish reck was amplified from 48 hpf cDNA and recombined into BamHI-XbaI digested pCS2 using In-fusion cloning ( Takara , Mountain View , CA ) with the following primers: reck-Fo: CTTGTTCTTTTTGCAGGATCCACCATGAGCGGGTGTCTCCAGATC; reck-Re: ACGACTCACTATAGTTCTAGAGTCAGAGGTCAGAGGTCAGGG . gpr124/125 hybrids were generated by recombining two PCR fragments overlapping by 15 bp derived from the above mentioned constructs into BamHI-XhoI digested pCS2 using In-fusion cloning ( Takara , Mountain View , CA ) . In hybrids 1 and 2 , the hinge corresponds to amino acid 742/743 of Gpr124 ( VLHP742/V743IYT ) and amino acid 739/740 of Gpr125 ( PLHP739/V740IYA ) . In hybrids 3 and 4 , the hinge corresponds to amino acid 1034/1035 of Gpr124 ( HHCF1034/K1035RDL ) and amino acid 1029/1030 of Gpr125 ( HHCV1029/N1030RQD ) . Splice-blocking morpholinos against erbb3b ( TGGGCTCGCAACTGGGTGGAAACAA ) ( Lyons et al . , 2005; Budi et al . , 2008 ) , sorbs3 ( TTTCCGACAGGGAAAGCACATACCC ) ( Malmquist et al . , 2013 ) , reck ( CAGGTAGCAGCCGTCACTCACTCTC ) ( Prendergast et al . , 2012 ) and gpr124 ( ACTGATATTGATTTAACTCACCACA ) were obtained from Gene Tools ( Eugene , OR ) and injected at the one-cell stage at 0 . 5 , 4 , 1 , and 2 ng , respectively , unless otherwise stated . Injection of a standard control morpholino ( up to 8 ng ) did not affect brain vasculature or DRG formation . Host one-cell stage Tg ( kdrl:ras-mCherry ) s896 embryos were injected with the indicated morpholinos . Donor Tg ( kdrl:GFP ) s843 and host embryos were dechorionated with pronase ( 1 mg/mL ) for 5 min at 28°C in 1/3 Ringer solution supplemented with penicillin ( 50 U/mL ) /streptomycin ( 50 µg/mL ) before being incubated in agarose-coated dishes in the same medium . Twenty to 50 cells were removed from donor embryos at mid-blastula stages and transplanted along the blastoderm margin of age-matched host embryos which were subsequently grown at 28 . 5°C until the indicated stages . The contribution of GFP+ transplanted cells was assessed using a Leica M165 stereomicroscope and EC position within mosaic vessels was determined using confocal microscopy . Contribution of cells of defined genotype to the tip cell position was calculated as a percentage of the total number of mosaic CtAs or ISVs . An enhancer from the fli1a gene was used to direct endothelial transgenic expression ( Covassin et al . , 2006 ) . Expressing cells were visualized by fusing dll4 and gpr124 coding sequences to EGFP by a self-cleaving viral 2A peptide sequence . Transient mosaic endothelial overexpression was obtained by co-injecting 25 pg of Tol2 transposase mRNA and 25 pg of the pTol2-fli1:egfp , pTol2-fli1:gpr124-2A-EGFP , pTol2-fli1:dll4-2A-EGFP , pTol2-fli1:EGFP-2A-gpr124 or pTol2-fli1:EGFP-2A-dll4 constructs . Manually dechorionated embryos were heat-shocked at 26 hpf for 50 min by transferring them into egg water pre-warmed to 38°C or incubated with inhibitors starting at the 16-somite stage until the indicated developmental stage . The following chemical inhibitors were used: LiCl ( 100 mM ) , 1-AKP ( 1-Azakenpaullone; 2 . 5 μM ) , SB 216763 ( 50 μM ) , BIO ( 0 . 5 μM ) . LiCl was diluted in egg water , other drugs were prepared in 100% DMSO and diluted to the indicated concentration with egg water . As a control , volume-matched DMSO solutions in egg water were used . RNA was extracted from 20 Tg ( kdrl:GFP ) s843 gpr124 mutants and 20 wild-type siblings , cDNAs were amplified , labeled with Cy3 ( gpr124 mutants ) or Cy5 ( wild-type siblings ) , and hybridized to the Agilent Zebrafish ( V3 ) , 4x44k Gene Expression Microarray . FLAG-Gpr124 and HA-Reck were generated by recombining two PCR fragments overlapping by 15 bp derived from the above mentioned constructs into BamHI-XhoI digested pCS2 using In-fusion cloning ( Takara , Mountain View , CA ) . The FLAG peptide was inserted between the amino acid 48/49 of zebrafish Gpr124 and the HA peptide between the amino acid 23/24 of zebrafish Reck . The mouse Fz4-GFP and human FLAG-Dvl2 constructs were obtained from Addgene ( Cambridge , MA ) . HEK 293T cells were transfected with Lipofectamine 2000 ( Life Technologies , Carlsbad , CA ) 1 day after seeding and were grown in IBIDI imaging chambers for 2 days before fixation for 15 min in 4% paraformaldehyde . Indirect immunofluorescence and proximity ligation assays were performed with the following antibodies: mouse monoclonal anti-FLAG M2 ( F1804; Sigma-Aldrich , St . Louis , MO ) diluted at 1/2500 , purified polyclonal rabbit anti-HA ( H6908; Sigma-Aldrich , St . Louis , MO ) diluted at 1/250 and Alexa-conjugated secondary antibodies ( Molecular Probes , Carlsbad , CA ) diluted at 1/5000 . PLAs were performed following the manufacturer's instructions ( Sigma-Aldrich , St . Louis , MO ) . DAPI nuclear counterstaining was performed for 2 min at 5 µg/mL . The dll4 probe was generated with the following primers: dll4-ISH-Fo:GCAGCTTGGCTCACCTTTCTC and dll4-ISH-Re: TAATACGACTCACTATAGGGAGTCCTTTCTCCTGATGCCTGC and T7 was used for transcription and digoxigenin labelling . For whole-mount in situ hybridization , embryos were fixed in 4% paraformaldehyde overnight at 4°C and processed as described previously ( Thisse and Thisse , 2008 ) . STF cells were transfected with FuGeneHD ( Promega , Madison , WI ) 1 day after seeding in a 96-well tray and were harvested 2 days later . Assays were performed in triplicate , and the ratio of activities of firefly luciferase ( expressed from a stably integrated reporter with seven tandem Lef/Tcf binding sites ) to Renilla luciferase was determined for each well . Wnt signaling components were expressed from a CMV promoter . The following amounts of plasmid DNA were transfected per well . Figure 4A: Renilla luciferase ( 0 . 5 ng ) , Lrp5 or Lrp6 ( 2 . 5 ng ) , Wnt7a or Wnt7b ( 20 ng ) , Fz4 ( 20 ng ) , Gpr124 ( 20 ng ) , Reck ( 20 ng ) . Figure 4B: Renilla luciferase ( 0 . 5 ng ) , GFP ( 20 ng ) , Wnt7a ( 20 ng ) , Gpr124 ( 20 ng when held constant [right panel] , or the indicated amount [left panel] ) , Reck ( 20 ng when held constant [left panel] , or the indicated amount [right panel] ) . Figure 4C and Figure 4—figure supplement 1A: Renilla luciferase ( 0 . 3 ng ) , GFP ( 1 . 5 ng ) , Lrp5 or Lrp6 ( 1 . 5 ng ) , Wnt7a ( 3 ng ) , Gpr124 ( 0 or 15 ng ) , Reck ( 0 or 15 ng ) , and Frizzled ( 15 ng ) . Figure 4D: Renilla luciferase ( 0 . 5 ng ) , GFP ( 20 ng ) , Gpr124 ( 0 or 20 ng ) , Reck ( 0 or 20 ng ) , Wnt , Norrin , or vector control ( 20 ng ) . The mouse Reck cDNA ( GE Dharmacon; clone BC138065 , Lafayette , CO ) differs from the reference sequence in the NCBI database by a single amino acid substitution ( Thr757 in the cDNA instead of Lys757 in the NCBI database ) ; human Reck has Thr at the corresponding location . Statistical analysis was performed using GraphPad software . Data presented in bar graphs represent mean ± SD . p-values were calculated by the one-way ANOVA ( post hoc Tukey's test ) and Student's t test for multiple and single comparisons of normally distributed data ( D'Agostino & Pearson omnibus normality test ) , respectively and by the Kruskal–Wallis ( post hoc Dunn's test ) and Mann–Whitney test for multiple and single comparisons of non-normally distributed data , respectively . p-values of tip cell genotype in mosaic vessels ( Figure 7 ) were determined by the exact Fisher test . ( *p < 0 . 05; **p < 0 . 01 ) . | Organs develop alongside the network of blood vessels that supply them with oxygen and nutrients . One way that new blood vessels grow is by sprouting out of the side of an existing vessel , via a process called angiogenesis . This process relies on signals that are received by the endothelial cells that line the inner wall of blood vessels , and that direct the cells to form a new ‘sprout’ , consisting of tip and stalk cells . In the developing brain , the Wnt/β-catenin signaling pathway helps direct the formation of blood vessels . In this pathway , a member of a protein family called Wnt signals to specific proteins on the surface of the cells lining the blood vessels . Much effort has gone into uncovering the identity of these proteins , with many studies looking at blood vessel development in the brain of mouse embryos . In this study , Vanhollebeke et al . turned to zebrafish embryos to uncover new regulators of angiogenesis and define their roles during the multi-step process of blood vessel development in the brain . A variety of experimental techniques were used to alter and study the activity of different Wnt signaling pathway components . These experiments revealed that the Wnt7a and Wnt7b proteins signal to an endothelial cell membrane protein complex containing the proteins Gpr124 and Reck . Vanhollebeke et al . then created ‘mosaic’ zebrafish embryos , which contained two genetically distinct types of cells—cells that were missing one of the components of Wnt/β-catenin signaling pathway , and wild-type cells . Visualizing the growth of the vessels showed that all the new blood vessels that sprouted had normal tip cells . However , the cells in the stalk of the sprout could be either normal or missing a signaling protein . These findings demonstrate that Wnt/β-catenin signaling controls the pattern of blood vessel development in the brain by acting specifically on the invasive behaviors of the tip cells of new sprouts , a cellular mechanism that allows efficient organ-specific control of vascularization . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | Tip cell-specific requirement for an atypical Gpr124- and Reck-dependent Wnt/β-catenin pathway during brain angiogenesis |
Fluoroquinolones ( FQ ) are the recommended antimicrobial treatment for typhoid , a severe systemic infection caused by the bacterium Salmonella enterica serovar Typhi . FQ-resistance mutations in S . Typhi have become common , hindering treatment and control efforts . Using in vitro competition experiments , we assayed the fitness of eleven isogenic S . Typhi strains with resistance mutations in the FQ target genes , gyrA and parC . In the absence of antimicrobial pressure , 6 out of 11 mutants carried a selective advantage over the antimicrobial-sensitive parent strain , indicating that FQ resistance in S . Typhi is not typically associated with fitness costs . Double-mutants exhibited higher than expected fitness as a result of synergistic epistasis , signifying that epistasis may be a critical factor in the evolution and molecular epidemiology of S . Typhi . Our findings have important implications for the management of drug-resistant S . Typhi , suggesting that FQ-resistant strains would be naturally maintained even if fluoroquinolone use were reduced .
The evolution of antimicrobial resistance in bacteria is driven by the pressure of sustained exposure to antimicrobials . Such strong selective pressure has stark short and long-term consequences , as evidenced by more than half a century of antimicrobial usage and resistance evolution ( Andersson and Hughes , 2010 ) . The development of specific antimicrobial resistance patterns within sentinel organisms has spawned the popular term ‘super-bug’ ( Foster , 2004; Anzaldi and Skaar , 2011 ) . This term is misleading in the sense that resistance to a specific antimicrobial typically confers a reduction in Darwinian fitness ( a fitness cost ) in the absence of the pressure induced by that antimicrobial ( Andersson and Levin , 1999; Andersson , 2003 ) . Nevertheless , bacterial mutations that appear to have a low fitness cost or no fitness cost are sometimes observed , and mutations that may induce a fitness benefit in the absence of antimicrobials have been described but are rare ( Macvanin et al . , 2003; Enne et al . , 2004; Luo et al . , 2005; Kassen and Bataillon , 2006; Rozen et al . , 2007; Marcusson et al . , 2009; Bataillon et al . , 2011; Miskinyte and Gordo , 2013 ) . Understanding the fitness effects of antimicrobial resistance evolution is crucial for controlling the spread of resistance , as the fitness cost induced by antimicrobial resistance is one of the few biological features of resistant organisms that can be leveraged against them . Over the past 20 years , as a consequence of resistance to multiple first-line antimicrobials in the Enterobacteriacea , clinicians have become progressively reliant on the fluoroquinolones for treating infections caused by this Gram-negative bacterial family ( Parry , 2004; Vinh et al . , 2011 ) . The increasing usage and dependency on the fluoroquinolones has coincided with resistance to fluoroquinolones becoming common within the Enterobacteriacea ( Le et al . , 2009; Butler , 2011 ) . This change is highly pertinent for diseases like typhoid , a severe invasive infection caused by the bacterium Salmonella enterica serovar Typhi ( Parry et al . , 2002 ) , as antimicrobial therapy is essential for treatment . Typhoid remains common in many developing countries , and increasing minimum inhibitory concentrations ( MICs ) to fluoroquinolones , such as ciprofloxacin and ofloxacin , correlate with increased fever clearance times and treatment failure ( Parry et al . , 2011 ) . The elevated MICs to fluoroquinolones observed in S . Typhi are a consequence of mutations in the DNA gyrase gene , gyrA , and secondary mutations in the topoisomerase gene , parC ( Chau et al . , 2007; Parry et al . , 2010 ) . Evidence of strong selection for these mutations can be observed within the phylogenetic structure of S . Typhi , as several different gyrA mutations have arisen independently in multiple lineages ( Roumagnac et al . , 2006; Holt et al . , 2008 ) . To determine how fluoroquinolone-resistance evolution in S . Typhi impinges on the relative fitness of this pathogen , we directly competed strains in a series of controlled experiments , using isogenic strains to isolate the fitness effects of specific mutations , and employing pyrosequencing for precise measurement of allele frequencies . In classical competition assays ( Lenski , 1991; Lenski et al . , 1998 ) , antimicrobial-susceptible and antimicrobial-resistant organisms are competed over many generations , and the frequencies of resistant and sensitive strains are compared at various time points . The relative fitness of the resistant strain to the sensitive strain can be calculated from the population trajectories observed in the experiment ( Macvanin et al . , 2003; Enne et al . , 2005; Gagneux et al . , 2006; Balsalobre and de la Campa , 2008 ) . The choice of bacterial strains is critical in performing this type of competitive growth assay . Competing genetically unrelated clinical isolates ( Laurent et al . , 2001; Wichelhaus et al . , 2002 ) , or strains that are otherwise imperfectly isogenic , may make it difficult to isolate the effects of a single mutation ( Enne et al . , 2005; Komp Lindgren et al . , 2005; Gagneux et al . , 2006; MacLean and Buckling , 2009; O’Regan et al . , 2010 ) . Furthermore , bacterial enumeration and selective culturing after serial dilutions are typically used to calculate population sizes ( Laurent et al . , 2001; Wichelhaus et al . , 2002; Macvanin et al . , 2003; Gagneux et al . , 2006; Rozen et al . , 2007; Balsalobre and de la Campa , 2008; MacLean and Buckling , 2009; Randall et al . , 2008 ) , and these methodologies can be affected by experimental variation or spontaneous mutations in the target gene ( s ) as a consequence of exposure to low levels of antimicrobial . To overcome these limitations , we performed competitive growth experiments with isogenic strains containing single and multiple mutations in fluoroquinolone target genes , and we used pyrosequencing to assay allele frequency , avoiding exposing the organisms to antimicrobials . As gyrA mutations and susceptibility to fluoroquinolones appear to be under strong selective pressure in S . Typhi , we assessed the biological fitness of S . Typhi strains with a relevant complement of mutations in the gyrA and parC genes . We demonstrate that a large proportion of clinically and epidemiologically relevant gyrA mutations induce significant fitness benefits in S . Typhi in the absence of antimicrobial pressure . We show that strong epistatic interactions between loci in the gyrA and parC genes of S . Typhi confer additional significant selective advantages that may be responsible for driving the evolution and current regional expansion of S . Typhi in the developing countries .
We constructed 12 individual S . Typhi mutants , seven of which have been isolated clinically , by introducing one or more single mutations into the gyrA and parC genes of a host S . Typhi strain by allelic exchange ( Turner et al . , 2006 ) . A strain description and the minimum inhibitory concentrations ( MICs ) against nalidixic acid and a range of fluoroquinolones are shown in Table 1 . The seven mutants with naturally occurring equivalents all demonstrated significant increases in MICs over the parent strain with all tested antimicrobials . The increases in MICs were comparable to those observed in clinical isolates with the corresponding natural mutations ( Chau et al . , 2007; Parry et al . , 2010 ) . Double mutants exhibited greater MICs than single mutants and the triple mutant exhibited the highest MIC to all tested fluoroquinolones ( Parry et al . , 2010 ) . The four strains without a naturally occurring counterpart also demonstrated higher MICs than the parent strain , and the control strain—a strain containing a mutation in the defunct aroC gene—demonstrated no significant difference from the parent S . Typhi strain ( Table 1 ) . 10 . 7554/eLife . 01229 . 003Table 1 . S . Typhi mutants constructed for this studyDOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 003S . Typhi strainGenotypeMinimum inhibitory concentrations ( μg/ml ) Nalidixic acidNorfloxacinOfloxacinCiprofloxacinGatifloxacinLevofloxacinParent BRD948ΔaroA , ΔaroC , ΔhtrA1 . 50 . 0640 . 0470 . 0080 . 0080 . 012DPT001SNP in ΔaroC ( codon 10 ) 1 . 50 . 0640 . 0470 . 0080 . 0080 . 012S83YSNP in gyrA ( codon 83 ) 2560 . 50 . 250 . 1250 . 1250 . 125S83FSNP in gyrA ( codon 83 ) 2560 . 750 . 380 . 1250 . 1250 . 125D87ASNP in gyrA ( codon 87 ) 480 . 750 . 190 . 0940 . 0640 . 064D87NSNP in gyrA ( codon 87 ) 480 . 750 . 250 . 1250 . 1250 . 125D87GSNP in gyrA ( codon 87 ) 480 . 750 . 250 . 1250 . 1250 . 25S80ISNP in parC ( codon 80 ) 30 . 190 . 0470 . 0160 . 0160 . 016D87G-S80ISNP in gyrA ( codon 87 ) and SNP in parC ( codon 80 ) 25610 . 250 . 1250 . 0940 . 094S83F-D87G2 SNPs in gyrA ( codons 83 and 87 ) 25610 . 380 . 190 . 250 . 25S83F-D87A2 SNPs in gyrA ( codons 83 and 87 ) 1921 . 50 . 380 . 250 . 380 . 25S83F-D87N2 SNPs in gyrA ( codons 83 and 87 ) 640 . 750 . 380 . 190 . 190 . 19S83F-D87G-S80I2 SNP in gyrA ( codons 83 & 87 ) and SNP in parC ( codon 80 ) 2562416823 We assessed the selective advantages/disadvantages of the 12 mutants relative to the parent S . Typhi strain through competitive growth experiments . Because ( i ) our ability to differentiate strains by bacterial culture on selective media was limited as a consequence of low MICs , and ( ii ) serial dilution and colony enumeration , with and without antimicrobial , generated extensive experimental variation between replicates ( Figure 1 ) , we developed three pyrosequencing assays to discriminate the parent S . Typhi strain from the single/double nucleotide variations in gyrA , the single mutation in parC , and the single mutation in aroC . 10 . 7554/eLife . 01229 . 004Figure 1 . Comparing two methods for calculation of allele frequencies . ( A ) Pyrosequencing-measured allele frequencies ( y-axis ) of a range of S83F/parent strain dilutions plotted against enumeration-measured frequencies ( x-axis ) ( n = 198 ) . A linear regression between the two variables ( solid black line ) explains 90% of the variation in the relationship between these two measurements . ( B ) The same 198 data points ( y-axis ) are shown plotted against the original bacterial dilution ratio ( x-axis ) . The broken line is the diagonal highlighting where predicted frequency and measured frequency would be identical . 18 measurement replicates were performed for each predicted frequency of S83F from 0 . 0 to 1 . 0 . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 004 To validate the pyrosequencing methodology , we combined known concentrations of bacterial cultures of the parent S . Typhi strain and the S83F mutant and compared bacterial dilution and enumeration against allelic frequency detection using pyrosequencing . Allele frequencies measured by bacterial enumeration and pyrosequencing demonstrated a strong linear relationship ( r2 = 0 . 90 , Figure 1 ) , with measurement by pyrosequencing exhibiting less variation among replicates than classical culture and enumeration ( p<4 × 10−4 for all nine starting dilutions between 0 . 1 and 0 . 9; 6 . 4 < F17 , 17 < 38 . 1 ) . These data indicate our pyrosequencing assay is an accurate and highly reproducible method for determining the specific allele frequencies in mixed bacterial suspensions over a range on concentrations . We combined the parent S . Typhi with each of the mutants described in Table 1 and performed 12 competitive growth experiments ( five replicates each ) , lasting between 152 and 158 generations ( 15 days ) . Selection coefficients ( s ) for the 12 mutants were defined as per-generation percentage reductions or increases in fitness , relative to the parent ( wild-type ) strain . For example , if a strain has an estimated s^=− . 03 , this means that on average this strain will produce 3% fewer surviving offspring than the parent strain , in one bacterial generation . The fitness coefficient ( w ) is defined as w = 1 + s . Maximum likelihood estimates ( MLE ) of the selection coefficients of two mutants ( D87N , S83Y ) were not statistically different from zero ( exhibiting neutrality ) , while selection coefficients for three mutants ( D87A , D87G , and the triple mutant S83F-D87G-S80I ) were slightly negative indicating that these mutants carried fitness costs ( Figure 2 ) . The largest fitness cost was observed in the triple mutant that demonstrated a 1% reduction in fitness per bacterial generation ( MLE s^=− . 010 ) . The remaining six mutants had statistically significant positive selection coefficients , ranging from 0 . 8% to 7 . 4% per bacterial generation . The six high-fitness mutants were two strains with a single mutation ( S83F and S80I ) and all four double mutants: S83F-D87A , S83F-D87G , S83F-D87N , and D87G-S80I . The single mutant with the highest selection coefficient was S83F ( MLE s^=+ . 013 ) and the double S83F-D87N mutant displayed the highest selection coefficient of all tested strains ( MLE s^=+ . 074 ) . 10 . 7554/eLife . 01229 . 005Figure 2 . Likelihood profiles for the selection coefficients from 12 competition experiments . Data generated by competing 12 S . Typhi mutants ( labeled at the top of each panel ) against the parent S . Typhi strain over approximately 150 generations . Open circles correspond to likelihood values over the entirety of the experiment ( primary y-axis ) ; the filled gray circles correspond to the maximum likelihood estimates ( MLE ) for the variance parameter σ ( secondary y-axis ) , describing the 24-hourly variance in both process and measurement . The MLE selection coefficient ( s^ ) is shown in the top right of each panel . Vertical dashed lines demark the 95% confidence intervals for the MLE s^ . Note the compressed x-axis scale in the bottom-right panel . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 005 To exclude the possibility of compensatory mutations arising during the experiments and affecting bacterial fitness ( Bataillon et al . , 2011; Sousa et al . , 2012 ) , the fitness coefficients for the first 5 days of each 15-day assay were recomputed ( Figure 3 ) . The expected pattern under a scenario of compensatory evolution would be slow evolution in the early phases of the competition experiment with more rapid selection in the later stages . With the exception of the S83Y mutant , none of the competition experiments exhibited this pattern; and for S83Y , both fitness coefficients had confidence intervals that included zero . Therefore , unless compensatory mutations occurred and became fixed in the very early stages of the competition experiments , compensation did not have an effect on our estimation of fitness coefficients . 10 . 7554/eLife . 01229 . 006Figure 3 . Fitness coefficients computed from 5 and 15 days of bacterial competition . Black boxes show fitness coefficients computed across the entire 15-day competition . White boxes show fitness coefficients computed from the first 5 days only . The ΔaroC F10T mutation is that of the control strain . Horizontal lines are 95% confidence intervals . In a situation of compensatory evolution , we would expect to see the white box to the left of the black box . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 006 All of the double mutants demonstrated highly increased fitness over the S . Typhi parent strain and all single mutants , suggesting that epistatic interactions among resistance loci may play a role in determining strain fitness . Defining fitness for the double mutants as wij = ( 1 + si ) ( 1 + sj ) + εij , we obtained MLEs for the epistasis parameters ( εij ) , all of which were statistically different from zero and positive , indicating that the combined fitness effect of two gyrA mutations ( or in one case , a gyrA and a parC mutation ) was greater than that predicted under a scenario of multiplicative non-epistasis ( Figure 4 ) . Notably , all three single mutations at codon 87 in the gyrA gene were selectively neutral or nearly-neutral unless in combination with the S83F mutation or the S80I parC mutation . The S83F and D87N mutations demonstrated the greatest degree of synergistic epistasis , with a 6 . 6% increase in fitness resulting from the epistatic interaction . The relationship among the various S . Typhi mutants evaluated in this study , their MICs , selection coefficients , and epistatic interactions are summarized in Figure 5 . 10 . 7554/eLife . 01229 . 007Figure 4 . Likelihood profiles for the epistasis coefficient ( ε^ ) from the four double mutant competition experiments . Open circles correspond to likelihood values; the filled gray circles correspond to the maximum likelihood estimate ( MLE ) for the variance parameter σ , describing the 24-hourly variance in both process and measurement . The MLE epistasis coefficient ε^ is shown in the top right of each panel . Vertical dashed lines demark the 95% confidence intervals for the MLE ε^ . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 00710 . 7554/eLife . 01229 . 008Figure 5 . Relationships among MICs , selection coefficients and epistasis parameters of S . Typhi mutants . Diagram depicts the interactions among MLE selection coefficients ( s^ ) ( x-axis ) , MICs to ciprofloxacin ( y-axis ) , and MLE epistasis coefficients ε^ . Black circles denote S . Typhi strains that have been isolated clinically , while gray circles denote S . Typhi strains that have not been isolated clinically . Lines correspond to epistatic interactions of the four double mutants , two of which have been isolated clinically ( black lines and ε^ value ) and two of which have not ( gray lines and ε^ value ) . The grayed upper half of graph highlights the current MIC breakpoint indicative of resistance and increasing risk of treatment failure ( >0 . 125 μg/ml ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 008 We also calculated the epistasis coefficient for the S83F-D87G-S80I triple mutant relative to the parent strain . The epistasis coefficient for this triple mutant was negative at −2 . 6% ( Figure 6 ) . However , this epistasis calculation assumes that a single interaction coefficient modulates the interaction of fitness effects among these three mutations . The actual interactions in nature would occur as single mutations emerging on the background of a double mutant . As two of the relevant double mutants were generated for our experiments , we were able to calculate the epistasis coefficient of S80I emerging onto a background of S83F-D87G and that of S83F emerging onto a background of D87G-S80I . The S83F-D87G mutant has been observed clinically and its interaction with the S80I mutation is antagonistic ( MLE ε^=− . 061 ) ; this is the most likely pathway to the triple mutant . The interaction between S83F and the laboratory-generated double mutant D87G-S80I is similarly negative ( MLE ε^=− . 051 ) . Likelihood profiles for the three possible epistatic interactions leading to the triple mutant are shown in Figure 6 . 10 . 7554/eLife . 01229 . 009Figure 6 . Likelihood profiles for the epistasis coefficient ( ε^ ) of three possible epistatic interactions that could have generated the triple-mutant S83F-D87G-S80I . The interaction types are described on the top of each panel . The left panel shows the epistatic interaction among three single mutations . The middle and right panels show the epistatic interaction between a single mutation and a double mutation ( joined by a hyphen ) . Open circles correspond to likelihood values; the filled gray circles correspond to the maximum likelihood estimate ( MLE ) for the variance parameter σ , describing the 24-hourly variance in both process and measurement . The MLE epistasis coefficient ε^ is shown in the top right of each panel . Vertical dashed lines demark the 95% confidence intervals for the MLE ε^ . DOI: http://dx . doi . org/10 . 7554/eLife . 01229 . 009
In exploring the Darwinian fitness of S . Typhi strains containing a range of clinically and epidemiologically relevant fluoroquinolone-resistance mutations , we endeavored to address some of the drawbacks that may hinder the calculation of accurate fitness coefficients in other bacterial systems . To this end , we generated mutants from the same parent through controlled site-directed mutagenesis , validating all nucleotide substitutions , in order to avoid the use of clinical isolates from unspecified genetic backgrounds . Additionally , we used pyrosequencing to determine allele frequencies in order to minimize assay fluctuation , bacterial enumeration error , and other biases that may result from exposing S . Typhi to antimicrobials during the experimental procedure . This pyrosequencing approach permitted a greater degree of precision than serial dilution and colony counting , and it can be readily applied to a range of experimental systems without the requirement of a phenotypic marker . Our experimental results run contrary to the dogma that antimicrobial-resistant organisms exhibit a selective disadvantage in the absence of antimicrobials . Notable exceptions to this rule include studies in quinolone-resistant Neisseria gonorrhoeae ( Kunz et al . , 2012 ) , sulphanomide-resistant E . coli ( Kassen and Bataillion , 2006 ) , and fluoroquinolone-resistant Campylobacter jejuni ( Luo et al . , 2005 ) Streptococcus pneumoniae ( Rozen et al . , 2007 ) and E . coli ( Marcusson et al . , 2009 ) , all of which described individual genotypes with both increased MIC and higher intrinsic fitness . Studies on Pseudomonas fluorescens ( Kassen and Bataillon , 2006; Bataillon et al . , 2011 ) have shown that a small fraction of laboratory-generated mutants can harbor both drug resistance and fitness benefits in the absence of antimicrobial pressure . Most recently , an investigation by Miskinyte and Gordo ( 2013 ) suggested that E . coli with antimicrobial resistance mutations have measurable fitness benefits inside macrophages . Our work adds to these finding by demonstrating a pathogen-drug combination in which the majority of resistant genotypes are associated with dramatic fitness benefits in the absence of antimicrobial pressure . Our findings have implications for the control of typhoid as enhanced fitness in the absence of antimicrobial pressure eliminates the option of prudent antimicrobial use as a public health strategy . Additionally , our results show that combinations of these mutations exhibit higher Darwinian fitness than would be expected under a scenario of independent fitness effects , showing that higher fitness in gyrA and parC mutants arises from synergistic epistasis . It has been shown previously that genome-wide epistatic interactions may aid the development of multidrug resistance ( Trindade et al . , 2009; Silva et al . , 2011 ) , but the described mutations in these studies are not found in genes that are directly related to antimicrobial resistance . The potential impact of our results must be viewed in the context of certain limitations of S . Typhi experimental systems and our current understanding of S . Typhi biology . S . Typhi is a human-restricted facultative intracellular organism , and the selection coefficients that we observed using an S . Typhi aro mutant strain in an in vitro media-derived competitive growth assay system may not be replicated in other experimental systems . The selected parent ( S . Typhi BRD948 ) is a well-studied non-invasive laboratory strain and was selected to avoid potential biological safety issues attributed to introducing known antimicrobial resistance mutations into an invasive human pathogen . If future experimentation were to be performed with invasive S . Typhi strains , this would provide better evidence that the in vitro fitness effects we observed would be similar to those observed in a natural epidemiological setting . However , there is currently no in vivo or in vitro experimental system outside humans or primates that accurately mimics an S . Typhi infection cycle . The options for future experimentation regarding the fitness of fluoroquinolone-resistance mutations in Salmonella are in vitro and ex vivo cellular systems or the available murine models . All options have their own inherent limitations with respect to typhoid fever in humans , but they also have the potential to provide informative data . S . Typhi invades intestinal epithelial cells ( M cells ) and uses the macrophage as a vehicle for systemic dissemination; therefore an epithelial cell ( Bishop et al . , 2008 ) or macrophage ( Segura et al . , 2004 ) invasion/replication assay may provide appropriate fitness measurements for S . Typhi mutants in an intracellular system . There are physiological limitations with such cellular systems such as uptake , antimicrobial exposure ( gentamycin ) , and cellular replication that may hinder experimental reproducibility . A more suitable approach may be the classical S . Typhimurium challenge model . Direct competition of attenuated S . Typhimurium mutants in a mouse model is well described , and gyrA mutations in S . Typhimurium could be generated and compared using in vivo competition assays . One potential limitation of this animal model in determining accurate selection coefficients is the brief duration of infection , which results in a small number of bacterial generations over which to observe fitness differences . The importance of replicating our finding in other systems is highlighted by two previous studies on S . Typhimurium that assessed fitness effects associated with mutations in gyrA , one of which showed a selective disadvantage of fluoroquinolone-resistant strains in the gut of chickens ( Giraud et al . , 2003 ) , and a second of which demonstrated a reduction in invasion measured in immortalized epithelial cells ( Fàbrega et al . , 2009 ) . The strains in both studies were generated by serial passage to select for resistant isolates , which may have induced a range of uncontrolled secondary mechanisms . An additional caveat in considering a suitable experimental model relates to our current understanding of S . Typhi epidemiology . Fitness advantages in a transmission context may not correspond to those observed in vitro , ex vivo , or in animal models . Genotype-by-environment interactions have been observed to have significant effects on bacterial fitness measurements ( Paulander et al . , 2009; Miskinyte and Gordo , 2013 ) , and it is not clear which environments play the largest real-life role in modulating the epidemiological fitness of different S . Typhi strains . For example , individuals that carry S . Typhi asymptomatically in their gallbladder are thought to play an important role in the maintenance and dissemination of S . Typhi strains ( Dongol et al . , 2012 ) , and S . Typhi is hypothesized to replicate extracellularly in the gallbladder ( Gonzalez-Escobedo et al . , 2011 ) . Therefore , experimental systems that mimic the replication of S . Typhi in the human gallbladder need to be developed to determine if this environment preferentially favors the onward transmission of specific genotypes . To advance our understanding of the impact of these gyrA/parC mutations on S . Typhi epidemiology , we suggest validating our findings in more illustrative physiological systems , such as the S . Typhimurium biliary carriage model ( Crawford et al . , 2010 ) or a human S . Typhi challenge model ( Pollard et al . , 2012 ) . Notwithstanding these experimental limitations , the positive selection coefficients we measure for our gyrA and parC mutations are consistent with the current understanding of the microbiology and molecular epidemiology of S . Typhi . The past two decades have seen the emergence of S . Typhi strains exhibiting reduced susceptibility to fluoroquinolones ( Wain et al . , 1997; Parry et al . , 1998 ) , resulting in the widespread distribution of these strains in almost all locations where typhoid is endemic . We now know that the majority of this epidemic is associated with one particular genotype ( H58 ) , which has swept across Asia and into Africa ( Holt et al . , 2010; Kariuki et al . , 2010; Holt et al . , 2011 ) , displacing other genotypes in the process . A potentially crucial factor catalyzing the spread of this genotype is an S83F mutation in the gyrA gene . While this mutation is not unique to this genotype ( Roumagnac et al . , 2006 ) , there is a strong association between the H58 strains and the S83F gyrA mutation . There are currently no data directly comparing S . Typhi genotype and disease outcome in typhoid patients; there is , however , a strong correlation between reduced susceptibility to fluoroquinolones ( caused by the S83F mutation ) and prolonged infection ( Parry et al . , 2011 ) . Our data suggest that the dominance of the H58 genotype , the current global clonal expansion , and the emergence of other strains with reduced susceptibility to fluoroquinolones may not stem solely from therapeutic fluoroquinolone usage . Indeed , positive selection of bacteria with gyrA mutations has been predicted by mechanisms inducing DNA supercoiling continuation via suboptimal topoisomerases and modifications that enhance the functionality of the gyrA promoter ( Marcusson et al . , 2009; Balsalobre et al . , 2011 ) . There are several outstanding questions regarding the impact of gyrA mutations on the evolution of S . Typhi that need to be addressed . ( 1 ) Why have certain high-fitness mutation combinations not arisen in natural populations of S . Typhi in the absence of fluoroquinolones ? It may simply be improbable for an individual organism to acquire two specific mutations at the same time , especially given the slightly deleterious effects of D87A and D87G . However , this would not explain the absence , until recently , of S83F-D87N ( s^=+ . 074 ) or of the S83F mutation itself ( s^=+ . 013 ) . Perhaps , for mechanistic reasons , these mutations are unlikely to emerge in the absence of fluoroquinolone pressure . ( 2 ) Why have S . Typhi strains with only certain gyrA/parC mutations been described in clinical typhoid ? We know that the S83F mutation has emerged and spread recently; therefore , we suggest that strains containing one or two additional resistance mutations may not have had enough time to achieve high frequency in the S . Typhi population . In addition , there is a general lack of systematic cross-sectional analyses for gyrA/parC mutations in S . Typhi . Data from the published studies known to us suggest that , with the exception of S83F and S83Y , the clinical frequencies of the fluoroquinolone-resistant mutants described here are below 3% ( Roumagnac et al . , 2006; Chau et al . , 2007; Parry et al . , 2010; Emary et al . , 2012 ) . However , because the majority of the bacterial isolates described in these studies are not recent , some of these genotypes may now be circulating at higher frequencies . ( 3 ) Why do S83F mutants dominate globally ? Our results suggest that the S83F and the S80I mutants would be positively selected in nature in the absence of compensatory mutations or any selective pressure induced by antimicrobial usage . Of these two mutations , it is S83F that exhibits both the greatest selection coefficient and the highest MIC to the fluoroquinolones , making the emergence or establishment of S80I variants evolutionary less likely . The S83F mutation provides the most probable primary foundation for the evolution of fluoroquinolone-resistant S . Typhi . Phylogenetic investigations have shown that the S83F mutation has arisen on multiple occasions in several lineages ( Roumagnac et al . , 2006; Holt et al . , 2008 ) . Additionally , the ability of S83F mutants to catalyze positive epistatic interactions with other otherwise neutral mutations suggests that new S83F combinations may be observed more frequently in the future ( Koirala et al . , 2012 ) . Typhoid is a disease that necessitates antimicrobial therapy and the results presented here have repercussions for typhoid therapy and control . Currently , the control of typhoid across Asia and Africa relies on fluoroquinolone treatment ( WHO , 2003 ) , often prescribed in endemic settings for any non-specific low-grade febrile disease . However , many locations are observing a preponderance of strains that have elevated MICs to fluoroquinolones , resulting from the gyrA and parC mutations we have assessed here ( Holt et al . , 2010; Kariuki et al . , 2010; Holt et al . , 2012 ) . These observations are of great concern as incremental increases in MICs to fluoroquinolones correspond directly with treatment failure and disease severity ( Parry et al . , 2011 ) . Furthermore , antimicrobial resistance and an inability to clear infection may have obvious consequences for prolonged asymptomatic carriage of these pathogens . As the majority of S . Typhi strains described here do not exhibit a fitness cost , an antimicrobial control strategy , including the withdrawal of fluoroquinolones from general usage within the population , is unlikely to reduce the population-level frequencies of antimicrobial resistance to these drugs . In fact , our results indicate that there would be a continued rise in the frequency of fluoroquinolone-resistance , even in the absence of sustained drug exposure . In conclusion , there is no other bacterial pathogen of which we are currently aware whose primary routes of drug resistance evolution are associated with such dramatic increases in intrinsic fitness . Our findings have important implications for the selection , maintenance , and treatment of drug-resistant members of the Enterobacteriacea , predicting that fluoroquinolone-resistant strains may be naturally maintained even if the use of this crucial group of antimicrobials is restricted .
The attenuated S . Typhi Ty2 strain BRD948 , containing deletions in the aroA , aroC and htrA genes , was the parent for all bacterial strains ( Tacket et al . , 1997 ) . S . Typhi BRD948 is a well-characterized laboratory S . Typhi strain , with the attenuating mutations in the aro locus making it auxotrophic for aromatic compounds . These mutations affect the ability to grow in the intracellular compartment due to limitation of exogenous aromatic metabolites at this site . This strain is safe for use in a containment level two laboratory and avoids considerations associated with the genetic manipulation and the introduction of antimicrobial resistance mutations into an invasive human pathogen ( Bishop et al . , 2008 ) . All genetic manipulations were performed using Luria–Bertani ( LB ) media , all competition assays were performed using minimal ( M9 ) media . As S . Typhi BRD948 is an aromatic auxotroph , growth media were supplemented with 40 mg/l each of l-phenylalanine and l-tryptophan , and 10 mg/l of p-aminobenzoic acid and 2 , 3-dihydroxybenzoic acid ( aro mix ) . When required , media were supplemented with chloramphenicol , ampicillin , or nalidixic acid , at concentrations of 15 , 50 and 20 mg/l , respectively . Fluoroquinolone susceptibility testing was performed by assessing the MICs for all strains against nalidixic acid , levofloxacin , ciprofloxacin , ofloxacin , and gatifloxacin by E-test on Mueller–Hinton media containing aro mix , following manufacturer’s recommendations ( AB Biodisk , Sweden ) . All mutations were constructed by allelic exchange using derivatives of the suicide vector pJCB12 containing DNA fragments incorporating point mutation ( s ) of gyrA , parC and aroC constructed by overlap extension PCR ( PfuUltra DNA polymerase [Stratagene , La Jolla , USA] ) ( Supplementary file 1 ) ( Turner et al . , 2006 ) . Fragments were ligated into pJCB12 using appropriate restriction enzyme sites ( New England Biolabs Ltd , USA ) , propagated in E . coli CC118 λpir , and introduced into S . Typhi BRD948 by electrotransformation using 0 . 1 cm cuvettes in a GenePulser XcellTM ( Bio-Rad Laboratories , Hemel Hempstead , UK ) at 1 . 6–1 . 8 kV , 25 µF and 200 O . Plasmid DNA was prepared with the plasmid midi kit ( QIAGEN , USA ) and genomic DNA was prepared with the Wizard genomic DNA purification kit ( Promega , USA ) unless stated otherwise . Chloramphenicol-resistant transformants were screened by PCR ( Platinum PCR Supermix’ [Invitrogen , USA] ) to identify successful recombinants . Recombinants were grown in the absence of antimicrobial selection to allow homologous recombination and derivatives were selected by growth in LB-broth containing chloramphenicol and ampicillin followed by plating on NaCl-free LB-agar supplemented with 5% sucrose . Appropriate colonies were selected and stored at −80°C until required . Mutants were screened for susceptibility to ciprofloxacin using antimicrobial discs or , for the parC and aroC mutants , by quantitative real-time PCR . The gyrA , gyrB and parC loci in all mutants were sequenced to ensure the correct nucleotide substitutions . Bacterial cultures were agitated and 100 µl aliquots were mixed with 900 µl of sterile phosphate buffered saline ( PBS ) . Bacterial cultures were diluted , with thorough mixing , to 10−8 dilutions in 900 µl of sterile PBS . For enumeration , 3 × 20 µl of the 10−5 , 10−6 , 10−7 , 10−8 dilutions were inoculated onto quadrants of LB aro mix plates with and without nalidixic acid , were appropriate . Plates were incubated overnight at 37°C and the dilution from each plate containing <20 colonies per 20 µl were counted; the median number of colonies was recorded . The number of bacterial generations was calculated from the number of cell divisions required to generate the stationary phase cfu/ml−1 from the starting inoculum . All experimental bacterial strains are shown in Table 1 . Individual S . Typhi mutants were grown overnight on LB aro mix agar without antimicrobial supplementation . Single colonies were picked and inoculated into 10 ml of M9 aro mix broth and incubated overnight ( 16 hr +/− 2 hr ) at 37°C with agitation . The bacterial growth at stationary phase was enumerated to ensure standardization in bacterial colony saturation . Through a series of growth dynamics experiments over 24-hr time periods we found that the parent S . Typhi strain and the mutants generated reproducible and comparable stationary phase bacterial counts . The concentration of bacterial cells was measured by OD600 and adjusted with M9 broth to match the parent S . Typhi strain , prior to mixing for competitive growth assays . 5 µl of the parent S . Typhi strain ( approximately 1 × 107 organisms ) was inoculated in 10 ml of M9 aro mix broth into an Erlenmeyer flask concurrently with the same concentration as a mutant . Both inocula were enumerated at the time of mixing to ensure accuracy ( 1:1 ratio ) . The bacterial mixture was incubated for a period of 24 hr ( ± 2 hr ) at 37°C with circular agitation ( speed 3 . 6 , Lab companion SI-300 shaking incubator , South Korea ) . The following day , 50 µl of the stationary phase bacterial culture was removed for colony counting , 1 ml was stored at −80°C for DNA extraction and 10 µl was transferred into a secondary sterile Erlenmeyer flask containing 10 ml of M9 aro mix . The bacterial culture was incubated as before , and after a 24-hr growth period , identical volumes were removed and processed as before . Each experiment continued for a period of 15 days . Bacterial cells from each time point were thawed and DNA was extracted using heat treatment and phenol-chloroform purification . Briefly , a 100 µl aliquot of the concluding bacterial culture was agitated and incubated at 100°C for 10 min before returning to ambient temperature . The resulting solution was centrifuged and an equal volume of phenol-chloroform was added . The mixture was vortexed and centrifuged at 13 , 200 rpm for 30 min in a benchtop microfuge ( Eppendorf , USA ) . The aqueous layer was removed , placed in a sterile microfuge tube and prepared immediately for pyrosequencing . The DNA from the competitive growth assays was PCR amplified ( Platinum PCR Supermix [Invitrogen , USA] ) in triplicate using biotinylated primer pairs targeting the region containing the SNP distinguishing the two organisms in the assay , that is mutations in gyrA , parC and aroC ( Supplementary file 1 ) . PCR amplifications were performed in 60 µl reactions containing 1 × NH4 buffer , 1 . 5 mM of MgCl2 , 200 µM of dNTP , 10 pM of each primer , 1 . 25 Units of Hotstart DNA polymerase ( Qiagen , USA ) and 5 µl of template DNA . Reactions were cycled once at 95°C for 15 min , followed by 30 cycles of 94°C for 1 min , 55°C for 1 min and 72°C for 1 min , with a final elongation of 72°C for 5 min . All PCR amplifications were visualized on 1% agarose gels prior to pyrosequencing . A pyromark Q96 ID DNA pyrosequencer ( Biotage , Sweden ) was used to detect the proportion of each allele in the competitive assays at each time point as per the manufacturer’s recommendations . PCR amplicons were combined with 56 µl of binding buffer and 4 µl of streptavidin sepharose beads . The resulting mixture was agitated for 5 min before denaturation in denaturation buffer and washing with the Vacuum Prep Tool ( Biotage , Sweden ) . DNA fragments were transferred into a 96-well plate containing 3 . 5 pmol of sequencing primer in 40 µl of annealing buffer and the DNA sequencing reaction was performed using the Pyro Gold Kit ( Biotage , Sweden ) . The ratio of the parent S . Typhi strain to the engineered mutant was determined by the ratio of the non-wild-type allele to wild-type allele at the known SNP position . The allelic quantification mode in the software PyromarkID v1 . 0 ( Biotage , Sweden ) was used to quantify the proportion of each allele at each time point . The accuracy of allele frequency measurement by the pyrosequencer was validated by comparing the predicted value of the S83F gyrA mutation and the parent strain . Single colonies of S . Typhi BRD948 and the engineered S83F mutant were cultured separately in 10 ml of M9 aro , overnight at 37°C with agitation . Bacterial concentrations ( OD600 ) were determined and adjusted to the same value with M9 aro . These cultures were then combined in ratios of 0:10 , 1:9 , 2:8 , 3:7 , 4:6 , 5:5 , 6:4 , 7:3 , 8:2 , 9:1 and 10:0 in a total volume of 200 µl . DNA was extracted as before from 100 µl of the mixed cultures . DNA was subjected to pyrosequencing and compared to the expected frequency of the mutant , as before . The bacterial mixtures were additionally enumerated by colony counting and compared to the expected frequency of the mutant and the pyrosequencing data . These experiments were replicated 18 times . The results were compared by linear regression and F-test to compare variances; statistical analyses were performed in MATLAB ( Mathworks , Natick , MA , USA ) . For each strain , five independent competition assays were performed over a 15-day period . The competition period was modeled and fit with a standard Wright-Fisher model consisting of a wild type of unit fitness ( w = 1 ) and a mutant with fitness equal to w = 1 + s , using the allelic frequencies generated by SNP-specific pyrosequencing . In this type of model , fitness differences are expressed on a per-generation basis , meaning that during one generation of bacterial replication a mutant strain is expected , on average , to generate 1 + s offspring for every one offspring generated by the wild-type ( parent ) strain; if the estimate of s is negative , the mutant is less fit than the wild type and generates , on average , fewer surviving offspring than the wild type . Using yt to denote the measured frequency of an allele at time t , and zt to denote the true allele frequency at time t , the likelihood function for a single 15-day competition experiment was defined as:∏t=115 ∫01f ( yt+1| zt , σ1 ) . g ( zt| yt , σ2 ) dzt , where g is the probability density function describing the combined measurement and sampling error during the pyrosequencing procedure , and f is the probability density describing the process error from time t to time t + 1 and the measurement error at time t + 1 . Because the probability of measurement error at time t + 1 will be included in the next term in the product—when we factor in the likelihood of observation yt+2 conditioned on yt+1—we can simply view f as describing process variation . Both the density functions f and g were modeled as normal distributions , truncated outside the closed interval [0 , 1] , and renormalized to integrate to unity . Because the allele frequencies did not approach within one standard deviation of the frequency boundaries zero and one , a normal distribution was an appropriate approximation of the binomial Wright–Fisher process for the density function f . As all of the allele frequency trajectories were quite regular , we set σ1 = σ2 in order to have a single variance parameter for the system describing the variation in allele frequency introduced in a 24-hr period . Likelihoods values were multiplied across the five replicates . In the likelihood expression above , the density function f depends on the fitness coefficient s via the mean of this normal distribution , which is ( 1 + s ) zt/ ( 1 + szt ) . Likelihood optimization was performed with a standard Nelder–Mead method ( C++ with GSL Library , http://gnu . org/software/gsl ) , and confidence intervals were obtained using likelihood profiles ( Figures 2 , 4 and 6 ) . For the analysis of epistasis parameters , the same likelihood equation was used , with the fitness coefficient of the double mutant defined as wij = ( 1 + si ) ( 1 + sj ) + εij; the parameters si and sj are the selection coefficients for strains with mutations i and j , respectively , and εij is the multiplicative epistasis parameter for the strain containing both mutations i and j . In this case , likelihoods were computed across 15 replicates—five replicates each of the two strains with single mutations , and five replicates of the strain with both mutations . The corresponding epistasis equation for a triple mutant is wijk = ( 1 + si ) ( 1 + sj ) ( 1 + sk ) + εijk; and in this case the epistatic interaction is defined as the fitness interaction among all three mutations . For the triple mutant , the epistatic interaction can also be modeled as the fitness interaction between a single new mutation emerging onto a genetic background already containing two mutations . In this case , the fitness is written down as wijk = ( 1 + sij ) ( 1 + sk ) + ε ( ij ) ( k ) . Both types of epistatic interactions for the triple mutant were considered , and estimates of the epistasis parameter are shown in Figure 6 . | The fluoroquinolones are a group of antimicrobials that are used to treat a variety of life-threatening bacterial infections , including typhoid fever . Before the introduction of antimicrobials , the mortality rate from typhoid fever was 10–20% . Prompt treatment with fluoroquinolones has reduced this to less than 1% , and has also decreased the severity of symptoms suffered by people with the disease . Now , however , the usefulness of many antimicrobials , including the fluoroquinolones , is threatened by the evolution of antimicrobial resistance within the bacterial populations being treated . Drug resistance in bacteria typically arises through specific mutations , or following the acquisition of antimicrobial resistance genes from other bacteria . It is thought that the frequent use of antimicrobials in human and animal health puts selective pressure on bacterial populations , allowing bacterial strains with mutations or genes that confer antimicrobial resistance to survive , while bacterial strains that are sensitive to the antimicrobials die out . At first it was thought that specific mutations conferring antimicrobial resistance came at a fitness cost , which would mean that such mutations would be rare in the absence of antimicrobials . Now , based on research into typhoid fever , Baker et al . describe a system in which the majority of evolutionary routes to drug resistance are marked by significant fitness benefits , even in the absence of antimicrobial exposure . Typhoid is caused by a bacterial pathogen known as Salmonella Typhi , and mutations in two genes—gyrA and parC—result in resistance to fluoroquinolones . Baker et al . show that mutations in these genes confer a measurable fitness advantage over strains without these mutations , even in the absence of exposure to fluoroquinolones . Moreover , strains with two mutations in one of these genes exhibited a higher than predicted fitness , suggesting that there is a synergistic interaction between the two mutations . This work challenges the dogma that antimicrobial resistant organisms have a fitness disadvantage in the absence of antimicrobials , and suggests that increasing resistance to the fluoroquinolones is not solely driven by excessive use of this important group of drugs . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2013 | Fitness benefits in fluoroquinolone-resistant Salmonella Typhi in the absence of antimicrobial pressure |
Daily rhythms of food anticipatory activity ( FAA ) are regulated independently of the suprachiasmatic nucleus , which mediates entrainment of rhythms to light , but the neural circuits that establish FAA remain elusive . In this study , we show that mice lacking the dopamine D1 receptor ( D1R KO mice ) manifest greatly reduced FAA , whereas mice lacking the dopamine D2 receptor have normal FAA . To determine where dopamine exerts its effect , we limited expression of dopamine signaling to the dorsal striatum of dopamine-deficient mice; these mice developed FAA . Within the dorsal striatum , the daily rhythm of clock gene period2 expression was markedly suppressed in D1R KO mice . Pharmacological activation of D1R at the same time daily was sufficient to establish anticipatory activity in wild-type mice . These results demonstrate that dopamine signaling to D1R-expressing neurons in the dorsal striatum plays an important role in manifestation of FAA , possibly by synchronizing circadian oscillators that modulate motivational processes and behavioral output .
Circadian ( ∼24 hr ) rhythms of behavior and physiology are regulated by a distributed system of cell-autonomous circadian oscillators located in the brain and in most peripheral organs and tissues ( Bell-Pedersen et al . , 2005; Mohawk et al . , 2012 ) . In mammals , a population of coupled circadian clock cells in the hypothalamic suprachiasmatic nuclei ( SCN ) function as a master pacemaker responsible for coordinating circadian oscillators elsewhere in the brain and other tissues with daily light–dark cycles ( Welsh et al . , 2010 ) . Circadian clocks in many tissues , including multiple brain regions outside of the SCN , can also be entrained by daily cycles of food availability , independently of the SCN pacemaker ( Stokkan et al . , 2001; Mistlberger , 2011; Verwey and Amir , 2011; Mohawk et al . , 2012 ) . In rat and mouse , this is readily demonstrated by restricting food access to the middle of the light period , when nocturnal rodents normally eat little and are inactive . This induces a marked shifting of circadian oscillators and organ functions to align with the new daily feeding time , while the SCN remains coupled to the light-dark ( LD ) cycle . In many species , this is also associated with the emergence of a daily bout of locomotor activity that anticipates meal time by 1–3 hr ( Boulos and Terman , 1980; Stephan , 2002; Mistlberger , 2011 ) . Remarkably , this so-called food anticipatory activity ( FAA ) exhibits formal properties of circadian clock control , yet persists robustly after removal of the SCN ( Stephan et al . , 1979; Boulos and Terman , 1980; Marchant and Mistlberger , 1997 ) . Anticipatory rhythms can also be induced by scheduled daily access to water , salt , palatable foods , an opportunity to mate , and psychostimulant drugs ( Mistlberger and Rusak , 1987; Mistlberger , 1994; Kosobud et al . , 1998; Iijima et al . , 2002; Mendoza et al . , 2005; Honma and Honma , 2009; Webb et al . , 2009; Hsu et al . , 2010a , 2010b; Gallardo et al . , 2012; Jansen et al . , 2012; Landry et al . , 2012; Mohawk et al . , 2013 ) . The ability of animals to coordinate activity and physiology with access to critical resources is obviously adaptive . Similar circadian processes in humans could create daily windows of vulnerability to drug seeking , overeating , and other addictive behaviors . A major knowledge gap in circadian neurobiology is the location of circadian oscillators that generate food ( and other reward ) anticipatory circadian rhythms . The stimuli and neural pathways that entrain these oscillators also remain unspecified . Conventional lesion experiments have ruled out a number of brain regions as the site of circadian oscillators necessary for FAA ( Mistlberger , 1994 , 2011; Davidson , 2009 ) . This includes the dorsomedial nucleus of the hypothalamus , an area that may nonetheless participate in the expression of daytime FAA by inhibiting output from the SCN pacemaker that normally opposes activity and promotes rest during the day in nocturnal rodents ( Acosta-Galvan et al . , 2011; Landry et al . , 2011 ) . Induction of anticipatory rhythms by a range of ingestive and non-ingestive reward stimuli suggest that activation of reward circuits in the brain may play a role , perhaps as a final common entrainment pathway . This is supported by evidence that circadian clock genes exhibit daily rhythms of expression in components of the reward system , including the dorsal striatum and nucleus accumbens ( Wakamatsu et al . , 2001; Angeles-Castellanos et al . , 2007; Verwey and Amir , 2011 ) . These rhythms are inverted by daytime feeding , and , in the dorsal striatum , can be induced or reset by dopaminergic stimuli ( Iijima et al . , 2002; Hood et al . , 2010; Natsubori et al . , 2013 , 2014 ) . Importantly , FAA rhythms can be shifted by a single injection of a dopamine D2 receptor agonist ( Smit et al . , 2013 ) and can be attenuated by dopamine D1 and D2 antagonists ( Liu et al . , 2012 ) . In addition , genetic deletion of factors that activate dopamine transmission ( e . g . , ghrelin receptors ) can attenuate FAA ( Blum et al . , 2009; LeSauter et al . , 2009; Lamont et al . , 2012 ) , but see ( Szentirmai et al . , 2010; Gunapala et al . , 2011; Patton et al . , 2013 ) while deletion of factors that suppress DA transmission ( e . g . , leptin and 5HT1c receptors ) can increase FAA ( Mistlberger and Marchant , 1999; Hsu et al . , 2010c; Ribeiro et al . , 2011 ) . Lesion experiments rule out the nucleus accumbens as the site of dopamine receptors that might be necessary for FAA ( Mistlberger and Mumby , 1992 ) , but comparable experiments of the dorsal striatum have not been reported . In this study , we used dopamine-deficient mice and dopamine receptor knockout mice to show that food-entrained circadian rhythms are markedly attenuated in the absence of D1 receptors but are spared in mice lacking D2 receptors and in mice that express dopamine only in the dorsal striatum . We further show that a D1 receptor agonist administered once daily in the light period can induce an anticipatory rhythm . Analyses of total daily activity , FAA as a proportion of daily activity , body temperature , and striatal clock gene expression suggest that the FAA phenotype of D1R KO mice may involve impaired synchronization of food-entrainable oscillators combined with an alteration in the strategy for maintaining metabolic homeostasis , favoring reduced total daily activity over nocturnal hypothermia .
To determine the significance of the dopamine system in mediating FAA , we tested mice lacking either dopamine D1 receptors ( D1R KO mice ) or dopamine D2R receptors ( D2R KO mice ) . First , we investigated whether mice lacking D2R ( Kelly et al . , 1997 ) showed impairments in FAA . We measured the baseline home-cage behavior using automated video-based behavior analysis ( Steele et al . , 2007 ) of D2R KO mice and wild-type ( WT ) littermates in their home-cage environment . Prior to any dietary intervention ( ‘day -7’ ) , both groups of mice demonstrated normal nocturnal activity waveforms with no significant differences in any 1-hr bin and no increase in activity in the hours before Zeitgeber Time ( ZT ) 8 ( on a 13:11 LD cycle , ZT 12 is lights-off by convention ) ( Figure 1A ) . Next , we tested their ability to time a daily 60% CR meal fed at ZT 8 every day for 28 days , recording their behavior weekly . Data were normalized by dividing the amount of high activity behavior in each hour by the total seconds of activity over the 24 hr video recording to express a fraction of high activity per hour . Measurements taken at 14 , 21 , and 28 days of CR revealed that both WT and D2R KO mice had a large increase in high activity ( hanging , jumping , walking , and rearing ) behaviors during the 3 hr preceding feeding time ( Figure 1B–D ) . In both D2R KO and WT mice fed 60% CR daily , we observed similar acquisition and maintenance of FAA defined as normalized high activity in the 3 hr before feeding ( ZT 5–8 ) ( Figure 1E ) . As expected , mice of either genotype with ad libitum ( AL ) access to food showed very little high activity behavior in the hours preceding scheduled feeding when they were given an additional food pellet as a control for handling and disturbance ( Figure 1F ) . These results suggest that D2R is not necessary for mediating FAA on a 60% CR meal . 10 . 7554/eLife . 03781 . 003Figure 1 . Activity of D2R KO mice and WT mice on 60% CR . ( A ) The fraction of all recorded frames within each 1-hr bin on day -7 when the mice were walking , hanging , jumping , or rearing . All mice were still on an ad libitum diet . ( B , C , D ) The fraction of high activity frames for D2R WT ( n = 12 ) and KO ( n = 8 ) mice in each 1-hr bin on days 14 , 21 , and 28 of CR . Arrows indicate the bin in which the calorie restricted meal was delivered ( ZT 8 ) . Shaded boxes represent lights-off and yellow boxes indicated lights on . ( E ) The fraction of high activity in the 3 hr before feeding time ( ZT 5–8 ) on days -7 , 0 , 7 , 14 , 21 , and 28 of the study for mice on CR diets . ( F ) The fraction of high activity in the 3 hr before feeding time ( ZT 5–8 ) for mice on ad libitum diets . There were no significant differences ( Mann–Whitney ) in fraction of high activity between D2R WT and KO mice . Median data are plotted with error bars indicating interquartile ranges . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 003 We also examined mice deficient in D1R ( Drago et al . , 1994 ) . The normalized activity waveform of D1R WT and D1R KO mice prior to any dietary intervention on day -7 revealed strong nocturnal activity in D1R KO and control mice , with D1R KO mice showing slightly increased nocturnal activity compared to WT controls ( Figure 2A , Figure 2—figure supplement 1 ) . Summation of time of high activity behaviors in the 11 hr of dark yielded a median value of 93 . 5 min ( min ) for D1R WT mice ( n = 16 ) compared with a value of 173 . 8 min for D1R KO ( n = 19 ) ; this difference was statistically significant ( p = 0 . 045 , Mann–Whitney Test ) . In comparison , activity values during the lights-on period were not significantly different ( median values of 44 . 2 and 51 . 3 min were observed for D1R WT and KO , respectively; p = 0 . 46 ) These results suggest that the D1R KO mice do not exhibit locomotor defects preventing nocturnal activity . 10 . 7554/eLife . 03781 . 004Figure 2 . Activity of D1R KO ( n = 18 ) mice and WT ( n = 16 ) mice on 60% CR . ( A ) The fraction of high activity within each 1 hr bin on day -7 during which all mice were on an ad libitum diet . ( B ) The fraction of high activity on day 14 , ( C ) day 21 , and ( D ) day 28 of CR . Arrows indicate the bin in which the meal was delivered ( ZT 8 ) . Shaded boxes represent lights-off and yellow boxes indicate lights on . ( E ) Summed normalized high activity in the 3 hr before feeding ( ZT 5–8 ) for days -7 , 0 , 7 , 14 , 21 , and 28 of the mice on CR diets . ( F ) Summed normalized high activity in the 3 hr before feeding ( ZT 5–8 ) for the mice on AL feeding schedules . Bars show medians and interquartile ranges . The statistical test used was Mann–Whitney , where * indicates p < 0 . 05 , ** indicates p < 0 . 01 , and *** indicates p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 00410 . 7554/eLife . 03781 . 005Figure 2—figure supplement 1 . High activity data for D1R knockout mice in seconds ( median +/- SEM ) . High activity behaviors of D1R KO and WT ( same mice as Figure 2 ) in terms of median seconds ( unnormalized ) for ( A ) day 7 , ( B ) day 0 , ( C ) day 7 , ( D ) day 14 , ( E ) day 21 , and ( F ) day 28 . ( G ) Amount of high activity in the 3 hr preceding scheduled feeding in seconds . ( H ) Amount of total high activity over 24 hr recordings in seconds . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 00510 . 7554/eLife . 03781 . 006Figure 2—figure supplement 2 . Individual mouse normalized high activity data from n = 6 , WT and mice on day 0 and day 21 of 60% CR diet . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 00610 . 7554/eLife . 03781 . 007Figure 2—figure supplement 3 . Individual mouse normalized high activity data from n = 6 , D1R KO mice on day 0 and day 21 of 60% CR diet . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 007 When placed on a timed 60% CR feeding schedule with a ZT 8 meal time , D1R WT mice showed notable increase in activity preceding scheduled meal time by 14 days of CR while D1R KO mice showed only a small increase in activity ( Figure 2B ) . On day 21 of CR , WT control mice exhibited a significant ( p < 0 . 05 ) increase in high activity compared to D1R KO mice in ZT 6 and 7 ( Figure 2C ) . This was observed again on day 28 with WT mice showing a significantly higher ( p < 0 . 01 ) fraction of high activity in each hourly bin from ZT 5–7 ( Figure 2D ) . Summation of the amount of normalized high activity in the 3 hr preceding scheduled meal access revealed that D1R KO mice had a stable and significant impairment in FAA ( p < 0 . 01 at day 14 , p < 0 . 001 at days 21 and 28 ) from day 14 of scheduled feeding onwards ( Figure 2E ) . As expected , both D1R KO and WT mice with AL access to food showed very little normalized high activity behaviors in ZT 5–8 with the exception of day 7 when D1R WT control mice showed a small increase in activity relative to D1R KO mice ( Figure 2F ) . We also plotted the median time ( in seconds ) of high activity in D1R KO and WT mice over the course of 28 days of CR and observed that D1R KO mice had reduced overall activity on a 60% CR diet after day 14 ( Figure 2—figure supplement 1 ) . Data from representative D1R WT and KO mice on days 0 and 21 are presented in Figure 2—figure supplements 2 and 3 , respectively . To confirm these findings using conventional methods for long-term continuous recording of circadian activity rhythms in mice , D1R KO ( n = 4 ) and WT ( n = 12 ) mice were housed individually in plastic cages with horizontal running discs , within isolation cabinets with motion sensors and controlled lighting ( LD 12:12 , ∼70 lux ) , temperature ( 22 ± 2°C range ) , and humidity ( 50% ) . During AL food access , activity measured by running discs and motion sensors was nearly indistinguishable in amount and timing ( % nocturnality ) in the KO and WT groups ( Figure 3A , D ) . When food ( powdered chow mixed with corn oil 20% by weight ) was gradually restricted to a 4-hr daily meal for 32 days , beginning at ZT 6 ( 6 hr after lights-on ) , WT mice exhibited a bout of disc running that began ∼2 hr prior to meal time and increased monotonically to a peak at meal time within a few days ( Figure 3B ) . Total daily running did not change . Relative to WT mice , D1R KO mice exhibited significantly less total daily running during food restriction , and a marked reduction in FAA , expressed both as 2-hr total counts and as a ratio relative to activity during the rest of the day , excluding meal time ( p < 0 . 05; Figure 3E ) . Equivalent effects were evident in activity measured by motion sensors ( not shown ) . KO mice weighed significantly less than WT mice prior to restricted feeding ( 15 . 2 ± 2 . 6 g vs 23 . 3 ± 2 . 3 g , p < 0 . 001 ) and showed equivalent changes in weight across the 32 days of restricted feeding , losing ∼15% body weight over the first week , regaining this by the end of the second week , and remaining stable until day 32 ( weight as a percent of baseline = 96 ± 3% vs 105 ± 7% in WT vs KO , respectively , p = 0 . 22 ) . 10 . 7554/eLife . 03781 . 008Figure 3 . Disc-running activity of D1R KO ( n = 4 , red curves and bars ) and WT/HT mice ( n = 12 , blue curves and bars ) during ad-lib and temporal food restriction schedules . ( A ) Group mean ( ±SEM ) waveforms of activity in 10 min bins during ad-lib food access . Data from each mouse are averages of the last 7 days prior to restricted feeding ( red dashed lines and bars ) . Lights-on ( ZT 0–12 ) is indicated by the yellow bar . ( B ) Group mean waveforms of activity during restricted feeding ( 4 hr daily access to a moderately high fat diet ) . ( C ) Group mean waveforms of activity during restricted feeding regular chow . ( D ) Total daily activity and nocturnality ratios of WT and KO mice during ad-lib food access . ( E ) Total daily activity , FAA ( 2-hr pre-meal ) counts , and FAA ratios ( 2-hr pre-meal counts divided by activity during lights-off ) during moderately high fat chow schedule . ( F ) The same metrics as panel E , during regular chow schedule . * denotes significant difference between WT and KO , p < 0 . 05 , 1-tailed . **denotes significant difference , p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 00810 . 7554/eLife . 03781 . 009Figure 3—figure supplement 1 . Actogram data for representative D1R knockout and control mice . Representative actograms of disc running activity in D1R WT ( A and C ) and D1R KO ( B and D ) mice . Each line represents 24 hr plotted in 10 min bins from left to right . Consecutive days are aligned vertically . Bins during which activity was registered are denoted by vertical deflections of varying height in proportion to the amount of activity . Lights-off is indicated by shading . Days of restricted feeding are indicated by the vertical thin bars to the immediate right or left of the actograms . Meal time hours are outlined by thin red lines . Food was a high fat chow ( A and B ) or regular chow ( C and D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 009 The diet of powdered chow mixed with corn oil was used to encourage food intake during the limited time of availability . To determine if the palatability or caloric density of the food influences the magnitude of FAA differentially in KO and WT mice , the diet was changed to powdered chow mixed with water . The mice were fed ad-lib for 22 days and were then restricted to a 4-hr daily meal for 74 days . KO mice continued to show significantly less total activity ( p < 0 . 01 ) and less FAA ( p < 0 . 01 ) than WT mice . FAA ratios in both groups were increased relative to the ratios evident when fed high fat chow , but remained significantly lower in KO mice ( p < 0 . 05 ) . In only one KO mice were the FAA counts and the FAA ratios significantly increased during the last 10 days of restricted feeding compared to the last 10 days of ad-lib food access ( counts: paired t ( 9 ) = 6 . 52 , p < 0 . 001; ratios t ( 9 ) = 5 . 22 , p = 0 . 005 ) , whereas the counts and ratios were significantly increased in all of the WT mice at p < 0 . 00001 . Differences between KO and WT mice were stable across 74 days on this restricted feeding schedule . Body weights remained stable at ∼90% of baseline in both groups . Data from representative D1R KO and WT mice are shown in Figure 3—figure supplement 1 . As D1R KO mice show low body weight on standard diets and have a decreased interest in feeding ( Drago et al . , 1994 ) , one possible explanation for their lack of FAA on a 60% CR diet consisting of standard chow ( 5001 rodent chow; LabDiet ) is that it was insufficiently palatable to induce wakefulness that is a prerequisite for FAA . Based on our prior work demonstrating that fatty foods are more potent inducers of FAA in mice than sugary foods ( Hsu et al . , 2010b; Gallardo et al . , 2012 ) , we tested whether D1R KO mice would show FAA for a 60% CR diet consisting of a more palatable , fat-rich diet . We fed D1R KO and WT controls a 60% CR diet of ‘breeder’ chow ( 5015 mouse diet; LabDiet ) , which has 25 . 3% calories from fat ( 5001 rodent chow has 13 . 5% calories from fat ) at ZT 8 daily . D1R KO mice failed to show FAA for breeder chow , exerting only 4–8% of their total daily high activity behaviors in the 3 hr preceding scheduled meal time ( Figure 4A , C ) . Interestingly , although WT mice showed FAA for breeder chow , it was attenuated compared to that observed with standard chow , as WT mice redistributed only ∼20% of high activity behaviors to the 3 hr preceding feeding ( Figure 4A , C ) compared to 30–40% on standard chow ( Figure 2E ) . We also tested an even higher fat content chow , rodent ‘high fat diet’ ( HFD ) , in which 60% of the calories come from fat . A 60% CR diet of HFD fed once daily at ZT 8 failed to induce FAA in D1R KO mice and induced a very modest FAA in WT controls , which allocated only about 10% of total high activity behaviors to the 3 hr preceding meal time ( Figure 4B , D ) . From these experiments , we concluded that D1R KO mice do not fail to anticipate scheduled meal time due to a lack of palatability of the CR food source . 10 . 7554/eLife . 03781 . 010Figure 4 . Higher fat content diet FAA studies in D1R KO mice . ( A ) Normalized high activity behavior of D1R KO ( n = 6 ) and control ( n = 8 ) mice on day 21 of 60% CR of breeder chow diet . ( B ) Normalized high activity behavior of D1R KO ( n = 11 ) and control ( n = 6 ) mice on day 21 of 60% CR on rodent high fat diet . ( C ) Normalized high activity in the 3 hr preceding scheduled meal time for mice on a diet of 60% CR breeder chow . ( D ) Normalized high activity in the 3 hr preceding scheduled meal time for mice on a diet of 60% CR high fat diet . The statistical test used was Mann–Whitney , where * indicates p < 0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 010 Because D1R KO mice failed to anticipate a variety of 60% CR meals fed once daily , we assessed whether this defect was due to an inability to cope with reduced calories . To address this issue , we employed an alternative assay for relating hunger status to activity levels by acutely fasting D1R KO and control mice . In previous studies of the home-cage behavioral response to acute fasting in WT C57BL/6J mice , mice increased their high-activity behaviors two-fold upon acute fasting as compared to AL food access ( Gallardo et al . , 2014 ) . We measured the total high-activity behavior of D1R KO and WT controls over a 3-day period . During days 1 and 2 , all mice had AL access to standard chow . As expected , D1R WT and D1R KO mice showed similar median levels of high activity behaviors over 3 days of AL diet ( Figure 5A ) . On the third day , mice were either maintained on an AL diet or had their food removed entirely ( ‘fasted’ ) . D1R KO mice showed a strong increase in high-activity behaviors upon acute food deprivation , increasing their activity to a greater extent than fasted WT mice , with mean values of 270 . 5 min and 376 . 9 min of high activity , respectively ( Figure 5B ) . To normalize these data , we plotted the ratio of activity on fasting day 3 over that of AL day 2 ( Figure 5C ) . This ratio is expected to be around one for mice that were not food deprived , in that the activity of an individual mouse should not vary much from day-to-day . Indeed , the amount of high activity on day 2 vs day 3 was close to one for both D1R WT ( 0 . 96 ) and D1R KO mice ( 0 . 92 ) ( Figure 5C ) . However , when food deprived , WT mice had an activity ratio of 1 . 9 and D1R KO mice had a ratio of 2 . 5 ( both of which were statistically significant when compared within genotype to AL , p < 0 . 01 Mann–Whitney ) . Importantly , this assay does not measure food timing , as these mice were naive to any feeding schedule . These results suggest that the D1R KO mice are capable of up-regulating activity in response to acute food deprivation ( i . e . , hunger ) , demonstrating an intact circuitry of detecting and responding to fasting and thus ruling out gross metabolic defects . 10 . 7554/eLife . 03781 . 011Figure 5 . Activity of acutely fasted D1R KO and WT mice . ( A ) Total number of seconds of high intensity activity ( walking , hanging , jumping , or rearing ) for D1R KO ( n = 6 KO ) and WT ( n = 12 ) mice on 3 consecutive days of ad libitum diet . ( B ) Total number of seconds of high intensity activity for D1R KO ( n = 7 ) and WT ( n = 14 ) mice on 3 consecutive days . On day 1 and day 2 all mice were on an ad libitum diet , but on the third day all mice were deprived of food . ( C ) The ratio of total seconds of high activity on day 3 divided by total seconds of high activity on day 2 . Bars show medians and interquartile ranges . The statistical test used was Mann–Whitney , where ** indicates p < 0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 011 Having noted that some D1R KO mice had a small amount of FAA on a 60% CR diet ( note the interquartile range in Figure 2B–E ) and that D1R KO mice increased their activity when acutely fasted , we sought to further prompt FAA timing in D1R KO mice . We intentionally cued the D1R KO and control mice to expect food by entering the room in which they were being video recorded and disturbing their cage lids 2 hr in advance ( ) of actual feeding time . We expected that this ‘cue’ would alert D1R KO mice and perhaps facilitate FAA . We cued D1R WT and KO mice on an AL diet to assay the amount of activity that occurs as the result of a disturbance independent of FAA . We observed a small increase in high activity behaviors lasting only 10 min post-cue in both D1R WT and D1R KO mice; there were no significant differences in the amount of activity induced in D1R WT and KO mice ( Figure 6A ) . D1R KO mice on a CR diet did not show a sustained increase in activity after being cued to feeding time; their activity lasted only ∼10 min subsequent to the cue ( Figure 6B ) . WT control mice on a CR diet showed a large increase in activity induced by the cue , or were already demonstrating FAA at the time of the cue , and sustained their increased activity for the following 2 hr ( Figure 6B ) . Summation of the time of high-activity behavior 2 hr prior to feeding demonstrated a significant increase in activity in WT mice when CR compared to AL feeding , but not in D1R KO mice ( Figure 5C ) . These results suggest that even cuing the D1R KO mice to the subsequent availability of food did not trigger FAA behaviors . 10 . 7554/eLife . 03781 . 012Figure 6 . Cued handling FAA . ( A ) High activity data ( in seconds ) in 5 min bins for D1R KO ( n = 7 ) and D1R WT ( n = 8 ) mice that were disturbed 2 hr prior to scheduled feeding . ( B ) High activity data ( in seconds ) in 5 min bins for timed , calorie restricted D1R KO ( n = 6 ) and D1R WT ( n = 7 ) mice . Mice were disturbed 2 hr prior to feeding . ( C ) Summed high activity data over the cued period . The statistical test used was Mann–Whitney , where * indicates p < 0 . 05 , ** indicates p < 0 . 01 , and *** indicates p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 012 Rodents exhibit FAA in response to a timed palatable mid-day meal even with AL access to regular chow ( Mistlberger and Rusak , 1987; Mendoza et al . , 2005 ) . This model does not involve substantial weight loss and mice , in particular , show FAA for fatty meals as opposed to sugary meals ( Hsu et al . , 2010b; Gallardo et al . , 2012 ) . To gain further insight into the role of the dopaminergic system in FAA , we next studied FAA in response to scheduled palatable meals . We fed a daily meal of rodent HFD at ZT 9 to D1R KO and WT male mice daily for 14 days while they retained AL access to standard chow . The palatable meal corresponded to about 30% of their total daily caloric intake . D1R KO mice , which often fail to eat from the wire food bin and need to be fed their standard chow on the cage floor , approached the HFD meal placed in their food bin immediately and started consuming it avidly . Both WT and D1R KO mice consumed their HFD meal in less than 1 hr , spending 10% of their food-bin entry time during the hour after palatable-diet feeding ( Figure 7A ) . We noted that preprandial food-bin entry , a hallmark of HFD meal entrainment ( Hsu et al . , 2010b ) , was elevated in D1R KO mice by ZT 8 on the 14th day of scheduled palatable meals ( Figure 7B ) . On day 14 , both groups of mice exhibited similar activity patterns over the course of the day , with small , but similar , increases in activity preceding their HFD meal ( Figure 7C ) . Summing high activity in the 3 hr prior to feeding ( ZT 6–9 ) shows a similar trend of increased fraction of high activity after 10 days of HFD meal treatment for both D1R KO and control mice ( Figure 7D ) . Given how poorly D1R KO mice expressed FAA for 60% CR meals , it is surprising that daily HFD meals elicited even modest FAA , leading us to retest FAA for 60% CR meals in the same cohort of mice . 10 . 7554/eLife . 03781 . 013Figure 7 . Activity of D1R KO ( n = 9–11 ) and WT ( n = 14 ) mice on a palatable meal schedule . ( A ) Fraction of time spent entering the food bin in the hour after feeding ( ZT 9 ) on days 4 , 0 , and 7 . ( B ) The fraction of normalized food bin entry and ( C ) normalized high activity in each 1-hr bin on day 14 . ( D ) Sum of normalized high activity in the 3 hr before feeding on days 4 , 0 , 7 , 10 , 11 , and 14 . There were no statistically significant differences between groups , Mann–Whitney . Arrows indicate the bin in which the palatable meal was delivered ( ZT 9 ) . Bars show medians and interquartile ranges . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 013 Subsequent to being maintained on a daily HFD meal schedule for 2 weeks , we converted D1R KO and control mice to a standard AL chow diet without palatable meal access for 1 week , days 14–21 ( Figure 8A ) . Then , we changed their diet to a daily scheduled 60% CR meal consisting of standard chow at ZT 9 on day 22 . On the first day of 60% CR , we did not observe any differences in behavior between D1R KO and control mice as neither group showed FAA , as expected ( Figure 8B ) . However , after just 1 week ( ‘day 28’ ) of 60% CR , the D1R KO mice that had previously been treated with HFD showed marked FAA for their daily feeding , with a notable FAA peak that was as high as the night time activity peak ( Figure 8B ) . There was no significant difference between the amount of FAA exhibited between D1R KO and WT mice on day 28 ( p = 0 . 734 , Mann–Whitney ) . Remarkably , by the 14th day of 60% CR ( ‘day 35’ of the experiment ) , the D1R KO mice no longer showed FAA ( Figure 8D , G ) and the difference with control mice was highly significant ( p = 0 . 0003 , Mann–Whitney ) . This lack of FAA in D1R KO mice persisted through days 42 and 49 , whereas control mice maintained on a CR diet continued to express robust FAA ( Figure 8E , F ) . We performed a Kruskal–Wallis test to determine whether there were significant within-genotype variations in this experiment and found that WT mice showed significant differences between FAA observed at day 22 and days 28 , 35 , 42 , and 49 , whereas D1R KO mice only showed significant differences in FAA between day 22 and day 28 but not any other day , indicating that the D1R KO mice transiently up-regulate FAA whereas control mice sustain this increase . We conclude that D1R-mediated signaling is not absolutely required for FAA . 10 . 7554/eLife . 03781 . 014Figure 8 . Activity of D1R KO ( n = 9 ) and WT ( n = 14 ) mice on a 60% CR meal pre-treated with 14 days of a palatable meal schedule . ( A ) A diagram representing the feeding schedule used in this study . ( B ) The fraction of high activity each 1-hr bin on day 22 , ( C ) day 28 , ( D ) day 35 , ( E ) day 43 , and ( F ) day 49 . Arrows indicate the scheduled feeding time . Shaded boxes represent lights-off while yellow represents lights on . ( G ) The fraction high activity in the 3 hr before feeding time ( ZT 6–9 ) on day 22 , 28 , 35 , 42 , 49 . Bars show medians and interquartile ranges . The statistical test used was Mann–Whitney , where * indicates p < 0 . 05 , ** indicates p < 0 . 01 , and *** indicates p < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 014 Caloric restriction can induce torpor in WT mice , which could mask entrainment of circadian oscillators by food . Visual observations at meal time provided no evidence for torpor in D1R KO mice . To rule out defects in thermoregulation as a reason why D1R KO mice do not show FAA , we placed D1R KO ( n = 7 ) and WT ( n = 7 ) mice at 30°C continuously . After 1 week at 30°C , we tested their behavioral response to 60% CR for the next 4 weeks and did not observe FAA in D1R KO mice and noted a marked reduction in FAA for the WT mice at this elevated temperature compared to ambient ( 22–24°C ) temperature ( Figure 9—figure supplement 1 ) . To address body temperature regulation more directly , we implanted miniature radiofrequency transmitters ( Minimitter ) for continuous recording of body temperature by telemetry in D1R KO ( n = 6 ) and WT ( n = 8 ) mice , which were then maintained on 60% CR feeding at ZT 8 . The 24-hr body temperature profiles showed identical waveforms for WT and KO groups before CR schedules were started ( ‘day -3’ ) with temperatures remaining close to 34 . 5°C during the daytime and increasing to up to ∼36°C at night-time ( Figure 9A ) . By day 14 of CR , there was a similar redistribution of mean temperature waveforms in both D1R KO and WT mice , starting with a slightly elevated postprandial temperatures and steady temperature declines in the middle of the dark phase ( ZT 19 ) that continues until about 6 hr before expected meal time ( Figure 9B ) . Despite similar profiles in temperature waveforms , we did note that D1R KO mice tended to show a slightly higher temperature during the entirety of the dark cycle compared to WT controls during CR ( Figure 9B–D , Figure 9—figure supplement 2 ) . We tested for an effect of genotype and day of measurement on nocturnal temperature , finding that there were no statistically significant differences between D1R WT and KO ( genotype accounts for 5 . 63% of the total variance; F = 1 . 15; p = 0 . 3044 ) but there was a significant effect of day of experiment ( interaction accounts for 4 . 43% of the total variance; F = 4 . 53; p = 0 . 0001; 2-way repeated measures ANOVA ) . However , there was no significant difference in average temperatures between D1R KO and controls during the 3 hr preceding feeding at any point during the experiment ( Figure 9E ) . We also examined the fluctuation in temperature , ‘the ΔT’ , in the 3 hr preceding meal time by subtracting the maximum and minimum temperature values within the FAA window ( Figure 9F ) . By this metric , D1R KO and WT mice show increasing ΔT following the start of CR feeding schedules with values peaking by day 7 ( Figure 9F ) . The D1R KO mice had a similar ΔT to control at all times except for days 14 and 17 , where this difference was statistically significant ( p < 0 . 05 , day 17 ) . Given that the ΔT is similar again by day 21 , we conclude that overall there are subtle differences in temperature entrainment in D1R KO mice but that overall their metabolism is effectively modulated by CR feeding; importantly , they do not fail to show FAA due to entering a torpor state . 10 . 7554/eLife . 03781 . 015Figure 9 . Body temperature measurements . ( A ) Mean ( ± SEM ) body temperature of WT and D1R KO mice 3 days prior to initiating CR , ( B ) day 14 , ( C ) day 17 , and ( D ) day 21 of CR . ( E ) Mean ( ±SEM ) body temperature overall for each day of measurement . ( F ) Mean change in temperature in the 3 hr prior to scheduled feeding . * indicates p < 0 . 05 , Mann–Whitney; n = 6 D1R KO and n = 8 WT . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 01510 . 7554/eLife . 03781 . 016Figure 9—figure supplement 1 . FAA study of D1R KO ( n = 7 ) and WT ( n = 7 ) mice at 30°C . After 1 week at 30°C mice were placed on timed 60% CR feeding . ( A ) Normalized high activity on day 14 and ( B ) day 28 . ( C ) Fraction of high activity in the 3 hr preceding expected meal time . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 01610 . 7554/eLife . 03781 . 017Figure 9—figure supplement 2 . Entrainment of behavior and neuronal activation from D1 agonist injection . Body temperature measurements during lights-off ( A ) for the entire dark cycle and ( B ) for the last 4 hr of the dark cycle . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 017 To determine where in the striatum D1R neurons are signaling FAA , we employed the dopamine-deficient ( DD ) mouse model . In this system , the rate-limiting enzyme for dopamine production , tyrosine hydroxylase ( TH ) , has a ‘lox-stop-lox’ cassette within the first intron of the Th gene and does not make any dopamine ( Zhou and Palmiter , 1995; Zhou et al . , 1995 ) . This lox-stop-lox cassette can be removed by the action of Cre-recombinase , allowing for region specific re-activation of this allele ( Szczypka et al . , 1999 , 2001; Hnasko et al . , 2006 ) . Cre-recombinase was encoded in canine adenovirus ( CAV ) , which infects neuronal terminals and undergoes retrograde transport . CAV-inoculated mice showed expression of dopamine restricted to neurons that innervate the dorsolateral striatum ( Figure 10A , B ) . Quantitative analysis of the TH straining revealed a significant difference in restoration when comparing dorsal vs ventral striatum or anterior commissure ( aca ) , which is always devoid of TH staining ( p = 0 . 018 , Kruskal–Wallis test; p = 0 . 0486 comparing dorsal vs ventral , p = 0 . 0112 for dorsal vs aca , and p = 0 . 999 for ventral vs aca , Dunn's post-test for multiple comparisons ) . To these virally rescued DD mice ( DD-VR ) , or controls ( WT mice injected with virus ) , we presented a 60% CR meal at ZT 9 for 21 days . Initially the activity waveforms of the DD-VR mice showed a trend toward increased activity at night ( Figure 10B ) , and they did not show appreciable FAA at the 7-day-time point ( p < 0 . 05 ) ( Figure 10C ) . However , by day 14 and onwards they showed a strong FAA response , eventually having almost all of their daily activity preceding meal time as did sham surgery controls ( Figure 10D ) . From this experiment we conclude that dopamine signaling in the dorsal striatum is sufficient for acquisition of FAA . 10 . 7554/eLife . 03781 . 018Figure 10 . Viral restoration of dopamine signaling in the dorsolateral striatum of dopamine-deficient mice . ( A ) Representative tyrosine hydroxylase staining in a dopamine-deficient dorsolateral viral restoration mouse . ( B ) Quantitation of tyrosine hydroxylase expression in dopamine deficient mice ( n = 5 ) . TH immune-stained striatal sections from DD-VR mice were analyzed with MacBiophotonics ImageJ software to measure fluorescence intensities in the dorsal striatum , ventral striatum , and also in the anterior part of the anterior commissure ( aca ) , a structure that is always devoid of TH staining . For each mouse fluorescence intensity values were divided by the size of the analyzed area to generate normalized fluorescence values . ( C ) Normalized high activity in control ( normal dopamine levels , n = 4 ) and dopamine-deficient viral restoration mice ( n = 7 ) on the first day of CR . ( D ) Normalized high activity on day 7 of CR and ( E ) day 14 of CR . ( F ) Summation of normalized high activity in the 3 hr preceding meal time over the course of the experiment . * indicates p < 0 . 05 , Mann–Whitney . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 018 Circadian rhythms in mammals are generated at the single-cell level by autoregulatory transcription–translation feedback loops involving so-called circadian clock genes and their protein products . A core loop is comprised of the clock genes Per1 , Per2 , Cry1 , and Cry2 , which are positively regulated by heterodimers of BMAL1 and CLOCK and negatively regulated by hetero- and homodimers of PER and CRY proteins . Circadian clock genes exhibit 24-hr rhythms of expression in the dorsal striatum ( Harbour et al . , 2013 ) , and these rhythms can be shifted by daytime feeding schedules ( Wakamatsu et al . , 2001 ) and dopamine agonists ( Hood et al . , 2010 ) . If attenuation of FAA rhythms in D1R KO mice is due to impaired entrainment of circadian oscillators in the dorsal striatum , then daily rhythms of clock gene expression in food-restricted D1R KO mice should be significantly attenuated or differently phased relative to WT mice . WT and KO mice were fed regular chow once daily in the light period ( ZT 6–10 ) for at least 30 days and were then euthanized at 1 of 4 time points ( ZT 0 , 6 , 12 , 18 ) for quantification of Per2 expression by quantitative , reverse-transcriptase PCR . A 2-way ANOVA revealed a significant effect of time of day ( F ( 4 , 30 ) = 5 . 07 , p = 0 . 0031 ) , genotype ( F ( 1 , 30 ) = 10 . 08 , p = 0035 ) , and interaction ( F ( 4 , 30 ) = 22 . 02 , p < 0 . 0001 ) in Per2 expression . WT mice exhibited a 24-hr rhythm with peak expression at ZT 12 ( lights-off ) , which is advanced by comparison with the daily rhythm previously reported in AL fed rodents ( e . g . , Harbour et al . , 2013 ) . KO mice exhibited a markedly attenuated daily rhythm of striatal Per2 expression , due primarily to reduced expression at ZT 12 ( Figure 11 ) . These results permit a substantive conclusion that the daily rhythm of Per2 expression evident in the dorsal striatum of WT mice anticipating a daily meal is dependent on dopamine signaling at D1 receptors . 10 . 7554/eLife . 03781 . 019Figure 11 . Per2 mRNA expression measured by quantitative reverse-transcriptase PCR at 4 times of day in D1R KO ( red dashed curve ) and WT mice ( blue curve ) fed for 4 hr daily at Zeitgeber Time 6 . n = 4 mice per group per time point . ANOVA confirms a significant effect of sample time in both groups . **denotes significant difference between KO and WT mice , p < 0 . 0001 . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 019 We next asked whether deliberate timed activation of D1R neurons is sufficient for behavioral entrainment in the absence of dietary restriction . We injected WT C57BL/6J male mice daily intraperitoneally with D1R agonist SKF-81297 for 12 days ( see top panel in Figure 12 ) . As controls , we injected additional mice daily with either water as a negative control for IP injection or caffeine as a negative control for activation of the locomotor system , since caffeine induces hyperactivity even in dopamine-deficient mice it is a dopamine-independent pathway ( Joyce and Koob , 1981; Kim and Palmiter , 2003 ) . The immediate response to injection was observed as an increase in activity in caffeine and SKF-81297 mice that was more prolonged than that of water-injected controls ( Figure 12A ) . This trend continued through day 12 , where water-injected mice showed a habituation to injection but caffeine-injected mice showed increased activity ( Figure 12B ) . On day 13 , when no injection was performed , leaving all mice undisturbed ( no cage changes ) , the SKF-injected mice showed increased activity compared to both water and caffeine-injected controls during the 3 hr after scheduled injection ( Figure 12C ) . We examined mice at days 0 , 7 , 12 , and 13 for evidence of behavioral entrainment to injection . On day 0 , there was no increase in activity preceding daily injection , as expected ( Figure 12D ) . After 7 days of injection , we observed a trend toward increased activity in SKF-injected mice that was not statistically significant ( p = 0 . 0513 , Kruskal–Wallis test ) ( Figure 12E ) . By days 12 and 13 , the fraction of high activity of SKF-injected mice was significantly greater than either that of water- or caffeine-injected mice ( day 12 , SKF-81297 vs water , SKF 81297 vs caffeine [p < 0 . 01]; day 13 , SKF 81297 vs water [p < 0 . 01] , SKF-81297 vs caffeine [p < 0 . 01] ) ( Figure 12F–H ) . 10 . 7554/eLife . 03781 . 020Figure 12 . Mice were injected i . p . daily with water ( n = 9 ) , caffeine ( n = 8–9 ) , or SKF-81297 ( n = 9–10 ) . ( A ) Seconds of high activity behavior in the 2 hr after injection on day 0 , the first day of injection , ( B ) day 12 , and ( C ) on day 13 when no injection was performed . ( D ) Fraction of high activity plotted in 1 hr bins after the first day of injection , ( E ) seventh , ( F ) 12th day of injection , and ( G ) 1 day after the last injection . ( H ) The sum of normalized high activity in the 3 hr preceding scheduled injection at each behavioral measurement . * indicates p < 0 . 05 , **p < 0 . 01 , Mann–Whitney . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 02010 . 7554/eLife . 03781 . 021Figure 12—figure supplement 1 . Mice were injected i . p . with either SKF-81297 ( n = 8–12 ) or water ( n = 8–12 ) for 14 days and deprived of food each day for 4 hr post-injection to prevent any drug-induced food consumption . ( A ) High activity in the 3 hr preceding daily injection on days 0 , 7 , and 14 of the experiment . ( B ) Quantitation of c-Fos immunostaining in the dorsal striatum of D1R WT ( n = 2 ) and D1R KO ( n = 2 ) . ( C ) A low magnification ( D ) and at 10× . ( E ) c-Fos immunostaining in the dorsal striatum of a D1R KO mouse , a low magnification ( F ) and at 10× . DOI: http://dx . doi . org/10 . 7554/eLife . 03781 . 021 We repeated this experiment , omitting caffeine , and prevented food intake for 4 hr post-injection to control for drug-induced feeding as a potential factor for inducing anticipatory activity in the experiment described above . To that end , mice were injected daily for 14 days with either SKF-81297 ( n = 12 ) or water ( n = 12 ) . We observed significant ( p < 0 . 05 ) behavioral entrainment by the 14th day of injection ( Figure 12—figure supplement 1 ) . To confirm that the SKF-81297 was activating D1R neurons , we injected n = 2 D1R WT and D1R KO mice with drug and euthanized them 1 hr after injection , processing their brains of c-Fos immunostaining . We noted a marked number of cells staining in the dorsal striatum of WT but not the D1R KO mice ( Figure 12—figure supplement 1 ) . From these experiments , we conclude that timed activation of D1R neurons is sufficient to induce moderate behavioral entrainment ( about 20% of total high activity behavior in a 3-hr pre-injection window ) . Moreover , this entrainment can persist in the absence of stimuli , as at the end of day 13 anticipatory activity occurs nearly 46 hr after the last injection of drug .
In rodents , behavioral anticipation of a regularly scheduled feeding time is mediated by circadian oscillators entrainable by feeding . Localization of these oscillators and their entrainment pathways has been an enduring challenge ( Davidson , 2009; Mistlberger , 2011 ) . Recent studies suggest a role for dopamine signaling in the expression and timing of food anticipatory rhythms in mice and rats ( Liu et al . , 2012; Smit et al . , 2013 ) . In this study , we provide comprehensive evidence supporting an important role for dopamine signaling via D1R in the expression of FAA in CR mice fed in the light period . Using video-based automated behavior analysis , motion sensors , and running discs , we demonstrated that anticipation of a daily meal is markedly attenuated in mice lacking D1R . By contrast , it is normal in mice lacking D2R and in dopamine-deficient mice with local production of dopamine in the dorsal striatum . The reduction in FAA in D1R KO mice is evident both in total counts and when normalized for total daily activity . We propose that attenuated FAA in D1R KO mice may involve three distinct processes; circadian oscillator entrainment , incentive motivation , and metabolic homeostasis . Circadian clocks regulate the timing of behavior , but can also regulate the peak level of activity by variations in clock amplitude . Mammalian circadian clocks are comprised of populations of coupled circadian oscillators ( clock cells ) ( Colwell , 2000; Mohawk and Takahashi , 2011 ) . Disruption of coupling within a population of circadian oscillators can flatten aggregate clock output and attenuate rhythm peaks . A circadian oscillator entrainment interpretation of attenuated FAA in D1R KO mice is suggested by two observations . First , in parallel with attenuated FAA rhythms , D1R KO mice also exhibited a marked attenuation of a daily rhythm of clock gene ( Per2 ) expression in the dorsal striatum . Second , a daily injection of the D1 agonist SKF-81257 was sufficient to induce a rhythm of anticipatory activity that was synchronized to the injection time and that persisted during a withdrawal day . These results converge on a hypothesis that dopamine signaling via D1R in the dorsal striatum entrains circadian oscillators that regulate the timing and amplitude of FAA . We speculate that a daily rhythm of dopamine release associated with food acquisition serves to coordinate the phase of circadian clock cells in the dorsal striatum , and that genetic deletion of D1R impairs this function , thereby reducing the amplitude of the population rhythm and the magnitude of the activity rhythm driven by these oscillators . In this respect , D1R signaling in food-entrainable oscillators of the dorsal striatum may be analogous to VPAC receptor signaling in the SCN , which is critical for coupling the population of circadian oscillators that constitute this light entrained circadian pacemaker ( Harmar et al . , 2002; Aton et al . , 2005 ) . Circadian clocks are thought to directly drive or gate output of neural systems that produce observable daily rhythms of rest and activity . Circadian clocks also generate internal time cues that animals can use to discriminate time of day; for example , to permit daily time-place learning , or time-compensated-sun-compass orientation ( Mistlberger , 1994 ) . A representation of circadian clock phase , by repeated association with a fixed daily meal time , could acquire salience as an incentive stimulus and generate a daily rhythm of incentive motivation manifest as FAA . Such a model could account for the ability of some animals to anticipate two or more meal times per day ( Biebach et al . , 1991; Van der Zee et al . , 2008; Luby et al . , 2012; Mistlberger et al . , 2012; Mulder et al . , 2013 ) . It is therefore possible that attenuated FAA in D1R KO mice , which did not express FAA even when aroused 2 hr prior to meal time ( Figure 6 ) , reflects impaired incentive motivation , due to a failure to attribute incentive salience to representations of circadian phase , due to degradation of the clock signals by striatal oscillator damping , or both . An unexpected finding was that while a palatable HFD did not improve FAA in CR D1R KO mice , the same diet provided once daily did induce a weak FAA equivalently in KO and WT mice with AL access to regular chow throughout the day . Furthermore , upon transfer to the CR schedule , the KO mice showed a transient FAA that dissipated over the course of the first week . These observations suggest that enhanced palatability can compensate for deficits in oscillator entrainment or incentive motivation , but that this effect is masked by CR . We speculate that the activity phenotype in D1R KO mice reflects an additional metabolic factor . Although D1R KO mice , like WT mice , exhibited hyperactivity during acute food deprivation they were significantly less active throughout the day and night during the chronic CR schedules . This was not due to metabolic collapse , given that during chronic restriction , D1R KO mice maintained nocturnal body temperatures relative to WT mice , which were hypothermic at night and early in the light period . These results suggest a different strategy used by WT and KO mice to maintain metabolic homeostasis during CR . The WT mice may conserve energy by reducing metabolism at night and expend energy by raising body temperature and locomotor activity in anticipation of a regular daily meal time . The KO mice , by contrast , switch from an initial hyperactivity to a long-term hypoactivity of sufficient magnitude to support a more normal body temperature despite restricted calorie intake . This strategy results in low levels of activity both at night and prior to meal time . Nonetheless , the reduction in FAA is significant even when normalized against total daily activity . Thus , the food anticipation phenotype in CR D1R KO mice may be explained by an alteration in the strategy to maintain metabolic homeostasis , which damps activity non-specifically throughout the 24-hr cycle and an impairment in circadian entrainment and/or incentive motivation processes that results in a disproportionate reduction of activity in anticipation of meal time . Another notable feature of the food entrainment phenotype in D1R KO mice is that despite the marked attenuation of FAA , the preprandial rise in body temperature is essentially normal ( Figure 9 ) . Feeding schedules entrain circadian oscillators in many body tissues and brain regions . There is at present no direct evidence for a master food-entrainable pacemaker analogous to the retinorecipient light-entrainable pacemaker in the SCN , which in mammals is indispensible for entrainment to light:dark cycles . Indeed hypophysectomy eliminates preprandial temperature increases but not FAA ( Davidson and Stephan , 1999 ) , a dissociation that we have also observed with mice showing temperature entrainment without FAA ( Gallardo et al . , 2012 ) and mice showing FAA without showing preprandial temperature increases ( Luby et al . , 2012 ) . Food entrainment may involve multiple parallel entrainment pathways , acting on a fully distributed , non-hierarchical system of circadian oscillators in local circuits , each responsible for generating food-entrained rhythms in tissue specific functions . The differential effect of the D1R KO on behavior and body temperature is consistent with this model . Evidence that daily rhythms of food anticipatory activity may reflect entrainment of circadian oscillations in the dorsal striatum by daily reward schedules and D1R signaling suggest new insights into the processes that may induce and sustain daily repetitive or even addictive behaviors .
These experiments were approved by the institutional animal care committees of California Institute of Technology ( protocol number 1567 ) ; Keck Science Department of Claremont McKenna College , Pitzer College , Scripps Colleges; California State Polytechnic University , Pomona ( 13 . 029 ) , University of Washington , Seattle; and Simon Fraser University . For experiments performed at Caltech ( Figures 1 , 2 , Figure 2—figure supplements 1 , 2 , Figures 4–9 , Figure 9—figure supplements 1 , 2 ) mice were maintained on a 13:11 Light:Dark cycle and their behavior was measured by computer vision of video recordings ( described below ) . For data collected at Simon Fraser University ( Figure 3 , Figure 3—figure supplement 1 , Figure 11 ) , FAA was measured using horizontal running discs and motion sensors and mice were maintained on 12:12 Light:Dark cycles . Experiments utilizing dopamine-deficient mice with viral restorations of TH were performed at the University of Washington ( Figure 10 ) with behavioral measurements using computer vision in mice maintained on 12:12 Light:Dark cycles . Pharmacological studies of dopamine receptor 1 activation with SKF-81297 were performed at the Keck Science Department ( Figure 12 ) and Cal Poly Pomona ( Figure 12—figure supplement 1 ) in mice maintained on 12:12 Light:Dark cycles . The D1R KO mice used in this study ( Drago et al . , 1994 ) had been backcrossed on to C57BL/6J for at least eight generations . DNA was extracted from tail clips and the Drd1 locus was examined by genotyped using the following primers: neomycin CACTTGTGTAGCGCCAAGTGC , drd1 TCCTGATTAGCGTAGCATGGAC , d1 GGTGACGATCATAATGGCTACGGG . DA-deficient mice were bred and reared as described previously ( Zhou and Palmiter , 1995 ) . The DD mice were maintained by daily injections of l-DOPA until they were old enough to undergo surgery to re-establish TH expression in the dorsal striatum . Mice were maintained on l-DOPA for 2 weeks after surgery and then monitored closely when l-DOPA injections were terminated . Mice that were able to maintain at least 80% of body weight were considered ‘rescued’ . As conducted previously ( Darvas and Palmiter , 2011 ) , TH immune-stained striatal sections from DD-VR mice were analyzed with MacBiophotonics ImageJ software to measure fluorescence intensities in the dorsal striatum , ventral striatum , and also in the anterior part of the anterior commissure ( aca ) , a structure that is always devoid of TH staining . For each mouse , fluorescence intensity values were divided by the size of the analyzed area to generate normalized fluorescence values . D2R knockout mice on a C57BL/6J background were obtained from Jackson laboratory ( stock # 003190 ) . WT C57BL/6J mice ( stock #000664 ) used for D1R pharmacology were purchased from Jackson labs . Prior to being placed on special feeding protocols all mice were single-housed with AL access to food ( Rodent Chow Type 5001; Lab Diet , St . Louis , MO ) and water in standard microisolator cages . Daily food intake was measured over a 48-hr period beginning at least 3 days after single-housing . For all experiments , the mass of food provided to CR mice in any 24-hr period was equal to 60–70% of the average daily intake during AL food availability . For all experiments , during scheduled feeding , food was provided in a restricted amount or for a restricted duration in the latter half of the daily light period . In different experiments , meal onset began at ZT 6 , 8 , or 9 ( i . e . , 6 , 4 , or 3 hr prior to lights-off ) , but was consistent for all groups and individual mice within experiments . Previous studies have shown that the timing of meal onset in the light period makes little to no difference in the duration and magnitude of food anticipatory activity , therefore the different meal times used in these experiments do not affect interpretation of the results ( Mistlberger , 1994; Mistlberger et al . , 2012 ) . In the CR experiments , the amount of food provided in any 24-hr period was equal to 60–70% of the average daily intake during AL food availability . D1R KO mice were fed their food allotment on the cage floor rather than in the wire food bin as we found that they were more likely to consume food when it was more readily accessible . For palatable meal feeding schedules , we used male mice only as female mice do not anticipate scheduled palatable meals robustly ( Hsu et al . , 2010b ) . HF group received a daily meal of 0 . 8 g of Bio-SERV ( Flemington , NJ ) high-fat diet , corresponding to ∼33% of total caloric intake . To assess activity , mice were video recorded for 23 . 5–24 hr . Dim red lighting was provided during the 11 hr dark cycle with red LED lights ( LEDwholesalers . com , Hayward , CA ) to allow acceptable contrast when recording during the night cycle . Video-based activity data were analyzed using HomeCageScan 3 . 0 ( Clever Systems , Inc; Reston , VA ) ; behavioral definitions were as described previously ( Steele et al . , 2007; Hsu et al . , 2010a ) . High intensity activity was defined as walking , jumping , rearing , and hanging behaviors . Data were normalized by dividing the number of seconds per hour of a high activity behavior ( e . g . , hanging , jumping , rearing , and walking ) by the total number of seconds engaged in that behavior across the ∼24 hr video recording . For Figure 3 only , data were collecting using running wheels and motion detectors as described previously ( Smit et al . , 2013 ) . Behavioral data were exported from HomeCageScan 3 . 0 as excel files , which were analyzed using MATLAB programs to sum , average , and visualize data . Statistical tests were performed using GraphPad InStat , and graphs were produced using GraphPad Prism . For comparisons of behavioral data , we used non-parametric analysis: Mann–Whitney for comparisons of two groups and Kruskal–Wallis ANOVA for >2 groups . Sample sizes for each experiment are indicated in the figure legends . For mice in Figure 3 ( Simon Fraser University ) , powdered food was provided . Following the baseline activity measurement period , food was presented at ZT 6 for 8 hr on the first RF day , with gradually decreasing windows of food access until stable weight was maintained on a 4-hr window of food access . Food under this RF procedure was 20% corn oil by weight mixed in with normal powdered chow . This medium high caloric formulation was chosen to help D1R KO mice maintain healthy weight under RF conditions . Data collection under these conditions continued for 30 days . FAA was defined as locomotor activity in the 4-hr window preceding onset of food availability . Both raw FAA counts and FAA ratios ( the ratio of activity in the FAA window to total activity in the day ) were analyzed . Average 24 hr waveforms were constructed for each RF condition for comparisons of activity profiles including onset times , peak FAA activity , and nocturnal activity profiles . For testing whether external temperature would modulate FAA in D1R KO mice , we housed D1R KO and WT mice at 30° for 1 week before beginning a 60% CR feeding schedule . Activity was measured weekly during this experiment . For temperature monitoring experiments , mice were implanted with Mini-Mitters ( Starr Life Science Corp . , Oakmont , PA ) subcutaneously . Mice were anesthetized using isoflurane gas . Post-operative monitoring and recovery time was provided for at least 7 days before putting mice on CR feeding schedules . Body temperature , but not activity levels , was measured in these mice . SKF-81297 and caffeine were purchased from Sigma ( St . Louis , MO ) ( Figure 12 ) or Tocris ( Bristol , United Kingdom ) ( Figure 12—figure supplement 1 ) . Both were dissolved in water and filter sterilized prior to being injected i . p . daily for 12 days . Caffeine was injected at a dose of 15 mg/kg and SKF-81297 was injected at 5 mg/kg . Total injection volumes were approximately 250 µl per mouse . For the mice in Figure 12—figure supplement 1 , we decreased the dose of SKF-81297 to 3 mg/kg due to seizures induced by the drug . c-Fos staining was done as described previously ( Gallardo et al . , 2014 ) . Mice were sacrificed by sodium pentobarbital injection immediately prior to tissue collection at ZT 18 , ZT 0 , ZT 6 , or ZT 12 . Tissue was rapidly extracted , frozen on dry ice , and stored at −80°C until RNA extraction . Tissue samples were later mechanically homogenized and RNA was extracted using Trizol Reagent ( Invitrogen ) according to manufacturer's specifications . RNA concentrations were obtained using a Qubit 2 . 0 Fluorometer ( Life Technologies , Carlsbad , CA ) and concentrations were adjusted to 50 ng/µl . cDNA was synthesized from 500 ng RNA using High Capacity Reverse Transcription Kit ( Life Technologies ) . Real-time PCR was performed using 2 µl cDNA with SYBR Green FastMix ( Quanta Biosciences , Gaithersburg , MD ) in a Step-One real time PCR system ( Life Technologies ) . Gene expression was normalized to total RNA input . Primers used for PCR were as follows: Per2 forward: ACCTCCCTGCAGACAAGAA , Per2 reverse: CTCATTAGCCTTCACCTGCTT . | If you have ever traveled a long distance by plane , you will likely be familiar with jet lag . This disorientating sensation occurs because our brains have ‘internal clocks’ that keep track of the day–night cycle and control when we feel most tired or most alert . Flying rapidly from one time zone to another causes this clock to fall out of sync with the local time . It then takes time for the brain's clock to slowly adjust by responding to the levels of light and dark in the new environment . Humans—and other animals , plants , and even algae—have similar internal clocks , which are used to control behavior and predict events , such as the timing of a meal . These clocks can be set based on previous experiences of when food has been available and can be independent of those that follow the daily cycle of light and dark . Mice , for example , have internal clocks that make them more active at night and sleep during the day . However , if food is only provided during the day—say , at 2 o'clock in the afternoon—hungry mice will quickly adjust when they are awake in order to get the food as soon it is provided . Also , for a few hours before their new feeding time the mice will tend to jump and move around more; this is known as ‘food anticipatory activity’ . Researchers have been studying this activity for around 40 years , but the specific regions of the brain and the processes that support these rhythms of feeding behavior remained unknown . Now , Gallardo et al . have shown that mice need dopamine—a neurotransmitter that is often called the brain's ‘feel-good chemical’—to maintain the internal clock that supports food anticipatory activity . Neurotransmitters are chemicals that carry signals between neurons; one neuron releases the chemical , and another detects it using proteins on the neuron's surface called receptors . Two main types of receptors—called D1 receptors and D2 receptors—detect dopamine . Gallardo et al . found that D1 receptors are important for maintaining feeding-related daily rhythms , but that D2 receptors are not . Additionally , dopamine only needs to be produced in a region of the brain called the dorsal striatum for food anticipatory activity to occur . This suggests that only D1 receptors in this region influence this activity , though there are many other regions of the brain that contain these receptors . The next challenge is to unravel the neural circuits that control food anticipation behavior . For example , what ‘tells’ the neurons in the dorsal striatum that an animal is hungry ? Which of the D1 receptor expressing neurons relay the information about the timing of food anticipatory behavior and to where ? Also , if a similar clock operates in humans , testing to see if it is misregulated in people with eating disorders could help us to better understand these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Dopamine receptor 1 neurons in the dorsal striatum regulate food anticipatory circadian activity rhythms in mice |
Tissue microenvironment functions as an important determinant of the inflammatory response elicited by the resident cells . Yet , the underlying molecular mechanisms remain obscure . Our systems-level analyses identified a duration code that instructs stimulus specific crosstalk between TLR4-activated canonical NF-κB pathway and lymphotoxin-β receptor ( LTβR ) induced non-canonical NF-κB signaling . Indeed , LTβR costimulation synergistically enhanced the late RelA/NF-κB response to TLR4 prolonging NF-κB target gene-expressions . Concomitant LTβR signal targeted TLR4-induced newly synthesized p100 , encoded by Nfkb2 , for processing into p52 that not only neutralized p100 mediated inhibitions , but potently generated RelA:p52/NF-κB activity in a positive feedback loop . Finally , Nfkb2 connected lymphotoxin signal within the intestinal niche in reinforcing epithelial innate inflammatory RelA/NF-κB response to Citrobacter rodentium infection , while Nfkb2−/− mice succumbed to gut infections owing to stromal defects . In sum , our results suggest that signal integration via the pleiotropic NF-κB system enables tissue microenvironment derived cues in calibrating physiological responses .
Tight regulation of inflammatory responses is important; uncontrolled inflammation underlies various human ailments , while insufficient responses limit host defense to pathogens . Tissue-resident cells those that participate in inflammatory immune activation also exhibit functional differences by adapting to the repertoire of cell-differentiating cues present in distinct microenvironments . Indeed , macrophages and dendritic cells present in different anatomic niche display heterogeneity in inflammatory signatures ( Iwasaki and Kelsall , 1999; Stout and Suttles , 2004 ) . Likewise , a requirement for CD40 , primarily involved in B-cell maturation , in inflammatory gene expressions in endothelial cells was documented ( Pluvinet et al . , 2008 ) . Similarly , lymph node inducing lymphotoxin-β receptor ( LTβR ) was shown to be critical for innate immune responses ( Spahn et al . , 2004; Wang et al . , 2010 ) . Yet , the cellular circuitry that functions at the intersection of tissue microenvironment derived signals and those impinged upon by pro-inflammatory cytokines or pathogen-derived substances remains obscure . The NF-κB family of transcription factors plays an essential role in activating pathogen-responsive gene-expression program in tissue-resident cells . In the canonical NF-κB pathway , inflammatory cues engage NEMO-IKK2 ( NEMO-IKKβ ) kinase complex to phosphorylate inhibitory IκB proteins , the major isoform being IκBα , bound to the cytoplasmic RelA:p50 NF-κB dimers . Signal-induced phosphorylation leads to proteasomal degradation of IκBs and release of RelA:p50 dimers into the nucleus . The nuclear RelA dimers activate the expressions of pro-inflammatory chemokine and cytokine genes as well as its own inhibitor IκBα , which ensures proper attenuation of inflammatory responses in a negative feedback loop . In contrast to canonical signaling , the non-canonical pathway transduces signals from cell-differentiating cues those engage BAFFR , CD40 , or LTβR . Non-canonical signaling involves NIK and IKK1 ( NIK-IKKα ) mediated phosphorylation of Nfkb2 encoded precursor p100 bound to RelB ( Sun , 2012 ) . Subsequent proteasomal processing removes the C-terminal inhibitory domain of p100 from RelB:p100 complex to generate RelB:p52 NF-κB dimer , which mediates the expressions of organogenic chemokine genes in the nucleus ( Bonizzi et al . , 2004 ) . Molecular interaction between the non-canonical signal transducer p100 and RelA has also been charted . In its homo-oligomeric form , termed IκBδ , p100 was shown to utilize its inhibitory domain to sequester a subpopulation of the RelA:p50 dimer ( Basak et al . , 2007; Savinova et al . , 2009 ) . LTβR through non-canonical NIK-IKK1 signal inactivates IκBδ to induce a weak yet sustained RelA:p50 activity . Conversely , RelA-induced synthesis of p100 and consequent accumulation of inhibitory IκBδ was shown to exert negative feedback limiting canonical RelA activity ( de Wit et al . , 1998; Legarda-Addison and Ting , 2007; Shih et al . , 2009 ) . In addition , an alternate RelA:p52 dimer has been reported which is thought to constitute a minor kappaB DNA binding activity ( Hoffmann et al . , 2003 ) . Crosstalk between apparently distinct cell signaling pathways is known to offer diversity in cellular responses . Despite these connectivities , a plausible role of signal integration via the NF-κB system in regulating inflammatory RelA NF-κB responses has not been investigated . In a multidisciplinary study combining biochemistry , genetics , and mathematical modeling , here , we characterized a duration code that determines stimulus-specific crosstalk between canonical and non-canonical signaling in fine-tuning inflammatory RelA NF-κB activity . Through such crosstalk , LTβR sustained TLR4 triggered RelA NF-κB responses by supplementing RelA:p52 NF-κB dimer in a positive feedback loop . Finally , we established the physiological significance of crosstalk control of RelA in intestinal epithelial cells ( IECs ) , where , the NF-κB system integrates gut microenvironment derived lymphotoxin signals through Nfkb2 to calibrate innate immune responses to Citrobacter rodentium .
Given the interconnectedness of the canonical and non-canonical arms ( see Introduction and Figure 1A ) , we asked if signal integration via the NF-κB system would allow cell-differentiating cues to modulate inflammatory RelA NF-κB responses . Mathematical reconstruction of dynamic networks illuminates emergent properties , such as crosstalk ( Basak et al . , 2012 ) . To explore crosstalk control , we developed a mathematical model , which we termed the NF-κB Systems Model v1 . 0 ( Appendix-1 ) , basing on previously published single NF-κB dimer model versions ( Hoffmann et al . , 2002; Basak et al . , 2007 ) . In our mathematical model , however , we depicted nuclear activation of both the major RelA:p50 dimer and RelA:p52 dimer , which is thought to constitute a minor RelA NF-κB activity . As described in the preceding single dimer models ( Hoffmann et al . , 2002; Basak et al . , 2007 ) , signal-responsive degradation and resynthesis of IκBα , IκBβ , IκBε , and inhibitory p100/IκBδ dynamically controlled RelA activity . The model was parameterized based on literature , our own measurements ( Appendix-1 , Appendix figures 1–5 , Supplementary files 1–3 ) , and fitting procedures . Simulating individual TNFR or LTβR regime , we could recapitulate experimentally observed strong , but temporally controlled , activation of RelA NF-κB complexes during canonical IKK2 signaling or the weak induction of RelA dimers during non-canonical NIK-IKK1 signaling , respectively ( Figure 1B , Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 05648 . 003Figure 1 . Computational simulations predicting a duration code underlying crosstalk control . ( A ) A current model for RelA NF-κB activation via the canonical ( IKK2 ) or the non-canonical ( NIK-IKK1 ) pathways , respectively . RelA dimers represent both RelA:p50 and RelA:p52 . Regulation of RelA NF-κB activities through crosstalk between these two pathways has not been addressed . ( B ) Computational simulations of nuclear RelA NF-κB induction by TNFR-induced IKK2 ( top , magenta ) or LTβR-induced NIK-IKK1 signals ( bottom , green ) . ( C ) A theoretical library of 356 distinct kinase activity profiles . ( D ) A schematic describing in silico crosstalk studies . The kinase inputs were fed into the model through the IKK2 or NIK-IKK1 or both the arms . The RelA NF-κB responses , quantified as baseline corrected total area under the respective activity curves , were used for computing crosstalk indexes . ( E ) Based on their respective crosstalk indexes , top 10% combinations of theoretical IKK2 and NIK1 activity profiles were identified and duration as well as amplitude of the associated crosstalk-proficient IKK2 ( top ) or NIK-IKK1 ( bottom ) profiles were plotted . ( F and G ) The IKK2 ( F ) or NIK-IKK1 ( G ) activities were monitored by incubating GST-IκBα or GST-IκBδ with NEMO or NIK co-immunoprecipitates derived from MEFs treated with IL-1β , LPS ( F , left and right panels ) , or αLTβR ( G ) , respectively . For IKK2 assay , co immunoprecipitated IKK1 and for NIK-IKK1 assay , actin present in the cell extracts was used as loading controls . ( H ) Computational simulations predicting augmented RelA activity in LPS+αLTβR ( right panel ) and a lack of crosstalk in IL-1+αLTβR ( left ) co-treatment regimes . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 00310 . 7554/eLife . 05648 . 004Figure 1—figure supplement 1 . Analyzing the composition of signal-induced NF-κB dimers . The dynamic profiles , and the composition of the nuclear RelA NF-κB DNA binding activities induced upon TNFR ( top panel ) or LTβR ( bottom ) engagement were analyzed in EMSA and supershift assay using the indicated antibodies . A longer exposure of EMSA gel was used in the bottom panel to reveal otherwise weak LTβR induced DNA binding activities . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 00410 . 7554/eLife . 05648 . 005Figure 1—figure supplement 2 . Mathematical modeling revealing a duration code underlying signaling crosstalk . ( A ) A schematic of a typical kinase activity profile . For the indicated values of ‘a’ , ‘b’ , and ‘h’ , a theoretical library of activity profiles , as presented in Figure 1C , was generated . ( B ) Representative long or short duration as well as low or high amplitude kinase profiles were selected from the library to simulate RelA activity in response to singular IKK2 or NIK-IKK1 inputs or co-treatment regimes . All 16 NF-κB activity plots were enclosed together in a blue rectangle . Crosstalk index for different IKK2 and NIK-IKK1 combinations have been indicated in the respective NF-κB plots which indicates a dominant role of duration of the activating kinases in crosstalk . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 00510 . 7554/eLife . 05648 . 006Figure 1—figure supplement 3 . Kinase assays probing cellular activation of NEMO-IKK2 and NIK-IKK1 . ( A ) NIK associates with and stimulates an NEMO-independent pool of IKK1 , which participates in non-canonical signaling . By immunoprecipitating NIK , we could reveal LTβR induced kinase activity in the immunopellet at 8 hr post-stimulation ( lane 2 and 3 ) that specifically phosphorylated GST-IκBδ ( lane 3 and 4 ) . The specificity of the NIK-IKK1 kinase assay was further confirmed using αLTβR treated Nik−/− ( lane 1 ) or Ikk1−/− ( lane 5 , 6 ) MEFs . ( B ) Quantification of the data presented in Figure 1F revealing IKK2 activation in response to IL-1 or LPS ( top ) and Figure 1G revealing NIK-IKK1 activation in response to αLTβR ( bottom ) in WT MEFs . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 006 Next , we examined potential crosstalk between IKK2 and NIK-IKK1 inputs in augmenting RelA NF-κB response in silico . To this end , we generated a theoretical library ( Shih et al . , 2009 ) of 356 kinase activity profiles , where each member possesses distinct peak onset time , peak amplitude , and duration ( Figure 1C , Figure 1—figure supplement 2A ) . To screen for permissive crosstalk conditions , we fed this library into the model through IKK2 or NIK-IKK1 or both the arms and iteratively simulated respective RelA activities . Then , we computed RelA responses in the co-treatment regime relative to individual cell stimulations to assign crosstalk indexes to different IKK2 and NIK-IKK1 combinations ( Figure 1D ) . Plotting the dynamic features of the crosstalk-proficient kinase inputs , we could reveal a critical duration threshold; where IKK2 activities sustained for more than 2 hr were more likely to engage into crosstalk for varied peak amplitudes and inputs with shorter duration were crosstalk inefficient ( Figure 1E and Figure 1—figure supplement 2B ) . Illustrating a similar but more elaborate duration control , NIK-IKK1 activities longer than 8 hr selectively participated into crosstalk with the canonical pathway . Inflammatory mediators activate canonical IKK2 with disparate temporal controls . Consistent to the prior report ( Werner et al . , 2008 ) , our kinase assay ( ‘Materials and methods’ ) revealed that IL-1β , an important pro-inflammatory cytokine , only transiently activates IKK2 in mouse embryonic fibroblasts ( MEFs ) ( left panel , Figure 1F ) . In contrast , bacterial LPS through TLR4-induced IKK2 activity that persisted above the basal level even at 24 hr post-stimulation ( right panel , Figure 1F , Figure 1—figure supplement 3B ) ( Covert et al . , 2005 ) . Mimicking prolonged signaling during cell-differentiation processes , LTβR engagement using agonistic αLTβR antibody led to sustained activation of the non-canonical NIK-IKK1 ( Figure 1G and Figure 1—figure supplement 3A , B ) . Using these experimental kinase activities as inputs , our computational simulations revealed insulation of IL-1R signaling from LTβR-mediated crosstalk ( left panel , Figure 1H ) , but robust crosstalk between TLR4 and LTβR that amplified late RelA response upon costimulation ( right panel , Figure 1H ) . Therefore , our mathematical analyses predicted that a duration code selectively engages long lasting canonical kinase activities into crosstalk with LTβR induced NIK signal to impart stimulus specificity . To experimentally verify stimulus specificity of crosstalk control , we measured nuclear RelA activities induced in MEFs by canonical or non-canonical inducers or co-treatment regime that concomitantly activated both the pathways . IL-1R signal , in parallel to transient IKK2 activation , elicited strong RelA activity at 30 min in EMSA that was largely attenuated within 1 hr , whereas , non-canonical LTβR signal only weakly induced RelA and RelB dimers those persisted even at 24 hr ( Figure 2A ) . Indeed , we were unable to detect any significant enhancement of RelA activity , relative to IL-1 induced peak , upon costimulation ( Figure 2A ) . In comparison , canonical TLR4 induced a temporally distinct RelA NF-κB activity with an early peak at 1 hr , subsequent descend and a progressively weakened late phase between 8 hr and 24 hr ( Figure 2B ) . Corroborating our mathematical prediction , concomitant LTβR signal sustained NF-κB response triggered by TLR4 ( Figure 2B ) . Signal integration via the NF-κB system synergistically enhanced TLR4-induced late RelA activity at 24 hr in the costimulation regime with mostly unaltered RelB response relative to solitary LTβR engagement ( Figure 2B , quantification and Figure 2C ) . Sequentially engaging MyD88 and Trif , TLR4 was shown to produce extended IKK2 activity ( Covert et al . , 2005 ) . To determine if the observed stimulus specificity of crosstalk is indeed due to the duration of IKK2 , we utilized Trif-deficient MEFs that only transiently activated IKK2 upon LPS treatment ( Figure 2D ) . Despite a functional non-canonical pathway ( Figure 2—figure supplement 1 ) , LTβR was restricted from crosstalk with TLR4 in Trif-deficient cells ( Figure 2E ) , thereby , suggesting a Trif-dependent mechanism that relies on the duration of canonical IKK2 in imparting stimulus specificity of crosstalk control . 10 . 7554/eLife . 05648 . 007Figure 2 . LTβR signal sustains TLR4 , but not IL-1R , induced RelA NF-κB response . ( A ) Nuclear NF-κB activities induced in MEFs by IL-1β or αLTβR or co-treatment were resolved in EMSA using a κB site containing DNA probe . The faster migrating complex , indicated with an arrowhead , consists of RelB and the slower migrating complex activated by both canonical or non-canonical signaling , denoted with an arrow , consists of RelA dimers . The compositions of the DNA binding complexes were ascertained in Figure 2C . Right , signal corresponding to RelA NF-κB activities were quantified and graphed relative to the respective IL-1 induced peak value . Data were expressed as mean of 3 quantified biological replicates ± SEM . ( B ) EMSA result , representative of three independent biological replicates , revealing augmented late NF-κB activities in the co-treatment regime as compared to cell treatment with LPS or αLTβR alone . Right , signal corresponding to RelA NF-κB activities were similarly quantified and graphed relative to LPS induced peak value . Note , late RelA activities in the LPS+αLTβR co-treatment regime were significantly augmented from of the LPS induced activities . ( C ) Supershift analysis distinguishing RelA and RelB dimers induced in MEFs treated with LPS ( L ) or αLTβR ( B ) or both ( LB ) for 24 hr . ( D ) Kinase assay revealing transient IKK2 activities in response to LPS in Trif-deficient MEFs . ( E ) EMSA data , representative of three independent experiments , revealing a lack of NF-κB crosstalk between TLR4 and LTβR in Trif-deficient MEFs . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 00710 . 7554/eLife . 05648 . 008Figure 2—figure supplement 1 . Analyzing crosstalk control in Trif-deficient MEFs . Immunoblot revealing p100 processing into p52 upon LTβR engagement in Trif-deficient MEFs to confirm a functional non-canonical pathway in these cells . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 008 Long-lasting kinase activities are expected to elicit sustained RelA responses . Then what might be the significance of signal integration via the NF-κB system ? Interestingly , computational simulations demonstrated only a muted increment in the RelA activity with increasing duration of IKK2 ( Figure 3A ) . LTβR induced NIK-IKK1 signal relieved this saturation to fully unravel the NF-κB activation potential of long duration IKK2 signals upon crosstalk . We postulated that the difference in the RelA responses induced by long-lasting IKK2 in the presence or absence of non-canonical signal would decode into differential gene activities . 10 . 7554/eLife . 05648 . 009Figure 3 . LTβR signal augments the late expressions of TLR4-induced NF-κB target genes . ( A ) Computational simulation revealing total RelA activities induced by IKK2 inputs of various durations , in the absence or presence of LTβR induced NIK-IKK1 . ( B ) Quantitative RT-PCR measuring early ( 1 hr ) and late ( 24 hr ) expressions of chemokine and cytokine genes in WT MEFs by IL-1R or LTβR or costimulation . ( C ) Gene-expression analyses similarly revealing early ( 3 hr ) and late ( 24 hr ) expressions of chemokine and cytokine genes in WT MEFs by TLR4 or LTβR or costimulation . In ( B ) and ( C ) , data are expressed as mean of 3 quantified biological replicates ± SEM . The statistical significance was determined using two-tailed Student's t-test . ( D ) LPS-induced genes , identified in representative microarray experiments at 24 hr post-stimulation , were ranked based on their normalized crosstalk score ( bottom panels ) , which reflects synergistic gene activation in the co-treatment regime as compared to individual cell treatments for a positive value . GSEA demonstrated statistically significant enrichment of NF-κB targets ( top ) , with enrichment score of 0 . 44 for WT MEFs , among genes positively controlled through crosstalk . Hits ( middle ) indicate NF-κB response genes . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 00910 . 7554/eLife . 05648 . 010Figure 3—source data 1 . List of LPS target genes positively regulated through crosstalk . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 01010 . 7554/eLife . 05648 . 011Figure 3—source data 2 . A pre-determined list of 290 NF-κB response genes used in GSEA . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 011 To evaluate the potential gene-effect of crosstalk , we measured the expressions of several known RelA target chemokine and cytokine genes using quantitative RT-PCR . Our analyses revealed that IL-1 treatment rapidly induces the expressions of TNF , IP-10 , and MIP-1α mRNAs within 1 hr with residual expressions at 24 hr post-stimulation ( Figure 3B ) . The levels of IL-1β and RANTES mRNAs were insensitive to IL-1 treatment . Also , weak LTβR signal alone did not significantly induce the expressions of these pro-inflammatory genes in MEFs . Indeed , LTβR costimulation was ineffective in augmenting IL-1 induced early or late expressions of the chemokine and cytokine genes ( Figure 3B ) . Solitary LPS treatment not only robustly induced the expressions of TNF , IP-10 , and MIP-1α mRNAs , but also led to late accumulation of IL-1β and RANTES mRNAs at 24 hr ( Figure 3C ) . Consistent to our hypothesis , LTβR costimulation prolonged TLR4-induced gene expressions with further augmented late , but not early , expressions of IL-1β , IP-10 , MIP-1α , and RANTES mRNAs . TNF mRNA levels were insensitive to crosstalk regulation ( Figure 3C ) . Furthermore , we used microarray analyses to compare global gene expressions activated by TLR4 or LTβR or both at 24 hr post-stimulation . Estimating normalized crosstalk scores ( bottom panel , Figure 3D , Figure 3—source data 1 , ‘Materials and methods’ ) , we could reveal a synergistic effect of LTβR on TLR4 stimulated late gene expressions in WT MEFs . Out of 943 LPS induced genes , however , a select set of 114 genes was further upregulated upon costimulation . Strikingly , gene set enrichment analysis ( GSEA ) ( Subramanian et al . , 2005 ) ( Figure 3—source data 2 , ‘Materials and methods’ ) demonstrated an enrichment of NF-κB targets among genes positively controlled through crosstalk ( middle and top , Figure 3D ) . We have also noted downregulation of several LPS-induced genes in the costimulation regime those appeared less likely to be NF-κB targets in GSEA . Taken together , these analyses substantiated an important function of prolonged RelA activity in crosstalk-amplification of TLR4-induced late expressions of NF-κB target genes , particularly those encode pro-inflammatory chemokines and cytokines . To determine the mechanism underlying signal integration via the NF-κB system , we individually perturbed 105 model parameters and quantified relative changes in the crosstalk index ( ‘Materials and methods’ ) . Our parameter sensitivity analysis identified the rate constant associated with the NF-κB-induced transcription of Nfkb2 as the most critical parameter underlying crosstalk control of RelA NF-κB activity ( Figure 4A ) . As such , Nfkb2 encodes for both NF-κB inhibitor and NF-κB precursor functions . Computationally simulating individual RelA:p50 and RelA:p52 nuclear activities in the LPS , αLTβR , or costimulation regimes , we could further suggest that the Nfkb2 precursor function in generating RelA:p52 dimer is important for augmenting RelA NF-κB activity during crosstalk in WT system ( Figure 4B ) . Consistently , our modeling analyses predicted complete abrogation of crosstalk in Nfkb2−/− cells ( Figure 4B ) . 10 . 7554/eLife . 05648 . 012Figure 4 . Signal generation of RelA:p52 NF-κB dimer underlies a pro-synergistic function of Nfkb2 . ( A ) Local sensitivity analysis revealing the effect of perturbation of the individual biochemical parameters on crosstalk between TLR4 and LTβR . ( B ) Computational simulations of total RelA:p50 and RelA:p52 activities between 8 and 24 hr in response to LPS , αLTβR , or both in WT or Nfkb2 deficient systems . ( C ) Immunoblot charting cellular abundance of NF-κB/IκB proteins during signaling . Right , signal corresponding to p100 and p52 levels at 24 hr post-stimulation were quantified and graphed . ( D ) Immunoblot of RelA co-immunoprecipitates , normalized for the RelA content , obtained using whole cells extracts derived from MEFs treated with LPS ( L ) or αLTβR ( B ) or both ( LB ) for 24 hr . The quantified data demonstrates the level of RelA-p100 or RelA-p52 complexes at 24 hr post-stimulation . ( E ) Supershift analysis revealing the composition of the RelA dimers induced upon indicated cell treatments for 24 hr . Right , signal corresponding to RelA:p50 or RelA:p52 NF-κB activities were quantified and graphed . ( F ) Representative immunoblot demonstrating an increase in the RelA protein level in MEFs , in parallel to phospho-IκBα accumulation , upon proteasome inhibition using MG-132 . ( G ) EMSA revealing RelA activities induced in WT or Nfkb2−/− MEFs upon indicated cell-stimulations by supershifting RelB . Below , quantified late ( 24 hr ) RelA activities were plotted for different genotypes subsequent to normalizing against the respective LPS induced early 1 hr peak . ( H ) mRNA analyses comparing late ( 24 hr ) expressions of chemokines/cytokines in Relb−/− ( top ) and Nfkb2−/− ( bottom ) MEFs upon indicated cell stimulations . Quantified data for both biochemical and gene-expression analyses presented in this figure are expressed as mean of 3 biological replicates ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 01210 . 7554/eLife . 05648 . 013Figure 4—figure supplement 1 . An important role of the non-canonical signal transducer Nfkb2 in crosstalk . ( A ) The level of Nfkb2 mRNA ( top ) and Nfkbia mRNA , which encodes IκBα ( bottom ) , was monitored in a time course during TLR4 or LTβR signaling or in the co-stimulation regime . Strong RelA activity , observed in EMSA , led to potent induction of both the NF-κB targets in response to LPS treatment or in the co-treatment regime . LTβR that only weakly activated RelA , only subtly altered the levels of these two mRNAs . Data are expressed as mean of 3 quantified biological replicates ± SEM . ( B ) EMSA revealing NF-κB DNA binding activities in Nfkb2−/− MEFs upon cell-treatment with indicated stimuli . Note constitutive RelB DNA binding activity that largely substituted LTβR induction of RelB dimers . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 013 Our biochemical studies revealed that LPS induces RelA NF-κB-driven transcription of Nfkb2 to produce p100 ( top panel , Figure 4C and Figure 4—figure supplement 1A ) , which was shown to oligomerize as NF-κB inhibitory IκBδ ( Savinova et al . , 2009; Shih et al . , 2009; Tao et al . , 2014 ) . Concomitant LTβR signal instead utilized TLR4-induced , newly synthesized p100 to potently generate p52 , thereby , neutralizing the inhibitory p100 function ( Figure 4C ) . p50 levels were not discernibly altered in these stimulation regimes in our experiments . Our immunoprecipitation based analysis demonstrated that NIK-IKK1 signal relieves RelA from p100/IκBδ-mediated inhibition during crosstalk ( Figure 4D ) . Intriguingly , LTβR costimulation of MEFs for 24 hr also produced ∼fourfold more RelA:p52 NF-κB dimer as compared to solitary LPS treatments ( Figure 4D ) . Limited transcriptional up-regulation of Nfkb2 by weak LTβR signal was correlated with only the modest p52 and RelA:p52 generation ( Figure 4C , D ) . In supershift assay , we could ascertain that RelA:p52 dimer generated upon LTβR costimulation appears as a strong nuclear DNA binding activity at 24 hr ( Figure 4E ) to supplement to the TLR4-induced RelA NF-κB responses . An important role of dimerization in stabilizing NF-κB monomers from degradation has been reported earlier ( Fusco et al . , 2008 ) . Interestingly , RelA protein rapidly accumulated in cells upon proteasome inhibition ( Figure 4F ) suggesting a robust constitutive degradation mechanism that offsets basal synthesis of RelA monomers in maintaining the steady-state level . Our study indicated that this enduring flux ensures copious supply of RelA to bind to de novo generated p52 , produced from the newly synthesized p100 during crosstalk . Next , our genetic analyses revealed that canonical RelA:p50 response to TLR4 signal , primarily controlled through classical IκBs , is largely intact in Nfkb2−/− with early induction and diminished late activities comparable to WT MEFs ( Figure 4G and Figure 4—figure supplement 1B ) . Consistent to the prediction based on computational modeling studies , a lack of RelA:p52 dimer generation in Nfkb2−/− cells , however , ablated LTβR-mediated enhancement of TLR4-induced late RelA DNA binding activity ( Figure 4G ) as well as crosstalk amplification of RelA target pro-inflammatory gene expressions ( bottom panel , Figure 4H ) . LTβR costimulation not only enhanced TLR4 induced RelA DNA binding but also activated RelB dimers through the non-canonical pathway . Prior reports have indicated cell-type specific inhibitory as well as activating role of RelB in chemokine gene expressions ( Weih et al . , 1996; Shih et al . , 2012 ) . Importantly , costimulation of Relb−/− MEFs led to similar hyperactivation of LPS-induced late expressions of IL-1β , IP-10 , and RANTES mRNAs as in WT cells ( compare top panel , Figure 4H with Figure 3C ) . Although , the crosstalk effect on MIP-1α expressions was somewhat muted owing to prolonged expression of this gene in Relb−/− MEFs in response to solitary LPS treatment . Therefore , our analyses confirmed that the precursor function encoded by Nfkb2 in generating RelA:p52 NF-κB dimer is critical for integrating lymphotoxin derived signals to the pro-inflammatory RelA NF-κB pathway . Our analyses also suggested that LTβR costimulation led to the hyperactivation of LPS-induced expressions of chemokine and cytokine genes in an Nfkb2-dependent manner with only a minor , if any , role for Relb . Given the computational prediction for an important role of NF-κB-induced transcription of Nfkb2 , we compared the inducible expression of the crosstalk mediator Nfkb2 in response to LPS or IL-1 to understand the molecular basis of stimulus-specific control . As opposed to rapid expression of Nfkbia mRNA , which encodes IκBα , LPS induced Nfkb2 mRNA with a delay ( top panel , Figure 5A ) . Similarly , chronic TNF treatment induced Nfkb2 mRNA in WT MEFs with an explicit 1 hr delay ( Figure 5B ) that was also observed earlier and incorporated in both the previous ( Basak et al . , 2007 ) as well as the current mathematical model versions . Analogous time lags were observed in the expression of several inflammatory genes those require additional chromatin modifications for the initiation of RelA-induced transcription ( Natoli et al . , 2005 ) . When Nfkb2 transgene was stably expressed in Nfkb2−/− MEFs from an exogenous NF-κB responsive promoter , Nfkb2 mRNA was readily induced by TNF without a delay ( Figure 5B ) . Remarkably , IL-1 treatment was ineffective in activating the expression of Nfkb2 mRNA in WT MEFs , despite the early induction of Nfkbia ( bottom , Figure 5A ) . Only upon eliminating the transcriptional delay , our mathematical model could simulate Nfkb2 mRNA induction by IL-1 treatment ( Figure 5C ) . Indeed , we could also experimentally rescue the defect in Nfkb2 mRNA induction by IL-1 treatment in the engineered Nfkb2−/− cell line , which expresses Nfkb2 transgene from the NF-κB-responsive promoter without the delay ( Figure 5D ) . Consistent to our computational identification that NF-κB inducible transcription of Nfkb2 is important , disruption of NF-κB-inducible synthesis by expressing p100 from a constitutive promoter in Nfkb2−/− MEFs abrogated crosstalk amplification of TLR4-induced late NF-κB activity by concomitant LTβR signal ( Figure 5E and Figure 5—figure supplement 1A ) . While the NF-κB-responsive expression of Nfkb2 transgene in Nfkb2−/− cells restored the crosstalk effect at the level of RelA NF-κB activation ( Figure 5E ) by potentiating RelA:p52 induction in LPS+αLTβR costimulation regime ( Figure 5F , Figure 5—figure supplement 1B ) . 10 . 7554/eLife . 05648 . 014Figure 5 . Induction of Nfkb2 expressions by canonical signal is required for crosstalk . ( A ) Relative levels of Nfkb2 mRNA and Nfkbia mRNA , which encodes IκBα , in WT MEFs during LPS or IL-1 signaling . ( B ) TNF induced delayed expression of Nfkb2 mRNA in WT MEFs and rapid production in an engineered Nfkb2−/− cell-line from an exogenous NF-κB dependent promoter . Data presented in ( A ) and ( B ) are expressed as mean of 3 quantified biological replicates ± SEM . ( C ) Simulations comparing IL-1-induced Nfkb2 mRNA expressions in the presence or absence of transcriptional delay . ( D ) Quantitative RT-PCR revealing IL-1 induced expression of Nfkb2 and Nfkbia mRNAs in the engineered Nfkb2−/− cell line with transgenic expressions of Nfkb2 from the NF-κB inducible promoter . Data are expressed as mean of 3 quantified biological replicates ± SEM . ( E ) EMSA comparing NF-κB activities induced in Nfkb2−/− MEFs expressing Nfkb2 from either a constitutive ( lane 1–7 ) or an NF-κB responsive ( lane 2–14 ) transgenic ( Tg ) promoter . ( F ) Supershift analyses comparing nuclear abundance of different NF-κB dimers activated upon costimulation with LPS+αLTβR for 24 hr in these two engineered Nfkb2−/− cell lines . ( G ) The engineered Nfkb2−/− cell line , which expresses Nfkb2 from the NF-κB-inducible promoter , was pretreated for 8 hr with αLTβR and subsequently treated with IL-1 . Supershifting RelB , representative RelA activities were captured in EMSA . ( H ) A graphical depiction of the proposed crosstalk control; two negative feedback loops coordinately attenuate TLR4 responses . However , one of these negative feedback loops is converted into a positive feedback loop by non-canonical signals to generate crosstalk at the level of RelA NF-κB activation . Magenta and green arrows indicate canonical IKK2 and non-canonical NIK-IKK1 inputs , respectively , and line thickness signifies relative strength of feedbacks . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 01410 . 7554/eLife . 05648 . 015Figure 5—figure supplement 1 . Induction of Nfkb2 expression by canonical signal is important for crosstalk . ( A ) Comparing LPS induced changes in Nfkb2 mRNA levels in two engineered Nfkb2−/− cell lines expressing Nfkb2-Tg from either a constitutive or an NF-κB responsive promoter . Our analyses confirmed a lack of LPS-inducible expression of Nfkb2 from the constitutive promoter , but 3 . 5 fold induced expression from the inducible promoter . ( B ) Immunoblot comparing p52 generation upon LPS+LTβR costimulation in these two engineered Nfkb2−/− cell lines . ( C ) Left panel , immunoblot comparing p52 production in WT cells upon 8 hr of αLTβR treatment with or without subsequent 3 hr IL-1 stimulation . Similarly , the engineered Nfkb2−/− cell line , which expresses Nfkb2 from the NF-κB responsive promoter , was subjected to IL-1 treatment subsequent to 8 hr of LTβR pre-treatment ( right panel ) . ( D ) Quantification of the data presented in Figure 5G and two more experimental replicates . Data are expressed as mean of 3 quantified biological replicates ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 015 Our results confirmed that signal induction of Nfkb2 is important for crosstalk and suggested that a promoter intrinsic delay necessitates persistent canonical signal for RelA-mediated induction of pro-synergistic Nfkb2 . Such delay encoding insulated IL-1R signaling , which transiently activates IKK2 and RelA by restricting Nfkb2 mRNA expressions , and accounted for the abrogated crosstalk in Trif-deficient MEFs that transiently activated the NF-κB pathway . Our studies also explained the requirement for the long-duration NIK-IKK1 signals in targeting this late acting p100 Nfkb2 feedback for RelA:p52 dimer generation . In contrast , IL-1 signal led to early induction of Nfkb2 mRNA expressions in the engineered cells , which inducibly express Nfkb2 transgene without the delay ( Figure 5D ) . Pretreatment of these engineered cells with αLTβR for 8 hr and subsequent IL-1 stimulation that effectively converged the non-canonical signal to IL-1-induced Nfkb2 feedback , potentiated p52 production ( Figure 5—figure supplement 1C ) and prolonged IL-1-induced RelA response ( Figure 5G and Figure 5—figure supplement 1D ) . Correlating with the early onset of Nfkb2 mRNA induction in response to IL-1 treatment , observed crosstalk effects were indeed obvious within 1 hr of IL-1 treatment in these engineered cells . In sum , we elucidate a crosstalk mechanism that discriminates between TLR4 engagement and concomitant cell activation through TLR4 and LTβR ( Figure 5H ) . Negative feedbacks by IκBα and p100/IκBδ coordinately terminate canonical TLR4 response . But , Nfkb2 functions pro-synergistically upon costimulation; in a positive feedback loop , non-canonical LTβR signal targets the newly synthesized p100 , abundantly produced by TLR4 , to potently generate p52 and RelA:p52 dimers in sustaining inflammatory RelA NF-κB responses . Importantly , RelA:p50 and RelA:p52 heterodimers were shown to share DNA binding and gene-expression specificities ( Siggers et al . , 2012; Zhao et al . , 2014 ) . Our experimental data also indicated that RelA:p52 dimer has comparable efficiency in inducing the expression of Nfkb2 mRNA as the RelA:p50 dimer ( Appendix-1 , Appendix figure 4C ) . Although , emergent crosstalk is expected to be controlled by several biochemical constrains , the transcriptional delay intrinsic to the Nfkb2 promoter appears to be critical for the duration code and thereby the stimulus specificity . In addition to its role in lymph node development during embryogenesis , recent studies have illustrated a requirement for LTβR in innate immune responses in adult mice . Disruption of LTβR signal using LTβR-Ig fusion protein was shown to compromise innate immune responses upon subsequent infection with C . rodentium , a natural mouse enteric pathogen that led to mortality ( Spahn et al . , 2004; Wang et al . , 2010 ) . IEC-specific deletion of LTβR similarly obliterated bacterial clearance ( Wang et al . , 2010 ) . The engagement of ligand-expressing RORγt+ innate lymphoid cell is thought to provide the critical lymphotoxin signal in colon during the course of bacterial infection ( Upadhyay and Fu , 2013 ) . A requirement of epithelial RelA activity in the chemokine gene expressions has been documented earlier ( Alcamo et al . , 2001 ) . Given our identification of a costimulatory function of LTβR in inflammatory RelA activation , we asked if signal integration via the NF-κB system could explain the epithelial requirement of LTβR in innate immune responses in vivo . First , a functional non-canonical pathway downstream of LTβR ( Figure 6—figure supplement 1A ) augmented the RelA activity induced by pathogen sensing TLR4 in otherwise hypo-responsive MSIE colon epithelial cell-line ( Figure 6A ) . Next , we biochemically analyzed NF-κB activation in IECs derived from WT mice intraperitoneally injected with antagonistic LTβR-Ig or a control-Ig 1 day prior to oral infection with C . rodentium . Upon colonization , Citrobacter initially triggered epithelial accumulation of p100 that was fully processed into p52 by day5 ( Figure 6B ) generating RelA:p52 dimer ( Figure 6C ) in control-Ig , but not LTβR-Ig , treated mice . Bacterial infection elicited RelA DNA binding activity in IECs that gradually accumulated in the nucleus ( Figure 6D , E ) with substantial contribution from RelA:p52 dimer along with RelA:p50 dimer at day5 post-infection ( Figure 6—figure supplement 1B ) . Our supershift analyses further confirmed complete absence of RelB containing NF-κB DNA binding activity in IECs derived from infected mice ( Figure 6E ) . Perturbing LTβR signal attenuated RelA NF-κB activation with more obvious defects at day5 ( Figure 6D ) . Likewise , pathogen-responsive RelA activation in IECs derived from Nfkb2−/− mice was severely weakened at day5 ( Figure 6F ) that led to significantly reduced expressions of the RelA target chemokines encoding KC and MIP-2α as compared to WT mice ( Figure 6G ) . Indeed , infected Nfkb2−/− mice exhibited diminished neutrophil recruitment in the lamina propria , as revealed by anti-myeloperoxidase immunostaining of the colon sections ( Figure 6H ) . Sustained epithelial RelA activity that relies on LTβR mediated processing of pathogen-induced p100 into p52 , therefore , mirrored our MEF-based analyses depicting crosstalk between canonical and non-canonical signaling . Collectively , our results connected the previously reported epithelial requirement of LTβR ( Wang et al . , 2010 ) and NIK ( Shui et al . , 2012 ) in innate immune response to the NF-κB system in reinforcing RelA activity through Nfkb2 mediated crosstalk control . Subdued epithelial NF-κB activation , and not hyper-induction , in IECs from infected Nfkb2−/− mice also suggested that a dominant precursor function of p100 supplying RelA:p52 dimer prolongs RelA response within the intestinal niche . 10 . 7554/eLife . 05648 . 016Figure 6 . Nfkb2 dependent LTβR crosstalk prolongs RelA NF-κB response in the colon of Citrobacter rodentium-infected mice . ( A ) EMSA data , representing two independent experiments , revealing LPS-induced total RelA NF-κB activities induced in MSIE colon epithelial cell line at 12 hr in the absence or presence of 1 μg/ml of αLTβR . ( B ) , ( C ) , and ( D ) WT mice were injected with control-Ig or LTβR-Ig ( n = 2 ) 1 day prior to infection with C . rodentium . IECs were isolated at day3 and day5 post-infection and analyzed for p52 and p100 levels by immunoblotting ( B ) , RelA:p52 complex formation by immunoprecipitation-based assay ( C ) or NF-κB DNA binding activities in EMSA ( D ) . OCT1 DNA binding activity served as loading control . ( E ) Supershift analyses revealing that exclusively RelA/NF-κB dimer are activated in IECs derived from mice infected with C . rodentium . ( F ) NF-κB activities induced in IECs derived from infected WT and Nfkb2−/− mice ( n = 2 ) were similarly measured . ( G ) Epithelial expressions of KC and MIP-2a mRNA derived from WT and Nfkb2−/− mice ( n = 5 ) at day5 post-infection . Data are expressed as mean of 3 quantified biological replicates ± SEM . The statistical significance was determined using two-tailed Student's t-test . ( H ) Representative data showing antimyeloperoxidase staining of neutrophils in colons of WT and Nfkb2−/− mice at day4 post-infection . Colon sections from three animals per set and five fields/section were used for quantification and presented as mean ± SEM . The panels with 40× magnification have been presented using scale bars that represent 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 01610 . 7554/eLife . 05648 . 017Figure 6—figure supplement 1 . Control of epithelial RelA NF-κB activation through signaling crosstalk . ( A ) Immunoblot showing functional non-canonical LTβR pathway generating p52 in MSIE colon epithelial cell line . ‘*’ indicates a non-specific protein band . ( B ) Supershift analyses to confirm the induction of both p50 as well as p52 containing RelA/NF-κB dimers in IECs derived from mice infected with C . rodentium . Right , quantified data representing the relative abundance of individual RelA/NF-κB complexes . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 017 While WT mice efficiently eliminated infections , increased fecal excretion of bacteria at day10 post-infection in Nfkb2−/− mice ( Figure 7A ) indicated an inadequacy in limiting local infection , thereby , correlating with the observed defects in the early innate inflammatory response in this knockout ( Figure 6 ) . Histological analysis of the shrunken colons ( Figure 7—figure supplement 1A , B ) derived from the infected Nfkb2−/− mice further revealed exacerbated damage with signatures of submucosal leukocyte infiltration ( Figure 7B ) . Breach in the intestinal barrier was accompanied by systemic bacterial dissemination with increased count in blood ( left panel , Figure 7C ) and liver ( right , Figure 7C ) . Finally , bacterial colitis induced in Nfkb2−/− mice resulted in significant body weight loss ( Figure 7D ) and onset of mortality as early as day10 post-infection ( Figure 7E ) . Next , we performed reciprocal bone marrow transfer experiments between WT and Nfkb2−/− mice to ensure that the observed sensitivity was not due to previously reported hematopoietic defects in Nfkb2−/− mice ( Caamaño et al . , 1998 ) . WT bone marrow cells in Nfkb2−/− recipients ( Figure 7—figure supplement 1C ) were unable to prevent the infection-related colon pathology ( Figure 7F ) , reductions in the body weight ( Figure 7G ) and morality ( Figure 7H ) . In contrast , WT recipients receiving either WT or Nfkb2−/− bone marrow resolved infections with comparable efficiencies ( Figure 7F–H ) . 10 . 7554/eLife . 05648 . 018Figure 7 . A protective role of pro-synergistic Nfkb2 in the non-hematopoietic compartment to Citrobacter infection . ( A ) and ( C ) Bacterial titers in the fecal homogenate ( A ) , blood ( C , left panel ) or spleen and liver homogenate ( C , right ) derived from WT and Nfkb2−/− mice ( n = 4 ) at day10 post-infection . Data are expressed as mean ± SEM . The statistical significance was determined using two-tailed Student's t-test . ( B ) and ( F ) H&E staining of the representative colon sections derived from WT and Nfkb2−/− mice at day10 ( n = 4 , five fields/section ) ( B ) or indicated bone marrow chimeras at day7 ( n = 2 , five fields/section ) ( F ) after inoculation . The panel shows 20× magnification with scale bars representing 200 μm . ( D ) and ( G ) Average change in the body weight of WT and Nfkb2−/− mice ( n = 7 ) ( D ) or indicated bone marrow chimeras ( n = 4 ) ( G ) upon infection . Data are expressed as mean ± SEM . The statistical significance was determined using two-tailed Student's t-test . ( E ) and ( H ) Survival rates of WT and Nfkb2−/− mice ( n = 7 ) ( E ) or indicated bone marrow chimeras ( n = 4 ) ( H ) infected with C . rodentium . The statistical significance was determined using log rank ( Mantel–Cox ) test . ( I ) A model depicting the proposed regulatory role of crosstalk in sustaining colonic inflammatory immune responses . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 01810 . 7554/eLife . 05648 . 019Figure 7—figure supplement 1 . Nfkb2 controls gut immune response to Citrobacter infection . ( A and B ) Comparison of colons from WT and Nfkb2−/− mice infected ( B ) with C . rodentium revealing shrinkage with signatures of hyperplasia in the absence of Nfkb2 . Colon from uninfected mice had no discernible differences ( A ) . ( C ) Representative FACS plots confirming repopulation of CD45 . 1 hematopoietic cells ( ∼90% ) from WT donor into Nfkb2−/− mice with CD45 . 2 congenic background and vice versa after 5 weeks of bone marrow transplant . DOI: http://dx . doi . org/10 . 7554/eLife . 05648 . 019 In addition to modulating RelA activity through crosstalk , Nfkb2 also mediates RelB:p52 activation in response to singular non-canonical stimuli . Heightened RelB:p52 activation in B-cells was shown to strengthen host defense in Otu7b−/− mice ( Hu et al . , 2013 ) . Our inability to rescue the infection-related mortality in Nfkb2−/− mice using WT hematopoietic cells indicated that the protective function of Nfkb2 lies within stromal cells . Despite the presence of mesenteric lymph nodes ( Lo et al . , 2006 ) , additional stromal defects in Peyer's patches in Nfkb2−/− mice ( Yilmaz et al . , 2003 ) could impair mucosal IgA responses . Although IgM levels were comparable , very low IgA levels at day7 , prior to the onset of mortality in Nfkb2−/− recipients , precluded bona fide comparisons in the radiation chimeras . Interestingly , adult mice , similar to those utilized in our study , lacking IgA or its transporter pIgR efficiently resolved Citrobacter infections ( Maaser et al . , 2004 ) . μMT mice depleted of peripheral B-cells presented few mucosal changes in the first 2 weeks of infection and were completely free of mortality ( Simmons et al . , 2003 ) . Finally , Rag1−/− mice succumbed to Citrobacter only late during infections in the 4th week ( Wang et al . , 2010 ) . Though , our study does not rule out crosstalk independent engagement of the Nfkb2 pathway in activating RelB:p52 dimer in the hematopoietic compartment at a later stage of infection , epithelial RelA NF-κB activation defects coupled to aggravated early colon pathology and early onset of mortality in Nfkb2−/− mice suggested that the stromal requirement of Nfkb2 , at least in part , lies within the intestinal epithelial cells in the initial events controlling early innate immunity and involves crosstalk regulation of RelA NF-κB activity ( Figure 7I , also discussed later ) .
The NF-κB system transduces signals from a variety of cell-activating stimuli . Our study suggested that such pleiotropic system enables tuning of cellular response to an instructive signal by microenvironment-derived additional costimulatory signals . However , a duration code selectively integrates costimulatory LTβR signal with the TLR4 pathway , insulating transient cytokine signaling , secondary to microbial infections , from crosstalk amplifications . Recent studies have identified important positive feedback regulations underlying dose threshold control of NF-κB response in B-cells ( Shinohara et al . , 2014 ) or in myeloid cells ( Sung et al . , 2014 ) . Our crosstalk study illuminated a role of a positive feedback loop in sustaining NF-κB response; where non-canonical LTβR signal prolonged canonical TLR4 response by targeting RelA induced p100 for the generation of RelA:p52 NF-κB dimer . Indeed , a dominant precursor function of p100 in producing RelA:p52 led to ablated crosstalk in Nfkb2−/− cells . As such , non-canonical signal generates RelB:p52 by removing the C-terminal domain of p100 present in the preexisting RelB:p100 dimeric complex ( Sun , 2012 ) or liberates RelA:p50 dimer from the p100/IκBδ-inhibited complexes ( Basak et al . , 2007 ) . Signal generation of the RelA:p52 dimer requires both canonical induction of p100/Nfkb2 expressions and concomitant processing of the newly synthesized p100 into p52 by non-canonical signal . Previous report demonstrating that p100 could be efficiently processed cotranslationally ( Mordmüller et al . , 2003 ) explained the requirement of canonical signals in amply synthesizing nascent p100 as a substrate for the abundant production of p52 subunit during crosstalk . More so , mature p52 could readily dimerize with RelA , despite a preference of full-length p100 for RelB binding ( Fusco et al . , 2008 ) . These observations along with our current study , thereby , elaborated a requirement of convergence of canonical and non-canonical signals in synergistically generating and activating RelA:p52 dimer . Nevertheless , consistent to the observed gene-effect of crosstalk in our study , a significant overlap between RelA:p50 and RelA:p52 dimers in DNA binding ( Siggers et al . , 2012 ) and pro-inflammatory gene expressions ( Hoffmann et al . , 2003 ) was reported earlier . Of note , our results rely on bulk measurements of signaling intermediates and deterministic modeling approaches . Given studies documenting cell-to-cell variations in signal-induced NF-κB responses ( Lee et al . , 2009 ) , it would be interesting to examine the potential implication of signal integration at the single cell level in determining cellular heterogeneity . TLR4 activation of epithelial RelA was implicated in the chemokine gene expressions and neutrophil recruitment upon bacterial infections ( Khan et al . , 2006 ) . Yet , epithelial LTβR ( Wang et al . , 2010 ) was also important for effective innate immune responses to Citrobacter . In our proposed model ( Figure 7I ) , we could clarify that LTβR provides a critical costimulatory signal through Nfkb2 to sustain RelA NF-κB response to pathogens in otherwise hyporesponsive colonic epithelial cells . Such signal integration ameliorated innate immune functions by enhancing pro-inflammatory gene expressions . Interestingly , p100−/− mice , which lacked the expression of p100 , but aberrantly produced p52 , revealed hyperplasia of gastric epithelial cells and elevated expressions of RelA target genes ( Ishikawa et al . , 1997 ) . In Nlrp12−/− mice , robust p52 generation in stromal cells through the non-canonical pathway led to colon cancer associated inflammation ( Allen et al . , 2012 ) , a hallmark for aberrant RelA activity . These studies indeed support a possible role of Nfkb2 in mucosal epithelial cells in strengthening RelA activity . LTβR engagement in the dendritic cells within colonic patches was shown to trigger IL-22 production by innate lymphoid cells involving cell–cell communications to potentiate gut immunity ( Tumanov et al . , 2011 ) . Although , a defect in the colonic patches in Nfkb2−/− mice could impair IL-22 mediated protective responses , our cell-intrinsic crosstalk model explained that the reported epithelial requirement of LTβR ( Wang et al . , 2010 ) is in sustaining RelA NF-κB response during bacterial infection . In future , tissue-specific knockouts may help to further distinguish between innate immune functions of Nfkb2 in different cell types . The underlying mechanism and biological functions of RelB:p52 dimer activated by non-canonical signal is well established . From the perspective of signaling crosstalk , our study offers a significant revision of our understanding of non-canonical signaling in amplifying canonical RelA responses . It extends the intriguing possibility that cell-differentiating cues present in the tissue microenvironment may play a more direct role , separate from merely determining the differentiation states of the resident cells , in calibrating innate immune responses by engaging into cell-autonomous signaling crosstalks . Future studies ought to further examine the signal integration via Nfkb2 in potentiating immune responses against other microbial pathogens . More so , the potential involvement of the deregulated crosstalk control in the pathophysiology of inflammatory disorders , particularly those involving gut , remains to be addressed .
Wild-type or gene-deficient C57BL/6 mice were housed at NII small animal facility and used in accordance with the IAEC guidelines . Primary MEFs were generated from E12 . 5–14 . 5 embryos . Late passage Trif-deficient and NIK-deficient MEFs have been described ( Basak et al . , 2007 ) . MSIE cell line was a gift from R . Whitehead , Ludwig Cancer Research . Mouse Nfkb2 was stably expressed from a promoter containing five tandem kappaB sites from HRS . puro or from a constitutive promoter from pBabe . puro retroviral constructs . Cells were stimulated using 0 . 3 μg/ml αLTβR ( a gift from J Browning and A Papandile , Biogen , Cambridge , MA , USA ) , 100 ng/ml recombinant LTα1β2 ( Sigma , St . Louis , MO , USA ) , 1 ng/ml TNF ( Roche , BASEL , Switzerland ) , 1 ng/ml IL-1β ( Biosource , Carlsbad , CA , USA ) , or 1 μg/ml LPS ( Enzo , NY , USA ) , either individually or in combination . EMSA , immunoblot analyses , and IKK assay have been described earlier ( Basak et al . , 2007 ) . Recombinant GST-IκBα ( 1-54aa ) used in IKK assay was from BioBharati Life Sciences , Kolkata , India . NIK was immunoprecipitated ( Cell Signaling Technology , Danvers , MA , USA ) from cytoplasmic extracts and immunopellets were examined for kinase activity using GST-IκBδ as substrate ( GST-p100406–899 , BioBharati Life Sciences , Kolkata , India ) . The gel images were acquired using PhosphorImager ( GE , Amersham , UK ) and quantified in ImageQuant . Immunoblotting of immunoprecipitates was done using TrueBlot ( eBioscience , San Diego , CA , USA ) . Total RNA was isolated using RNeasy Kit ( Qiagen , Venlo , Netherlands ) . For microarray analysis , labeling , hybridization of RNA samples to the Illumina MouseRef-8 v2 . 0 Expression BeadChip , data processing and quantile normalization was performed by Sandor Pvt . Ltd ( Hyderabad , India ) . We have considered genes that are induced at least 1 . 3 fold by LPS at 24 hr in representative data sets and has a detection p-value < 0 . 05 for LPS , αLTβR and co-treatment regimes . Next , LPS response genes in WT MEFs were ranked based on the merit of their normalized crosstalk scores , which is defined below and also described earlier ( Zhu et al . , 2006 ) . Crosstalk score = [ ( Δco-treatment − ( ΔLPS + ΔαLTβR ) ) /0h_int] , where 0h_int indicates the signal intensity of a given gene in untreated cells and Δtreatment signifies the differences in signal intensities between treated and untreated cells , Normalized crosstalk score = Crosstalk Score * [{ ( ΔL + ΔB ) ) /0h_int/| ( ΔL + ΔB ) ) /0h_int|] , As implied , positive crosstalk scores signify hyperactivation , whereas negative crosstalk scores imply diminished gene expressions in the co-treatment regime as compared to cell treatment with the individual stimuli . The ordered gene set was examined in GSEA v2 . 0 . 12 ( Broad Institute at MIT ) ( Subramanian et al . , 2005 ) . The MIAME version of the microarray data set discussed in this publication are available on NCBI Gene Expression Omnibus ( accession number GSE62301 ) . For quantitative RT-PCR , total RNA was reverse transcribed with Transcriptor cDNA kit and amplified using Sybr Green PCR mix ( Roche , Mannheim , Germany ) in ABI7500 cycler . The relative gene expressions were quantified using ΔΔCT method upon normalizing to β-actin mRNA level . Absolute quantification was done using plasmid DNA constructs encoding respective genes as standards and normalized to express as gene/actin mRNA level . Sex matched , 8 to 10 week old mice , fasting for 8 hr , were orally gavaged with 1 . 2 × 1010 cfu of C . rodentium strain DBS100 ( ATCC 51459 ) . In certain instances , mice were intraperitoneally injected with 200 μg of murine LTβR-IgG1 fusion protein or MOPC21 isotype control ( Biogen Idec ) 1 day prior to infection , as described ( Wang et al . , 2010 ) . IECs , isolated following published procedure ( Greten et al . , 2004 ) , were utilized for biochemical analyses . For histology , dissected colons were fixed in 10% neutral buffered formalin . Paraffin-embedded tissue sections were stained with anti-myeloperoxidase antibody ( Pierce , Waltham , Massachusetts , USA ) for neutrophil recruitment or with Hematoxylin and Eosin ( H&E ) for tissue pathology evaluation . Fecal samples were weighed , homogenized , and serially diluted homogenates were plated on MacConkey agar ( HiMedia , Mumbai , India ) to score for C . rodentium . Similarly , spleens and livers were aseptically removed and assessed for bacterial load . For bone marrow chimera experiment , recipient WT or Nfkb2−/− mice were lethally irradiated and marrow cells from the indicated donor mice were transferred . After 6–8 weeks , mice were infected . The NF-κB Systems Model v1 . 0 was simulated in Matlab ( v . 2012b , Mathworks , Natick , MA , USA ) using the ode15 s ( Basak et al . , 2007 ) . A detailed description of the model has been provided in the Appendix-1 . To estimate crosstalk sensitivity , each parameter values were individually increased and decreased by 10% , euclidean distances were used to determine the resultant changes in the crosstalk indexes as compared to the unperturbed system , averaged for a given parameter and normalized to nominal crosstalk index . Data are expressed as mean of 3–5 quantified biological replicates ± SEM . Statistical significance was calculated by two-tailed Student's t-test . For survival curves , log rank ( Mantel–Cox ) test was conducted . This article additionally contains ( i ) 10 figure supplements associated with the main text , ( ii ) three Supplementary tables ( Supplementary files 1–3 ) and five Appendix figures associated with the an Appendix file ( Appendix-1 ) , which provides a detailed description of the mathematical model , as well as a file describing Matlab source codes . | The innate immune system is the body's first line of defense against infection and disease . Innate immune cells are found in every tissue type , poised to respond immediately to damaged , stressed , or infected host cells . When innate immune cells recognize any injury or infection , one of the first things they do is trigger the inflammatory response . Fluid and other immune cells then move from the blood into the injured tissues . This movement can cause redness and swelling . But the response helps to establish a physical barrier against the spread of infection , promotes the elimination of both invading microbes and damaged host cells , and encourages the repair of the tissue . Inflammation is tightly controlled . If the response is too weak , it could leave an individual prone to serious infection . On the other hand , excessive inflammation can severely damage healthy cells and tissues . Inflammation is regulated differently in different tissue types , and the environment within the tissue itself influences the activity of local innate immune cells and the inflammatory response . However , the molecular mechanisms responsible for receiving and interpreting the signals derived from the host tissue remain unknown . Now , Banoth et al . , have revealed that the integration of inflammation-provoking signals , such as injury or infection and cues from the tissue environment occurs via the so-called ‘NF-κB signaling system’ . NF-κB is a protein found in almost all cell types , and when activated it is able to switch on the expression of many different genes . Banoth et al . explain that signal integration via the NF-κB system enables cues from the tissue environment to tune a cell's responses . Further experiments confirmed the importance of this signal integration by showing how a signal coming from intestinal tissue can influence the activity of innate immune cells and inflammation in the gut . These findings suggest that a breakdown in the NF-κB signaling system's ability to integrate multiple signals , including those derived from the tissue environment , may be responsible for many inflammatory disorders , and in particular those that involve the gut . Future work is now needed to explore this possibility . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"immunology",
"and",
"inflammation"
] | 2015 | Stimulus-selective crosstalk via the NF-κB signaling system reinforces innate immune response to alleviate gut infection |
Somatic copy number alterations ( CNAs ) are a hallmark of cancer , but their role in tumorigenesis and clinical relevance remain largely unclear . Here , we developed CNApp , a web-based tool that allows a comprehensive exploration of CNAs by using purity-corrected segmented data from multiple genomic platforms . CNApp generates genome-wide profiles , computes CNA scores for broad , focal and global CNA burdens , and uses machine learning-based predictions to classify samples . We applied CNApp to the TCGA pan-cancer dataset of 10 , 635 genomes showing that CNAs classify cancer types according to their tissue-of-origin , and that each cancer type shows specific ranges of broad and focal CNA scores . Moreover , CNApp reproduces recurrent CNAs in hepatocellular carcinoma and predicts colon cancer molecular subtypes and microsatellite instability based on broad CNA scores and discrete genomic imbalances . In summary , CNApp facilitates CNA-driven research by providing a unique framework to identify relevant clinical implications . CNApp is hosted at https://tools . idibaps . org/CNApp/ .
The presence of somatic copy number alterations ( CNAs ) is a ubiquitous feature in cancer . In fact , the distribution of CNAs is sufficiently tissue-specific to distinguish tumor entities ( Ried et al . , 2012 ) , and allows the identification of tumors responsive to particular therapies ( Cairncross et al . , 2013; Davoli et al . , 2017 ) . Moreover , high levels of CNAs , which result from chromosome instability , are generally associated with high-grade tumors and poor prognosis ( Hieronymus et al . , 2018; Sansregret et al . , 2018; Smith and Sheltzer , 2018; Stopsack et al . , 2019 ) . Two main subtypes of CNAs can be discerned: broad CNAs , which are defined as whole-chromosome and chromosomal arm-level alterations , and focal CNAs , which are alterations of limited size ranging from part of a chromosome-arm to few kilobases ( Krijgsman et al . , 2014; Zack et al . , 2013 ) . Recently , it has been uncovered that while focal events mainly correlate with cell cycle and proliferation markers , broad aberrations are mainly associated with immune evasion markers , suggesting that tumor immune features might be determined by mechanisms related to overall gene dosage imbalance rather than specific actionable genes ( Buccitelli et al . , 2017; Davoli et al . , 2017; Taylor et al . , 2018 ) . Furthermore , it has been shown that broad CNAs involving whole-chromosome arms may confer high risk of lethal disease in prostate cancer ( Stopsack et al . , 2019 ) . Nevertheless , the precise role of CNAs in tumor initiation and progression , as well as their clinical relevance and therapeutic implications in most cancer types remain still poorly understood . Interpretation and visualization of CNAs is time-consuming and very often requires complex analyses with clinical and molecular information . Well-established CNA algorithms , such as the gold-standard circular binary segmentation , define the genomic boundaries of copy number gains and losses based on signal intensities or read depth obtained from array comparative genomic hybridization and SNP-array or next-generation sequencing data , respectively ( Olshen et al . , 2004 ) . However , the tumor-derived genomic complexity may cause an under- or overestimation of CNAs . This complexity is represented by tumor purity , tumor aneuploidy , and intratumor heterogeneity , which imply high levels of subclonal alterations . Thus , recent segmentation methods improved the accuracy to identify copy number segments in tumor samples either by considering the B allele frequency ( BAF ) , such as ExomeCNV ( Sathirapongsasuti et al . , 2011 ) , Control-FREEC ( Boeva et al . , 2012 ) and SAAS-CNV ( Zhang and Hao , 2015 ) , or through adjusting by sample purity and ploidy estimates , such as GAP ( Popova et al . , 2009 ) , ASCAT ( Van Loo et al . , 2010 ) and ABSOLUTE ( Carter et al . , 2012 ) . However , the state-of-the-art computational approach for CNA analysis in cancer is GISTIC2 . 0 ( Mermel et al . , 2011 ) , which is a gene-centered probabilistic method that enables to define the boundaries of recurrent putative driver CNAs in large cohorts ( Beroukhim et al . , 2010 ) . Nevertheless , despite ongoing progress on identifying CNAs , to our knowledge none of the existing software packages is readily available for integrative analyses to unveil their biological and clinical implications . To address this issue , we developed CNApp , the first open-source application to quantify CNAs and integrate genomic profiles with molecular and clinical variables . CNApp is a web-based tool that provides the user with high-quality interactive plots and statistical correlations between CNAs and annotated variables in a fast and easy-to-explore interface . In particular , CNApp uses purity-corrected genomic segmented data from multiple genomic platforms to redefine CNA profiles , to compute CNA scores based on the number , length and amplitude of broad and focal genomic alterations , to assess differentially altered genomic regions , and to perform machine learning-based predictions to classify tumor samples . To exemplify the applicability and performance of CNApp , we used publicly available segmented data from The Cancer Genome Atlas ( TCGA ) to ( i ) measure the burden of global , broad , and focal CNAs as well as generate CNA profiles in a pan-cancer dataset spanning 33 cancer types , ( ii ) identify cohort-based recurrent CNAs in hepatocellular carcinoma and compare them with previously reported data , and ( iii ) assess predicting models for colon cancer molecular subtypes and microsatellite instability status based on CNA scores and specific genomic imbalances . CNApp is hosted at https://tools . idibaps . org/CNApp/ and the source code is freely available at GitHub ( Franch-Expósito , 2020 ; copy archived at https://github . com/elifesciences-publications/CNApp ) .
CNApp comprises three main sections: 1- Re-Seg and Score: re-segmentation , CNA scores computation , variable association and survival analysis , 2- Region profile: genome-wide CNA profiling , CNA frequencies , correlation profiles and descriptive regions , and 3- Classifier model: machine learning classification model predictions ( Figure 1 ) . Each of these sections and their key functions are described below . The input file consists of a data frame with copy number segments provided by any segmentation algorithm . Mandatory fields and column headers are sample name ( ID ) , chromosome ( chr ) , start ( loc . start ) and end ( loc . end ) genomic positions , and the log2 ratio of the copy number amplitude ( seg . mean ) for each segment . Section one incorporates the correction for tumor purity ( i . e . fraction of tumor cells in the sample ) to measure the actual magnitude of CNAs . Thus , when available , the input file will also include sample purity estimations ( purity ) and BAF values ( BAF ) , which correct the accuracy of CNA calls and provide copy number neutral loss-of-heterozygosity ( CN-LOH ) events . Ploidy values , if known , might also be indicated as an independent variable . Annotation of variables can be included in the input file ( tagged in every segment from each sample ) or by uploading an additional file indicating new variables per sample . First , we evaluated the ability of CNApp to analyze and classify cancer types according to CNA scores , and assessed whether CNApp was able to reproduce specific CNA patterns across different cancer types . To do so , by using CNApp default parameters we obtained re-segmented data , CNA scores and cancer-specific CNA profiles for 10 , 635 tumor samples spanning 33 cancer types from the TCGA pan-cancer dataset . The distribution of BCS , FCS and GCS confirmed the existence of distinct CNA burdens across cancer types ( Figure 2A—source data 1 ) . While cancer types such as acute myeloid leukemia ( LAML ) , thyroid carcinoma ( THCA ) or thymoma ( THYM ) showed low levels of broad and focal events ( GCS median values of −1 . 67 for LAML , −1 . 68 for THCA , and −1 . 52 for THYM ) , uterine carcinosarcoma ( UCS ) , ovarian cancer ( OV ) and lung squamous cell carcinoma ( LUSC ) displayed high levels of both types of genomic imbalances ( GCS median values of 2 . 55 , 2 . 44 , and 0 . 97 for UCS , OV , and LUSC , respectively ) . Some cancer types displayed a preference for either broad or focal CNAs . For example , kidney chromophobe ( KICH ) tumors showed the highest levels of broad events ( median BCS value of 27 ) , while focal CNAs in this cancer type were very low ( median FCS value of 49 ) . In contrast , breast cancer ( BRCA ) samples displayed high FCS values ( median FCS value of 150 ) , while BCS values were only intermediate ( median BCS value of 7 ) . Overall correlations between CNA scores were assessed by computing Spearman’s rank test , obtaining values of 0 . 59 between BCS and FCS , 0 . 90 between BCS and GCS , and 0 . 85 between FCS and GCS . In addition , we further assessed the correlation between BCS and FCS for each individual BCS value . While tumors with low BCS displayed a positive correlation between broad and focal alterations , tumors did not maintain such correlation in higher BCS values ( Figure 2—figure supplement 1A and B ) . This correlation between BCS and FCS is maintained across the 33 cancer types ( Figure 2—figure supplement 1C ) . Subsequent analysis aimed at generating genome-wide patterns for each cancer type based on chromosome-arm genomic windows and the overall corresponding frequencies . In agreement with previous studies ( Beroukhim et al . , 2010 ) , cancer type-specific patterns of genomic gains and losses determined the tissue-of-origin ( Figure 2B ) . Additionally , we found that chromosome arms altered in more than 25% across all samples were 1q , 7 p , 7q , 8q and 20q for copy number gains , and 8 p and 17 p for copy number losses . Conversely , chromosome arms affected by CNAs in less than 10% of all cancer types included chromosome arms 2q and 19 p ( Figure 2C ) . By using a subset of 20 out of the 33 cancer types for which tumor type information was available , we asked CNApp to compute the average arm-region for each cancer type to assess if they clustered according to their CNA profiles ( Figure 2—figure supplement 2A ) . Our analysis showed that correlation values resulting from Pearson’s test hierarchically clustered according to the tissue-of-origin from the tumor . Gastrointestinal ( colon , rectum , stomach and pancreatic ) , gynecological ( ovarian and uterine ) and squamous ( cervical , head and neck , and lung ) cancers clustered together based on specific CNA profiles for each group ( Figure 2D ) . Intriguingly , correlation profiles using 5 Mb windows and only considering focal alterations showed a very similar degree of clustering based on the tissue of origin ( Figure 2—figure supplement 2B&C ) . Next , we attempted to test the ability of CNApp to identify recurrent broad and focal CNAs in a large cohort , and to assess the impact of the customizable parameters to describe CNA profiles . For that reason , we chose to perform CNA analysis of 370 samples from TCGA corresponding to the Liver Hepatocellular Carcinoma ( LIHC ) cohort . The pattern of recurrent broad and focal CNAs identified by GISTIC2 . 0 in the TCGA study ( Ally et al . , 2017 ) was similar to earlier reports , confirming the suitability of this cohort and the consistent identification of a CNA profile for hepatocellular carcinoma ( HCC ) ( Chiang et al . , 2008; Guichard et al . , 2012; Schulze et al . , 2015; Totoki et al . , 2014; Wang et al . , 2013 ) . By applying the default parameters of CNApp to the LIHC dataset and selecting chromosome arms as genomic regions to assess broad events , we consistently found copy number gains at 1q ( 56% ) and 8q ( 46% ) , and copy number losses at 8 p ( 62% ) and 17 p ( 47% ) as the most frequent alterations ( Figure 3A ) . These CNAs are the same as those identified by GISTIC2 . 0; however , frequencies were slightly lower ( Supplementary file 1 ) . Similarly , GISTIC2 . 0 detected significant gains with rates between 25–40% on eight additional chromosome-arms , including 5 p , 5q , 6 p , 20 p , 20q , 7 p , 7q , and 17q , which were also identified by CNApp , but in 20–30% of the samples . Likewise , GISTIC2 . 0 detected significant broad deletions at a frequency between 20% and 40% on 18 additional chromosome-arms , of which 4q , 6q , 9 p , 13q , 16 p , and 16q losses were observed at frequencies ≥ 20% by CNApp , and the rest of them displayed rates between 10% and 20% . Therefore , the identification of CNAs in CNApp is very consistent with those described by GISTIC2 . 0 . Differences in frequencies might be expected due to the lower copy number amplitude thresholds used by GISTIC2 . 0 in comparison with CNApp default cutoffs ( |0 . 1| vs . |0 . 2| , corresponding to ~2 . 14/1 . 8 copies vs . 2 . 3/1 . 7 copies , respectively ) . Indeed , previous reports analyzing CNAs in other HCC cohorts and using greater copy number thresholds , showed frequencies of alterations much more similar to those estimated by CNApp ( Chiang et al . , 2008; Guichard et al . , 2012; Schulze et al . , 2015; Wang et al . , 2013 ) . To assess the impact of modifying CNApp amplitude thresholds , we analyze the same dataset dropping the cutoff to |0 . 1| . By doing so , the overall number of broad CNAs increased , reaching frequency values similar or even higher than those reported by GISTIC2 . 0 ( Figure 3B and Supplementary file 1 ) . Of note , such drop from |0 . 2| to |0 . 1| might result in the identification of subclonal genomic imbalances , which are very frequent in tumor samples ( McGranahan and Swanton , 2017 ) , and it would also be of utility to compensate for low tumor purities . To evaluate the impact of other customizable parameters of CNApp to the results , we also tested whether the identification of broad events was affected by: ( i ) the relative length to classify a segment as arm-level alteration , and ( ii ) the re-segmentation provided by CNApp . As expected , increasing the percentage of chromosome arm required to classify a CNA segment as arm-level ( from ≥50% to≥70% ) or skipping the re-segmentation step led to an underestimation of some broad events , whereas decreasing the percentage of chromosome arm ( from ≥50% to≥40% ) resulted in the opposite ( Figure 3—figure supplement 1A–C and Supplementary file 1 ) . As far as focal CNAs are concerned , CNApp and GISTIC2 . 0 use different strategies to quantify their recurrence . Therefore , the comparison between the two methods was evaluated in a more indirect manner . GISTIC2 . 0 generates minimal common regions ( also known as ‘peaks’ ) that are likely to be altered at high frequencies in the cohort , which are scored using a Q-value and may present a wide variety of genomic lengths ( Mermel et al . , 2011 ) . Instead , CNApp allows dividing the genome in windows of different sizes , calculating average copy number amplitudes for all segments included within each window . We reasoned that considering the length of GISTIC2 . 0 reported ‘peaks’ , CNApp might also be capable of identifying recurrent focal altered regions by dividing the genome in smaller windows . To test our hypothesis , we asked CNApp to calculate the frequency of focal gains and losses by dividing the genome by sub-cytobands . As a result , CNApp consistently localized the most frequently altered sub-cytobands , including gains at 1q21 . 3 ( 25% ) , 8q24 . 21 ( 17% , MYC ) , 5p15 . 33 ( 13% , TERT ) , 11q13 . 3 ( 12% , CCND1/FGF19 ) and 6p21 . 1 ( 11% , VEGFA ) , and losses at 13q14 . 2 ( 20% , RB1 ) , 1p36 . 11 ( 18% , ARID1A ) , 4q35 . 1 ( 17% , IRF2 ) and 9p21 . 3 ( 14% , CDKN2A ) , which are in agreement with previous studies in HCC ( Figure 3C and Supplementary file 2 ) ( Chiang et al . , 2008; Guichard et al . , 2012; Schulze et al . , 2015; Wang et al . , 2013 ) . Compared to GISTIC2 . 0 , CNApp reported 14 of the 27 significant amplifications and 14 of the 34 significant deletions at rates > 10% , and the remaining alterations displaying rates between 4–10% ( Supplementary file 3 ) ( Wang et al . , 2013 ) . Most importantly , regions with the highest frequency detected by CNApp showed a good match with lowest GISTIC2 . 0 Q-residual values , indicating that the most significant ‘peaks’ identified by GISTIC2 . 0 were actually included in the most recurrently altered sub-cytobands reported by CNApp . Recurrent focal alterations occur at lower frequencies than broad events ( Beroukhim et al . , 2010 ) . Previous studies describing the genomic landscape of HCC mostly focused on focal high-amplitude CNAs ( >3 copies for gains and <1 . 3 copies for losses ) , thus reporting lower frequencies than those estimated by CNApp using default thresholds ( Chiang et al . , 2008; Guichard et al . , 2012; Schulze et al . , 2015 ) . In our analysis , excluding the low-level alterations and evaluating only the moderate and high-amplitude events ( ≥3 and≤1 copies ) , amplifications reached maximum rates of 11% , whereas high-level losses only reached ~2% ( Figure 3D and Supplementary file 2 ) . Top recurrent focal gains involved sub-cytobands 1q21 . 3 ( 11% ) , 8q24 . 21 ( 11% , MYC ) , 11q13 . 3 ( 7% , CCND1/FGF19 ) , and 5p15 . 33 ( 5% , TERT ) . Recurrent losses estimated at ~2% of the samples included 13q14 . 2 ( RB1 ) , 9p21 . 3 ( CDKN2A ) , 4q35 . 1 ( IRF2 ) , and 8p23 . 1 . Although slight discrepancies between frequencies might be explained by minimal variability in the copy number threshold , CNApp results are highly consistent with previous reports ( Chiang et al . , 2008; Guichard et al . , 2012; Schulze et al . , 2015 ) . In order to assess genomic differences between independent sample sets , CNApp determines significance based on regression analysis and statistical tests such as Student's t-test or Fisher's exact test , which allow association of genomic regions based on the seg . mean value or the presence of alterations with specific samples , respectively . To test the suitability of our method , we analyzed a subset of 100 randomly selected samples from the LIHC ( n = 50 ) and COAD ( n = 50 ) TCGA cohorts and we compared our results with those obtained by the recently published tool CoNVaQ ( Larsen et al . , 2018 ) . Fourteen CNAs ( 5 gains and 9 losses ) showing significant differences between COAD and LIHC were identified by CoNVaQ with a Q < 0 . 05 . The lengths of these CNAs ranged from 38 to 133 Mb , with an average of 67 Mb , suggesting that they were mainly broad events . Consequently , in CNApp , we selected genomic windows corresponding to chromosome arms , and used the default CNA cut-offs ( i . e . |0 . 2| ) and maximum number of base pairs allowed for merging adjacent segments ( i . e . 1 , 000 , 000 bp ) for comparison . By doing so , 9 out of 14 events identified by CoNVaQ were also detected by CNApp with an adjusted p<0 . 05 ( Fisher’s exact test ) . For the remaining five events , four of them were found with an adjusted p=0 . 06–0 . 1 , thus suggesting a good correlation between CNApp and CoNVaQ when considering broad CNAs . A proposed taxonomy of colorectal cancer ( CRC ) includes four consensus molecular subtypes ( CMS ) , mainly based on differences in gene expression signatures ( Guinney et al . , 2015 ) . Briefly , CMS1 includes the majority of hypermutated tumors showing microsatellite instability ( MSI ) , high CpG island methylator phenotype ( CIMP ) , and low levels of CNAs; CMS2 and CMS4 typically comprise microsatellite stable ( MSS ) tumors with high levels of CNAs; and finally , mixed MSI status and low levels of CNAs and CIMP are associated with CMS3 tumors . A representative cohort of 309 colon cancers from the TCGA Colon Adenocarcinoma ( COAD ) cohort ( Cancer et al . , 2012 ) with known CMS classification ( CMS1 , N = 64; CMS2 N = 112; CMS3 N = 51; CMS4 N = 82 ) and MSI status ( MSI , N = 72; MSS , N = 225 ) was analyzed by using CNApp . In agreement with Guinney and colleagues , survival curves generated by CNApp indicated that CMS1 patients after relapse showed the worst survival rates as compared to CMS2 patients ( Figure 4—figure supplement 1A ) ( Guinney et al . , 2015 ) . Next , we asked CNApp to perform the re-segmentation step using the default copy number thresholds and excluding segments smaller than 500 Kbp to avoid technical background noise . Then , broad CNAs were considered to generate genomic region profiles using chromosome-arm windows . As expected , the CNA frequency plot displayed the most commonly altered genomic regions in sporadic CRC ( Figure 4—figure supplement 1B ) ( Camps et al . , 2008; Cancer et al . , 2012; Meijer et al . , 1998; Nakao et al . , 2004; Ried et al . , 1996 ) . Most frequently altered chromosome arms included gains of 7 p , 7q , 8q , 13q , 20 p , and 20q , and losses of 8 p , 17 p , 18 p , and 18q , occurring in more than 30% of the samples ( Figure 4A ) . On the other hand , focal CNA patterns were obtained by generating genomic profiles by sub-cytobands . Of note , five out of six losses and five out of 18 gains were also identified by GISTIC2 . 0 in the COAD TCGA cohort ( Cancer et al . , 2012 ) . Subsequently , we performed integrative analysis of genomic imbalances , CMS groups , and CNA scores . By using CNApp , we assessed whether CNA scores were able to classify colon cancer samples according to their CMS . While BCS established significant differences between CMS paired comparisons ( p≤0 . 0001 , Student’s t-test ) , FCS poorly discerned CMS1 from three and CMS2 from 4 ( Figure 4B—source data 1 and Figure 4—figure supplement 1C—source data 1 ) . Thus , we reasoned that broad CNAs rather than focal were able to better discriminate between different CMS groups . In fact , the distribution of CMS groups based on BCS resembled the distribution of somatic CNA counts defined by GISTIC2 . 0 ( Guinney et al . , 2015 ) . Next , we integrated the BCS and the CMS groups with the microsatellite status . Our results showed an average BCS of 1 . 51 ± 2 . 11 and 10 . 25 ± 5 . 92 for MSI ( N = 72 ) and MSS ( N = 225 ) tumors , respectively . By applying CNApp's Classifier model to the COAD cohort , we chose MSI and MSS status ( 72 MSI and 225 MSS samples ) . Global accuracy in average for the 50-permutation in the RandomForest model was 82 . 2% . Re-classification of MSI and MSS samples was successfully achieved and distribution of BCS values across MSI and MSS-predicted samples is plotted in Figure 4—figure supplement 1D . Intersection between predicted MSI and MSS groups resulted in a BCS value of 4 . 75 , and further analysis by ROC curve implementation between MSI and MSS groups with BCS values as classifier proxy resulted in an AUC of 0 . 917 . In addition , the intersection between sensitivity and specificity from ROC analysis resulted in a BCS value of 3 . 5 ( Figure 4—figure supplement 1E ) . Altogether , we decided to implement a BCS value of 4 as a threshold to re-classify samples according to their MSI status . In order to validate that a BCS value of 4 was able to predict microsatellite status , we used a completely independent CRC cohort ( Berg et al . , 2019 ) . The validation dataset consisted of segmented data from 147 samples with known microsatellite status and CMS annotation ( MSI , N = 28; MSS , N = 119; CMS1 , N = 27; CMS2 , N = 61; CMS3 , N = 29; CMS4 , N = 30 ) . We then ran CNApp to obtain BCS values for each sample and applied our BCS of 4 as the threshold to re-classify samples ( i . e . MSI if BCS ≤4 , and MSS if BCS >4 ) using the Classifier model . By doing so , we obtained a global accuracy of 81% , which is similar to the accuracy obtained in the COAD dataset ( 82 . 2% ) , thus suggesting that a BCS equal to four might be considered as a universal threshold to determine microsatellite status in CRC . Applying the cutoff of 4 to the COAD cohort , 186 out of 225 ( 83% ) of MSS tumors showed BCS values greater than 4 ( Figure 4C ) . In contrast , 39 ( 17% ) MSS tumors showed a BCS of 4 or lower , corresponding to three CMS1 , six CMS2 , 18 CMS3 and 12 CMS4 tumors , further demonstrating the existence of MSS tumors with a very low CNA burden . On the other hand , seven MSI tumors showed BCS higher than 4 . Among them , five samples displayed genomic imbalances typically associated with the CRC canonical pathway , including a focal amplification of MYC , unveiling tumors with co-occurrence of MSI and extensive genomic alterations ( Trautmann et al . , 2006 ) . Our dataset comprised nine out of 51 CMS3 tumors with MSI . Intriguingly , two of them showed focal deletions on chromosome two involving MSH2 and MSH6 , suggesting the inactivation of these mismatch repair genes through a focal genomic imbalance . In fact , 46% of CMS3 MSS tumors showed BCS below 4 , in agreement with the finding that CMS3 tumors display low levels of somatic CNAs . Moreover , CNApp enabled the identification of possible sample misclassifications by integrating CMS annotation and BRAF-mutated sample status . As expected , CMS1 cases were enriched for BRAF mutation , although two CMS4 samples also showed mutations in BRAF . One of these samples showed a BCS of 11 , displaying canonical CNAs . In contrast , the other CMS4 BRAF-mutated sample showed MSI and a BCS of 0 , similar features as CMS1 . Likewise , four BRAF-wt samples , classified within the CMS4 group , displayed MSI and a BCS of 0 , thus being candidates to be labeled as CMS1 based on the levels of CNAs ( Figure 4D ) . These disparities are of utmost importance since recent studies reported that high copy number alterations correlate with reduced response to immunotherapy ( Davoli et al . , 2017 ) . Importantly , it has been suggested that MSI status might be predictive of positive immune checkpoint blockade response in advanced CRC , probably due to the low levels of CNA usually presented by MSI tumors ( Le et al . , 2015 ) . We then asked CNApp to compare differentially represented genomic regions between all CMS groups based on a Student's t-test or Fisher's test with adjusted p-value . By applying a Student's t-test , we observed that CMS1 resembled CMS3 , except for the gain of chromosome seven and the loss of 18q , which were regions commonly altered in CMS3 samples with BCS above 4 ( adjusted p≤0 . 001 , Student's t-test ) ( Figure 4E ) . Even though only subtle CNA differences between CMS2 and CMS4 were identified , the loss of 14q was significantly more detected in CMS2 ( 42% ) than in CMS4 ( 17 . 1% ) ( adjusted p≤0 . 005 , Student's t-test ) . The gain of 12q was more frequently associated with CMS1 than CMS2 ( adjusted p≤0 . 005 , Student's t-test ) , in agreement with previous studies reporting that the gain of chromosome 12 is associated with microsatellite unstable tumors ( Figure 4E ) ( Trautmann et al . , 2006 ) . Intriguingly , the gain of the chromosome arm 20q alone mimicked the distribution of somatic CNAs defined by GISTIC2 . 0 across consensus subtype samples ( Figure 4—figure supplement 1F ) ( Guinney et al . , 2015 ) . Finally , applying machine learning-based prediction models to classify samples by the most discriminative descriptive regions across CMS groups ( i . e . 13q , 17 p , 18 , and 20q ) , CNApp reached 55% of accuracy to correctly predict CMS . In fact , the occurrence of these genomic alterations was able to differentiate CMS2 from CMS4 with an accuracy of 70% , and CMS1 from CMS3 with a 72 . 3% accuracy . As expected , this set of genomic alterations distinguished CMS1 from CMS2 samples with an accuracy of 95% .
Here , we present CNApp , a web-based computational tool that provides a unique framework to comprehensively analyze and integrate CNAs associated with molecular and clinical variables , assisting data-driven research in the biomedical context . Although CNApp has been developed using segmented genomic copy number data obtained from SNP-arrays , the software is also able to accommodate segmented data from any next-generation sequencing platform . CNApp transforms segmented data into genomic profiles , allowing sample-by-sample comparison and the assessment of differentially altered genomic regions , which can then be selected by the user to assess classifier variables by computing machine learning-based models . Importantly , besides identifying the impact of specific CNAs , CNApp provides the unique opportunity to establish associations between the burden of genomic alterations and any clinical or molecular variable . To do so , CNApp calculates CNA scores , a quantification of the broad ( BCS ) , focal ( FCS ) and global ( GCS ) levels of genomic imbalances for each individual sample . The fact that high levels of aneuploidy may correlate with tumor immune evasion markers and that CNA burdens may be an independent prognostic factor for cancer-specific lethality in some cancers exemplifies the potential association of CNA scores with clinical outcomes ( Buccitelli et al . , 2017; Davoli et al . , 2017; Hieronymus et al . , 2018; Taylor et al . , 2018 ) . To note , CNA scores are calculated after an optional process of re-segmentation that enables to redefine CNA boundaries and to adjust sample-specific copy number thresholds by correcting for tumor purity estimates . In agreement with recently reported findings ( Beroukhim et al . , 2010; Hoadley et al . , 2018; Taylor et al . , 2018 ) , CNApp was benchmarked by analyzing 10 , 635 samples spanning 33 cancer types from the TCGA pan-cancer dataset , and was able to cluster major tumor types according to CNA patterns . Moreover , the software successfully reproduced the well-characterized genomic profile of HCC and CRC , considering both broad and focal events , demonstrating the reliability of CNApp in identifying regions encompassing the most recurrent CNAs ( Ally et al . , 2017; Cancer et al . , 2012 ) . Finally , applying CNApp to the TCGA colon cancer sample set , for which MSI status and CMS classification was well annotated , we determined that a BCS value of 4 discriminates MSI from MSS tumors with high accuracy , reinforcing the utmost significance of quantifying the CNA burdens . Most importantly , due to the inverse correlation between MSI and aneuploidy in CRC , our results suggest that this BCS value could be established as a cutoff to define the edge between low and high aneuploid tumors . In fact , while high aneuploid tumors show poor response to immunotherapy , it has been suggested that CMS1 microsatellite unstable tumors are likely to show a positive response to immune checkpoints inhibitors ( Kalyan et al . , 2018; Le et al . , 2015 ) . However , BCS was not associated with overall survival in patients after relapse ( data not shown ) . Moreover , specific genomic regions defined by CNApp contributed to classify the CMS groups , confirming the functional importance of specific genomic imbalances in the pathogenesis of this disease and providing insights into the classification of CRC based on CNA profiles . In summary , although our results ought to be further validated in independent cohorts , here we show that CNApp enables not only the fundamental analysis of CNA profiles , but also the functional understanding of CNAs in the context of clinical outcome and their potential use as biomarkers , thus becoming an asset to the cancer genomics community .
CNApp can be accessed at https://tools . idibaps . org/CNApp/ . It was developed using Shiny R package ( version 1 . 1 . 0 ) , from R-Studio ( Shiny , RRID:SCR_001626 ) ( Chang et al . , 2018 ) . The tool was applied and benchmarked while using R version 3 . 4 . 2 ( 2017-09-28 ) -- ‘Short Summer’ . List of packages , libraries and base coded are freely available at GitHub , and instructions for local installation are also specified . The amount of non-aberrant cell admixture may differ between cancer samples , necessitating separate adjustment of copy number thresholds for each assayed sample . To homogenize the analysis , purity estimations are used by CNApp to apply a baseline adjustment of the seg . means before the subsequent CNA calling by CNA amplitude thresholds . This adjustment allows for subsequent sample-to-sample comparison . Seg . mean values ( n ) by sample ( x ) , when purity ( r ) available , are re-computed into new seg . mean values ( N ) as follows:N ( x ) =2n ( x ) + ( r ( x ) -1 ) When purity is not provided , 100% purity ( r = 1 ) is assumed . Minimal purity accepted is 40% ( r = 0 . 4 ) , setting to 40% all purities below . Threshold of 40% also sets the negative capping value ( C ) as the maximal loss possible when heterozygous deletion happens ( from 2 copies to one copy ) :C = log212·0 . 4= -2 . 32 Segments resulting from re-segmentation ( or original segments from input file when re-segmentation is skipped ) are classified in chromosomal , arm-level and focal events by considering the relative length of each segment to the whole-chromosome or chromosome arm . Using default parameters , segments are tagged as chromosomal when 90% or more of the chromosome is affected; as arm-level when 50% or more of the chromosome arm is affected; and as focal when affecting less than 50% of the chromosome arm . Percentages for relative lengths are customizable . Broad ( chromosomal and arm-level ) and focal alterations are then weighted according to their amplitude values ( seg . mean ) and taking into account copy number amplitude ranges defined by CNA calling thresholds . Default amplitude thresholds to define these ranges , CNA levels and their corresponding absolute copy number are presented in Table 1 . In order to take into account the relative chromosome length from each segment when FCS is computed , coverage punctuation for focal events is implemented as specified in Table 2 . Broad CNA Score ( BCS ) : for a total N of broad events in a sample ( x ) , it equals to the summation of segments weights ( A ) in that corresponding sample and being i the corresponding segment:BCS ( x ) =∑i=1NAi Focal CNA Score ( FCS ) : same as in BCS , with an additional pondering value L included to the summation , which captures the relative size of the chromosome-arm coverage of each focal CNA ) :FCS ( x ) =∑i=1NAi·Li Global CNA Score ( GCS ) : for a sample x , it is calculated as the summation of normalized BCS and FCS values , where meanBCS and meanFCS stand for mean values of BCS and FCS from total samples , respectively , and sdBCS and sdFCS stand for standard deviation values of BCS and FCS from total samples , respectively:normBCSx=BCSx-meanBCSsdBCS normFCSx=FCSx-meanFCSsdFCSGCSx=normBCS ( x ) +normFCS ( x ) Correlation between broad , focal and global CNA scores ( BCS , FCS and GCS , respectively ) was assessed by computing Spearman’s rho statistic . The TCGA pan-cancer dataset including 10 , 635 samples spanning 33 cancer types was used to perform these associations . This data was downloaded from the Genomic Data Commons Data Portal ( GDC Data Portal , RRID:SCR_014514 ) ( Grossman et al . , 2016 ) . To evaluate the applicability of the CNA scores when it comes to sample CNA level assessment , we performed correlation analyses between CNA scores and the altered genome fraction . BCS and FCS were specifically correlated to the altered genome fraction by broad ( i . e . , chromosomal and arm-level events ) and focal alterations , respectively . To do so , we applied the re-segmentation procedure and CNA scores computation on a TCGA dataset comprised by 10 , 635 primary tumors from 33 TCGA projects analyzed by Affymetrix 6 . 0 SNP-array and DNACopy ( DNAcopy , RRID:SCR_012560 ) ( Venkatraman and Olshen , 2007 ) . Global altered genome fraction ( altFract ) was computed , for each sample ( x ) , by the summation of all copy number events lengths ( l ) and divided by total length of human genome ( hgLength ) . altFract ( x ) =∑i=1NlihgLength Altered genome fractions for broad and focal events were computed , for each sample ( x ) , by the summation of broad and focal alterations lengths ( l ) , respectively , and divided by total length of human genome ( hgLength ) . Broad altFract ( x ) =∑i=1Nli ( Broad ) hgLength Focal altFract ( x ) =∑i=1Nli ( Focal ) hgLength Correlation tests were performed by applying Spearman’s rho statistic to estimate a rank-based measure of association . Annotated variables from input file are statistically associated with CNA scores computed by CNApp in Re-Seg and Score section . Different association tests are applied according to variable class ( i . e . categoric or numerical ) , as described in Table 3 . Both parametric and non-parametric tests are computed to present p-values in order to assess statistical significance for each association . Survival analysis by Kaplan-Meier was performed using survival and survminer R packages ( CRAN , RRID:SCR_003005 ) . Region profiling section allows genome segmentation analysis by user-selected windows ( i . e . arms , half-arms , cytobands , sub-cytobands , and 40 Mb till 1 Mb ) . In order to do that , windows files were generated for each option and genome build ( hg19 and hg38 ) . Cytobands file cytoBand . txt from UCSC page and for both genome builds was used as mold to compute regions ( Tyner et al . , 2017 ) . Segmented samples are transformed into genome region profiles using genomic windows selected by user . Segments from each sample are consulted to assess whether or not overlap with the window region . Thus , window-means ( W ) are computed for each genomic window by collecting segments ( t ) overlapping with window-region ( i ) . Segments with loc . start or loc . end position falling within the region are collected , as well as those segments embedding the entire region . At this point , the summation of each segment-mean ( S ) corrected by the relative window-length ( L ) affected by the segment length ( l ) is performed:W ( i ) =∑t=1nSt·ltL ( i ) Potential descriptive regions between groups defined by the annotated variables provided in the input file can be studied and p-values are presented to evaluate significance in differentially altered regions between those groups . The alterations can be considered as ( 1 ) numerical continuous ( seg . mean values ) and ( 2 ) categorical variables ( gains , losses and non-altered ) . In the first case , to assess statistical significance between groups Student’s T-test is applied , whereas in the second situation the significance is assessed by applying the Fisher’s exact test . False discovery rate ( FDR ) adjustment is performed using the Benjamini-Hochberg ( BH ) procedure in both cases and corrected p-values ( Adj . p-value ) or non-corrected p-values ( p-values ) are displayed by user selection . Correlation and clustering methods can be applied into heatmap plots in CNApp Region profile section . Accepted correlation methods are Pearson’s , Spearman’s and Kendall’s tests . Hierarchical cluster analysis is applied when clusters are computed in heatmaps . We used the randomForest R package ( RandomForest Package in R , RRID:SCR_015718 ) ( Liaw and Wiener , 2002 ) to compute machine learning classifier models . Variables to define sample groups must be selected , as well as at least one classifier variable . Model construction is performed 50-times and training set is changed by iteration . In order to compute model and select training set , multiple steps and conditions have to be accomplished: After model computation , contingency matrix with prediction and reference values by group is created to compute accuracy , specificity and sensitivity by group . | In most cases , human cells contain two copies of each of their genes , yet sometimes this can change , an effect called copy number alteration ( CNA ) . Cancer is a genetic disease and thus , studying the DNA from tumor samples is crucial to improving diagnosis and choosing the right treatment . Most tumors contain cells with CNAs; however , the impact of CNAs in cancer progression is poorly understood . CNAs can be studied by examining the genome of tumor cells and finding which regions display an unusual number of copies . It may also be possible to gather information about different cancer types by analyzing the CNAs in a tumor , but this approach requires the analysis of large amounts of data . To aid the analysis of CNAs in cancer cells , Franch-Expósito , Bassaganyas et al . have created an online tool called CNApp , which is able to identify and count CNAs in genomic data and link them to features associated with different cancers . The hope is that a better understanding of the effect of CNAs in cancer could help better diagnose cancers , and improve outcomes for patients . Potentially , this could also predict what type of treatment would work better for a specific tumor . Besides , by using a machine-learning approach , the tool can also make predictions about specific cancer subtypes in order to facilitate clinical decisions . Franch-Expósito , Bassaganyas et al . tested CNApp using previously existing cancer data from 33 different cancer types to show how CNApp can help the interpretation of CNAs in cancer . Moreover , CNApp can also use CNAs to identify different types of bowel ( colorectal ) cancer in a way that could help doctors to make decisions about treatment . Together these findings show that CNApp provides an adaptable and accessible research tool for the study of cancer genomics , which could provide opportunities to inform medical procedures . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"tools",
"and",
"resources",
"genetics",
"and",
"genomics"
] | 2020 | CNApp, a tool for the quantification of copy number alterations and integrative analysis revealing clinical implications |
PHF13 is a chromatin affiliated protein with a functional role in differentiation , cell division , DNA damage response and higher chromatin order . To gain insight into PHF13's ability to modulate these processes , we elucidate the mechanisms targeting PHF13 to chromatin , its genome wide localization and its molecular chromatin context . Size exclusion chromatography , mass spectrometry , X-ray crystallography and ChIP sequencing demonstrate that PHF13 binds chromatin in a multivalent fashion via direct interactions with H3K4me2/3 and DNA , and indirectly via interactions with PRC2 and RNA PolII . Furthermore , PHF13 depletion disrupted the interactions between PRC2 , RNA PolII S5P , H3K4me3 and H3K27me3 and resulted in the up and down regulation of genes functionally enriched in transcriptional regulation , DNA binding , cell cycle , differentiation and chromatin organization . Together our findings argue that PHF13 is an H3K4me2/3 molecular reader and transcriptional co-regulator , affording it the ability to impact different chromatin processes .
PHF13 , also known as Survival time associated PHD finger in Ovarian Cancer 1 ( SPOC1 ) , is a chromatin affiliated protein that is conserved from zebra fish to humans . PHF13 has been shown to modulate various processes including development ( Bördlein et al . , 2011 ) , DNA damage ( Frohns et al . , 2014; Mund et al . , 2012 ) , cell cycle ( Kinkley et al . , 2009 ) , antiviral host cell response ( Schreiner et al . , 2013 ) and higher order chromatin structure ( Kinkley et al . , 2009; Mund et al . , 2012 ) , underlining its biological importance . Furthermore , PHF13 expression and chromatin localization are temporally regulated ( Kinkley et al . , 2009 ) and its misregulation correlates with malignant phenotypes ( Mohrmann et al . , 2005 ) and defective differentiation ( Bördlein et al . , 2011 ) . Together these observations argue that altered PHF13 expression has consequential outcomes and further underscores the necessity of understanding the molecular interplay and contexts governing PHF13 function . To date many chromatin binding domains recognizing specific histone posttranslational modifications , have been identified . Included in these are PHD domains which are recruited predominantly to methylated lysine residues ( He et al . , 2013; Li et al . , 2006; Mansfield et al . , 2011; Peña et al . , 2006; Shi et al . , 2006; Wysocka et al . , 2006; Xie et al . , 2012 ) with a few exceptions ( Ali et al . , 2012; Hu et al . , 2009; Lan et al . , 2007; Lange et al . , 2008; Mansfield et al . , 2011; Org et al . , 2008; Tsai et al . , 2010 ) . The ability of different PHD domain containing proteins to recognize the same modified residue argues that additional factors may regulate their recruitment . Aside from differences in binding affinities , the sensitivity to adjacent sequence modifications also contributes to their specificity ( Ali et al . , 2013; Fiedler et al . , 2008; Gatchalian et al . , 2013; Iberg et al . , 2008; Ramon-Maiques et al . , 2007; Vermeulen et al . , 2007; Yuan et al . , 2012 ) indicating that many PHD domains read the local combinatorial chromatin environment . While the molecular readers containing PHD domains interact selectively with chromatin , they do so albeit with relatively weak affinities ( Musselman and Kutateladze , 2011 ) . To achieve a more stable chromatin association , many chromatin readers contact chromatin in a multivalent fashion , via multiple chromatin binding modules ( Adams-Cioaba et al . , 2012; Lange et al . , 2008; Liu et al . , 2013; Patel et al . , 2013; Rothbart et al . , 2013; Ruthenburg et al . , 2011 ) or in complex with other chromatin readers , which cooperatively read multivalent chromatin signatures ( Ballare et al . , 2012; Nayak et al . , 2011 ) . The majority of characterized PHD domain proteins are known to be affiliated with chromatin modulating complexes ( Morra et al . , 2012; Todd and Picketts , 2012; Wysocka et al . , 2006 ) and co-regulate important chromatin processes , including epigenetic programming ( Lan et al . , 2007; Wen et al . , 2010 ) , transcription ( Fortschegger and Shiekhattar , 2011 ) , DNA repair ( Li et al . , 2013; Mund et al . , 2012 ) , differentiation ( Bördlein et al . , 2011; Gatchalian et al . , 2013 ) , cell cycle ( Kinkley et al . , 2009; Lim et al . , 2013 ) and higher chromatin order ( Papait et al . , 2008 ) . Underscoring their importance in the co-regulation of chromatin function , the dysregulation or aberrant fusion of several PHD domain containing proteins has been shown to lead to genomic instability and cancer ( Wang et al . , 2009 ) . PHF13 contains a single C-terminal PHD domain , which we demonstrate biochemically and structurally , is a molecular reader of H3K4me2/3 . Additionally , we show that PHF13 directly interacts with DNA via a centrally located domain , indicating that it can form multivalent chromatin interactions . These interactions were confirmed by peptide binding assays , gel shift assays , ChIP sequencing and x-ray crystallography allowing us to additionally map these interactions and identify key cis-acting molecular determinants that affect PHF13’s affinity for chromatin . Furthemore , we utilized mass spectrometry , size exclusion chromatography and co-immunoprecipitation experiments to identify Polycomb repressive complex 2 ( PRC2 ) and RNA polymerase II ( RNA PolII ) complexes as novel PHF13 chromatin interaction partners . Consistently , PHF13 ChIP sequencing targets co-occurred with CpG rich DNA , H3K4me2/3 , PRC2 and the hypophosphorylated , serine 5 and serine 7 phosphorylated forms of RNA PolII in murine embryonic stem cells ( mESCs ) . PHF13 depletion in mESCs resulted in the reduced binding of SUZ12 and RNA PolII S5P to H3K4me3 and H3K27me3 and altered gene expression of a fraction of PHF13 bound genes . Genes that were up regulated upon PHF13 knockdown were enriched in H3K4me2/3 , Polycomb and RNA PolII S5P , while down regulated genes were enriched in H3K4me2/3 and RNA PolII S2P and S5P . Finally , PHF13 target genes were enriched in the functional categories of transcription regulation , cell cycle , chromosome organization and differentiation , consistent with earlier publications describing a role of PHF13 in these processes ( Bördlein et al . , 2011; Kinkley et al . , 2009; Mohrmann et al . , 2005; Mund et al . , 2012; Schreiner et al . , 2013 ) . Together , these findings argue that PHF13 is a transcriptional co-regulator and a novel H3K4me2/3 molecular reader .
We have previously demonstrated via differential nuclear fractionation experiments that PHF13 is predominantly affiliated with the chromatin fraction of nuclear lysate , implicating a role in chromatin function ( Kinkley et al . , 2009 ) . Therefore , to gain clearer insight into whether PHF13 contacts chromatin directly , we explored the ability of GST-PHF13 and different GST-PHF13 deletion mutants to interact with recombinant mono-nucleosomes ( Figure 1A–C ) . Mono-nucleosomes were generated using recombinant histone octamers and either a 200 bp DNA fragment ( Figure 1B ) or a 151 bp DNA fragment ( Figure 1C ) to recapitulate mono-nucleosomes with or without linker DNA . GST-PHF13 was found to very efficiently shift reconstituted mono-nucleosomes with DNA overhangs , similar to ISWI ( positive control ) and in contrast to GST ( Figure 1B ) . We also noted that the free DNA in the reaction was notably absent in GST-PHF13 lanes in comparison to ISWI and GST , suggesting that PHF13 may also interact with DNA . To further map PHF13’s interaction with mono-nucleosomes we analyzed the ability of different PHF13 fragments to shift mono-nucleosomes devoid of DNA overhangs ( Figure 1C ) . Surprisingly , the PHD domain of PHF13 was not found to interact with the recombinant mono-nucleosomes and the interaction was mapped to the middle region of PHF13 ( 101–200; Figure 1C ) . These observations indicate that linker DNA is not necessary for PHF13 to affiliate with the nucleosomes and that the middle 100 aa of PHF13 is capable of forming a direct contact with recombinant mono-nucleosomes . Again we noted the lack of free DNA in the PHF13 101–200 lanes suggesting that this region may interact with free and nucleosomal complexed DNA . To test this idea and if a direct interaction with DNA exists , we performed DNA electrophoretic mobility shift assays ( EMSA ) using full-length GST-PHF13 and GST-PHF13 deletion fragments ( Figure 1D–E ) . Since PHF13 lacks a predicted DNA binding domain , no DNA sequence specificity could be inferred . Therefore the EMSA’s were performed using two random and unrelated DNA fragments , a 248 bp DNA ( Figure 1D ) or a 40 bp DNA fragment ( Figure 1E ) . Increasing amounts of GST-PHF13 , GST-ΔPHD and ACF1 ( positive control ) strongly retarded the electrophoretic mobility of the DNA , in contrast to GST alone ( Figure 1D ) , indicating that PHF13 can directly interact with DNA independent of its PHD domain . Consistent with these and our previous observations , we mapped the DNA binding region to the middle 100aa ( 101–200 ) of PHF13 , whereas the N- and C-terminal 100 amino acids of PHF13 did not shift the DNA ( Figure 1E ) . Together , these findings support that PHF13 can directly contact chromatin via a direct interaction with DNA . 10 . 7554/eLife . 10607 . 003Figure 1 . PHF13 binds to DNA and recombinant nucleosomes . ( A ) Schematic of the putative domain structure of PHF13 . ( B ) Mononucleosome EMSA using recombinant mononucleosomes reconstituted on a 200 bp DNA fragment ( 20 nM ) and increasing concentrations of GST ( 150 nM , 1200 nM ) , GST-ISWI ( 28 nM , 226 nM ) and GST-PHF13 ( 70 nM , 140 nM , 280 nM , 560 nM ) . ( C ) Mononucleosome EMSA using recombinant mononucleosomes reconstituted on a 151 bp fragment ( 20 nM ) and increasing concentrations ( 80 , 160 , 320 nM ) of GST-1-100 , GST-101-200 , GST-201-300 , GST-PHD and PWWP ( positive control ) . ( D ) EMSA: 248 fM of P32 radioactively labeled 248 bp DNA with increasing concentrations of GST ( 37 . 5 , 150 , 300 nM ) GST-PHF13 ( 17 , 34 , 68 , 102 and 135 nM ) and GST-ΔPHD ( 18 . 5 , 37 , 74 , 111 , 148 nM ) and ACF1 ( 10 . 5 , 21 , 31 . 5 and 42 nM ) . ACF1 served as a positive control . ( E ) EMSA: 10 nM Cy5 labeled 40 bp DNA with increasing concentrations ( 40 , 80 , 160 nM ) of GST , GST-1-100 , GST-101-200 , GST-201-300 , GST-PHD and GST-PWWP ( positive control ) . The input DNA and mononucleosomes are indicated in B–E . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 003 The inability of PHF13’s PHD domain to interact with recombinant mono-nucleosomes which are devoid of histone post-translational modifications , suggested that it may selectively interact with modified histone tails . Consistently , previous mass spectrometry and bioinformatic predictions indicated that PHF13 may interact with H3K4me3 ( Nikolov et al . , 2011; Ruthenburg et al . , 2007; Slama and Geman , 2011 ) . To address this possibility , we evaluated the ability of full-length GST-PHF13 and different deletion fragments to directly interact with either non-modified or differentially methylated histone peptides or with H3K4me3 from nuclease digested chromatin lysate ( Figure 2 ) . As postulated both GST-PHF13 and the GST-PHF13 fragment containing the PHD domain ( 201–300 ) co-precipitated with the H3K4me1 , H3K4me2 and H3K4me3 peptides but not with the un-modified H3K4me0 peptide ( Figure 2A ) , explaining the inability of the PHD domain to interact with and shift the recombinant mono-nucleosomes . In contrast , GST-only and the GST-PHF13 fragments lacking the PHD domain ( 1–100 and 101–200 ) were incapable of co-precipitating with the biotinylated histone peptides . These data strongly support that PHF13’s PHD domain interacts with methylated histone H3K4 and suggest that it has a stronger binding preference for H3K4me2 and H3K4me3 , in comparison to H3K4me1 ( Figure 2A ) . 10 . 7554/eLife . 10607 . 004Figure 2 . PHF13’s PHD domain specifically interacts with H3K4me2/3 ( A ) Biotinylated histone peptide pull down . Equivalent amounts ( 1 µg ) of GST , GST-PHF13 and GST-PHF13 deletion fragments 1–100 , 100–200 and 200–300 aa were incubated with 1 µg of differentially modified biotinylated histone peptides and streptavidin beads . Co-precipitation of GST-proteins was analyzed using a GST specific antibody . ( B ) Fluorescence Polarization Assay . The dissociation constant of the PHF13-PHD only protein with differentially methylated H3 and H4 peptides . ( C ) GST pull down of H3K4me3 from nuclease digested chromatin lysates using GST-alone , GST-PHF13 and the indicated GST-PHF13 deletion and point mutant proteins . Precipitation of H3K4me3 was analyzed with a specific antibody . Amount of GST proteins were controlled by Coomassie staining . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 004 To further evaluate the specificity of PHF13’s PHD domain for H3K4me2/3 in relation to other methyl lysine residues and to calculate approximate binding constants we employed fluorescence polarization analysis using the PHD domain only of PHF13 and differentially modified histone H3 peptides ( Figure 2B ) . PHF13’s PHD domain bound strongest to H3K4me3 ( Kd = 88 . 5 ± 20 . 6 µM ) , with a slightly reduced affinity to H3K4me2 ( Kd = 122 ± 29 . 4 µM ) and with a ~4-fold reduced affinity to H3K4me1 ( Kd = 332 . 4 ± 82 . 6 µM ) . In contrast to H3K4 , no detectable binding was observed for other tri-methylated histone peptides demonstrating a strong preference and specificity of the PHF13 PHD domain for H3K4me2/3 ( Figure 2B ) . Finally , to explore whether PHF13’s PHD domain can interact with H3K4me3 in a native chromatin context , we analyzed the ability of the different recombinant PHF13 proteins to precipitate H3K4me3 from nuclease digested Hela chromatin lysates ( Figure 2C ) . These experiments showed that PHF13 and the deletion fragment containing its PHD domain ( 201–300 ) were capable and sufficient to precipitate H3K4me3 , an interaction that was lost by deletion of the PHD domain , specific point mutations in the PHD domain ( M246A and W255A ) that are predicted to disrupt the PHD domain structure , and in fragments not containing the PHD domain ( 1–100 and 101–200 ) . Together these findings demonstrate that PHF13’s PHD domain is a specific H3K4me2/3 molecular reader and together with its DNA binding ability that PHF13 can interact with chromatin in a direct and multivalent manner . To gain structural insight into this specific recognition of histone H3K4me3 by the PHD domain of PHF13 , we determined the crystal structures of both the apo-PHF13 PHD domain ( aa 226–280 ) and the PHF13 PHD-H3K4me3 peptide complex ( Figure 3A-C ) . The diffraction data and refinement statistics for these structures can be found in Figure 3—source data 1 . Similar to other PHD domains ( Adams-Cioaba and Min , 2009; Musselman and Kutateladze , 2009 ) , the PHD domain of PHF13 folds into a canonical Cys4-His-Cys3 ( or C4HC3 ) motif , that coordinates two zinc ions with a well-conserved globular domain ( Figure 3A ) . In the complex structure of PHF13 PHD-H3K4me3 , we can see that the histone H3K4me3 peptide is bound against the central double-strand anti-parallel β-sheet of the PHF13 PHD domain and completes a three-stranded β-sheet ( Figure 3A ) . The first four residues of the H3K4me3 peptide are embedded in an exposed binding groove on the surface of the PHF13 PHD domain ( Figure 3B–C ) . Similar to other methyl lysine binding proteins ( Adams-Cioaba and Min , 2009 ) , the trimethyl lysine 4 ( K4me3 ) is bound in an aromatic cage formed by residues F241 , M246 and W255 of PHF13 ( Figure 3B–C ) . In addition , the K4me3 residue also forms two main chain hydrogen bonds with M246 ( Figure 3C ) . The first residue alanine ( Ala1 ) of the H3K4me3 peptide is anchored in a small and secluded pocket created by I247 , V268 , P269 , E270 and F272 , and the backbone amine group of Ala1 forms a hydrogen bond with the backbone carbonyl oxygen of P269 and E270 in PHF13 ( Figure 3C ) . The limited length between the aromatic cage and the Ala1 binding pocket and the restricted nature of the Ala1 binding pocket determines that the PHF13 PHD domain selectively recognizes only methylated histone H3K4 . Consistently , PHF13 PHD domain was not co-precipitated by biotinylated H3K4me0 ( Figure 2A ) , nor did it interact with recombinant mononucleosomes ( Figure 1B ) or H3K4me0 fluorescent peptides ( Figure 2B ) . The Arg2 residue in the H3K4me3 peptide is bound by two backbone hydrogen bonds with E248 ( Figure 3B–C ) . Mutation of any of the residues affiliated with the H3K4me3 or H3R2 binding pockets ( T234A , F241A , M246A , W255A and E248A ) abrogated binding of PHF13’s PHD domain to H3K4me3 ( Figure 2C and 3D ) . 10 . 7554/eLife . 10607 . 005Figure 3 . Crystal structure of PHF13’s PHD domain . ( A–B ) The crystal structure of apo-PHF13 PHD domain ( A ) and PHF13 PHD finger in complex with H3K4me3 ( B ) . ( C ) Electrostatic surface potential of PHF13’s PHD finger in complex with H3K4me3 . Dashed lines represent intermolecular hydrogen bonds . ( D ) Fluorescence Polarization Assay - The dissociation constant of PHF13 and PHF13 PHD point mutants with H3K4me3 . ( E ) Isothermal titration Calorimetry ( ITC ) - Binding affinity of PHF13’s PHD domain for H3R2K4me3 ( left panel ) or H3R2me2aK4me3 ( right panel ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 00510 . 7554/eLife . 10607 . 006Figure 3—source data 1 . Data collection and refinement statistics – Detailed specifications of the data obtained from the crystallization of Apo-PHF13-PHD and the PHF13-PHD-H3K4me3 complex . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 00610 . 7554/eLife . 10607 . 007Figure 3—source data 2 . PHF13 PHD domain binding to a differentially methylated histone peptide tail chip . ( A ) Binding of His-PHF13-PHD-only to a histone peptide array was detected with an anti-His antibody . Boxes denote the positive controls ( 12 x histidine ) and an interaction with H3K4me3 when the neighboring R2 is not di-methylated . ( B ) Peptide key of the spotted differentially modified histone peptides on the histone peptide array . The length , position and sequence of the spotted peptides are annotated in the table . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 00710 . 7554/eLife . 10607 . 008Figure 3—figure supplement 1 . Comparison of the H3R2 binding pocket of different PHD domains . ( A ) PHF13-H3K4me3 . PDB code: 3O7A ( B ) RAG2-H3R2me2aK4me3 . PDB code: 2V83 ( C ) BPTF-H3K4me3 . PDB code: 2F6J , D ) ING2-H3K4me3 . PDB code: 2G6Q . The PHD domains are displayed in surface charge representation model and the histone peptides are shown in stick models . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 008 The surface charge representation shows that Arg2 sits in a shallow negatively charged pocket , reminiscent of the Arg2 binding in the ING2 and BPTF2 PHD domain-H3K4me3 complex structures , but in contrast to the RAG2-H3K4me3 complex structure , in which Arg2 is bound in a relatively hydrophobic pocket ( Figure 3C and Figure 3—figure supplement 1 ) . This indicates that similar to ING2 and BPTF , that Arg2 methylation should diminish PHF13 binding to H3K4me3 , whereas it has been shown to enhance RAG2 binding ( Iberg et al . , 2008; Yuan et al . , 2012 ) . Therefore to address this prediction and test the influence of Arg2 methylation on PHF13’s PHD domain H3K4me3 binding ability we performed isothermal titration calorimetry ( ITC; Figure 3E ) . Our binding results show that the binding of PHF13’s PHD domain to H3K4me3 was diminished by the simultaneous asymmetric dimethylation of Arg2 ( H3R2me2aK4me3 ) . Consistently , the binding of PHF13’s PHD domain to H3K4me3 on a differentially modified histone peptide chip was also significantly reduced in the presence of dimethyl Arg2 ( Figure 3—source data 2 ) . Together , these results confirm that the PHD domain of PHF13 is an H3K4me3 reader and demonstrate the structural relationship between them . Furthermore , they identify key residues within PHF13’s PHD domain involved in mediating this molecular interaction as well as inhibiting modifications of neighboring histone residues that opposes their interaction . To gain additional information about PHF13 chromatin interactions , we immunoprecipitated PHF13 from the nuclease digested chromatin fraction of mESCs and analyzed the co-immunoprecipitating proteins in comparison to an IgG control by mass spectrometry ( Figure 4A and Figure 4—source data 1 ) . Independent PHF13 IPs were digested by trypsin or LysC ( Figure 4—figure supplement 1 ) prior to MS analysis and the co-precipitating proteins were combined ( Figure 4A and Figure 4—source data 1 ) . Similar complexes were retrieved regardless of whether the results were pooled ( Figure 4A and Figure 4—source data 1 ) or examined individually ( Figure 4—figure supplement 1 ) . For all interactions where only a single unique peptide was detected the mass spectrometry profile is shown , to validate their identity ( Figure 4—figure supplement 2–4 ) . The mass spectrometry findings confirmed previous observations that PHF13 interacts with DNA damage response ( DDR ) proteins , ATM and TRIM 28 ( Mund et al . , 2012 ) and revealed that PHF13 interacts with an RNA polymerase II/ splicing complex as well as with several components of PRC2 , namely , SUZ12 , RBBP4 and RBBP7 ( Figure 4A ) . Analysis of the associated functional terms affiliated with PHF13 interacting proteins , indicated a role in RNA metabolic processes , gene expression , RNA and nucleic acid binding ( Figure 4—source data 2 ) . To reproduce these observations we performed reciprocal co-immunoprecipitation in mESCs which confirmed that PHF13 , SUZ12 and RNA PolII S5P interact with each other and that they all co-precipitate with H3K4me3 and H3K27me3 ( Figure 4B ) . To provide additional evidence that these proteins exist in a common chromatin molecular complex we performed size exclusion chromatography from nuclease digested chromatin fraction of mESCs to look for fractions where PHF13 coexists with PRC2 , RNA PolII , H3K4me3 and H3K27me3 ( Figure 4C ) . PHF13 was present in three different compositions , a very high molecular weight fraction ( fractions 14–16 ) where it co-eluted with RNA PolII ( S2P , S5P and S7P ) H3K4me3 and H3K27me3 , a high-mid molecular weight fraction ( fractions 18–26 ) where it co-eluted with RNA PolII ( S2P , S5P and S7P ) , PRC2 ( EZH2 and SUZ12 ) , H3K4me3 and H3K27me3 and in mid-low molecular weight fractions ( 28–32 ) where it co-eluted only with RNA PolII ( Figure 4C ) . Together , these findings raise the possibility that PHF13 co-exists with different RNA PolII complexes , i . e . some containing PRC2 and histones and some that do not . To address this possibility we pooled the very high ( 13–17 ) , high-mid ( 18–26 ) and mid-low ( 27–33 ) fractions and examined whether PHF13 could co-precipitate RNA PolII , PRC2 , H3K4me3 and H3K27me3 from these different fractions ( Figure 4D ) . PHF13 interacted with RNA PolII S2P and S7P in all three pooled fractions ( 13–17 , 18–26 and 27–33 ) where as it interacted with RNA PolII S5P , H3K4me3 and H3K27me3 only in the larger molecular weight fractions ( 13–17 and 18–26 ) . Furthermore , PRC2 predominantly co-precipitated with RNA PolII , PHF13 , H3K4me3 and H3K27me3 in fractions 18–26 . Together , these findings reveal that PHF13 interacts with RNA PolII in different complex constellations and support the idea that PHF13 may play a role in gene regulation . 10 . 7554/eLife . 10607 . 009Figure 4 . PHF13 interacts with RNA polymerase II and PRC2 complexes . ( A ) A String functional protein association network ( http://string-db . org/ ) of PHF13 chromatin interactions as identified by mass spectrometry . High confidence interactions were selected ( 0 . 9 ) , and revealed an interaction network with DDR proteins , PRC2 and RNA polymerase II complex . ( B ) Immunoblot of co-immunoprecipitating interactions from control IgG , PHF13 , SUZ12 , PolII S2P and PolII S5P immunoprecipitations from nuclease digested chromatin fraction of E14 mESC cells . ( C ) Immunoblot of even fraction numbers collected from a superose 6 size exclusion chromatography run using nuclease digested chromatin fraction of E14 mESC cells . The co-elution of RNA PolII with PRC2 , H3K4me3 , H3K27me3 and PHF13 in fractions 18–26 is denoted by red hatched box . ( D ) Immunoblot of interacting proteins from an IgG and PHF13 immunoprecipitation of the pooled chromatography fractions 13–17 , 18–26 and 27–33 . Inputs for SUZ12 and H3K27me3 are separately boxed due to the fact that they represent a separate exposure in relation to the IPs . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 00910 . 7554/eLife . 10607 . 010Figure 4—source data 1 . Mass spectrometry table of PHF13 interacting proteins . Table represents the proteins identified by mass spectrometry obtained from PHF13 immunoprecipitations from nuclease digested chromatin lysates from E14 ESCs that were either trypsin digested or LysC digested . Shown is the number of unique peptides identified as well as the posterior error probability ( PEP ) or q-value . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01010 . 7554/eLife . 10607 . 011Figure 4—source data 2 . Associated functional terms of mass spectrometry interacting proteins . Shown are the top biological processes and molecular functions and their false discovery rates ( FDR ) , determined using by the string functional protein association network ( http://string-db . org/ ) analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01110 . 7554/eLife . 10607 . 012Figure 4—figure supplement 1 . String protein functional association networks for PHF13 interactions when digested by Trypsin ( A ) or by LysC ( B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01210 . 7554/eLife . 10607 . 013Figure 4—figure supplement 2 . Mass spectrometry profiles of PHF13 interacting proteins where only one unique peptide was identified . Manually verified MS/MS spectra of unique peptides from indicated proteins with the annotated b- and y- ion series of the higher-energy collision dissociation ( HCD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01310 . 7554/eLife . 10607 . 014Figure 4—figure supplement 3 . Mass spectrometry profiles of PHF13 interacting proteins where only one unique peptide was identified . Manually verified MS/MS spectra of unique peptides from indicated proteins with the annotated b- and y- ion series of the higher-energy collision dissociation ( HCD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01410 . 7554/eLife . 10607 . 015Figure 4—figure supplement 4 . Mass spectrometry profiles of PHF13 interacting proteins where only one unique peptide was identified . Manually verified MS/MS spectra of unique peptides from indicated proteins with the annotated b- and y- ion series of the higher-energy collision dissociation ( HCD ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 015 Our results strongly suggest that PHF13 binds nucleosomes marked by H3K4me2/3 via its C-terminal PHD domain – an interaction that can be further stabilized by PHF13’s ability to bind DNA and chromatin affiliated PRC2 and RNA PolII . Thus , we expect that PHF13 should co-localize with H3K4me2/3 , PRC2 ( SUZ12 and EZH2 ) , RNA PolII and certain DNA sequence motifs in vivo and genome wide . To test this idea we identified PHF13 bound regions by chromatin immunoprecipitation followed by sequencing ( ChIPseq ) and quantified their overlap with regions marked or bound by H3K4me1 , H3K4me2 , H3K4me3 , H3K9me3 , H3K27me3 , SUZ12 , EZH2 and the different hypophosphorylated and phosphorylated forms of RNA PolII ( Figure 5 ) obtained from publicly available ChIPseq datasets for mESCs ( Materials and methods ) . We identified 17 , 937 PHF13 bound regions and confirmed a few of them by ChIP qPCR ( Figure 5—figure supplement 1A ) . PHF13 bound regions strongly overlap with H3K4me3 ( 87% ) and H3K4me2 ( 71% ) , whereas this overlap was decreased with H3K4me1 ( 39%; Figure 5A ) . Combined with the biochemical and structural findings , these observations support that PHF13 is a bona fide H3K4me2/3 reader . 10 . 7554/eLife . 10607 . 016Figure 5 . ChIPseq shows a genome wide overlap with methylated H3K4 , DHS , CpG islands , PRC2 and RNAPolII . ( A ) Venn diagrams for the overlap of called-peaks . ( B ) Heatmaps for the ChIPseq signal centered around PHF13 peaks . Shown are two groups of peaks that are the result of a k-means clustering of all ChIPseq signals except the one from PHF13 , the DHS signal , and the CpG content . Above the heatmap the average signal for the two groups is plotted . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01610 . 7554/eLife . 10607 . 017Figure 5—figure supplement 1 . ChIP qPCR analysis of PHF13 ChIPseq targets . PHF13 was chromatin immunoprecipitated from formaldehyde fixed E14 mESCs and analyzed for PHF13 binding at target or negative control regions in comparison to IgG binding . Shown is a representative ChIP qPCR , standard deviations are qPCR technical replicates . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01710 . 7554/eLife . 10607 . 018Figure 5—figure supplement 2 . RFAT motif finder of PHF13 ChIP sequencing peak . The RSAT motif finder identified 21 recognizable motifs based on PHF13 ChIP sequencing peaks . These motifs show that sequences rich in CpG peak at the position of PHF13 peaks and extend left and right of the peak . In contrast AT rich sequences are depleted at the peak of PHF13 . Shown are the identified motif sequences and their distribution around the PHF13 peak . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 018 Similarly , we examined the overlap of PHF13 bound regions with H3K9me3 and H3K27me3 repressive modifications ( Figure 5A ) . Not surprisingly , we observed essentially no overlap with H3K9me3 , a mark that is mutually exclusive to H3K4me3 in mESCs except at imprinted genomic regions ( Dindot et al . , 2009 ) . In contrast however , we did find a modest overlap with H3K27me3 repressive marks ( 10% ) . Similar to H3K27me3 we observed a modest but relevant overlap of PHF13 peaks with EZH2 ( 16% ) and SUZ12 ( 25% ) whereas 5% of EZH2 and 11% of SUZ12 peaks overlapped with PHF13 . These findings demonstrate that PHF13 also co-localizes at a subset of Polycomb demarcated loci and is consistent with PHF13 interacting with PRC2 . Next we examined the overlap of PHF13 ChIPseq peaks with the hypophosphorylated , serine 2 , serine 5 and serine 7 phosphorylated forms of RNA PolII ( Figure 5A ) . We observed a substantial overlap between PHF13 targets and hypophosphorylated ( 69% ) , serine 5 phosphorylated ( 76% ) and serine 7 phosphorylated ( 66% ) forms of RNA PolII , whereas the serine 2 phosphorylated form only showed a modest overlap ( 11% ) . These observations argue that PHF13 interacts with promoter affiliated RNA polymerase II and not the elongating form , consistent with the fact that PHF13 is enriched at H3K4me3 ( promoter mark ) and not with H3K36me3 ( found in the gene body ) . Finally , given the in vitro DNA binding activity of PHF13 ( Figure 1 ) , we also asked the question whether PHF13 bound regions coincide with DNAse I hypersensitive ( DHS ) sites ( Figure 5A ) . We observed a significant overlap between PHF13 targets and DHS sites ( 78% ) which is consistent with an in vivo DNA binding activity to nucleosome free and/or complexed DNA . To further discern whether PHF13 recognizes specific DNA motifs , in silico analysis was performed using RSAT Peak motifs ( Thomas-Chollier et al . , 2012 ) . This analysis found that PHF13 bound regions were enriched at CpG rich motifs and depleted at AT rich sequences ( Figure 5—figure supplement 2 ) . This observation prompted us to additionally quantify the overlap of PHF13 targets with annotated CpG islands ( Figure 5A ) , which similarly revealed a significant overlap ( ~54% ) . Together , these findings implicate CpG rich DNA in PHF13 chromatin recruitment and/or stabilization at H3K4me2/3 . The relationship of PHF13 at active and repressed promoters was further visualized in the UCSC genome browser ( Figure 5—figure supplement 1B ) . To better understand these PHF13 chromatin contexts we employed the k-means clustering approach to separate PHF13 bound regions by the patterns of histone modifications and chromatin binding proteins ( Figure 5B; see Materials and methods for details ) . The results of this analysis ( Figure 5B ) show two distinct regimes . The first regime is a group of PHF13 bound regions containing H3K4me2/3 , H3K27me3 , PRC2 ( SUZ12 and EZH2 ) , RNA PolII S5P , CpG enrichment and is depleted of RNA PolII S2P and DNase I hypersensitivity . The absence of RNA PolII S2P and DNase I hypersensitivity and the presence of PRC2 and H3K27me3 is consistent with these regions being not accessible and repressed . Interestingly , this regime is reminiscent of bivalent promoters which are demarcated by the presence of both H3K4me3 and H3K27me3 , Polycomb , high CpG content and specifically the serine 5 phosphorylated form of RNA polymerase II ( Brookes et al . , 2012; Lynch et al . , 2012; Orlando et al . , 2012; Thurman et al . , 2012; Wachter et al . , 2014 ) . Furthermore , it is worthy to note that bivalent promoters represent only a small fraction of H3K4me3 and H3K27me3 positive promoters , and might reflect the approximate 10% representation observed here . The second larger group contains H3K4me2/3 , RNA PolII S2P , RNA PolII S5P , DHS sites , CpG enrichment and a notable absence of H3K27me3 and PRC2 , indicating that these regions are accessible and expressed ( Antequera and Bird , 1999; Orlando et al . , 2012; Tazi and Bird , 1990; Thurman et al . , 2012 ) . In all cases , with the exceptions of H3K27me3 and RNA PolII S2P , the levels peak at the position where PHF13 levels are highest , indicating a co-localization of the signals . These findings indicate that CpG rich sequences , RNA PolII S5P and H3K4me2/3 coexist at PHF13 targets and therefore likely cooperate in PHF13 recruitment and/or function at active and repressed chromatin states . To gain insight about the functional role of genes targeted by PHF13 , we performed an overrepresentation analysis of associated functional terms . We defined PHF13 target genes ( 10 , 826 ) by intersecting a window of +/- 1500 base pairs around their transcription start sites ( TSS ) with PHF13 bound regions ( Figure 6—source data 1 ) . The overrepresentation analysis revealed that PHF13 targets are functionally enriched in transcription , cell cycle , DNA repair , chromatin organization and developmental processes ( Figure 6A and Figure 6—source data 2 ) consistent with previous reports on PHF13 functions ( Bördlein et al . , 2011; Kinkley et al . , 2009; Mund et al . , 2012 ) . 10 . 7554/eLife . 10607 . 019Figure 6 . PHF13 ChIP and RNA sequencing overlap and associated functional terms . ( A ) Gene set enrichment analysis of genes overlapping with PHF13 ChIPseq peaks located within +/- 1500 bp of their TSSs . ( B ) Venn diagrams showing the overlap of PHF13 ChIPseq target genes and the genes that were significantly up or down regulated ( with an adjusted p-value less than 0 . 05 ) after PHF13 knockdown . ( C ) Gene ontology analysis of the significantly differentially expressed genes that were up or down regulated ( with an adjusted p-value of less than 0 . 05 ) after PHF13 knockdown . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 01910 . 7554/eLife . 10607 . 020Figure 6—source data 1 . PHF13 ChIPseq targets- PHF13 ChIPseq peaks in mouse ES cells located +/- 1500 bp of the TSSs were used to identify 10 , 826 PHF13 target genes . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 02010 . 7554/eLife . 10607 . 021Figure 6—source data 2 . David GO analysis of PHF13 ChIPseq targets- PHF13 ChIP sequencing target genes were analysed by david and resulted in the following biological processes , molecular functions and cellular components being over represented . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 02110 . 7554/eLife . 10607 . 022Figure 6—source data 3 . PHF13 shRNA RNAseq targets- PHF13 shRNA depletion for 12 days led to 1 , 386 genes being significantly up or down with an adjusted p-value less than 0 . 05 in mouse ES cells . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 02210 . 7554/eLife . 10607 . 023Figure 6—source data 4 . David GO analysis of PHF13 regulated genes- Down regulated genes in mESCs after PHF13 knockdown were analyzed by David functional annotation bioinformatic microarray analysis and returned the following biological processes , molecular functions and cellular components as being over represented . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 02310 . 7554/eLife . 10607 . 024Figure 6—source data 5 . David GO analysis of PHF13 regulated genes- Up regulated genes in mESCs after PHF13 knockdown were analyzed by David functional annotation bioinformatic microarray analysis and returned the following biological processes , molecular functions and cellular components as being over represented . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 02410 . 7554/eLife . 10607 . 025Figure 6—figure supplement 1 . qPCR validation of PHF13 regulated genes . ( A ) mRNA expression levels of PHF13 and up and down regulated genes in wild type control transduced E14 mESCs ( ShC 12d ) or in wild type uninduced mESCs ( 48h –Dox and 11d – Dox ) in comparison to PHF13 depleted ( Sh3 transduced ) E14 mESCs or induced mESCs ( 48 hr + Dox and 11d + Dox ) . Shown are representative qPCRs , standard deviations are qPCR technical replicates . ( B ) Immunoblot of PHF13 knockdown efficiency ( sh3 ) in comparison to shControl in E14 ES cell lysates , 12 days post transduction and selection . Tubulin served as a loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 025 Since PHF13 interacts with RNA PolII in different constellations , with and without PRC2 ( Figure 4 ) and since the ChIP sequencing results indicated that PHF13 localized to the promoters of both active and repressed genomic regions ( Figure 5 ) , we suspected that PHF13 may have a transcriptional co-regulatory function and that its depletion in mESCs may positively or negatively influence gene expression . To address this possibility we performed RNAseq on mESCs that were transduced with a PHF13 specific shRNA or a scrambled shRNA control ( Figure 6B and C ) . The efficiency of the knockdown was controlled 12 days later by qPCR and western blot and showed an approximate 80% reduction ( Figure 6—figure supplement 1 ) . Differential gene expression analysis showed a total of 1386 genes ( Figure 6—source data 3 ) that were up or down regulated at an adjusted p-value less than 0 . 05 . Of these 807 genes went down and 579 genes went up after PHF13 depletion , supporting the idea that PHF13 can promote both gene activation and repression . Several up and down regulated genes identified by RNA sequencing were confirmed by qPCR in control and PHF13 shRNA depleted E14 mESCs and in a Doxycyclin inducible PHF13 shRNA mESC cell line that expresses a completely independent shRNA ( Figure 6—figure supplement 1 ) . Of the 1386 genes that were affected by PHF13 depletion , 845 genes were also identified as PHF13 targets by the ChIPseq ( Figure 6B ) supporting a direct correlation between the occupancy of PHF13 at these targets and their expression levels . Of these 845 genes , 487 were upregulated and 358 were downregulated . Gene set enrichment analysis of the genes affected by PHF13 depletion further revealed a similar overrepresentation of associated functional terms to ChIPseq and mass spectrometry targets and identified transcription , cell cycle , chromosome organization , differentiation , DNA binding , RNA binding and chromatin binding ( Figure 6C and Figure 6—source data 4 and 5 ) . Together these findings support that PHF13 interacts with transcriptional regulatory proteins , that it binds to genes influencing transcription , cell cycle , chromosome organization and differentiation and that its depletion alters genes with similarly associated functional terms , all of which is consistent with a transcriptional co-regulatory role of PHF13 . To further explore the relationship between the differentially expressed genes and the ChIPseq levels of PHF13 , H3K4me1 , H3K4me2 , H3K4me3 , H3K27me3 , EZH2 , SUZ12 , and the differently phosphorylated forms of RNA PolII we compared their normalized levels to the corresponding levels in the unchanged genes ( Figure 7A ) . We found that the genes that were both up- and down- regulated had significantly higher PHF13 enrichment than the genes that were unchanged ( pvalue < 0 . 0001; Wilcoxon rank sum and signed rank test ) reflecting a causal relationship of PHF13 binding and function . Furthermore , both the genes that were up- and down- regulated had significantly higher levels of H3K4me1 , H3K4me2 , H3K4me3 and RNA PolII and significantly lower levels of H3K27me3 in comparison to the unchanged genes . This reflects the fact that PHF13 target genes are transcriptionally more active than the non-changed genes which are predominantly repressed and is consistent with PHF13 specificity for H3K4me2/3 . The genes that decrease in expression after PHF13 depletion had significantly more H3K4me3 and RNA PolII S2P than the genes that increased after PHF13 depletion , indicating that they are highly expressed . In contrast , the genes that increase after PHF13 depletion have significantly higher levels of H3K4me1 , H3K4me2 , H3K27me3 , EZH2 and SUZ12 and significantly less RNA PolII S2P , than the genes that decrease after PHF13 depletion indicating that they are normally silenced . Interestingly , we also observed significantly higher levels of H3K4me3 on these genes in comparison to the non-changed genes , albeit with significantly lower H3K4me3 levels than on the genes that decreased upon PHF13 depletion . We interpret the co-occurrence of H3K4me2/3 , H3K27me3 and PRC2 ( SUZ12 and EZH2 ) , with high levels of PolII S5P and low levels of PolIIS2P to represent bivalent domains . Consistently , these chromatin features resemble the first PHF13 cluster identified k-means clustering ( Figure 5B ) . 10 . 7554/eLife . 10607 . 026Figure 7 . PHF13 modulates recruitment and/or stabilizes RNA PolII S5P and RNA PolII S5P / PRC2 complexes at H3K4me1/2/3 demarcated chromatin . ( A ) PHF13 targets that go up or down after depletion are under wild-type conditions enriched or depleted of PRC2 , respectively and are enriched in H3K4me1/2/3 and RNA PolII . Differential transcript expression upon PHF13 knockdown in murine ES cells . For each ChIPseq track the normalized signal around the transcription start site is shown for transcripts , which are down-regulated ( blue ) , not changed ( white ) , up-regulated ( red ) at an adjust p-value threshold of 0 . 05 . The bars above each boxplot denote statistical significant higher signal in the down-/up-regulated genes compared to the unchanged transcripts and in the down regulated genes compared to the up regulated genes . The bars below each boxplot denote statistical significant lower signals in the down-/up-regulated genes compared to the unchanged transcripts and in the down regulated genes compared to the up regulated genes . The stars above and below the bars denote the significance threshold , which is for one star 0 . 01 , two 0 . 001 and three 0 . 0001 . P-values were calculated with the Wilcoxon rank sum test . ( B ) Immunoblots of co-precipitating proteins from IgG , PHF13 , SUZ12 , PolIIS2 and PolIIS5 immunoprecipitations performed in nuclease digested chromatin lysates from wild type ( - Doxycyclin ) or PHF13 depleted ( + Doxycyclin ) mESCs . ( C ) Left Panel: PHF13 interacts with RNA PolII ( S5P and S7P ) and most probably as of yet unknown transcriptional activating chromatin modulating complexes . Furthermore , PHF13 simultaneously binds to H3K4me3 and CpG rich DNA at the promoters of active genes . This in turn , acts to either recruit and/or stabilizes these complexes at active promoters , promoting transcription . Right Panel: PHF13 interacts with RNA PolII S5P and PRC2 . PHF13 simultaneously binds to bivalent CpG rich , H3K4me2/3 and H3K27me3 enriched promoters of silenced genes . This in turn acts to recruit and/or stabilizes PRC2 and RNA PolII S5P at bivalent promoters , thereby promoting gene repression . DOI: http://dx . doi . org/10 . 7554/eLife . 10607 . 026 The observations that PHF13 interacts with PRC2 and RNA polymerase II , that they co-localize together at both active and inactive gene promoters and that PHF13 depletion alters the gene expression of a subset of these has lead us to propose that PHF13 acts as a transcriptional co-regulator . In an effort to better understand how PHF13 is influencing transcription , we employed a Doxycyclin inducible PHF13 shRNA mESC cell line . Using this tool we immunoprecipitated PHF13 , SUZ12 , RNA PolII S2P and RNA PolII S5P from wild type ( - Dox ) and PHF13 depleted ( + Dox ) nuclease digested chromatin lysates . We then checked whether PHF13 depletion interfered with the co-precipitation of PHF13 , SUZ12 , RNA PolII S2P , RNA PolII S5P , H3K4me3 and H3K27me3 , from each of these immunoprecipitations ( Figure 7B ) . As expected PHF13 knockdown ( + Dox ) resulted in a reduction in the total levels of PHF13 and in its reduced precipitation and co-precipitation by SUZ12 , RNA PolII S2P and RNA PolII S5P . Interestingly , in addition we observed that SUZ12 , RNA PolII S2P and RNA PolII S5P also showed reduced co-precipitation of H3K27me3 and that SUZ12 and RNA PolII S5P , but not RNA PolII S2P showed a reduced co-precipitation with H3K4me3 ( Figure 7B ) , a finding that is consistent with the fact that RNA Poll S2P is not normally localized at H3K4me3 demarcated promoters . Furthermore , RNA PolII S5P additionally showed a reduced co-precipitation of PRC2 ( SUZ12 and EZH2 ) , whereas this interaction was unaffected in RNA PolII S2P ( Figure 7B ) . These findings suggest that PHF13 can act as a scaffold stabilizing RNA Poll S5P and PRC2 at bivalent H3K4me3 and H3K27me3 demarcated promoters and similarly stabilizes RNA Poll S5P likely with other transcriptional activating complexes at active H3K4me3 demarcated promoters . These observations indicate an influential role of PHF13 in targeting or stabilizing transcriptional activating and repressing complexes at H3K4me2/3 containing promoters and thereby impacting transcriptional activity .
In this manuscript we demonstrate that PHF13 is a novel H3K4me2/3 reader and effector . We show that its affiliation with H3K4me2/3 and CpG rich DNA sequences facilitates its specific recruitment to repressed or active chromatin regions , interactions that are conceivably strengthened by additional interactions with PRC2 and RNA PolII . PHF13 depletion disrupts the interactions between PRC2 , RNA PolII S5P , H3K4me3 and H3K27me3 , indicating that they are in common complexes and that PHF13 is integral to their localization at active and bivalent promoters . Furthermore , PHF13 depletion alters the gene expression of a subset of genes affiliated with these proteins consistent with a co-transcriptional role . The PHD domain of PHF13 displays a specific albeit weak binding affinity to H3K4me2/3 , a phenomenon that is common to many PHD domains ( Musselman and Kutateladze , 2009; Musselman and Kutateladze , 2011 ) and indicates that additional chromatin interactions are required for this interaction to be stable under physiological conditions . In line with this , we show that PHF13 can directly affiliate with nucleosome free or complexed DNA via a centrally located domain and with chromatin affiliated PRC2 and RNA PolII , arguing that PHF13 contacts chromatin in a multivalent fashion . The ability to simultaneously bind DNA , H3K4me2/3 and other chromatin affiliated proteins increases PHF13’s binding avidity for chromatin , thereby stabilizing weak H3K4me2/3 interactions . Such a feature is presumably beneficial for PHF13’s ability to recognize DNA damage and perform its DNA repair functions ( Mund et al . , 2012 ) as well as its ability to impact transcription . Analysis of PHF13 ChIPseq targets identified that PHF13 localizes to at least 2 distinct chromatin landscapes ( Figure 5C ) . A Polycomb enriched chromatin environment and an H3K4me3 active chromatin state . Interestingly , PHF13 ChIPseq peaks overlapped predominantly with H3K4me2 ( and to a lesser extent with H3K4me3 ) in Polycomb enriched chromatin and predominantly with H3K4me3 ( and to a lesser extent with H3K4me2 ) at active promoters . This sub-clustering of PHF13 peaks may indicate that PHF13’s functional impact on chromatin depends on whether it interacts with H3K4me2 or H3K4me3 . Consistently , genes that showed increased expression after PHF13 depletion also showed the highest levels of H3K4me2 , H3K27me3 and Polycomb , in comparison to those that decreased after PHF13 depletion which showed the highest levels of H3K4me3 and RNA PolII . In addition , in silico DNA motif analysis performed on PHF13 ChIP sequencing peaks revealed that it is affiliated with CpG rich DNA and a substantial fraction of PHF13 targets overlapped with unmethylated CpG islands at both active and Polycomb repressed promoters . It is noteworthy to mention that many of the factors mediating Polycomb recruitment to unmethylated CpG islands remain elusive . Recently it has been demonstrated that Pcl3 , a protein that is important for stem cell self-renewal , is capable of recruiting PRC2 to CpG islands at a subset of targets via an interaction with SUZ12 ( Hunkapiller et al . , 2012 ) . The co-occurrence of PHF13 and Polycomb in genes that were de-repressed after PHF13 depletion , and the fact that PHF13 interacts with PRC2 ( Figure 4 ) make it tempting to speculate that PHF13 may as well be involved in the recruitment of chromatin effector complexes to specific H3K4me2/3 CpG enriched regions . Consistent with such a possibility , PHF13 depletion reduced the affiliation of SUZ12 and RNA PolII S5P with both H3K4me3 and H3K27me3 arguing for an integral involvement of PHF13 in these interactions . The comparative analysis of the ChIPseq and RNAseq experiments ( Figure 6B and C ) indicates that PHF13 binds and regulates targets involved in transcription , cell cycle , chromosome organization , development and DNA binding . A role for PHF13 in differentiation , higher chromatin order and cell cycle has been previously described ( Bördlein et al . , 2011; Kinkley et al . , 2009; Mund et al . , 2012 ) . What is newly identified is a role of PHF13 in the regulation of gene expression or transcription . Consistent with this possibility , several H3K4me2/3 readers have been previously reported to either positively or negatively impact transcription ( Fortschegger and Shiekhattar , 2011; Shi et al . , 2006 ) . Perhaps the strongest line of support in favor of a transcriptional co-regulatory function is PHF13’s affiliation with RNA polymerase II . PHF13 showed a strong correlation with promoter associated RNA polymerase II ( namely hypophosphorylated , serine 5 phosphorylated and serine 7 phosphorylated ) consistent with its specificity for H3K4me3 . Nevertheless the strongest relationship was observed with RNA PolII S5P which is found at both active and polycomb repressed promoters ( Figure 5B ) . RNA PolII S5P regulates efficient transcription and enhances splicing ( Hsin et al . , 2014 ) , implicating PHF13 in these chromatin functions . Consistent with this possibility , we observed changes in gene expression upon PHF13 depletion ( Figure 6B and C and Figure 6—figure supplement 1 ) as well as interactions with splicing factors by mass spectrometry analysis of PHF13 chromatin interaction partners ( Figure 4A and Figure 4—figure supplement 1 and 2 ) . While the interaction between PHF13 and splicing factors was not confirmed , based on these collective observations it warrants future efforts . Another interesting parallel is that RNA PolII S5P interacts with the H3K27me3 methyltransferase complex PRC2 ( Brookes et al . , 2012 ) , as well as with H3K4me3 ( Set1A , Set1B and MLL ) methyltransferases ( Hsin and Manley , 2012; Hughes et al . , 2004; Skalnikova et al . , 2008 ) . Since the depletion of PHF13 leads to both a reduction of SUZ12 and RNA PolII S5P association with H3K27me3 and H3K4me3 , it is possible that PHF13 also associates with H3K4me3 methyltransferases at active promoters in RNA PolII complexes similar to its affiliation with PRC2 and RNA PolII at repressed promoters . While this is currently just speculation , it is an interesting hypothesis and based on the current observations deserves future investigation . A model depicting PHF13’s ability to affiliate H3K4me3 , DNA , PRC2 and RNA PolII to co-regulate gene expression is shown in Figure 7C . In this model we proposed that direct chromatin interactions with H3K4me3 and CpG rich DNA as well as indirect chromatin interactions with RNA PolII S5P ( potentially also hypomethylated and S7 phosphorylated RNA PolII ) and as of yet unknown transcriptional activating complexes , leads to their chromatin recruitment and or stabilization at active promoters to promote transcriptional activity . Likewise , direct chromatin interactions of PHF13 with H3K4me2/3 and CpG rich DNA and indirect chromatin interactions facilitated through PRC2 and RNA PolII S5P complexes leads to the recruitment and or stabilization of repressive PRC2 transcriptional complex at bivalent promoters , and thereby co-regulates gene repression . Together our results demonstrate that PHF13 can contact chromatin in a multivalent fashion via direct interactions with DNA and H3K4me2/3 and indirectly by affiliation with Polycomb and RNA PolII complexes . ChIP sequencing revealed that in vivo PHF13 associates with distinct chromatin landscapes and that it shows a substantial overlap with H3K4me2/3 , CpG rich DNA and hypo- , Ser5 , and Ser7- phosphorylated RNA PolII , and a modest overlap with a subset of bivalent PRC2 targets . Furthermore , its depletion lead to a reduced interaction of RNA PolII S5P and SUZ12 with H3K4me3 and H3K27me3 and with each other , supporting that there are common complexes between these proteins and that PHF13 may act as a scaffold or bridging factor . Finally , PHF13 depletion in mESCs lead to both the up and down regulation of a subset of target genes functionally enriched in transcription , cell cycle , chromatin organization , DNA binding and developmental processes , explaining in part its ability to modulate such processes . Taken , together these data support that PHF13 is an H3K4me2/3 molecular reader and transcriptional co-regulator of H3K4me3 active and H3K4me3/H3K27me3 bivalent promoters .
HeLa cells obtained from ATCC were used for GST-pull down experiments and were grown in DMEM supplemented with 1x Penicillin/Streptomycin , 10% FBS ( Biochrome ) , 1x hepes and 1x sodium pyruvate . All other experiments were performed with mouse embryonic stem cells ( mESCs ) grown in Glassgow MEM ( Sigma G5154 ) , 20% FBS Hyclone – ThermoScientific SV30160 . 03 , 1x Glutamax , 1x non-essential amino acids , 1x sodium pyruvate and 1 . 2 ml β-mercaptomethanol ( Gibco 31350010 ) . E14 mESCs were obtained originally from ATCC and were given to us by Prof . Luciano DiCroce . E14 mESCs were grown on gelatin coated plates . Inducible PHF13 shRNA mESCs are primary mESCs isolated from mice blastocytes and were purchased from Artemis Pharmaceuticals Gmbh . They were grown similar to E14 mESCs except that they were not grown on gelatin coated plates . Early passages required feeder cells but this was gradually eliminated by 10–15 passages . Both mESCs cell cultures were periodically stained for alkaline phosphatase activity ( Millipore SCR004 ) to ensure they were significantly enriched ( >90% ) in stem cells and contained limited differentiated cells . All cell lines were routinely tested to ensure the absence of mycoplasma ( Thermofisher Scientific C7028 ) . DNA EMSA reactions were performed using a 248 bp DNA fragment radio-labeled by PCR , as previously described ( Hartlepp et al . , 2005 ) or a 40 bp Cy5-end labeled fragment ( 5'-Cy5-CCTGGAGAATCCCGGTGCCGAGGCCGCTCAATTGGTCGTA-3'; Eurofins MWG ) . The Cy5-labeled DNA was incubated with recombinant proteins in 1X binding buffer ( 50 mM Hepes pH7 . 6 , 50 mM NaCl , 2 mM MgCl2 , 10% glycerol and 100 ng/µl BSA ) for 20 min on ice and was analyzed by native PAGE . The label was detected with a Fujifilm Phosphoimager FLA-3000 . Mononucleosomes for EMSAs were generated by salt gradient dialysis as previously described ( Dyer et al . , 2004 ) using recombinant Drosophila histone octamers and either a 200 bp DNA fragment or a 151 bp DNA fragment , both comprising derivatives of the Widom 601 nucleosome positioning sequence . The DNA fragments were released by restriction enzyme digest with NotI and XmaI , respectively , from plasmids comprising 12 repeats of the 200 bp fragment ( Huynh et al . , 2005 ) or four repeats of the 151 bp fragment ( Mueller-Planitz et al . , 2013 ) . The XmaI ends of the 151 bp fragment were labeled with dCTP-Cy5 and Klenow polymerase . The mononucleosomes EMSA reactions were performed and analyzed like DNA EMSAs as described above . Unlabeled mononucleosomes were visualized by ethidium bromide staining . Chromatography was performed using nuclease digested chromatin lysates obtained from 14 ( 15cm ) plates of E14 ES cells and run on a Superose 6 10/300 GL column ( GE Healthcare ) in a buffer composed of 50 mM Tris-HCl ( pH 7 . 5 ) and 100 mM NaCl with a flow rate of 0 . 3 ml/min . Fractions were collected every 2 min . Chromatin immunoprecipitation was performed using a standard technique . In brief E14 ES cells were grown in the presence of LIF and 20% FBS . The cells were collected and cross-linked for 15 min with 1% formaldehyde . The fixation reaction was quenched by the addition of glycine ( 0 . 125 M ) for 10 min . The cells were then lysed on ice for 10 min in 1 . 3 ml of ChIP buffer , 1 volume of SDS buffer ( 100 mM NaCl , 50 mM Tris-HCl pH 8 , 5 mM EDTA , 0 . 2% NaN3 and 0 . 5% SDS ) : 0 . 5 volume of Triton dilution buffer ( 100 mM Tris-HCl pH8 . 6 , 100 mM NaCl , 5 mM EDTA , 0 . 2% NaN3 and 5% Triton-X 100 ) . The lysate was passed through a syringe multiple times and then sonicated in Bioruptor at high frequency for 2 x 6 cycles . The lysate was then centrifuged at 13 , 000 rpm for 30 min to pellet debris and a protein determination was measured . 1 mg of chromatin was then incubated with 5 µl of PHF13 antibody ( CR56 ) or rabbit IgG control antibody overnight at 4°C under rotation . 30 µl of protein A beads were then added to the reaction and allowed to incubate at 4°C for 2h . The beads were then washed by successive low salt ( 0 . 1% SDS , 1% Triton X , 2 mM EDTA , 20 mM Tris-HCl and 150 mM NaCl ) , high salt ( 0 . 1% SDS , 1% Triton X , 2 mM EDTA , 20 mM Tris-HCl and 500 mM NaCl ) and LiCl washes and then eluted for 3h at 65°C with 110 µl of elution buffer ( 1% SDS and 100 mM NaHCO3 ) . The RNA and proteins were then digested with RNAse A and Proteinase K and the DNA purified and sequenced . Total cellular RNA was collected from wild-type E14 ES cells or from PHF13 depleted ES cells using RNeasy plus mini-kit ( Qiagen ) according to manufactures instructions . 1 . 5 µg of total RNA was then depleted of ribosomal RNA using the Ribo zero magnetic kit ( Epicenter ) according to the manufacturer’s directions . The rRNA depleted RNA was then measured by Qubit and 20 ng was used to generate the RNA library using ScriptSeq v2 RNA-Seq library preparation kit ( Epicenter ) according to the manufacturer’s instructions . Accession numbers are from the short read archive ( http://www . ncbi . nlm . nih . gov/sra ) . ChIPseq TargetAccession numberReferenceH3K4me1SRX122631Xiao et al . , 2012H3K4me1SRX185842Mouse ENCODEH3K4me1SRX186789Mouse ENCODEH3K4me2SRR414940Xiao et al . , 2012H3K4me3SRX026272Marks et al . , 2012H3K4me3SRX149184Ng et al . , 2013H3K4me3SRX149188Ng et al . , 2013H3K4me3SRX185845Mouse ENCODEH3K4me3SRX186795Mouse ENCODEH3K27me3SRX006976Marks et al . , 2012H3K27me3SRX026276Marks et al . , 2012H3K27me3SRX122629Xiao et al . , 2012EZH2SRR501769Marks et al . , 2012SUZ12SRX206420Morey et al . , 2013DHSSRX191012Mouse ENCODEH3K9me3SRX112916Marks et al . , 2012H3K9me3SRX186790Mouse ENCODEPolII-8WG16SRR391036 , SRR391037Brookes et al . , 2012PolII-S2PSRR391034 , SRR391038 , SRR391039Brookes et al . , 2012PolII-S5PSRR391032 , SRR391033 , SRR391050Brookes et al . , 2012PolII-S7PSRR391035 , SRR391040Brookes et al . , 2012 | In human and other eukaryotic cells , DNA is packaged around proteins called histones to form a structure known as chromatin . Chemical tags added to the histones alter how the DNA is packaged and the activity of the genes encoded by that DNA . For example , many active genes are packaged around histone H3 proteins that have “Lysine 4 tri-methyl” tags attached to them . Another protein that is associated with chromatin is called PHF13 and it has several roles , including repairing damaged DNA . However , it was not known whether PHF13 binds to chromatin via the chemical tags , or in another way . Ho-Ryun , Xu , Fuchs et al . used several biochemical techniques in mouse and human cells to explore how PHF13 specifically interacts with chromatin . These experiments showed that PHF13 binds specifically to DNA and to two types of methyl tags ( lysine 4-tri-methyl or lysine 4-di-methyl ) . These chemical tags are predominantly found at active promoters as well as at a small subset of less active promoters known as bivalent promoters . PHF13 interacted with other proteins on the chromatin that are known to either drive or repress gene activity and it’s depletion affected the activity of many genes . Whether PHF13 increased or decreased gene activity depended on whether it was bound to active or bivalent promoters . The active promoters targeted by PHF13 had higher numbers of the tri-methyl tags whereas the di-methyl tags were more common on the bivalent promoters . These findings provide preliminary evidence that a protein binding to different methyl tags in the same place on histone H3 can have opposite effects on gene activity . Ho-Ryun , Xu , Fuchs et al . now intend to find out more about the other proteins that interact with PHF13 on chromatin . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | PHF13 is a molecular reader and transcriptional co-regulator of H3K4me2/3 |
Injured mature CNS axons do not regenerate in mammals . Deletion of PTEN , the negative regulator of PI3K , induces CNS axon regeneration through the activation of PI3K-mTOR signaling . We have conducted an extensive molecular dissection of the cross-regulating mechanisms in axon regeneration that involve the downstream effectors of PI3K , AKT and the two mTOR complexes ( mTORC1 and mTORC2 ) . We found that the predominant AKT isoform in CNS , AKT3 , induces much more robust axon regeneration than AKT1 and that activation of mTORC1 and inhibition of GSK3β are two critical parallel pathways for AKT-induced axon regeneration . Surprisingly , phosphorylation of T308 and S473 of AKT play opposite roles in GSK3β phosphorylation and inhibition , by which mTORC2 and pAKT-S473 negatively regulate axon regeneration . Thus , our study revealed a complex neuron-intrinsic balancing mechanism involving AKT as the nodal point of PI3K , mTORC1/2 and GSK3β that coordinates both positive and negative cues to regulate adult CNS axon regeneration .
Injuries of mature central nervous system ( CNS ) axons result in loss of vital functions due to the failure of CNS axons to regenerate ( Schwab and Bartholdi , 1996; Goldberg et al . , 2002b; Fitch and Silver , 2008 ) . We and other investigators have found that the activation of phosphatidylinositol 3-kinase ( PI3K ) by the deletion of phosphatase and tensin homolog ( PTEN ) induces CNS axon regeneration through the activation of mammalian target of rapamycin ( mTOR ) signaling ( Park et al . , 2008; Liu et al . , 2010 ) . PI3K is a lipid kinase , which can be activated by receptor tyrosine kinase ( RTK ) and subsequently phosphorylates phosphatidylinositol 4 , 5-bisphosphate ( PIP2 ) in the lipid membrane to produce phosphatidylinositol ( 3 , 4 , 5 ) -triphosphate ( PIP3 ) . PIP3 in turn recruits AKT , a member of the AGC family of serine/threonine kinases , to the membrane to be activated by phosphorylation at T308 via phosphoinositide-dependent kinase-1 ( PDK1 ) ( Manning and Cantley , 2007 ) . PTEN is a lipid phosphatase that converts PIP3 to PIP2 , and thus inhibits the activation of downstream effectors of PI3K . The PI3K-AKT pathway is the main effector by which RTKs promote cell survival and growth in response to growth factor signaling ( Song et al . , 2012a ) . One of the multiple AKT downstream effectors is the tuberous sclerosis complex ( TSC1/TSC2 ) , the negative regulator of mTOR complex 1 ( mTORC1 ) . Thus AKT activation removes the inhibition of TSC and activates mTORC1 . The PI3K-AKT-mTORC1 pathway is a master regulator of protein synthesis and cellular growth ( Manning and Cantley , 2007; Laplante and Sabatini , 2012 ) . The two best characterized substrates of mTORC1 have been suggested as mediators of mTORC1’s prominent roles in regulating cell growth , size , proliferation , motility and survival . One of these is the eukaryotic initiation factor 4E-binding proteins ( 4E-BPs ) ; their phosphorylation releases inhibition of eukaryotic translation initiation factor 4E and initiates cap-dependent translation . The other is the ribosomal protein S6 kinases ( S6Ks ) ; their phosphorylation is critical for mRNA biogenesis , translation initiation and elongation ( Hay and Sonenberg , 2004 ) . Phosphorylation of ribosome protein S6 ( pS6 ) by S6K has often served as a marker for mTORC1 activation . Interestingly , S6K also functions as feedback inhibition of RTK/PI3K signaling , which balances mTORC1 activation ( Radimerski et al . , 2002; Um et al . , 2004; Yang et al . , 2014 ) . Both 4E-BP inhibition and S6K activation promote protein synthesis , but 4E-BPs act specifically to control cell proliferation ( cell number ) and S6Ks preferentially regulate cell growth ( cell size ) ( Dowling et al . , 2010; Ohanna et al . , 2005 ) . We previously reported that , although S6K1 activation , but not 4E-BP inhibition , is sufficient for axon regeneration , 4E-BP inhibition is necessary for PTEN deletion-induced axon regeneration ( Yang et al . , 2014 ) . Phosphorylation and inhibition of another AKT substrate , glycogen synthase kinase 3β ( GSK-3β ) , is critical for neuronal polarization , axon branching and axon growth ( Kim et al . , 2011b ) . However , how GSK-3β regulates peripheral axon regeneration is controversial ( Saijilafu et al . , 2013; Zhang et al . , 2014a; Gobrecht et al . , 2014 ) and its function in CNS axon regeneration remains to be determined . Although the mechanism is unclear , PI3K also activates mTOR complex 2 ( mTORC2 ) in a ribosome-dependent manner ( Zinzalla et al . , 2011 ) , which in turn phosphorylates AKT at S473 ( pS473 ) ( Sarbassov et al . , 2005; Guertin et al . , 2006; Hresko and Mueckler , 2005 ) . pS473 enhances phosphorylation of AKT-T308 ( Yang et al . , 2002; Scheid et al . , 2002 ) , and inhibition of S473 phosphorylation by destroying mTORC2 decreases AKT-T308 phosphorylation ( Sarbassov et al . , 2005; Carson et al . , 2013; Yuan et al . , 2012; Hresko and Mueckler , 2005; Guertin et al . , 2009 ) . mTORC2 is involved in cell survival and actin cytoskeleton dynamics ( Jacinto et al . , 2004 ) and , like mTORC1 , plays a role in lipogenesis and adipogenesis ( Yao et al . , 2013; Lamming and Sabatini , 2013 ) . It is not clear how these two mTOR complexes interact to determine multiple cellular events , and this information is particularly lacking for axon regeneration . AKT appears to be the nodal point that acts as the key substrate of both PI3K-PDK1 and PI3K-mTORC2 , and is also the critical upstream regulator of mTORC1 and GSK3β . By exploiting the anatomical and technical advantages of retinal ganglion cells ( RGCs ) and the crushed optic nerve ( ON ) as an in vivo model , we elucidated the important roles of AKT , mTORC1/2 and GSK3β in adult CNS axon regeneration . Understanding this cross-regulating mechanism should provide promising therapeutic targets for CNS injuries .
To determine the role of AKT in axon regeneration , we generated adeno-associated virus 2 ( AAV2 ) vectors containing Myr-3HA-AKT1 , 2 and 3 to express membrane-bound constitutively active forms of HA-tagged AKTs in vivoWe and other investigators have demonstrated a specific tropism of AAV2 for RGCs after intravitreal injection ( Park et al . , 2008; Pang et al . , 2008; Hu et al . , 2012; Boye et al . , 2013; Yang et al . , 2014 ) . We made AAV2-AKT viruses containing triple mutant capsid ( Y444 , 500 , 730F ) to take the advantage of the high RGC transduction efficiency of capsid-mutated AAV2 ( Petrs-Silva et al . , 2011 ) : transduction exceeded 80% of RGC , based on the ratio of HA ( transgene tag ) to Tuj1 ( antibody for RGC marker , β-III tubulin ) positive cells in flat-mount retinas ( Figure 1A , B ) . We also confirmed that all three isoforms of AKT produced the expected changes in RGCs: significantly increased levels of pAKT-S473 and pS6 ( mTORC1 activation marker ) , indicating activation of AKT and mTORC1 ( Figure 1A , B , note that pAKT-T308 antibody did not effectively immunostain retina ) . Western blot analysis of retina lysates showed comparable expression of AKT isoforms ( HA levels ) , but levels of pT308 and pS6 were significantly higher in retinas transfected by AKT3 than by AKT1 ( Figure 1C , D ) . This difference indicates the greater activity of AKT3 in retina . We noticed that both pAKT-T308 and pAKT-S473 antibodies can only be used to detect AKT1 and AKT3 reliably in Western blot , whereas the pAKT2-S474 specific antibody readily detected p-AKT2 ( Figure 1C ) . 10 . 7554/eLife . 14908 . 003Figure 1 . Overexpression of three constitutively active AKT isoforms in RGCs . ( A ) Confocal images of flat-mounted retinas showing co-labeling of HA tag , Tuj1 , pAKT-S473 and their merged images , and pS6 in a separate retina sample . Scale bar , 20 µm . ( B ) Quantification of HA , pAKT-S473 or pS6 positive RGCs . Data are presented as means ± s . e . m , n=6 . ( C ) Western blot of retina lysates from three biological replicates showing expression levels of HA-AKT isoforms , and phosphorylation levels of AKT-T308 , AKT-S473 and S6 . ( D ) Quantification of Western blots . *p<0 . 05 . Data are presented as means ± s . e . m , n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 003 We then performed ON crush in wild type ( WT ) mice 2 weeks after intravitreal injection of these AAV-AKTs . RGC axons that regenerated through the lesion site were labeled anterogradely by intravitreal injection of the tracer Alexa 488-conjugated cholera toxin β ( CTB ) , and imaged and quantified in ON longitudinal sections at 2 weeks post-crush ( Figure 2—figure supplement 1 ) . Interestingly , the three isoforms of AKT elicited different patterns of axon regeneration: AKT2 and AKT3 caused significantly more axon regeneration than AKT1; AKT2 induced the longest axon growth ( Figure 2A , B ) . The differing capabilities of the three isoforms of AKT in axon regeneration could not be explained simply by their expression levels , which HA staining showed to be comparable ( Figure 1C , D ) . The higher pT308 and pS6 levels and more potent axon regeneration induced by AKT3 suggest that its function in retina differs from those of AKT1 and AKT2 . Similar to their effects on axon regeneration , all three isoforms increased RGC survival to more than 40% based on Tuj1 staining in retinal flat-mounts , at least a 1-fold increase over WT mice; and significantly more RGCs survived in mice injected with AAV-AKT3 than with AKT1 ( Figure 2C , D ) . This series of experiments demonstrated that the downstream effector of PI3K , AKT , promotes both RGC survival and ON regeneration , and that there are significant differences among the three AKT isoforms . 10 . 7554/eLife . 14908 . 004Figure 2 . Differential effects of three AKT isoforms on axon regeneration and RGC survival . ( A ) Confocal images of ON longitudinal sections showing regenerating fibers labeled with CTB-Alexa 488 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( B ) Quantification of regenerating fibers at different distances distal to the lesion site . Data are presented as means ± s . e . m , n=10–20 . ( C ) Confocal images of flat-mounted retinas showing co-labeling of HA-AKTs and Tuj1 , 2 weeks after ON crush . Scale bar , 20 µm . ( D ) Quantification of surviving RGCs , represented as percentage of Tuj1 positive RGCs in the injured eye , compared to the intact contralateral eye . Data are presented as means ± s . e . m , n=8–10 . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 00410 . 7554/eLife . 14908 . 005Figure 2—figure supplement 1 . Illustration of the regenerating axon quantification procedure . Perpendicular lines on the ON sections were drawn distal to the crush site in increments of 250 µm till 1000 µm , then every 500 µm till no fibers were visible . The regenerating fibers that crossed the perpendicular lines were counted . The width of the nerve ( R ) was measured at the point ( d ) at which the counts were taken and used together with the thickness of the section ( t = 8 μm ) to calculate the number of axons per µm2 area of the nerve . The total number of axons per section was calculated based on the formula: ∑ad= πr2 * ( axon number ) / ( R*t ) , which then was averaged over 3 sections per animal . All CTB signals that were in the range of intensity that was set from lowest intensity to the maximum intensity after background subtraction were counted as individual fibers by Nikon NIS Element R4 software . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 00510 . 7554/eLife . 14908 . 006Figure 2—figure supplement 2 . Endogenous expression of three AKT isoforms in RGCs . ( A ) In situ hybridization images of retina sections showing expression of three AKT isoforms in ganglion cell layer ( GCL ) . Scale bar , 20 µm . ( B ) Western blots of retina lysates from three biological replicates showing expression of three AKT isoforms . ( C ) Confocal images of flat-mounted retinas showing co-labeling of HA tag and Tuj1 . Scale bar , 20 µm . ( D ) RGC-specific ribosome-associated translating mRNA purified by Ribo-IP from three replicates of WT mice . ( E ) RNA-seq showing expression levels of three AKT isoforms in RGCs . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 006 Consistent with a previous report ( Yang et al . , 2015 ) , in situ hybridization and Western blot detected all three isoforms of AKT in retina ( Figure 2—figure supplement 2A , B ) . Since AKT3 is the predominant isoform in adult mouse brain ( 50% AKT3 , 30% AKT1 and 20% AKT2 ) ( Easton et al . , 2005 ) , we determined whether this is also the case in RGCs . We employed the RiboTag mice , which are generated by knocking in the HA-tagged ribosome protein Rpl22 ( Rpl22HA ) to the endogenous Rpl22 allele , immediately after the floxed endogenous Rpl22 gene ( Sanz et al . , 2009 ) . After intravitreal injection of AAV2-Cre , Rpl22HA was expressed specifically in RGCs after Cre-mediated deletion of endogenous Rpl22 ( Figure 2—figure supplement 2C ) . This allowed us to selectively immunoprecipitate ( IP ) RGC-specific ribosomes ( Ribo-IP ) in situ ( Doyle et al . , 2008; Heiman et al . , 2008; Sanz et al . , 2009 ) and acquire high quality RGC-specific ribosome-associated translating mRNA from mouse retinas ( Figure 2—figure supplement 2D ) . We then used three biological replicates of RGC-translating mRNAs from WT mice ( 8–10 mice/replicate ) to perform RNA deep sequencing; each library acquired about 25 million reads and 93% of total reads aligned to unique genes in the mouse genome ( mm9 ) by RUM ( Grant et al . , 2011 ) . Measurement of transcript abundance in fragments per Kb of exon per million fragments mapped ( FPKM ) ( Mortazavi et al . , 2008 ) indicated that AKT1 and AKT3 were the two major isoforms of AKT in RGCs; each represented 45% of total AKT , a vivid contrast to AKT2 ( 10% ) ( Figure 2—figure supplement 2E ) . AKT1 and AKT3 are expressed equally in RGCs , but AKT3 is activated more potently in retina and induces better axon regeneration than AKT1 . We therefore focused on AKT3 to investigate which of its domains is critical for axon regeneration . We generated three AKT3 mutants and verified their activities in RGCs after AAV infection: AKT3 kinase dead ( KD ) mutant ( K177M ) , T305A/S472A ( AA ) mutant ( equal to T308 and S473 in AKT1 , T308 and S473 are used throughout the paper for convenience unless AKT3 is specified ) and single T305A mutant were all equally expressed in RGCs and , as we expected , all three mutants showed significantly lower phosphorylation of T308 , S473 and S6 than WT AKT3 ( Figure 3 ) . Consistent with their loss of AKT kinase function , the three mutants did not induce axon regeneration ( Figure 4A , B ) , indicating that the phosphorylation of AKT-T308 and the kinase activity of AKT are critical for axon regeneration . 10 . 7554/eLife . 14908 . 007Figure 3 . Over expression of three AKT3 mutants in RGCs . ( A ) Confocal images of flat-mounted retinas showing co-labeling of HA tag , Tuj1 , pAKT-S473 and their merged images , and pS6 in a separate retina sample . Scale bar , 20 µm . ( B ) Western blot of retina lysates from three biological replicates showing expression level of HA-AKT , and phosphorylation level of AKT-T308 , AKT-S473 and S6 . ( C ) Quantification of Western blots . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 00710 . 7554/eLife . 14908 . 008Figure 4 . Phosphorylation of AKT3-T305 is necessary but phosphorylation of AKT3-S472 by mTORC2 is inhibitory for axon regeneration . ( A ) Confocal images of ON longitudinal sections showing lack of regenerating fibers 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( B ) Quantification of regenerating fibers at different distances distal to the lesion site . Data are presented as means ± s . e . m , n=6–8 . ( C ) Confocal images of ON longitudinal sections showing regenerating fibers labeled with CTB 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( D ) Quantification of regenerating fibers at different distances distal to the lesion site . *p<0 . 05 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=20–30 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 00810 . 7554/eLife . 14908 . 009Figure 4—figure supplement 1 . AAV-Cre-mediated RGC-specific deletion of Rictor , Rptor , Mtor or Gsk3a did not induce axon regeneration . ( A ) Confocal images of ON longitudinal sections showing regenerating fibers labeled with CTB , 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( B ) Quantification of regenerating fibers at different distances distal to the lesion site . Data are presented as means ± s . e . m , n=6 . **p<0 . 01 , ***p<0 . 001DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 00910 . 7554/eLife . 14908 . 010Figure 4—figure supplement 2 . Translocation of AKT in ON . ( A ) Confocal images of ON longitudinal sections showing axons co-labeled with HA tag and Tuj1 . Scale bar , 100 µm . ( B ) Confocal images of ON longitudinal sections showing regenerating fibers co-labeled with HA tag and CTB , 2 weeks after ON crush . Scale bar , 100 µm . *crush site . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 01010 . 7554/eLife . 14908 . 011Figure 4—figure supplement 3 . The effects of AKT3 mutants on RGC survival . ( A ) Confocal images of flat-mounted retinas showing Tuj1 positive RGCs , 2 weeks after ON crush . Scale bar , 20 µm . ( B ) Quantification of surviving RGCs , represented as percentage of Tuj1 positive RGCs in the injured eye , compared to the intact contralateral eye . *p<0 . 05 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=8–10 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 011 We were surprised to find , however , that the AKT3-S472A mutant produced a small but statistically significant increase in axon regeneration at 500 µm distal to the crush site and a trend toward increase at other distances , compared to WT AKT3 ( Figure 4C , D ) , suggesting that pS473 inhibits the effect of AKT on axon regeneration . This contradicts the common theme that phosphorylation of both T308 and S473 is additive to AKT activity ( Bhaskar and Hay , 2007; Yang et al . , 2002; Scheid et al . , 2002 ) . To confirm this unanticipated finding , we tested whether mTORC2 also is inhibitory for AKT-induced axon regeneration , since S473 of AKT is phosphorylated by mTORC2 ( Sarbassov et al . , 2005; Guertin et al . , 2006; Hresko and Mueckler , 2005 ) . We blocked mTORC2 genetically by deleting its essential component , Rictor ( rapamycin-insensitive companion of mTOR ) ( Guertin et al . , 2006; Laplante and Sabatini , 2012 ) . We injected AAV-AKT3 and AAV-Cre together into one eye of Rictor floxed mice and compared ON regeneration to that elicited by the injection of AAV-AKT3 alone into the contralateral eye . Consistent with our AKT3-S472A result , AKT3 and Rictor KO produced even more extensive and lengthier axon regeneration than AKT3 alone ( Figure 4C , D ) . This was not due to an additive effect because Rictor KO alone did not cause any axon regeneration ( Figure 4—figure supplement 1 ) , but presumably through inhibition of S473 phosphorylation . Indeed , both immunostaining and Western blot analysis confirmed that AKT3-S472A mutant and Rictor deletion significantly decreased pS473 levels ( Figure 5A , B , F ) . The enhanced axon regeneration was not due to increased expression of AKT3 , which was slightly decreased in Rictor KO mice ( Figure 5B , D ) . It was also not due to enhanced activation of AKT or mTORC1 , because levels of pT308 and pS6 were also decreased ( Figure 5B , E , G ) . This is consistent with the positive effect of pS473 on T308 phosphorylation ( Sarbassov et al . , 2005; Carson et al . , 2013; Yuan et al . , 2012; Hresko and Mueckler , 2005; Guertin et al . , 2009; Yang et al . , 2002; Scheid et al . , 2002 ) . Surprisingly , mTORC2 and pS473 had a negative effect on GSK3β-S9 phosphorylation; deletion of Rictor significantly increased pGSK3β-S9 level ( Figure 5B , H ) . AKT3-S472A mutant also increased GSK3β-S9 phosphorylation although with large variation . Rictor KO alone did not increase pGSK3β-S9 compared to WT mice ( Figure 5C ) , suggesting that AKT kinase activity is required for GSK3β phosphorylation . Since GSK3β-S9 phosphorylation has been suggested to promote peripheral axon regeneration ( Saijilafu et al . , 2013; Zhang et al . , 2014a ) , inhibition of AKT-S473 phosphorylation may increase CNS axon regeneration through enhanced inactivation of GSK3β . Taken together , our results demonstrate that , in contrast to pAKT-T308 , mTORC2 and pAKT-S473 inhibit GSK3β phosphorylation , which may contribute to their negative effect on axon regeneration . 10 . 7554/eLife . 14908 . 012Figure 5 . Increased phosphorylation of GSK3β-S9 after blocking AKT-S473 phosphorylation in RGCs . ( A ) Confocal images of flat-mounted retinas showing co-labeling of HA tag , Tuj1 , pAKT-S473 and their merged images , and pS6 in a separate retina sample . Scale bar , 20 µm . ( B ) Western blots of retina lysates from three biological replicates showing expression level of HA , and phosphorylation levels of AKT-T308 , AKT-S473 , S6 and GSK3β-S9 . ( C ) Western blots of retina lysates from three biological replicates showing phosphorylation levels of GSK3β-S9 . ( D-H ) Quantification of Western blots . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=3 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 012 Since retina flat-mount preparations revealed HA signals in RGC cell bodies as well as proximal axons , we investigated the distribution of AKT in ON . All three AKT isoforms and AKT3-S472A mutant were translocated into ON . Intriguingly however , the distribution in ON of the three AKT3 mutants ( KD , AA and T305A ) that did not induce axon regeneration was significantly limited ( Figure 4—figure supplement 2A ) , suggesting a local function of axonal AKT in axon regeneration . Consistent with this idea , AKT3 was also present in regenerating axons ( Figure 4—figure supplement 2B ) . Consistent with their loss of kinase function , the three AKT3 mutants ( AKT3-KD , AA and T305A mutants ) significantly decreased survival of axotomized RGCs compared to WT AKT3 ( Figure 4—figure supplement 3 ) . Neither the AKT3-S472A mutant nor Rictor deletion significantly changed RGC survival ( Figure 4—figure supplement 3 ) , indicating that increased axon regeneration is not due to increased neuron survival . Taken all together , these results show that AKT kinase activity is required for both axon regeneration and RGC survival; and that phosphorylation of T308 by PI3K-PDK1 and phosphorylation of S473 by mTORC2 play opposite roles in GSK3β phosphorylation/inhibition and axon regeneration . We next tested the influence of mTORC1 on AKT3-induced axon regeneration . Previous studies of mTORC1 function in axon regeneration relied on pharmacological inhibition of mTORC1 by rapamycin , which generated inconsistent or contradictory results ( Christie et al . , 2010; Abe et al . , 2010; Park et al . , 2008 ) . To definitively determine the role of mTORC1 , we disrupted mTORC1 by deleting Rptor ( regulatory associated protein of mTOR ) in RGCs specifically . Rptor is unique to mTORC1 , and its deletion totally blocks mTORC1 activity ( Laplante and Sabatini , 2012; Guertin et al . , 2006 ) . We injected AAV-AKT3 and AAV-Cre together into one eye of Rptor floxed mice and compared the ON regeneration to that obtained with injection of AAV-AKT3 alone into the contralateral eye . Rptor deletion alone did not result in axon regeneration ( Figure 4—figure supplement 1 ) ; it significantly decreased AKT3-induced axon regeneration ( Figure 6A , B ) . The two best-characterized substrates of mTORC1 , S6K1 and 4E-BP , are necessary for axon regeneration ( Yang et al . , 2014; Hu , 2015 ) , we tested whether the dominant negative mutant of S6K1 ( S6K1-DN ) and constitutively active mutant of 4E-BP1 ( 4E-BP1-4A ) also inhibit AKT3-induced axon regeneration . Similar to Rptor deletion , S6K1-DN and 4E-BP1-4A also significantly decreased AKT3-induced axon regeneration ( Figure 6A , B ) , which further confirms the essential role of mTORC1 in AKT-induced axon regeneration . Since we found that mTORC1 is necessary but that mTORC2 is inhibitory for AKT-induced axon regeneration , we asked what the combined effect would be of blocking both mTORC1 and mTORC2 by deleting Mtor itself . We again injected AAV-AKT3 and AAV-Cre together into one eye of Mtor floxed mice and compared the ON regeneration to that obtained by injecting AAV-AKT3 alone into the contralateral eye . Deletion of Mtor itself did not result in axon regeneration ( Figure 4—figure supplement 1 ) but significantly inhibited AKT3-induced axon regeneration ( Figure 6A , B ) . This result is additional evidence for mTORC1’s essential role in axon regeneration . 10 . 7554/eLife . 14908 . 013Figure 6 . mTORC1 and its downstream effectors are essential for AKT3-induced axon regeneration . ( A ) Confocal images of ON longitudinal sections showing regenerating fibers labeled with CTB 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( B ) Quantification of regenerating fibers at different distances distal to the lesion site . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=8–10 . ( C , D ) Western blots of retina lysates from three biological replicates showing expression levels of HA-AKT3 , and phosphorylation levels of S6 . ( E ) Confocal images of flat-mounted retinas showing pS6 levels and Tuj1 positive RGCs , 2 weeks after ON crush . Scale bar , 20 µm . ( F ) Quantification of surviving RGCs , represented as percentage of Tuj1 positive RGCs in the injured eye , compared to the intact contralateral eye . Data are presented as means ± s . e . m , n=8–14 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 013 Consistent with its key role in protein synthesis , mTORC1 inhibition by deletion of Rptor or Mtor , or by overexpression of S6K1-DN and 4E-BP1-4A , significantly decreased the AAV-AKT3 expression ( Figure 6C , D ) . However , RGC survival was not changed by these manipulations ( Figure 6E , F ) , which suggests that the low level of AAV-AKT3 expression in these conditions is sufficient for its neuroprotection function in retina . The inhibitory effect of Rptor/Mtor KO or S6K1-DN mutant on pS6 levels ( Figure 6C–F ) implies that downregulation of mTORC1 activity is responsible for the decreased axon regeneration . We next investigated the influence of another substrate of AKT , GSK3β . GSK3β has been suggested to negatively regulate mammalian peripheral ( Saijilafu et al . , 2013; Zhang et al . , 2014a ) and CNS axon regeneration ( Dill et al . , 2008 ) , although contradictory result has also been reported ( Gobrecht et al . , 2014 ) . To definitively determine the role of GSK3β in CNS axon regeneration , we used AAV-GSK3β-S9A to express a GSK3β mutant that cannot be phosphorylated and inhibited by AKT . GSK3β-S9A significantly blocked AKT3-induced axon regeneration ( Figure 7A , B ) . Combining Rptor deletion with GSK3β-S9A overexpression blocked both mTORC1 activation and GSK3β inhibition , which almost totally prevented AKT3-induced axon regeneration ( Figure 7A , B ) . These results suggest that phosphorylation and inhibition of GSK3β and activation of mTORC1 are two parallel signal pathways downstream of AKT3 that act together to influence axon regeneration . 10 . 7554/eLife . 14908 . 014Figure 7 . GSK3β phosphorylation and inhibition by AKT is necessary and sufficient for axon regeneration . ( A , C , E ) Confocal images of ON longitudinal sections showing regenerating fibers labeled with CTB 2 weeks after ON crush . Scale bar , 100 µm . *crush site . ( B , D , F ) Quantification of regenerating fibers at different distances distal to the lesion site . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=8–15 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 01410 . 7554/eLife . 14908 . 015Figure 7—figure supplement 1 . AKT3 expression and RGC survival after GSK3β manipulation . ( A ) Western blots of retina lysates from three biological replicates showing expression level of HA-AKT3 . ( B ) Quantification of Western blots . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . Data are presented as means ± s . e . m , n=3 . ( C ) Confocal images of flat-mounted retinas showing individual labeling of pGSK3β , pS6 and Tuj1 , 2 weeks after ON crush . Scale bar , 20 µm . ( D ) Quantification of surviving RGCs , represented as percentage of Tuj1 positive RGCs in the injured eye , compared to the intact contralateral eye . Data are presented as means ± s . e . m , n=8–10 . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 015 To determine whether GSK3β inhibition is sufficient for axon regeneration , we injected AAV-Cre into the eyes of Gsk3b floxed mice to delete Gsk3b specifically in RGCs . As Figure 7C , D shows , Gsk3b KO alone induced a small but significant amount of axon regeneration , in dramatic contrast to the deletion of Gsk3a , which yielded no axon regeneration . Deletion of both Gsk3a and Gsk3b did not have an additive effect on axon regeneration ( Figure 4—figure supplement 1 ) . Since both AKT effectors , mTORC1 and GSK3β , are necessary for axon regeneration and either activation of mTORC1 effector S6K1 ( Yang et al . , 2014 ) or Gsk3b deletion is sufficient to promote axon regeneration , we next tested whether these two pathways have an additive effect . This experiment showed enhancement of axon regeneration when S6K1 constitutively active mutant ( S6K1-CA ) was expressed in Gsk3b KO mice ( Figure 7C , D ) , consistent with the idea that mTORC1 and GSK3β act in parallel , synergistic pathways downstream of AKT for axon regeneration . Since blocking AKT-S473 phosphorylation increased GSK3β-S9 phosphorylation/inactivation ( Figure 5B , G ) , we speculated that WT AKT alone only partially inhibits GSK3β and that AKT3 activation together with Gsk3b deletion would further increase axon regeneration . Indeed , regenerating axons were more numerous and longer in Gsk3b KO mice injected with AAV-AKT3 ( Figure 7E , F ) , indicating incomplete inhibition of GSK3β by WT AKT3 and/or AKT-independent GSK3β activity , which acted additively with AKT-dependent GSK3β activity on axon regeneration . Consistent with this inference , AKT-independent GSK3β inactivation has been recognized in peripheral axon regeneration ( Zhang et al . , 2014a ) . Although AKT3 levels were variable after Gsk3b and/or Rptor manipulation ( Figure 7—figure supplement 1A , B ) , RGC survival induced by AKT3 was unchanged ( Figure 7—figure supplement 1C , D ) , indicating sufficient AKT3 expression in retina for neuroprotection . In fact , even though expression of AKT3 was also lower in Gsk3b KO mice , axon regeneration was enhanced ( Figure 7E , F ) . Thus , the changes in axon regeneration depend on signaling cross-talk and alterations of pS6 and pGSK3β-S9 ( Figure 7—figure supplement 1C ) , but not on the absolute levels of AKT3 .
Although AKT has been linked with axon growth in vitro ( Shi et al . , 2003; Jiang et al . , 2005; Markus et al . , 2002 ) and axon regeneration in vivo ( Song et al . , 2012b; Namikawa et al . , 2000; Kim et al . , 2011a ) , there is surprisingly little information about the specificity of AKT isoforms and the role of different phosphorylation sites of AKT ( T308 and S473 ) in axon regeneration . In contrast to AKT1 and AKT2 , which are widely expressed , AKT3 is the predominant isoform in brain ( Easton et al . , 2005 ) . Deletion of AKT1 reduces whole body size and deletion of AKT2 results in diabetes-like syndrome ( Cho et al . , 2001a; Cho et al . , 2001b ) . Deletion only of AKT3 specifically reduces brain size ( Easton et al . , 2005 ) , indicating a specific role of AKT3 in regulation of CNS growth . Interestingly , AKT2 and AKT3 , but not AKT1 , regulate survival and growth of cultured hippocampal neurons ( Diez et al . , 2012 ) . Our analysis of the expression levels of the three AKT isoforms in RGCs and their distinct roles in axon regeneration provides additional in vivo evidence of the unique properties of AKT3 in CNS: AKT3 activation promotes significantly greater RGC survival and ON regeneration than AKT1 , presumably through its unique ability to activate mTORC1 ( higher pS6 ) in retina . This is also true in brain as AKT3 KO , but not AKT1 deletion , decreases pS6 significantly ( Easton et al . , 2005 ) . AKT3 may also selectively activate unknown , neuronal-specific signaling molecules . The partial functional redundancy of AKT isoforms , however , makes identification of these molecules difficult ( Dummler and Hemmings , 2007 ) . We found that mTORC1 inhibition ( Rptor or Mtor deletion , S6K1-DN or 4E-BP1-4A over-expression ) decreased AKT3 expression ( Figure 6C , D ) . Decreased AKT3 by itself may not account for reduced axon regeneration , however , because the remaining amounts enabled RGCs to survive ( Figure 6E , F ) . Although we cannot exclude the possibility that the decrease in AKT3 contributed to the reduced axon regeneration , we consider the significant decreases in mTORC1 signaling and protein synthesis to be the main reason for the inhibition of axon regeneration . This notion receives additional support from our observation of other conditions in which axon regeneration was enhanced despite downregulation of AKT3 ( Figure 4C , D and Figure 7E , F ) , which suggests that high levels are not required for axon regeneration . Based on our previous demonstration of the necessary role of 4E-BP and S6K1 ( Yang et al . , 2014 ) , we therefore conclude that mTORC1 is critically important for axon regeneration and that the effects on axon regeneration are due to the signaling alterations , rather than to the changed levels of AKT expression per se . Our observation of enhanced axon regeneration by AKT3-S472A mutant and Rictor KO suggests that pS473 either inhibits phosphorylation of T308 directly or allows AKT to activate distinct substrates that are different from those of pT308 , which are inhibitory for axon regeneration . Our finding that pS473 enhances , rather than inhibits , phosphorylation of T308 , implies that different substrates or differentially regulated substrates by pT308 and pS473 , must contribute to the difference in axon regeneration . The inhibitory effect of mTORC2 and pAKT-S473 on GSK3β phosphorylation is striking and very interesting . The significantly increased pGSK3β-S9 after blocking mTORC2 and S473 phosphorylation suggests that GSK3β is one of the AKT effectors that are differentially regulated by pAKT-T308 and pAKT-S473 . Although previous in vitro study showed unchanged GSK3β phosphorylation after Rictor KO and reduced S473 phosphorylation ( Guertin et al . , 2006 ) , the liver or muscle-specific Rictor deletion increases GSK3β phosphorylation ( Kumar et al . , 2008; Yuan et al . , 2012 ) . The results of our studies using GSK3β-S9A mutant and Gsk3b KO mice provide additional evidence of the inhibitory role of GSK3β in axon regeneration and definitively resolve the contradictory information in the literature ( Christie et al . , 2010; Abe et al . , 2010; Saijilafu et al . , 2013; Dill et al . , 2008; Gobrecht et al . , 2014 ) . In vitrokinase assay will be necessary to definitively demonstrate the roles of pAKT-T308 and pAKT-S473 in GSK3β-S9 phosphorylation , We have shown that pAKT-T308 and pAKT-S473 regulate the inhibition and phosphorylation of retinal GSK3β in opposite directions . This process acts in parallel with another AKT downstream effector , mTORC1 , to regulate CNS axon regeneration ( Figure 8 ) . Interestingly , downstream effectors that are specific to mTORC2-pAKT-S473 but not to pAKT-T308 have been proposed ( Guertin et al . , 2006; Yang et al . , 2006; Jacinto et al . , 2004; 2006; 2011 ) , consistent with the notion that pAKT-S473 regulates β-cell proliferation whereas pAKT-T308 controls β-cell size ( Gu et al . , 2011; Hashimoto et al . , 2006 ) . The balance between pro-regeneration pT308 and anti-regeneration pS473 forms of AKT appears to be critical . Our observation that Rictor deletion induces more potent axon regeneration with a smaller decrease in pS473 and pT308 than AKT3-S472A mutant ( Figure 5 ) may indicate a better therapeutic strategy for manipulating AKT signaling to promote axon regeneration . It is also worth noting that , because Rictor deletion did not totally block AKT-S473 phosphorylation , other kinases in addition to mTORC2 may contribute in retina . Moreover , because other kinases have been suggested to phosphorylate GSK3β-S9 in addition to AKT ( Tsujio et al . , 2000; Fang et al . , 2000; Armstrong et al . , 2001 ) , it will be important to investigate whether these kinases are also regulated by mTORC2 in a similar way as AKT . Since mTORC2 and pAKT-S473 are necessary for PTEN deletion-induced tissue overgrowth in drosophila eyes ( Hietakangas and Cohen , 2007 ) and development of prostate cancer in mouse ( Guertin et al . , 2009 ) , targeting/blocking mTORC2 may allow us to boost PTEN/AKT’s regeneration-promoting effect while at the same time minimizing its deleterious tumorigenic effect . 10 . 7554/eLife . 14908 . 016Figure 8 . A schematic illustration depicting the interplay between AKT , mTORC1/2 and GSK3β in CNS axon regeneration . The predominant isoform of AKT in brain and retina , AKT3 , generates the most robust axon regeneration . Its function in axon regeneration is positively regulated by the PI3K/PDK1 pathway through phosphorylation of T308 ( T305 in AKT3 ) , and negatively regulated by the PI3K/mTORC2 pathway through phosphorylation of S473 ( S472 in AKT3 ) , through at least partially regulation of GSK3β phosphorylation and inhibition . Both AKT downstream effectors , activation of mTORC1 and phosphorylation/inhibition of GSK3β , synergistically promote axon regeneration; inhibition of GSK3β alone is also sufficient for axon regeneration . The question mark represents other upstream regulators of GSK3β that may also promote axon regeneration . Green color-coated molecules are pro-axon regeneration and dark color-coated molecules are anti-axon regeneration . DOI: http://dx . doi . org/10 . 7554/eLife . 14908 . 016 Apparently , neuronal survival is a prerequisite for axon regeneration . But we and others did not find that increased neuron survival was invariably linked to proportionately greater axon regeneration ( Benowitz et al . , 2015 ) . This is consistent with findings in other systems . For example , most corticospinal neurons exhibit long term survival after transection in the spinal cord ( Nielson et al . , 2010; 2011 ) , but they fail to regenerate axons ( Schwab and Bartholdi , 1996; Goldberg et al . , 2002b; Fitch and Silver , 2008 ) . The 20% of RGCs that normally survive ON crush in mice can be increased significantly by inhibition of apoptosis , deleting tumor suppressor genes or by manipulating ER stress pathways , but these manipulations do not necessarily induce ON regeneration ( Park et al . , 2008; Hu et al . , 2012; Goldberg et al . , 2002a ) . This observation indicates that axon regeneration requires neuronal intrinsic growth stimulators that are distinct from neuronal surviving factors . Thus we consistently found that , although manipulation of mTOR complexes and GSK3β significantly changed axon regeneration , RGC survival induced by AKT remained the same . We could not exclude the possibility that changing RGC survival contributed to a change in axon regeneration , but no convincing evidence proves a direct causative relationship between these two events . The available evidence , therefore , supports the idea that the intrinsic signaling events after AKT activation and the involvement of its upstream or downstream signaling effectors are directly related to intrinsic growth control of neurons , and that these signaling pathways are distinct from or overlap only partially , signaling necessary for survival . There are far fewer regenerating axons than surviving RGCs , however , suggesting that only a small percentage of RGCs are regenerating and different subtypes of RGCs have different regeneration abilities ( Duan et al . , 2015 ) . In-depth understanding of the mechanisms of this difference will be required to maximize RGC axon regeneration . Increasing evidence has demonstrated the importance of localized protein synthesis in peripheral axon regeneration ( Willis and Twiss , 2006; Jung et al . , 2012; Perry and Fainzilber , 2014 ) . Intra-axonal translation has recently been demonstrated in mature mouse hippocampus ( Baleriola et al . , 2014 ) and , more intriguingly , certain mRNA species and additional components of translation machinery , including pS6 and 4E-BP1 , have been detected in regenerating axons in rat spinal cord ( Kalinski et al . , 2015 ) . Since we also observed that regeneration-promoting WT AKTs and AKT-S473A mutant were localized in ON whereas non-regeneration AKT mutants were excluded from ON , it will be very intriguing to investigate the significance of axonal AKT activation in CNS axon regeneration , especially its effect on axonal protein synthesis . In summary , our genetic manipulations in RGCs have established that the activation of mTORC1 and inhibition of GSK3β are two critical pathways downstream of AKT that act in parallel and synergistically to promote CNS axon regeneration ( Figure 8 ) . The opposite effects of mTORC1 and mTORC2 on axon regeneration suggest that a balancing mechanism exists downstream of the critical growth-promoting signal PI3K and that AKT integrates both positive and negative signals through phosphorylation of T308 and S473 and their specific roles in downstream effectors mTORC1 and GSK3β to control CNS axon regeneration ( Figure 8 ) . Interestingly , mTORC1/S6K also functions as feedback inhibition of PI3K signaling ( Laplante and Sabatini , 2012 ) , which also balances AKT activation and axon regeneration ( Yang et al . , 2014 ) . Thus , it is reasonable to expect that the increased understanding of the complicated cross-regulation and feedback-control mechanisms presented by our studies will eventually lead to safe and effective therapeutic strategies for CNS injury .
Rictorflox/flox , Rptorflox/flox , Mtorflox/flox and RiboTagmice with C57BL/6 background and C57BL/6 WT mice were purchased from Jackson Laboratories ( Bar Harbor , Maine ) . Gsk3bflox/flox and Gsk3aflox/flox mice with C57BL/6 background were originally developed by Dr . Jim Woodgett ( Doble et al . , 2007; Patel et al . , 2008 ) and were acquired from Dr . Thomas Force . We crossed them to generate Gsk3a/bflox/flox mice . All experimental procedures were performed in compliance with animal protocols approved by the IACUC at Temple University School of Medicine . For all surgical and treatment comparisons , control and treatment groups were prepared together in single cohorts , and the experiment repeated at least twice . pcDNA3-Myr-HA-AKT1 ( #9008 ) , pcDNA3-Myr-HA-AKT2 ( #9016 ) , pcDNA3-Myr-HA-AKT3 ( #9017 ) and pcDNA3-HA-GSK3β-S9A ( #14754 ) were obtained from Addgene . We used overlap PCR method to produce AKT3 mutants AKT3-T305A , AKT3-S472A , AKT3-T305A/S472A ( AKT3-AA ) , AKT3-K177M ( AKT3-KD ) and subcloned the WT AKT1-3 and AKT3 mutants into an AAV backbone that contained CBA promoter with Myr-3xHA tag at the N-terminus to get AAV-Myr-3HA-AKT1-3 and AKT3 mutants . We generated AAV-3HA-GSK3β-S9A similarly and AAV-Cre , AAV-S6K1-DN , AAV-S6K1-CA and AAV-4E-BP1-4A were described before ( Yang et al . , 2014 ) . The detailed procedure has been described previously ( Hu et al . , 2012; Yang et al . , 2014 ) . Briefly , AAV plasmids containing the transgenes were co-transfected with pAAV2 ( pACG2 ) -RC triple mutant ( Y444 , 500 , 730F ) ( Petrs-Silva et al . , 2011; Wang et al . , 2014; Zhang et al . , 2014b ) and the pHelper plasmid ( Stratagene , La Jolla , California ) into HEK293T cells . 72 hr after transfection , the cells were lysed to release the viral particles , which were precipitated by 40% polyethylene glycol and purified by cesium chloride density gradient centrifugation . The fractions with refractive index from 1 . 370 to 1 . 374 were taken out for dialysis in MWCO 7000 Slide-A –LYZER cassette ( Pierce , Thermo Fisher Scientific , Waltham , Massachusetts ) overnight at 4°C . The AAV titers that we used for this study were in the range of 1 . 5–2 . 5 x 1012 genome copy ( GC ) /ml determined by real-time PCR . Mice were anesthetized by xylazine and ketamine based on their body weight ( 0 . 01 mg xylazine/g+0 . 08 mg ketamine/g ) . For each AAV intravitreal injection , a micropipette was inserted into the peripheral retina of 3 week-old mice just behind the ora serrata , and advanced into the vitreous chamber so as to avoid damage to the lens . Approximately 2 µl of the vitreous was removed before injection of 2 µl AAV into the vitreous chamber . ON crush was performed 2 weeks following AAV injection: the ON was exposed intraorbitally and crushed with a jeweler’s forceps ( Dumont #5; Fine Science Tools , Forster City , California ) for 5 s approximately 0 . 5 mm behind the eyeball . Care was taken not to damage the underlying ophthalmic artery . Eye ointment containing neomycin ( Akorn , Somerset , New Jersey ) was applied to protect the cornea after surgery . 2 µl of cholera toxin β subunit ( CTB ) conjugated with fluorescence Alexa- 488 ( 2 μg/μl , Invitrogen ) was injected into the vitreous chamber 2 days before sacrificing the animals to label the regenerating axons in the optic nerve . Animals were sacrificed by CO2 and fixed by perfusion with 4% paraformaldehyde in cold PBS . Eyes with the nerve segment still attached were dissected out and post-fixed in the same fixative for another 2 hr at room temperature . Tissues were cryoprotected through increasing concentrations of sucrose ( 15%-30% ) and optimal cutting temperature compound ( OCT ) ( Tissue Tek , Sakura Finetek , Torrance , California ) . They were then snap-frozen in dry ice and serial longitudinal cross-sections ( 8 µm ) were cut and stored at –80°C until processed . Retinas were dissected out from 4% PFA fixed eyes and washed extensively in PBS before blocking in staining buffer ( 10% normal goat serum and 2% Triton X-100 in PBS ) for 30 min . Mouse or rabbit neuronal class ß-III tubulin ( clone Tuj1 , 1:500; Covance , Conshohocken , Pennsylvania ) , rat HA ( clone 3F10 , 1:200 , Roche , Basel , Switzerland ) , phospho-S6-Ser240/244 antibody ( 1:200 , #5364 , Cell Signaling , Danvers , Massachusetts ) , phospho-AKT-Ser473 ( 1:200; #4058 , Cell Signaling ) and phospho-GSK-3β ( Ser9 ) ( 1:100; #9323 , Cell Signaling ) were diluted in the same staining buffer . Floating retinas were incubated with primary antibodies overnight at 4°C and washed three times for 30 min each with PBS . Secondary antibodies ( Cy2 , Cy3 or Cy5-conjugated ) were then applied ( 1:200; Jackson ImmunoResearch , West Grove , Pennsylvania ) and incubated for 1 hr at room temperature . Retinas were again washed three times for 30 min each with PBS before a cover slip was attached with Fluoromount-G ( Southernbiotech , Birmingham , Alabama ) . Retinas were dissected out from ice-cold PBS perfused eyes and homogenized and lysed in RIPA buffer ( 50 mM Tris HCl pH8 . 0 , 150 mM NaCl , 1% NP-40 , 0 . 5% sodium deoxycholate , 0 . 1% SDS , 5 mM sodium pyrophosphate , 10 mM sodium fluoride , 1 mM sodium orthovanadate , protease inhibitors cocktail ) on ice for 30 min . The homogenates were centrifuged at 12 , 000 g for 20 min; supernatants were subjected to electrophoresis with 10% SDS-PAGE . After gel-transference , nitrocellulose membranes were blocked with Odyssey blocking buffer ( LI-COR , Lincoln , Nebraska ) for 1 hr before incubation with primary antibody at 4°C overnight . After washing three times for 10 min with PBS , the membranes were incubated with secondary antibodies ( IRDye 680RD goat-anti-mouse IgG or IRDye 800CW goat-anti-rabbit IgG , LI-COR ) at room temperature for 1 hr . The membranes were then washed three times for 10 min with PBS and scanned with Odyssey CLx ( LI-COR ) . The images were analyzed with Image Studio ( LI-COR ) . The primary antibodies used were mouse HA ( clone 16B12 , 1:2000 , Covance ) , mouse Anti-β-Actin ( clone AC-15 , 1:2000 , Sigma , St Louis , Missouri ) , and antibodies from Cell Signaling: rabbit phospho-AKT Thr308 ( 1:500 , #2965 ) , rabbit phospho-AKT Ser473 ( 1:1000 , #4058 ) , rabbit phospho-AKT2 Ser474 ( 1:1000 , #8599 ) , rabbit phospho-S6 Ribosomal Protein Ser240/244 ( 1:1000 , #5364 ) , rabbit phospho-GSK-3β Ser9 ( 1:500; #9323 ) , mouse S6 Ribosomal Protein ( 1:500 , #2317 ) and mouse GSK-3β ( 1:500 , #9832 ) . After adult 8-week old mice were perfused with ice-cold 4% PFA/PBS , eyeballs were dissected out and fixed in 4% PFA/PBS at 4°C overnight . The eyeballs were dehydrated with increasing concentrations of sucrose solution ( 15%-30% ) overnight before embedding in OCT on dry ice . Serial cross sections ( 12 µm ) were cut with a Leica cryostat and collected on Superfrost Plus Slides . The sections were washed twice for 10 min in DEPC-treated PBS and permeabilized twice in 0 . 1% Tween/PBS for 10 min . After blocking at 50°C for 1 hr with hybridization buffer ( 50% formamide , 5 x SSC , 100 μg/ml Torula Yeast RNA , 100 μg/ml Wheat Germ tRNA , 50 μg/ml heparin and 0 . 1% Tween in DEPC H2O ) , the sections were hybridized with 2 µg biotin-labeled antisense probes at 50°C overnight . The sections were washed three times at 55°C for 10 min with hybridization buffer , 0 . 1% Tween/PBS , and then blocked in PBS blocking buffer containing 0 . 1% BSA and 0 . 2% TritonX-100 . The hybridized probes were detected by Streptavidin-AP-conjugate ( Roche ) , and revealed by chromogenic substrate NBT/BCIP ( Roche ) . Mouse AKT1 , 2 , 3 probe sequences were from Allen Brain Atlas ( http://mouse . brain-map . org/ ) . For RGC counting , whole-mount retinas were immunostained with the Tuj1 antibody , and 6–9 fields were randomly sampled from peripheral regions of each retina . The percentage of RGC survival was calculated as the ratio of surviving RGC numbers in injured eyes compared to contralateral uninjured eyes . For axon counting , the number of CTB- labeled axons was quantified as described previously ( Leon et al . , 2000; Park et al . , 2008; Yang et al . , 2014 ) . Briefly , we counted the fibers that crossed perpendicular lines drawn on the ON sections distal to the crush site in increments of 250 µm till 1000 µm , then every 500 µm till no fibers were visible ( Figure 2—figure supplement 1 ) . The width of the nerve ( R ) was measured at the point ( d ) at which the counts were taken and used together with the thickness of the section ( t = 8 μm ) to calculate the number of axons per µm2 area of the nerve . The formula used to calculate is ∑ad= πr2 * ( axon number ) / ( R*t ) . The total number of axons per section was then averaged over 3 sections per animal . All CTB signals that were in the range of intensity that was set from lowest intensity to the maximum intensity after background subtraction were counted as individual fibers by Nikon NIS Element R4 software . The investigators who counted the cells or axons were blinded to the treatment of the samples . Three groups of RiboTag mice ( 8–10 mice/group ) were intravitreally injected with AAV2-Cre four weeks before sacrifice and removal of retinas . Ribo-IP was performed according to the published protocol ( Sanz et al . , 2009 ) . Briefly , for each replicate , 10–16 pooled retinas were homogenized and lysed in 1 ml homogenization buffer ( 50 mM Tris pH7 . 4 , 100 mM KCl , 12 mM MgCl , 1% NP-40 , 1 mM DTT , 100 μg/ml cyclohexamide , 1 mg/ml heparin , Protease Inhibitor Cocktail ( Sigma ) and RNasin Ribonuclease Inhibitor ( Promega Corp . , Madision , Wisconsin ) in RNase-free H2O ) on ice for 10 min and centrifuged at 4°C for 10 min at 12 , 000 g . The supernatant was collected and incubated at 4°C for 4 hr with 10 µg mouse HA antibody , after which 400 μl Dynabeads Protein G ( Life Technologies , Frederick , Maryland ) were added and incubation continued at 4°C overnight . Dynabeads were washed three times for 10 min with high salt buffer ( 50 mM Tris pH7 . 4 , 300 mM KCl , 12 mM MgCl , 1% NP-40 , 1 mM DTT and 100 μg/ml Cyclohexamide in RNase-free H2O ) before RNA extraction with RNeasy Micro Kit ( QIAGEN , Hilden , Germany ) . About 200ng total RNA generated from each group was used for RNA-seq , which was done at the University of Pennsylvania Next-Generation Sequencing Core . Briefly , Ribo-IP RNA samples from three biological replicates went through polyA selection before the generation of strand-specific RNA-seq libraries with Illumina TruSeq Stranded Total Kit and quality assessment with Agilent BioAnalyser and Kapa BioSystems Library Quant Kit . Pooled libraries that have been individually labeled were sequenced to 100 bp reads from one end of the insert using a HiSeq2000 sequencer . Each library acquired about 25 million reads and 93% of total reads aligned to unique genes in the mouse genome ( UCSC mm9 ) by RUM ( Grant et al . , 2011 ) . The 'raw' data ( reads per transcript ) were quantile normalized within groups with quantile normalization ( GCRMA ) to remove non-biological variability . Data are presented as means ± s . e . m and Student’s t-test was used for two-group comparisons and One-way ANOVA with Bonferroni’s post hoc test was used for multiple comparisons . | The central nervous system consists of the neurons that make up the brain and spinal cord . An important part of a neuron is the long , slender projection along which electrical signals travel , called the axon . In the central nervous system of mammals , damaged axons cannot regrow , which is why spinal injuries or optic nerve injuries can result in life-long neuronal deficits . Recent studies have found that activating a particular signaling pathway in central nervous system neurons causes their axons to regenerate . A key protein in this pathway is called AKT . Several signaling cascades are triggered by AKT to regulate cell survival and growth , but it was not known how the different branches of the AKT pathway are involved in axon regeneration . Miao , Yang et al . have now investigated AKT’s role in axon regeneration using a range of approaches to manipulate signaling in damaged mouse neurons . This revealed that a particular form of AKT ( called AKT3 ) causes damaged axons to regenerate to a greater extent than other forms of this protein . This response depends on two parallel pathways: one in which AKT3 activates a protein complex called mTORC1 , and one where AKT3 inhibits a protein called GSK3β . In addition , another protein complex called mTORC2 , which is closely related to mTORC1 , helps to inhibit the activity of AKT3 on GSK3β and hence inhibits axon regeneration . These findings reveal that a complex balancing mechanism , with AKT at its center , coordinates the many signals that regulate axon regeneration . Future studies into this system could ultimately help to develop new treatments for brain and spinal injuries . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | mTORC1 is necessary but mTORC2 and GSK3β are inhibitory for AKT3-induced axon regeneration in the central nervous system |
Synthetic strategies for optically controlling gene expression may enable the precise spatiotemporal control of genes in any combination of cells that cannot be targeted with specific promoters . We develop an improved genetic code expansion system in Caenorhabditis elegans and use it to create a photoactivatable Cre recombinase . We laser-activate Cre in single neurons within a bilaterally symmetric pair to selectively switch on expression of a loxP-controlled optogenetic channel in the targeted neuron . We use the system to dissect , in freely moving animals , the individual contributions of the mechanosensory neurons PLML/PLMR to the C . elegans touch response circuit , revealing distinct and synergistic roles for these neurons . We thus demonstrate how genetic code expansion and optical targeting can be combined to break the symmetry of neuron pairs and dissect behavioural outputs of individual neurons that cannot be genetically targeted .
Genetic code expansion offers the ability to introduce , during translation , functionalities into proteins beyond those offered by the set of canonical , naturally occurring amino acids . This is achieved by providing orthogonal translation components: an orthogonal aminoacyl-tRNA synthetase ( aaRS ) paired with a cognate orthogonal tRNA , and assignment of a codon to determine the site of incorporation of a non-canonical amino acid ( ncAA ) ( Chin , 2017 ) . The genetic code of several single and multicellular systems has been expanded to date ( Brown et al . , 2018a ) . Caenorhabditis elegans was the first multicellular organism in which genetic code expansion was established , and the incorporation of ncAA has been demonstrated in multiple C . elegans tissues ( Greiss and Chin , 2011; Parrish et al . , 2012 ) . Currently available orthogonal systems , however , often lack the efficiency required to fully harness the possibilities offered by genetic code expansion , especially in multicellular systems . A large variety of non-canonical amino acids are available for in vivo use , among them photocaged amino acids , which can be used to confer light-activated control over the function of proteins in a host organism ( Courtney and Deiters , 2018 ) . Sequence-specific DNA recombinases , such as Cre and FLP , have become transformative for controlling gene expression in many systems , including worms , flies , and vertebrates , by irreversibly activating or deactivating target gene expression . Indeed , the photo-induced reconstitution of split recombinases ( Schindler et al . , 2015; Kawano et al . , 2016 ) might be used to optically target recombinase activity to a single cell . However , the sensitivity of this technique can be hampered by high background levels of light-independent recombinase activity . Moreover , these systems are induced using blue light , rendering their use incompatible with most optogenetic and imaging applications , and requiring cells to be kept in the dark , complicating their use in transparent model systems . Recently , genetic code expansion methods were used to create photocaged versions of Cre in cultured mammalian cells and zebrafish embryos ( Luo et al . , 2016; Brown et al . , 2018b ) . In photocaged Cre , an orthogonal aaRS/tRNACUA pair is used to replace a catalytically critical tyrosine or lysine residue in the active site of Cre with a photocaged tyrosine or lysine residue , resulting in a catalytically inactive enzyme ( Luo et al . , 2016; Brown et al . , 2018b ) . The photocaging group can then be removed through a short exposure to 365 nm UV light , activating Cre recombinase , which in turn activates expression of its target genes . We set out to create an optimised photocaged Cre system that enables the optical activation of gene expression in single cells in an animal . The precise spatiotemporal control of gene expression offered by photocaged Cre would have a wide range of applications by allowing the expression of genes of interest in single cells or combinations of cells . Applied to control , for example , optogenetic channels or other neuromodulatory proteins , it could be a transformative tool in neurobiology . Defining the molecular and cellular basis of behaviour is a fundamental challenge in neuroscience . C . elegans contains 302 neurons , for which the physical connectivity has been determined ( White et al . , 1986 ) . This physical map provides a conceptual framework within which to design experiments to understand the signal processing properties of neural circuits that define behaviour . Central to this endeavour are methods for the spatiotemporal control of gene expression that enable precise perturbations to be effected in individual cells or combinations of cells and thereby advance our understanding of cellular contributions to circuit function . Two thirds of the neurons in C . elegans exist in bilaterally symmetric ( left and right ) pairs , and similar gross morphological symmetries are seen in the nervous systems of many other animals . A central question is whether and how this morphological symmetry is broken to produce functional asymmetry ( Troemel et al . , 1999; Wes and Bargmann , 2001; Suzuki et al . , 2008; Hobert et al . , 2002 ) . Tissue-specific promoters ( alone or in combination ) can be used to target groups of cells . However , this approach offers only limited precision , and in many cases appropriate promoters are unavailable; especially in scenarios when investigators aim to limit expression to a single cell or a defined subset of cells . This is particularly limiting in neuroscience studies where the goal is to couple neuronal activity of single neurons or a defined subset of neurons with functional output . Where specific promoters for targeting individual neurons are unavailable , laser ablation can be used to remove specific neurons , but this technology is not compatible with studying intact circuits . Optogenetic approaches using targeted illumination have been applied to stimulate groups or pairs of cells , and track resulting behavioural outputs in freely moving C . elegans ( Leifer et al . , 2011; Stirman et al . , 2011; Kocabas et al . , 2012 ) . However , without a means of precisely defining expression of optogenetic tools , this approach is technically very challenging as it requires specialised software and hardware tools for real-time tracking and targeting and does not lend itself to isolating single cells in neuronal pairs . Similar obstacles are encountered in other systems , where the solution often also requires technically challenging approaches utilising targeted illumination ( Chen et al . , 2018 ) . Here we describe the optimisation of genetic code expansion for ncAA incorporation and the optimisation of a photocaged Cre recombinase variant ( optPC-Cre ) in C . elegans . The combination of these advances allows us to optically control gene expression with unprecedented precision and over 95% efficiency in a population of animals ( compared to less than 1% efficiency before optimisation ) . By using a microscope-mounted 365 nm laser , we photo-activate Cre and thereby control gene expression in single cells ( Figure 1 ) . We name this approach Laser Targeted Activation of Cre-Lox recombination ( LaserTAC ) . LaserTAC is fast and within the technical capabilities of any lab currently able to perform laser ablation , allowing easy generation , in a single session , of dozens of animals with defined expression patterns for use in downstream experiments . We demonstrate the utility of LaserTAC by using it to target the expression of optogenetic channels to individual C . elegans touch sensory neurons within a left/right pair ( PLML and PLMR ) . These individual neurons cannot be targeted by other methods , and our approach allows us to study their contribution to behaviour for the first time . Our results reveal that the PLM neurons act in synergy to produce a robust touch response requiring input from both neurons . Furthermore , the individual neurons within this pair make asymmetric contributions to the touch response , suggesting distinct roles for PLMR and PLML in the habituation to repeated stimulation .
Efficient genetic code expansion depends on the ability of the orthogonal aminoacyl-tRNA synthetase to aminoacylate its cognate tRNACUA , which in turn delivers the ncAA to the ribosome for incorporation in response to the UAG amber stop codon ( Figure 1A ) . The charging of the ncAA onto the tRNACUA occurs mainly in the cytoplasm , and efficient charging is therefore dependent on cytoplasmic aminoacyl-tRNA synthetase availability . The orthogonal pyrrolysyl-tRNA synthetase ( PylRS ) /tRNA ( Pyl ) pair from archeal Methanosarcina species is the most versatile and widely used pair for genetic code expansion ( Wan et al . , 2014; Brown et al . , 2018a ) . The PylRS/tRNA ( Pyl ) pair is functional in C . elegans , and variants exist that recognise photocaged amino acids ( Greiss and Chin , 2011; Gautier et al . , 2010 ) . When expressed in eukaryotic cells , PylRS is localised predominantly to the nucleus due to a stretch of positively charged amino acids in the N-terminal domain , which is interpreted as a nuclear localisation sequence by the eukaryotic nuclear import machinery . Accordingly , addition of a rationally designed strong nuclear export sequence ( S-NES ) to the PylRS N-terminus was reported to significantly increase the efficiency of ncAA incorporation in mammalian cells by increasing the amount of cytoplasmic PylRS ( Nikić et al . , 2016 ) . To test this approach in C . elegans , we added the reported S-NES to the N-terminus of PCKRS , a PylRS variant optimised for incorporating photocaged lysine ( PCK ) ( Gautier et al . , 2010; Figure 2A ) . We used the ubiquitous C . elegans promoter sur-5p ( Yochem et al . , 1998 ) to drive S-NES::PCKRS expression and rpr-1p ( Parrish et al . , 2012 ) to drive tRNA ( Pyl ) CUA expression . To assay PCK incorporation efficiency , we used a dual-colour incorporation sensor; a ubiquitously expressed GFP::mCherry fusion , with the two fluorophore coding sequences separated by an amber stop codon ( Greiss and Chin , 2011; Figure 2B ) . For this reporter , translation in the absence of PCK leads to termination at the amber stop codon , resulting in production of only GFP . Conversely , in the presence of PCK , tRNA ( Pyl ) CUA is charged with the ncAA and acts as a nonsense suppressor , resulting in production of the full-length GFP::mCherry fusion protein , which includes a nuclear localisation sequence and an HA tag at its C-terminus . However , we did not detect any significant difference in incorporation between worms expressing unmodified PCKRS and S-NES::PCKRS . In fact , incorporation levels when using the S-NES::PCKRS construct appeared to be lower than with unmodified PCKRS ( constructs ‘no NES’ and ‘S-NES’; Figure 2C , Figure 2—figure supplement 1A ) . To test whether S-NES::PCKRS and unmodified PCKRS localise as predicted , we expressed S-NES::PCKRS::GFP and PCKRS::GFP fusion proteins . As expected , the S-NES::PCKRS::GFP fusion localised to the cytoplasm , while unmodified PCKRS::GFP was almost entirely nuclear , reflecting that the S-NES was able to efficiently shift PCKRS localisation from the nucleus to the cytoplasm ( constructs ‘no NES’ and ‘S-NES’ Figure 2D , Figure 2—figure supplement 1B ) . Therefore , we hypothesised that the S-NES tag itself might impinge on PCK incorporation efficiency . We next performed a mini-screen of three nuclear export sequences from human proteins: p120cts-NES , PKIα-NES , and Smad4-NES , which are demonstrated to act as NESs in C . elegans ( Yumerefendi et al . , 2015 ) . All tested NESs achieved nuclear export of the synthetase PCKRS ( Figure 2D , Figure 2—figure supplement 1B ) . However , the NES variants differentially affected incorporation efficiencies: when we co-expressed NES::PCKRS variants with tRNA ( Pyl ) and the GFP::mCherry incorporation reporter , we found that the p120cts-NES variant reduced incorporation efficiency below the level of unmodified PCKRS , whereas the PKIα-NES and the Smad4-NES robustly increased PCK incorporation efficiency ( Figure 2C , Figure 2—figure supplement 1A ) . We selected the two most efficient PCKRS variants for subsequent optimisation experiments . We then turned to the optimisation of tRNA ( Pyl ) . tRNA ( Pyl ) contains secondary structure elements not present in canonical mammalian tRNAs . These features reduce compatibility with the endogenous translational machinery , which limits ncAA incorporation in mammalian cells . However , ncAA incorporation efficiency in cultured mammalian cells can be improved by introducing mammalian tRNA elements into the archaeal tRNA ( Pyl ) or by using engineered mitochondrial tRNAs ( Serfling et al . , 2018 ) . To investigate whether the same approach might improve ncAA incorporation efficiency in other eukaryotes , we tested an improved tRNA variant in the C . elegans system . Specifically , we co-expressed the mitochondrial tRNA based variant C15 , previously validated in mammalian cells ( Serfling et al . , 2018 ) together with the GFP::mCherry incorporation reporter , and the two most efficient synthetase variants described above ( namely Smad4-NES::PCKRS or PKIα-NES::PCKRS ) . We found that the presence of the C15 variant significantly increased incorporation efficiency ( Figure 2C , E , F and G , Figure 2—figure supplement 1A , C ) . Compared to the unmodified PCKRS/tRNA ( Pyl ) pair , the Smad4-NES::PCKRS/tRNA ( C15 ) and PKIα-NES::PCKRS/tRNA ( C15 ) pairs improved incorporation efficiency by more than 50-fold; from <0 . 1% to 4 . 6% and 4 . 4% , respectively ( Figure 2F ) . We performed all further experiments using the Smad4-NES::PCKRS/tRNA ( C15 ) pair . We next applied our improved incorporation system to express photocaged Cre recombinase ( PC-Cre ) in the N2 wildtype C . elegans laboratory strain . Cre recombinase can be photocaged by replacing K201 , a lysine residue in the Cre active site , critical for enzyme activity , with PCK ( Gibb et al . , 2010; Luo et al . , 2016; Figure 3A and B ) . PC-Cre has previously been expressed in both mammalian cell culture and zebrafish embryos using transient transfection and direct mRNA/tRNA injections ( Luo et al . , 2016; Brown et al . , 2018b ) . To direct incorporation of PCK , we generated a PC-Cre construct where we replaced the lysine codon at position 201 in the Cre sequence with an amber stop codon ( TAG ) . The PC-Cre construct was expressed in an artificial operon with GFP , allowing us to visualise its expression . We used as a Cre recombinase target gene a fluorescently tagged channelrhodopsin ChR2::mKate2 , separated from its promoter by a transcriptional terminator flanked by loxP sites ( Figure 3C ) . We generated transgenic animals containing all genetic components: the optimised orthogonal synthetase/tRNA pair Smad4-NES::PCKRS/tRNA ( C15 ) , the PC-Cre construct , and the floxed Cre target ChR2::mKate2 . All protein coding components were driven by a glr-1p promoter , an orthologue of human GRIA1 ( glutamate ionotropic receptor AMPA type subunit 1 ) . This promoter is active in glutamatergic neurons , including command interneurons ( Maricq et al . , 1995; Figure 3—figure supplement 1A ) . We grew animals on PCK from the L1 larval stage and activated PC-Cre by UV illumination when they had reached the L4 larval stage ( after 2 days ) . 24 hr after activation , we saw strong expression of ChR2::mKate2 in neurons expressing PC-Cre . In contrast , we saw no expression of ChR2::mKate2 in animals which had undergone UV illumination without prior feeding on PCK , and in animals fed on PCK but which had not undergone UV illumination ( Figure 3—figure supplement 1B ) . Furthermore , as expected for a membrane channel , the red fluorescence of ChR2::mKate2 was localised at the cellular membrane . We observed red fluorescence only in cells expressing the glr-1p promoter , as evidenced by overlap with glr-1p-driven expression of GFP . Using the optimised Smad4-NES::PCKRS/tRNA ( C15 ) expression system , we observed expression of the target locus , indicative of PC-Cre photoactivation , in 63% of animals 24 hr after uncaging . We saw no red fluorescence without UV treatment . The 63% activation rate was a vast improvement relative to the 1% we observed with unmodified PCKRS and tRNA ( Pyl ) ( Figure 3D ) . To further improve the photoactivation method , we next turned to PC-Cre optimisation . When using amber stop codons for ncAA incorporation , the tRNACUA competes with endogenous release factors . This competition results in a mixture of full-length polypeptides , with ncAA incorporated , as well as truncated polypeptides due to translational termination at the amber stop codon . As assessed with the GFP::mCherry incorporation reporter , we show that the percentage of full-length protein is between 4% and 5% when using the improved Smad4-NES::PCKRS/tRNA ( C15 ) incorporation system ( Figure 2F ) . It is likely that , similar to the fluorescent reporter , the majority of translation events of the PC-Cre mRNA will also terminate at the internal amber stop codon , even in the presence of PCK . A Cre protein truncated at the amber stop codon in position 201 is missing the majority of its active site , but the parts of Cre responsible for the protein-protein interaction required for active Cre tetramer formation are still present , as are large parts of the DNA binding interface ( Guo et al . , 1997 ) . We initially based PC-Cre on previously reported Cre constructs , which contain an N-terminal SV40 NLS in addition to the internal NLS native to Cre ( Luo et al . , 2016; Brown et al . , 2018b; Le et al . , 1999; Macosko et al . , 2009 ) , thus ensuring localisation of the enzyme to the nucleus . We cannot exclude the possibility that truncated Cre , which locates to the nucleus due to the NLSs upstream of position 201 , may interfere with tetramer formation or DNA binding . We aimed to utilise the C . elegans nuclear import machinery to enrich for full-length product in the nucleus . Since Cre acts on nuclear DNA , we reasoned that we could boost PC-Cre activity by lowering the fraction of truncated PC-Cre while increasing the fraction of full-length PC-Cre present in the nucleus . To this end , we removed both NLSs upstream of position 201 by introducing a R119A mutation disabling the internal NLS sequence ( Le et al . , 1999 ) and by removing the N-terminal SV40 NLS . To restore nuclear import for full-length PC-Cre only , we attached the strong C . elegans EGL-13 NLS ( Lyssenko et al . , 2007 ) to the Cre C-terminus to create optimised PC-Cre ( optPC-Cre ) ( Figure 3E ) . When we compared optPC-Cre to PC-Cre , we indeed found that target gene activation was significantly improved: from 63% for PC-Cre to 97% for optPC-Cre ( Figure 3D ) . We saw no expression of the ChR2::mKate2 target gene without PCK or UV activation ( Figure 3D and F ) . We next performed behavioural assays to assess whether our system allowed optical control of neurons . After activation of optPC-Cre , we waited 24 hr to achieve sufficient expression of ChR2::mKate2 , which we confirmed by visual inspection under a dissecting fluorescence microscope . When we exposed worms to 470 nm blue light to activate ChR2 , we observed clear reversals in worms expressing ChR2::mKate2 in the presence of the ChR2 cofactor all-trans-retinal ( ATR ) . In animals expressing ChR2::mKate2 in the absence of ATR , we did not observe such reactions ( Figure 3G , Figure 3—figure supplement 1C , Video 1 and Video 2 ) . The induced behaviour concurs with previous studies showing that optogenetic activation of neurons expressing ChR2 from the glr-1p promoter induces backward movement ( Schultheis et al . , 2011 ) . When we compared strains expressing PC-Cre and optPC-Cre , respectively , we saw the strongest and most robust responses in the strains where optPC-Cre was used to control expression of ChR2::mKate2 , followed by strains using non-optimised PC-Cre , but using the efficient Smad4-NES::PCKRS/tRNA ( C15 ) incorporation system . Animals expressing unmodified PCKRS and non-optimised Cre showed almost no response to optogenetic activation ( Figure 3G , Figure 3—figure supplement 1C ) . Taken together , we conclude that optPC-Cre is an effective tool with which to control optogenetic channel expression for the purpose of optical control of neuronal activity . The use of light to control activity of Cre recombinase should allow precise spatial control of Cre-dependent DNA recombination . We tested the precision of uncaging by using a microscope mounted 365 nm laser to target individual cells in the touch response circuitry . C . elegans has six mechanosensory neurons for the perception of soft touch: AVM , ALML , and ALMR , which are located in the anterior of the worm , and PVM , PLML , and PLMR , which are located in the posterior ( Figure 4A ) . Activation of the touch receptor neurons results in avoidance behaviour . Worms respond with backward movement to an anterior touch , and with forward movement to a posterior touch ( Chalfie et al . , 1985 ) . Previous studies using targeted ChR2 illumination in individual animals likewise showed that optogenetic activation of the anterior neurons AVM and ALM results in reversals , while activation of the posterior neurons PVM and PLM results in forward movement . Concurrent activation of all six neurons results in reversals in the majority of cases . Repeated activation leads to a reduced response due to habituation ( Leifer et al . , 2011; Stirman et al . , 2011; Schild and Glauser , 2015 ) . We aimed to investigate the posterior PLM neurons by using LaserTAC to selectively express the optogenetic channel Chrimson ( Schild and Glauser , 2015 ) in both PLM neurons , or in PLML and PLMR individually . To this end , we generated strains with Smad-4-NES::PCKRS and optPC-Cre expression driven by the mec-7p promoter , which is active in all six mechanosensory neurons ( Mitani et al . , 1993 ) . To express Chrimson , we used a Chrimson::mKate2 fusion gene separated from the pan-neuronal maco-1p promoter by a transcriptional terminator flanked by loxP sites ( Figure 4—figure supplement 1A ) . We chose the panneuronal maco-1p promoter , an orthologue of human MACO1 ( macoilin 1 ) ( Arellano-Carbajal et al . , 2011 ) , because in contrast to mec-7p , it shows increasing expression from the L4 larval to the adult stage , the age when we aimed to induce Chrimson::mKate2 expression ( Figure 4—figure supplement 1B ) . To activate expression of the Chrimson channel in the PLM neurons , we mounted age-synchronised L4 stage animals , grown on PCK from L1 , on a microscope slide and used a Micropoint Galvo system to deliver a short 2–4 s 365 nm pulse to individual cells , targeting the nucleus . After uncaging , we transferred the worms to an NGM agar plate without PCK for recovery and to allow for induced Chrimson::mKate2 expression ( Figure 4B ) . After 12–24 hr , we observed Chrimson::mKate2 expression that was only present in laser-targeted cells . We were able to selectively activate optPC-Cre in both PLM neurons , as well as in PLML or PLMR alone ( Figure 4C–E ) . We confirmed that the presence of the PCK incorporation machinery and optPC-Cre did not affect the function of the touch sensory neurons by assaying the animals response to soft touch ( Chalfie et al . , 2014; Figure 4—figure supplement 1C , D ) . To determine the behavioural response to PLM stimulation , we prepared animals expressing Chrimson::mKate2 in either both PLM neurons , or in PLML or PLMR alone . After optPC-Cre activation , we grew the worms on plates supplemented with ATR for 24 hr to allow expression of the optogenetic channel and confirmed the desired expression pattern under a fluorescence dissecting microscope . We then activated Chrimson::mKate2 expressing neurons by illumination for 1 s every 30 s with 617 nm at the maximum power setting of 74 mW/cm2 . During the assay , the animals were moving freely on the plate and we illuminated the entire plate . We found that illumination of animals expressing Chrimson::mKate2 in both PLM neurons ( ‘2PLM’ ) triggered a robust response with the worms initiating forward movement ( Figure 5A , B , Figure 5—figure supplement 1 , Video 3 ) . Conversely , mock-treated animals ( ‘0PLM’ ) grown on PCK and ATR but without activation of optPC-Cre showed no response ( Figure 5—figure supplement 1 , Video 4 ) . Animals expressing Chrimson::mKate2 only in the PLML or the PLMR neurons respectively also showed a clear reaction to stimulation ( Figure 5A , B , Figure 5—figure supplement 1 , Video 5 and Video 6 ) . For all three expression patterns ( 2PLM , PLML , PLMR ) , we observed a gradual reduction in the response with increasing number of stimulations , both in the fraction of animals responding and in the speed upon stimulation , consistent with habituation to the stimulus ( Stirman et al . , 2011 ) . We found that the 2PLM animals responded stronger to stimulation than animals expressing Chrimson::mKate2 in either PLML or PLMR alone , both in regard to the fraction of animals responding , and by velocity ( Figure 5A , B ) . We then proceeded to more closely investigate the responses to activation of PLML and PLMR individually . The members of the PLM neuron pair , PLML and PLMR , show a striking asymmetry in their connections within the worm connectome ( Figure 5C; Chalfie et al . , 1985 ) . PLMR forms chemical synapses connecting it to 10 downstream neurons including the AVA and AVD interneuron pairs , which are involved in driving backward locomotion . In addition to chemical synapses , PLMR also forms gap junctions to PVCR , PHCR , and LUAR . PLML , on the other hand , is connected to the touch response circuit exclusively through gap junctions , formed with the interneurons PVCL , PHCL , and LUAL . In fact , PLML does not form any chemical synapses , with the exception of a single connection to HSNL , a neuron involved in the control of egg laying . Asymmetric wiring of PLML and PLMR has been proposed to influence the C . elegans tap response , a withdrawal reflex that integrates information from the anterior and posterior mechanosensory neurons . In worms with ablation of both PLM neurons and the single PLMR neuron , a tap stimulus increased backward movement . In contrast , ablating PLML had no effect ( Wicks and Rankin , 1995 ) . As we saw no difference between PLML and PLMR in the response to strong stimulation ( Figure 5A , B ) , we decided to test the animals' responses to weaker stimulation of the PLM neurons . For this , we reduced the duration of the 617 nm light pulse from 1 s to 0 . 1 s and delivered varying low intensities , starting at 3 mW/cm2 , the lowest setting that was possible with our setup . While the 0 . 1 s , 3 mW/cm2 stimulus was sufficient to elicit a response in animals expressing Chrimson in both PLMs , neither PLML nor PLMR animals reacted , indicating a synergistic effect of PLML and PLMR and the requirement for activation of both neurons to elicit a reaction in response to weak stimuli ( Figure 5D–F , Figure 5—figure supplement 2 ) . When we increased illumination intensity , we saw a striking difference appear between the responses to PLML and PLMR stimulation , respectively . Animals expressing Chrimson in PLMR showed a robust reaction to stimulation at 14 mW/cm2 and higher . In contrast , the response of PLML animals was consistently below that of PLMR or 2PLM animals ( Figure 5D–F , Figure 5—figure supplement 2 ) . Only at the highest intensity setting of 74 mW/cm2 did we observe animals of all three expression patterns responding robustly to the stimulus . The observed differences in response for PLML and PLMR are not due to differential expression of Chrimson::mKate2 ( Figure 5—figure supplement 3 ) , therefore they are likely due to the asymmetric connectivity of the two neurons within the C . elegans nervous system .
We demonstrate a robust , improved method for highly efficient ncAA incorporation in C . elegans . Our optimised ncAA incorporation system will provide the means to introduce a wide range of new chemical functionalities into proteins , such as crosslinkers , bioorthogonal groups , and post-translational modifications , and serve as a catalyst for new approaches to manipulate and understand cellular biology in C . elegans ( Chin , 2017; Davis and Chin , 2012; Nguyen et al . , 2018; Lang and Chin , 2014; Young and Schultz , 2018; Zhang et al . , 2015; Baker and Deiters , 2014 ) . Furthermore , since our optimisation approach is based on components that are not specific to C . elegans , we expect that it will be easily exportable to other organisms . The genetic code expansion approach relies on the use of an amber stop codon to direct incorporation of ncAA , which means that endogenous TAG codons may be inadvertently suppressed by the orthogonal system , albeit likely at low levels . Interestingly , C . elegans and other organisms have evolved natural surveillance mechanisms to mitigate against the read-through of stop codons ( Arribere et al . , 2016 ) , which will also mitigate against any potential effects caused by our system . Indeed , we find that the function of the mechanosensory neurons we assay is unaffected by the presence of the orthogonal ncAA incorporation machinery . We demonstrate the utility of our system by creating LaserTAC , which is based on an optimised photoactivatable Cre recombinase and offers several unique advantages for optical control of gene expression . First , the recombinase activity is tightly controlled; the presence of the photocaging group completely blocks activity and upon photoactivation , highly efficient wildtype Cre is generated . Second , following photoactivation , the photocaged amino acid required for amber stop codon suppression can be removed . Thus translation of full-length Cre ceases , leaving worms devoid of photocaged Cre for subsequent biological studies ( Maywood et al . , 2018 ) . Third , due to the stability of the photocaging group at longer wavelengths , our photoactivatable Cre is fully compatible with imaging and optogenetic methods that use visible light . We used LaserTAC to switch on optogenetic channel expression in individual neurons within a bilaterally symmetric pair that cannot be targeted by genetic means . Switching on gene expression using LaserTAC is fast and easily enables the preparation of dozens of animals for downstream experiments . Light activation requires only a microscope-mounted laser , generally used for ablation in most C . elegans laboratories , modified to emit light at 365 nm . Generating the required transgenic strains is straightforward and at present involves the introduction of the genetic constructs , followed by genomic integration . The power of the method we present here lies in the versatility it offers , not only in respect to the ability to investigate single neurons . Once a transgenic strain is made , any combination of neurons targeted by the promoter that was used can in principle be investigated . This removes the requirement for the creation of a potentially large numbers of lines by allowing multiple conditions to be investigated using a single strain . A further advantage offered by our approach is the ability of precise temporal control , which is generally not available using promoter-based systems . We demonstrate the capabilities of LaserTAC by targeting expression of the optogenetic channel Chrimson to the mechanosensory PLM neurons . This enables us to analyse the individual contributions of the neurons PLML and PLMR to the C . elegans tail touch circuit response . Since we target expression of the optogenetic channel , we can perform behavioural assays by globally illuminating a plate of freely moving animals . We found that PLML and PLMR show some redundancy in triggering the tail touch response after strong stimulation . However , using weaker stimuli , we found that only stimulation of both neurons elicits a response , indicative of a synergistic rather than a simply redundant role for PLML and PLMR . Interestingly , our results demonstrate asymmetric contributions of PLML and PLMR to the touch response . Considering the electrical and chemical connectome of this circuit ( Figure 5C ) , our results are consistent with a model where the touch response is controlled by two parallel pathways that rely on gap junctions or chemical synapses , respectively . The first pathway involves direct connections of both PLM neurons to the PVC interneurons via gap junctions . Activation of the PVCs through these gap junctions triggers forward motion . The PVCs form connections with forward command interneuron AVB as well as inhibitory connections with backward command interneurons AVA/AVD . The second pathway involves chemical synaptic transmission from PLMR onto AVA and AVD , inhibiting backward locomotion . Thus , when only PLMR is stimulated , both pathways are activated , resulting in a stronger response , whereas when PLML alone is stimulated only the gap junction pathway is activated , resulting in a weaker response . Unlike the PLMs , the PVCs are connected to each other by gap junctions and may thus act as electrical sinks for signals from the PLMs . This could help to explain the requirement for stronger stimulation of the single neurons , especially for PLML , to trigger a response as compared to the stimulation of both PLMs simultaneously . Stimulation of both PLM neurons together results in full activation of the gap junction pathway by activating both PVC neurons , which , together with the direct inhibition of AVA/AVD by PLMR , results in full activation of the tail touch motor response . The strong synergistic effect we observe with the simultaneous stimulation of PLML and PLMR indicates a requirement for the stimulation of both PVC neurons . While we focused here on expression of optogenetic channels in C . elegans neurons , simple extensions of our method will enable the cell-specific expression of other desired transgenes of interest . By inserting loxP sites into genomic loci using CRISPR/Cas9 , it should be possible to control , at single-cell resolution , the expression of any endogenous protein , such as neuropeptides , receptors , innexins , proteins involved in synaptic transmission , etc . Since the components we employ are functional in other C . elegans tissues ( Greiss and Chin , 2011 ) and indeed in other model systems such as zebrafish , fruit fly , mouse , cultured cells , and ex vivo tissues , we anticipate that our method will have broad applicability beyond the nervous system and beyond C . elegans .
All expression plasmids were generated from pENTR plasmids using the Gateway system ( Thermo Fisher Scientific ) . All plasmids are described in Supplementary file 1 . Strains were maintained under standard conditions unless otherwise indicated ( Brenner , 1974; Stiernagle , 2006 ) . Transgenic worms were generated by biolistic bombardment using hygromycin B as a selection marker ( Greiss and Chin , 2011; Radman et al . , 2013; Davis and Greiss , 2018 ) in either N2 or smg-2 ( e2008 ) genetic background . The smg-2 ( e2008 ) background lacks a functional nonsense-mediated decay machinery and was used to increase levels of reporter mRNA ( Greiss and Chin , 2011 ) . Gamma-irradiation to generate the integrated line SGR56 from SGR55 was carried out by Michael Fasseas ( Invermis/Magnitude Biosciences ) . After integration , the line was backcrossed into N2 and subsequently maintained on standard NGM without added hygromycinB . SGR56 was backcrossed twice , SGR96 was backcrossed four times . All non-integrated lines were maintained on NGM supplemented with hygromycin B ( 0 . 3 mg/ml; Formedium ) . Strains used in this paper are listed in the key resources table . All imaging was carried out on a Zeiss M2 imager . Worms were mounted on glass slides for imaging and immobilised in a drop of M9 supplemented with 25 mM NaN3 . Fluorescent images were analysed using ImageJ software . To determine nuclear to cytoplasmic ratios of NES::PCKRS::GFP variants , mean fluorescence intensity was measured for a region of interest within the nucleus and divided by the mean intensity of a region in the cytoplasm proximal to the nucleus . To compare Chrimson::mKate2 levels in the PLML and PLMR neurons , the threshold function was used to create regions of interest encompassing the neurons and the mean intensity of those regions was taken . PCK ( Gautier et al . , 2010 ) was custom synthesised by ChiroBlock GmbH , Germany . PCK-NGM plates were prepared by dissolving PCK powder in a small volume of 0 . 02 M HCl and adding the solution to molten NGM . The HCl in the NGM was neutralised by addition of equimolar amounts of NaOH as previously described ( Davis and Greiss , 2018 ) . Animals were age synchronised by bleaching ( Stiernagle , 2006 ) and added to PCK-NGM plates as L1 larvae . Food was then added to the PCK-NGM plates in the form of solubilised freeze-dried OP50 ( LabTIE ) reconstituted according to the manufacturer’s instructions . Synchronised populations were grown on PCK-NGM plates until the young adult stage and washed off plates using M9 buffer supplemented with 0 . 001% Triton-X100 . Worms were settled , supernatant was removed , and worms were resuspended in undiluted 4× LDS loading buffer ( Thermo Fisher Scientific ) supplemented with NuPAGE Sample Reducing Agent ( Thermo Fisher Scientific ) . Lysis was performed by a freeze/thaw cycle followed by 10 min incubation at 95°C . Samples were run on precast Bolt 4–12% gels ( Thermo Fisher Scientific ) for 19 min at 200 V . Proteins were transferred from the gel onto a nitrocellulose membrane using an iBlot2 device ( Thermo Fisher Scientific ) . After transfer , the membrane was blocked using 5% milk powder in PBST ( PBS + 0 . 1% Tween-20 ) for 1 hr at room temperature . Incubation with primary antibodies was carried out in PBST + 5% milk powder at 4°C overnight . Blots were washed 6 × 5 min with PBST + 5% milk powder before incubating with secondary antibody for 1 hr at room temperature . Antibodies used are listed in the key resources table . Pierce ECL Western Blotting Substrate ( Thermo Fisher Scientific ) or SuperSignal West Femto chemiluminescent Substrate ( Thermo Fisher Scientific ) were used as detection agent . For quantitative blots , a C-DiGit Blot Scanner ( LI-COR ) was used , and intensities were analysed using ImageStudio software . Worms were age synchronised and grown on PCK-NGM plates for 48 hr , washed onto unseeded 6 cm NGM plates , and illuminated for 5 min , 5 mW/cm2 in a 365 nm CL-1000L crosslinker ( UVP ) as previously described ( Davis and Greiss , 2018 ) . After uncaging , worms were transferred to seeded NGM plates and scored for expression of the target gene 24 hr later by counting the animals showing expression of the red fluorescent marker . 400 μM FUDR was added to plates after uncaging to prevent hatching of F1 progeny and thus aid scoring of animals expressing the target gene . All plates were scored twice , independently by two people . Scoring was performed blind . Experiments were performed three times , each with two independent transgenic lines . Significance tests were carried out using Welch’s t test as a pairwise comparison between each condition at each concentration using GraphPad Prism 8 software . Worms were grown on 4 mM PCK from the L1 stage . For uncaging , worms were mounted on a 3% agar and immobilised with 25 mM NaN3 in M9 buffer . Targeting was performed on a Zeiss M2 Imager using a Micropoint Galvo module ( Andor Technology/Oxford Instruments ) , fitted with a 365 nm dye cell . Neurons were identified using GFP as a guide , which was co-expressed with photocaged Cre , and the nucleus targeted with the laser . Each region was swept thrice using 10 repeat firing . The manual attenuator was set to full power and the Andor software attenuator to a power of 34 ( we chose the power setting so that partial bleaching of GFP could be observed during Micropoint firing ) . After uncaging , the coverslip was removed and worms were washed off the pad onto seeded 6 cm plates using M9 + 0 . 001% Triton-X . Mock laser treatment control worms were mounted and recovered similar to experimental worms but without exposure to the Micropoint laser . Immediately after uncaging , worms were transferred to NGM plates supplemented with ATR ( 30 μl 5 mM ATR dissolved in ethanol were added to the lawn of a seeded 6 cm NGM plate ) or control NGM-only plates . Worms were left to recover on plates for 24 hr after uncaging . Plates for behavioural assays were made approximately 1 hr before use . Fresh 6 cm NGM plates were seeded with a 20 μl drop of dilute OP50 , prepared from freeze-dried OP50 ( LabTIE ) . For this , the freeze-dried bacteria were reconstituted according to the manufacturer’s instructions and a 40× dilution of the reconstituted bacteria was used to spot the plates . The lawn was then surrounded by a copper ring ( inner diameter 17 mm ) to prevent worms from leaving the camera field of view . For experiments on PLM , expression of the target gene after optPC-Cre laser activation was visually confirmed using a Leica M165FC fluorescence microscope fitted with a 2 . 0× objective and animals showing the correct pattern ( either both neurons or individual neurons , depending on the uncaging experiment ) were picked onto the behaviour plates . For experiments using the glr-1p promoter , animals were randomly picked . After transfer , worms were left to acclimatise on behaviour plates for at least 10 min . Behavioural assays were carried out using a WormLab system ( MBF Bioscience ) . Chrimson or Channelrhodopsin2 were activated using either the integrated 617 nm LED or 470 nm LED of the WormLab platform at full power ( 74 mW/cm2 or 150 mW/cm2 , respectively ) , unless otherwise stated . Tracking of worm behaviour was carried out using WormLab software ( MBF Bioscience ) . For each worm , tracks detected by the WormLab software were consolidated into a single track . Data points where individual animals were not clearly identified by the software were discarded . For all behavioural experiments , worms were grouped into cohorts . A cohort consisted of 4–10 animals undergoing the same treatment simultaneously on the same plate . Behavioural experiments on worms expressing ChR2 from the glr-1p promoter were carried out by exposing worms to 1 s of 470 nm illumination at 150 mW/cm2 . Reversals were counted manually . The reversal percentages were determined for each replicate . Significance between conditions was tested using a Mann–Whitney U test . Experiments using repeated stimulation of animals expressing Chrimson::mKate2 in the PLM neurons were carried out by exposing worms to repeated 1 s long stimulations with 617 nm light ( 74 mW/cm2 ) at 30 s intervals . Raw speed data was binned and averaged at 1 s intervals for each animal . Worms were grouped by cohort . To determine the fraction of animals responding to a stimulus , worms with an average speed ≥100 μm s–1 in the 2 s following stimulation were scored as responders . The 100 μm s–1 threshold was selected as it constitutes the standard deviation of the mean speed of unstimulated worms . For each timepoint , the percentage of worms responding was determined for each cohort , constituting one replicate . Comparison between conditions was carried out by two-way ANOVA across all stimuli . Experiments comparing conditions across different power levels for animals expressing Chrimson::mKate2 in the PLM neurons were carried out by subjecting each cohort of worms to 0 . 1 s stimulation with 617 nm at the indicated power levels . The percentage of worms responding for each cohort was calculated after first normalising the data by subtracting the mean speed for each individual worm in the 0 . 5 s immediately prior to stimulation . Then the mean velocity for each individual worm was determined for the response peak between 0 . 2 s and 1 . 2 s after stimulation . Worms were scored as responders if the mean normalised velocity was ≥ 100 μm s–1 . When calculating the mean normalised velocity across experiments , the mean velocity among all animals within a cohort was determined first , followed by calculating the average of the mean velocities across cohorts . Significance for both analyses ( fraction responding/intensity and mean velocity/intensity ) was determined by two-way ANOVA . All statistical analysis of data was carried out in GraphPad Prism 8 . Statistical tests are stated in the figure legends . All values of n in this article , unless otherwise stated , refer to biological replicates . For behavioural experiments , we define as biological replicates cohorts of worms that were assayed on different days . A cohort of worms consisted of animals undergoing treatment and assayed together on the same plate . | Animal behaviour and movement emerges from the stimulation of nerve cells that are connected together like a circuit . Researchers use various tools to investigate these neural networks in model organisms such as roundworms , fruit flies and zebrafish . The trick is to activate some nerve cells , but not others , so as to isolate their specific role within the neural circuit . One way to do this is to switch genes on or off in individual cells as a way to control their neuronal activity . This can be achieved by building a photocaged version of the enzyme Cre recombinase which is designed to target specific genes . The modified Cre recombinase contains an amino acid ( the building blocks of proteins ) that inactivates the enzyme . When the cell is illuminated with UV light , a part of the amino acid gets removed allowing Cre recombinase to turn on its target gene . However , cells do not naturally produce these photocaged amino acids . To overcome this , researchers can use a technology called genetic code expansion which provides cells with the tools they need to build proteins containing these synthetic amino acids . Although this technique has been used in live animals , its application has been limited due to the small amount of proteins it produces . Davis et al . therefore set out to improve the efficiency of genetic code expansion so that it can be used to study single nerve cells in freely moving roundworms . In the new system , named LaserTAC , individual cells are targeted with UV light that ‘uncages’ the Cre recombinase enzyme so it can switch on a gene for a protein that controls neuronal activity . Davis et al . used this approach to stimulate a pair of neurons sensitive to touch to see how this impacted the roundworm’s behaviour . This revealed that individual neurons within this pair contribute to the touch response in different ways . However , input from both neurons is required to produce a robust reaction . These findings show that the LaserTAC system can be used to manipulate gene activity in single cells , such as neurons , using light . It allows researchers to precisely control in which cells and when a given gene is switched on or off . Also , with the improved efficiency of the genetic code expansion , this technology could be used to modify proteins other than Cre recombinase and be applied to other artificial amino acids that have been developed in recent years . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"tools",
"and",
"resources",
"neuroscience"
] | 2021 | Precise optical control of gene expression in C elegans using improved genetic code expansion and Cre recombinase |
We studied the role of the synaptic ribbon for sound encoding at the synapses between inner hair cells ( IHCs ) and spiral ganglion neurons ( SGNs ) in mice lacking RIBEYE ( RBEKO/KO ) . Electron and immunofluorescence microscopy revealed a lack of synaptic ribbons and an assembly of several small active zones ( AZs ) at each synaptic contact . Spontaneous and sound-evoked firing rates of SGNs and their compound action potential were reduced , indicating impaired transmission at ribbonless IHC-SGN synapses . The temporal precision of sound encoding was impaired and the recovery of SGN-firing from adaptation indicated slowed synaptic vesicle ( SV ) replenishment . Activation of Ca2+-channels was shifted to more depolarized potentials and exocytosis was reduced for weak depolarizations . Presynaptic Ca2+-signals showed a broader spread , compatible with the altered Ca2+-channel clustering observed by super-resolution immunofluorescence microscopy . We postulate that RIBEYE disruption is partially compensated by multi-AZ organization . The remaining synaptic deficit indicates ribbon function in SV-replenishment and Ca2+-channel regulation .
Encoding and processing of sensory information in the ear and the eye rely on ribbon synapses . Described in the 1960s as an electron dense structure tethering a halo of vesicles ( Sjostrand , 1958; Smith and Sjostrand , 1961 ) , the function of the synaptic ribbon has remained enigmatic despite decades of work ( recent reviews in Lagnado and Schmitz , 2015; Moser and Vogl , 2016; Safieddine et al . , 2012; Wichmann and Moser , 2015 ) . Approaches to ribbon function included studies that employed natural variation of ribbon size or abundance during diurnal cycle or hibernation ( Hull et al . , 2006; Mehta et al . , 2013 ) , photoablation ( Mehta et al . , 2013; Snellman et al . , 2011 ) and genetic manipulation ( Dick et al . , 2003; Frank et al . , 2010; Jing et al . , 2013; Khimich et al . , 2005; Lv et al . , 2016; Maxeiner et al . , 2016; Sheets et al . , 2011; Van Epps et al . , 2004 ) . Mutations initially focused on the presynaptic scaffold protein bassoon that is required for ribbon anchorage to the AZ ( Dick et al . , 2003; Khimich et al . , 2005 ) via interaction with RIBEYE ( tom Dieck et al . , 2005 ) . However , bassoon also exerts direct effects on AZ function ( Davydova et al . , 2014; Hallermann et al . , 2010; Mendoza Schulz et al . , 2014 ) and , hence , distinguishing direct effects of bassoon deletion and those caused by ribbon loss remained challenging ( Jing et al . , 2013 ) . RIBEYE-disruption turned out to be difficult: it is transcribed from the same gene as CtBP2 , an essential transcription factor , disruption of which causes embryonic lethality ( Hildebrand and Soriano , 2002 ) . Complete abolition of RIBEYE was hard to achieve in zebrafish ( Lv et al . , 2016; Van Epps et al . , 2004 ) given their duplicated genome . In fact , despite targeting both ribeye genes , RIBEYE immunofluorescence remained present in the retina and hair cells displayed ‘ghost ribbons’: structures recognized by a synaptic vesicle-halo but lacking electron density ( Lv et al . , 2016 ) . Complete disruption of RIBEYE expression and lack of retinal ribbons were recently reported in a mouse knock-out of the RIBEYE-specific exon ( Maxeiner et al . , 2016 ) . This study proved that RIBEYE is required for ribbon formation in the mammalian retina and the observed ribbon loss grossly impaired glutamate release from bipolar cell terminals . The key conclusion was that ribbons help to couple voltage-gated Ca2+-channels to vesicular release sites to enable tight , so-called Ca2+-nanodomain control of exocytosis ( Maxeiner et al . , 2016 ) , that was previously reported for ribbon synapses of ear and eye ( Bartoletti et al . , 2011; Brandt et al . , 2005; Graydon et al . , 2011; Jarsky et al . , 2010; Johnson et al . , 2017; Pangršič et al . , 2015; Wong et al . , 2014 ) . By employing the most specific , yet chronic , manipulation of the ribbon , this functional study on ribbonless retinal rod bipolar cells also confirmed that RIBEYE/the ribbon promotes a large complement of vesicular release sites . However , the electrophysiology was performed on rod bipolar cells while the molecular anatomy ( immunofluorescence ) focused on rod photoreceptors . Since the structure and function of ribbons formed at these two different cell types are distinct , a simple structure-function model was not easy to derive from this study . Moreover , the consequences of ribbon loss remained to be investigated at the systems level . Here , we studied the effects of RIBEYE-disruption on synaptic sound encoding in the cochlea . Combining assessments of the molecular anatomy from electron and fluorescence microscopy with cell and systems physiology , we revealed a role for the synaptic ribbon in organizing the topography of the IHC AZ , in Ca2+-channel regulation and in vesicle replenishment . In summary , we demonstrate that the synaptic ribbon is important for sound encoding at high rates and with temporal precision at IHC synapses .
We first employed immunohistochemistry to study IHCs of 3-week-old RIBEYE knock-out mice ( RBEKO/KO ) , in which the unique A-domain exon of RIBEYE was deleted by Cre-mediated excision ( described in Maxeiner et al . , 2016 ) . Next to the A-domain , RIBEYE contains a B-domain that is largely identical to the transcription factor CtBP2 , which is spared by the genetic manipulation and used as a target in immunohistochemistry of ribbons and nuclei ( Figure 1A , B; Khimich et al . , 2005 ) . Synaptic ribbons of IHC afferent synapses were identified as presynaptic RIBEYE/CtBP2-immunofluorescent spots in wild-type ( Figure 1B , RBEWT/WT ) and heterozygous ( Figure 1—figure supplement 1 , RBEWT/KO ) mice . Their number per IHC did not change in the heterozygous condition ( 15 . 5 ± 0 . 7 , S . D . = 1 . 58; n = 50 cells , N = 4 for RBEWT/KO vs . 15 . 7 ± 1 . 1 , S . D . = 2 . 19; n = 39 cells , N = 3 for RBEWT/WT at P21 ) , while their intensity was significantly reduced ( in arbitrary units: 3 . 4 ± 0 . 7 , S . D . = 1 . 78; n = 600 spots for 40 cells , N = 3 in RBEWT/KO vs . 5 . 1 ± 1 . 1 , S . D . = 2 . 23; n = 411 spots for 29 IHCs , N = 3 in RBEWT/WT; p<0 . 0001 , Mann-Whitney-Wilcoxon test; Figure 1—figure supplement 1A , B ) . RBEKO/KO IHCs lacked synaptic RIBEYE/CtBP2 immunofluorescence spots ( Figure 1B ) , while immunolabeling of nuclear CtBP2 remained present , corroborating previous findings in the retina ( Maxeiner et al . , 2016 ) . The number of afferent synapses per IHC was determined by the count of postsynaptic densities ( PSDs ) identified as PSD-95 immunofluorescent spots ( Figure 1C , D , E ) and was unchanged when RIBEYE was removed ( 13 . 7 ± 0 . 8 , S . D . = 2 . 04; n = 56 cells , N = 4 in RBEKO/KO vs . 12 . 9 ± 0 . 6 , S . D . = 2 . 13; n = 55 cells , N = 5 in RBEWT/WT ) . Bassoon ( Figure 1C ) and RIM2 ( Figure 1D ) , both presynaptic scaffold proteins ( Khimich et al . , 2005;Jung et al . , 2015a ) , remained present at the ribbonless afferent synapses of RBEKO/KO IHCs ( marked by PSD-95 ) . The scaffold protein piccolino , the short isoform of piccolo ( Regus-Leidig et al . , 2013 ) that is present in cochlear and retinal ribbons ( Khimich et al . , 2005; tom Dieck et al . , 2005; Regus-Leidig et al . , 2013 ) , was absent from afferent synapses of RBEKO/KO IHCs . However , piccolo immunofluorescence was present in the vicinity of afferent synapses likely marking the long form piccolo at the efferent presynaptic AZs ( Figure 1E , see schematic in Figure 1A ) . The PSD areas were calculated by fitting a 2-dimensional Gaussian function to each PSD-95 immunofluorescent spot , revealing a significant increase in the RBEKO/KO condition ( 2 . 82 ± 0 . 09 µm² , S . D . = 1 . 25; n = 178 spots , N = 3 vs . 1 . 74 ± 0 . 05 µm² , S . D . = 0 . 58; n = 163 spots , N = 3 in RBEWT/WT IHCs; p<0 . 0001 , Mann-Whitney-Wilcoxon; Figure 1F ) . In order to study the effects of RIBEYE deletion on the ultrastructure of afferent IHC synapses , we performed transmission electron microscopy on random sections and electron tomography . Random ultrathin ( 70–75 nm ) sections prepared from P21 mice ( two animals per genotype ) after aldehyde fixation and conventional embedding procedures showed that IHCs from RBEKO/KO mice completely lack synaptic ribbons , while RBEWT/WT and heterozygous RBEWT/KO typically display one ribbon per AZ ( Figure 2A–C ) . Interestingly , ribbons of RBEWT/KO IHCs were smaller in height , width and area compared to RBEWT/WT IHC synaptic ribbons ( Figure 2—figure supplement 1A–C; ribbon height: 118 . 32 ± 3 . 17 nm , S . D . = 31 . 84 nm; n = 101 ribbons , N = 2 for RBEWT/KO vs . 197 . 09 ± 4 . 36 nm , S . D . = 44 . 93 nm; n = 106 ribbons , N = 2 for RBEWT/WT; ribbon width: 119 . 80 ± 6 . 23 nm , S . D . = 62 . 27 nm for RBEWT/KO vs . 168 . 34 ± 6 . 83 nm , S . D . = 70 . 27 nm for RBEWT/WT; ribbon area: 11 . 5e3 ± 6 . 2e2 nm2 , S . D . = 6 . 3e3 nm² for RBEWT/KO vs . 25 . 4e3 ±1 . 1e2 nm2 , S . D . = 1 . 1e3 nm² for RBEWT/WT; p<0 . 0001 , Mann-Whitney-Wilcoxon test for all ) agreeing with the significantly reduced ribbon immunofluorescence intensity in the RBEWT/KO condition ( see above and Figure 1—figure supplement 1A–B ) . Random sections of synaptic contacts of RBEKO/KO mice ( Figure 2C ) often showed more than one presynaptic density ( PD ) , each associated with a cluster of synaptic vesicles ( henceforth considered individual AZs ) . The multiple AZs typically faced one continuous PSD , which is different from the synapses of immature IHC synapses that show multiple appositions of pre- and postsynaptic densities ( Sendin et al . , 2007; Wong et al . , 2014 ) . Moreover , we found more than one PD per synaptic contact in IHCs of older RBEKO/KO mice ( Figure 2E , F; 6 weeks and 8 months , respectively ) , arguing against a delayed synaptic maturation to be the cause of the phenotype . Sections from tangential cuts of the synapse ( Figure 2D ) , reconstructions from serial ultrathin sections ( Figure 2G , G’ ) and quantifications of random sections ( Figure 2H ) corroborated the notion of multiple small ribbonless AZs at the synaptic contacts of RBEKO/KO IHCs . Analysis based on serial 3D reconstructions of synaptic contacts of RBEKO/KO IHCs from P21 animals showed on average 1 . 92 ± 0 . 34 PDs ( S . D . = 1 . 16; n = 17 serial 3D reconstructions , N = 2 ) and 20 . 58 ± 2 . 98 total SVs per contact , S . D . = 10 . 34 ( Figure 2I ) . The lateral extent of the individual PDs , determined in random sections , was comparable between RBEKO/KO and RBEWT/WT synapses ( 129 . 89 ± 2 . 53 nm , S . D . = 26 . 26 nm; n = 108 PDs , N = 2 for RBEKO/KO vs . 129 . 35 ± 4 . 89 nm , S . D . = 50 . 86 nm; n = 108 PDs , N = 2 for RBEWT/WT; p=0 . 92 , NPMC test ) , while that of RBEWT/KO was enlarged ( Figure 2M; 157 . 64 ± 7 . 19 nm , S . D . = 72 . 24 nm; n = 101 PDs , N = 2; p=0 . 0004 for comparison to RBEWT/WT , NPMC test ) . PSDs tended to be increased in length at RBEKO/KO synapses compared to RBEWT/WT PSDs and were significantly larger than RBEWT/KO PSDs ( Figure 2N; 623 . 77 ± 26 . 70 nm , S . D . = 264 . 33 nm; n = 98 PSDs , N = 2 for RBEKO/KO vs . 555 . 91 ± 22 . 24 nm , S . D . = 236 . 42 nm; n = 113 PSDs , N = 2 for RBEWT/WT vs . 521 . 34 ± 24 . 20 nm , S . D . = 242 . 03 nm; n = 100 PSDs , N = 2 for RBEWT/KO; p=0 . 01 for RBEKO/KO vs . RBEWT/KO , Tukey’s test ) , which is consistent with the greater area of PSD-95 immunofluorescent spots in the knock-out condition ( Figure 1E ) . In the following , we characterized the populations of presynaptic SVs in random sections of vertically-cut IHC synapses . We counted membrane-proximal SVs ( MP-SVs , ≤25 nm distance between SV membrane and plasma membrane , laterally within 80 nm of the PD , yellow in Figure 2J–L ) as well as ribbon-associated SVs ( RA-SVs , first layer of SVs around the ribbon within 80 nm , green in Figure 2J , K ) or ‘PD-associated’ SVs ( PDA-SVs , ribbonless AZs: SVs within 80 nm distance of the PD and not falling into the MP-SV pool ( see above ) , green in Figure 2J , L ) . We found both MP-SVs ( Figure 2O; 1 . 92 ± 0 . 09 , S . D . = 0 . 93; n = 108 AZs , N = 2 for RBEKO/KO vs . 2 . 99 ± 0 . 12 , S . D . = 1 . 18; n = 101 AZs , N = 2 for RBEWT/KO vs . 2 . 77 ± 0 . 12 , S . D . = 1 . 18; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 for RBEKO/KO vs . RBEWT/WT , NPMC test ) and PDA-SVs ( Figure 2P; 4 . 12 ± 0 . 15 , S . D . = 1 . 50; n = 108 AZs , N = 2 for RBEKO/KO vs . 10 . 09 ± 0 . 27 , S . D . = 2 . 75; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 , Tukey’s test ) of the individual ribbonless IHC AZs of RBEKO/KO mice to be significantly fewer than the corresponding number of MP-SVs and RA-SVs counted at RBEWT/WT AZs . The fraction of PDA-SVs relative to the total number of SVs at RBEKO/KO AZs was less than that of RA-SVs at RBEWT/WT AZs ( Figure 2Q; 0 . 67 ± 0 . 02 , S . D . = 0 . 16; n = 108 AZs , N = 2 for RBEKO/KO vs . 0 . 78 ± 0 . 01 , S . D . = 0 . 08; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 , NPMC test ) . Consequently , we observed an increase in the fraction of MP-SVs at RBEKO/KO AZs ( Figure 2Q; 0 . 33 ± 0 . 02 , S . D . = 0 . 16; n = 108 AZs , N = 2 for RBEKO/KO vs . 0 . 22 ± 0 . 01 , S . D . = 0 . 08; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 , NPMC test ) . In line with the decreased ribbon size of RBEWT/KO AZs , we found a reduced number of RA-SVs , indicating a hypomorphic phenotype upon the loss of one allele of the RIBEYE gene . The SV diameter was unchanged for all three genotypes when jointly considering SVs of all categories in random sections ( Figure 2—figure supplement 1D; 39 . 59 ± 0 . 21 nm , S . D . = 5 . 37 nm; n = 108 AZs , N = 2 for RBEKO/KO vs . 40 . 53 ± 0 . 14 nm , S . D . = 4 . 44 nm; n = 101 AZs , N = 2 for RBEWT/KO vs . 41 . 80 ± 0 . 13 nm , S . D . = 4 . 79 nm; n = 106 AZs , N = 2 for RBEWT/WT; p=0 . 30 , NPMC test ) . However , we found a subtle but significant SV-diameter reduction in RBEKO/KO and RBEWT/KO for MP-SVs ( Figure 2—figure supplement 1E; 39 . 29 ± 0 . 34 nm , S . D . = 4 . 82 nm; n = 108 AZs , N = 2 for RBEKO/KO vs . 41 . 79 ± 0 . 26 nm , S . D . = 4 . 53 nm; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 , NPMC test and 40 . 29 ± 0 . 25 nm , S . D . = 4 . 40 nm; n = 101 AZs , N = 2 for RBEWT/KO vs . RBEKO/KO; p=0 . 03 , NPMC test ) and for RA-/PDA-SVs ( Figure 2—figure supplement 1F; 39 . 72 ± 0 . 27 nm , S . D . = 5 . 61 nm; n = 108 AZs , N = 2 for RBEKO/KO vs . 41 . 81 ± 0 . 15 nm , S . D . = 4 . 86 nm; n = 106 AZs , N = 2 for RBEWT/WT; p<0 . 0001 , NPMC test and RBEKO/KO vs . 40 . 63 ± 0 . 17 nm , S . D . = 4 . 45 nm; n = 101 AZs , N = 2 for RBEWT/KO; p=0 . 003 , NPMC test and RBEWT/KO vs . RBEWT/WT; p=0 . 02 , NPMC test ) . Next , to capture the synapses in a near-to-native state and to evaluate vesicle tethering , we performed electron tomography on 250 nm thick sections that were prepared with high-pressure freezing and freeze-substitution ( HPF/FS ) of organs of Corti from P21 mice ( Figure 3 ) . Tomography confirmed the absence of synaptic ribbons and the presence of multiple AZs per contact , each with a clear PD ( Figure 3B , D , F ) . However , we note that the 250 nm thick sections did typically not fully cover the synaptic contact , which leads to an underestimation for the total number of SVs particularly for the spatially extended RBEKO/KO synapses . The PDs appeared roundish in the RBEKO/KO with MP-SVs closely arranged around the PD as found at the more elongated RBEWT/WT AZs ( Figure 3C , D ) . For the tomograms , we followed the definition of MP-SV pool according to the 2D-random sections ( Figure 2 ) , but in addition we measured the MP-SVs also in a maximum distance of 50 nm from the AZ membrane and ≤100 nm from the PD ( Figure 3—figure supplement 1 and supplementary file 1 ) . This was motivated by the presence of long tethers connecting SV and AZ membrane and was previously introduced ( Jung et al . , 2015a ) . Further , we distinguished between tethered and non-tethered SVs ( Figure 3G , Figure 3—figure supplement 1A ) . There , we focused our analysis on tethers to the ribbon/PD , plasma membrane and those interconnecting two adjacent SVs ( Figure 3H , I , M , N ) . We found a significant reduction in the number of MP-SVs per AZ in RBEKO/KO IHCs ( Figure 3J; RBEKO/KO = 6 . 30 ± 0 . 86 , S . D . = 2 . 87 MP-SVs; n = 11 AZs , N = 3 vs . RBEWT/WT = 8 . 70 ± 0 . 82 , S . D . = 2 . 45 MP-SVs; n = 9 AZs , N = 3; p=0 . 04 , Mann-Whitney-Wilcoxon test; significant also by the second analysis method: Figure 3—figure supplement 1B , supplementary file 1 ) , while the fraction of tethered MP-SVs ( No . of tethered MP-SVs/No . of all MP-SVs ) was not significantly altered ( Figure 3K; RBEKO/KO = 0 . 75 ± 0 . 07 , S . D . = 0 . 24; n = 11 AZs , N = 3 vs . RBEWT/WT = 0 . 65 ± 0 . 06 , S . D . = 0 . 18; n = 9 AZs , N = 3; p=0 . 30 , t-test; Figure 3—figure supplement 1C , supplementary file 1 ) . The majority of the MP-SVs were tethered via a single tether in both RBEWT/WT and RBEKO/KO IHCs . The fraction of MP-SVs with multiple ( ≥2 ) tethers was significantly larger in RBEKO/KO IHCs ( Figure 3K; single-tethered MP-SVs: RBEKO/KO = 0 . 55 ± 0 . 06 , S . D . = 0 . 19; n = 11 AZs , N = 3 vs . RBEWT/WT = 0 . 61 ± 0 . 06 , S . D . = 0 . 17; n = 9 AZs , N = 3; p=0 . 81; multiple-tethered MP-SVs: RBEKO/KO = 0 . 20 ± 0 . 05 , S . D . = 0 . 15; n = 11 AZs , N = 3 vs . RBEWT/WT = 0 . 04 ± 0 . 02 , S . D . = 0 . 05; n = 9 AZs , N = 3; p=0 . 01 , Tukey’s test; Figure 3—figure supplement 1C , supplementary file 1 ) . Further , and in line with analysis of random sections , the number of PDA-SVs per RBEKO/KO AZ was smaller than that of RA-SVs at RBEWT/WT AZs ( Figure 3O; RBEKO/KO: 9 . 30 ± 1 . 13 , S . D . = 3 . 74 PDA-SVs; n = 11 AZs , N = 3 vs . RBEWT/WT: 30 . 33 ± 3 . 00 , S . D . = 9 . 01 RA-SVs; n = 9 AZs , N = 3; p<0 . 0001 , Mann-Whitney-Wilcoxon test; Figure 3—figure supplement 1 , supplementary file 1 ) . However , the fraction of PDA-SVs tethered to the PD was not different from that of RA-SVs tethered to the ribbon ( Figure 3P; RBEKO/KO: 0 . 80 ± 0 . 06 , S . D . = 0 . 19 tethered PDA-SV fraction; n = 11 AZs , N = 3 vs . RBEWT/WT: 0 . 70 ± 0 . 06 , S . D . = 0 . 17 tethered RA-SV fraction; n = 9 AZs , N = 3; p=0 . 12 , t-test; Figure 3—figure supplement 1F , supplementary file 1 ) . Finally , tomography indicated unchanged SV diameters at RBEKO/KO AZs ( Figure 3L , Q; MP-SV diameter: 50 . 17 ± 0 . 90 nm , S . D . = 2 . 95 nm; n = 11 PDs , N = 3 for RBEKO/KO vs . 47 . 81 ± 0 . 60 nm , S . D . = 1 . 70; n = 9 ribbons , N = 3 for RBEWT/WT; p=0 . 06 , Mann-Whitney-Wilcoxon test , Figure 3—figure supplement 1D , supplementary file 1; RA/RA-SV diameter: 49 . 71 ± 0 . 83 nm , S . D . = 2 . 75; n = 11 PDs , N = 3 for RBEKO/KO vs . 49 . 80 ± 0 . 78 nm , S . D . = 2 . 35; n = 9 ribbons , N = 3 for RBEWT/WT; p=0 . 71 , Mann-Whitney-Wilcoxon test; Figure 3—figure supplement 1G , supplementary file 1 ) . We presume that differences in the comparison of RBEKO/KO and RBEWT/WT between the random section and electron tomography analysis primarily reflects the larger number of AZ analyzed by the former approach . We then used confocal and stimulated emission depletion ( STED ) super-resolution immunofluorescence microscopy in order to study the abundance and spatial organization of presynaptic CaV1 . 3 Ca2+-channels ( Neef et al . , 2018 ) , which contribute more than 90% of the voltage-gated Ca2+-influx into IHCs ( Platzer et al . , 2000; Brandt et al . , 2003; Dou et al . , 2004 ) . Organs of Corti from 3-week-old RBEKO/KO and RBEWT/WT mice were processed in parallel for immunohistochemistry and imaging . CaV1 . 3 Ca2+-channels remained clustered at RBEKO/KO AZs and were identified as CaV1 . 3 labeling juxtaposed to PSD-95 immunofluorescent spots ( Figure 4A ) . In order to analyze the spatial organization of synaptic Ca2+-channels , we performed 3-color , 2D-STED immunofluorescence imaging for CaV1 . 3 , bassoon ( as a PD-marker ) , and PSD-95 . While more than 80% of the RBEWT/WT synapses showed the typical stripe-like co-alignment of CaV1 . 3 and bassoon immunofluorescence ( Neef et al . , 2018 ) , imaging of RBEKO/KO synapses indicated a high prevalence ( over 70% ) of smaller , rounder and often several Ca2+-channel clusters and PDs per synaptic contact ( Figure 4B , C ) . We then quantified stripe-like clusters by measuring their long and short axis using 2D Gaussian fits and found no differences between RBEKO/KO and RBEWT/WT AZs ( Figure 4D ) . Finally , we quantified the number of CaV1 . 3-immunofluorescent structures per contact ( as indicated by PSD-95 immunofluorescence ) . While more than 80% of RBEWT/WT synapses displayed a single cluster , over 60% of the RBEKO/KO synapses contained two or more Ca2+-channel clusters ( Figure 4E ) . Hence , the average number of CaV1 . 3-immunofluorescent structures was significantly higher at RBEKO/KO synapses compared to RBEWT/WT ( 2 . 06 ± 0 . 09 , S . D . = 1 . 16; n = 178 spots , N = 3 vs . 1 . 16 ± 0 . 03 , S . D . = 0 . 38; n = 183 spots , N = 2; p<0 . 0001 , Mann-Whitney-Wilcoxon test ) and we likely underestimated this difference due to the low resolution of 2D-STED in the z-axis . In summary , our results indicate that RIBEYE-disruption transforms the single ribbon-type AZ into a complex presynaptic organization with multiple conventional-like AZs facing the postsynaptic bouton . Next , we combined whole-cell patch-clamp with confocal Ca2+-imaging of IHCs to study Ca2+-influx at the whole IHC and single synapse levels using 5 mM [Ca2+]e to augment the signal to noise . Using step-depolarizations in conditions that isolated the Ca2+-current ( see Materials and methods ) , we probed the amplitude and voltage-dependence of IHC Ca2+-influx ( Figure 5A ) . The amplitude of Ca2+-influx ( Figure 5Ai; for Ca2+-current density , see Figure 5Aii ) was unaltered in RBEKO/KO IHCs ( -151 ± 12 . 9 pA , S . D . = 59 pA; n = 21 IHCs , N = 8 in RBEKO/KO vs . -161 ± 15 . 4 pA , S . D . = 71 pA; n = 21 IHCs , N = 9 in RBEWT/WT; p=0 . 62 , t-test ) , in agreement with findings in retinal bipolar neurons ( Maxeiner et al . , 2016 ) but in contrast to our previous findings in ribbon-deficient IHCs of bassoon mutant mice ( Khimich et al . , 2005; Frank et al . , 2010; Jing et al . , 2013 ) . Kinetics of Ca2+-channel activation were unchanged ( Figure 5B ) , whereas inactivation kinetics were slightly faster in the RBEKO/KO IHCs ( smaller residual Ca2+-current at 200 ms of depolarization ( normalized to the peak current ) : 0 . 82 ± 0 . 007 , S . D . = 0 . 02; n = 10 IHCs , N = 5 for RBEKO/KO vs . 0 . 85 ± 0 . 01 , S . D . = 0 . 04; n = 11 IHCs , N = 7 , in the RBEWT/WT condition; p=0 . 017 , Mann-Whitney-Wilcoxon test; Figure 5D ) . When analyzing the voltage-dependence of Ca2+-channel activation ( Figure 5C ) , we found a small ( 2 mV ) but significant depolarizing shift of the potential of half-maximal Ca2+-channel activation , Vh ( Figure 5Ci , −22 . 96 ± 0 . 43 mV , S . D . = 2 . 39 mV; n = 21 IHCs , N = 8 in RBEKO/KO vs . −25 . 04 ± 0 . 65 mV , S . D . = 2 . 98 mV; n = 21 IHCs , N = 9 in RBEWT/WT; p=0 . 017 , t-test ) . When analyzed in a smaller data set recorded in 2 mM [Ca2+]e the depolarized Vh-shift did not reach statistical significance ( data not shown ) . The average voltage-sensitivity of activation ( slope factor k ) was not altered ( 5 mM [Ca2+]e: p=0 . 67 , t-test , Figure 5Cii ) . Together , this suggests a RIBEYE/ribbons-mediated regulation of IHC Ca2+-channels affecting their voltage-range of operation as well as their inactivation kinetics . We then used the low-affinity Ca2+-indicator dye Fluo-4FF ( 800 µM ) to study Ca2+-influx at individual IHC AZs ( Frank et al . , 2009 ) using a spinning-disk confocal microscope that allows rapid registering and recording of the majority of the IHC synapses ( Figure 6A , Ohn et al . , 2016 ) . We chose conditions in which the Ca2+-indicator fluorescence approximates synaptic Ca2+-influx ( Frank et al . , 2009; Ohn et al . , 2016 ) and henceforth refer to synaptic Ca2+-influx when describing observations based on hotspots of Ca2+-indicator fluorescence at the basolateral IHC membrane . Prior to analysis of synaptic Ca2+-influx , we imaged fluorescently-conjugated CtBP2-binding peptide ( Zenisek et al . , 2004 ) , which bound to the ribbon-occupied AZs in RBEWT/WT IHCs while it only caused nuclear and diffuse cytosolic fluorescence in the ribbonless RBEKO/KO IHCs ( Figure 6B ) . We then employed ramp-depolarizations to assess amplitude and voltage-dependence of Ca2+-influx at the synapses located in the subnuclear , basal part of the IHCs ( Figure 6A ) . We found comparable maximal amplitudes of the baseline-normalized fluorescence change ( ΔF/F0 , 0 . 88 ± 0 . 08 , S . D . = 0 . 66; n = 61 AZs in 15 IHCs , N = 7 for RBEKO/KO vs . 0 . 85 ± 0 . 08 , S . D . = 0 . 68; n = 78 AZs in 15 IHCs , N = 8 for RBEWT/WT; p=0 . 20 , Mann-Whitney-Wilcoxon test; Figure 6C ) . This is compatible with an unaltered number of synaptic Ca2+-channels at AZs of RBEKO/KO IHCs and consistent with our observations of normal whole-cell Ca2+-current amplitudes . As previously reported ( Frank et al . , 2009; Ohn et al . , 2016 ) , there was a substantial variation of the maximal ΔF/F0 among the AZs , which was also comparable between AZs of both genotypes ( c . v . = 0 . 75 for RBEKO/KO vs . c . v . = 0 . 80 for RBEWT/WT ) . Next , we analyzed the voltage-dependence of activation for the synaptic Ca2+-influx as previously described ( Ohn et al . , 2016 ) . Analysis of fractional activation revealed a depolarized shift in Vh by on average 5 mV in RBEKO/KO IHCs ( −22 . 76 ± 1 . 25 mV , S . D . = 9 . 26 mV; n = 55 AZs in 15 IHCs , N = 7 for RBEKO/KO vs . −27 . 37 ± 0 . 90 mV , S . D . = 7 . 48 mV; n = 68 AZs in 15 IHCs , N = 8 for RBE WT/WT; p=0 . 0029 , t-test; Figure 6D , Di ) , while the slope factor of voltage-dependent activation was unaltered ( p=0 . 42 , t-test , Figure 6Dii ) . Such a shift in the operating range of synaptic Ca2+-influx is expected to alter spontaneous and sound-evoked transmitter release ( see below and Ohn et al . , 2016 ) . Finally , we studied the spatial extent of the synaptic Ca2+-signals and estimated Full Width Half Maximum ( FWHM ) by fitting 2D Gaussian functions to the hotspots of Ca2+-indicator fluorescence and found a greater spread of Ca2+-signals at RBEKO/KO AZs ( Figure 6E , F; long axis ( L . A . ) = 1317 ± 49 nm , S . D . = 384 nm , short axis ( S . A . ) = 906 ± 36 nm , S . D . = 284 nm; n = 61 AZs in 15 IHCs , N = 7 vs . L . A . = 1083 ± 33 nm , S . D . = 283 nm; ( p=0 . 00016 , t-test ) , S . A . = 793 ± 27 nm , S . D . = 233 nm , ( p=0 . 0029 , t-test ) ; n = 74 AZs in 15 IHCs , N = 8 for RBEWT/WT ) . This larger spread of the presynaptic Ca2+-signals is in agreement with the presence of several CaV1 . 3-immunofluorescent clusters at RBEKO/KO synapses . In order to exclude lower IHC Ca2+-buffering to contribute to the observed larger spread of presynaptic Ca2+-signals , we performed semi-quantitative immunofluorescence analysis for the three major cytosolic Ca2+-buffers , the EF-hand Ca2+-binding proteins parvalbumin-α , calretinin and calbindin-28k ( Pangršič et al . , 2015 ) . We did not find any significant differences in their immunofluorescence intensity between IHCs of both genotypes ( in arbitrary units , parvalbumin intensity: 2 . 24 ± 0 . 15 , S . D . = 1 . 04 for RBEKO/KO vs . 1 . 88 ± 0 . 15 , S . D = 1 . 01 for RBEWT/WT , p=0 . 08; calbindin intensity: 0 . 82 ± 0 . 06 , S . D . = 0 . 43 for RBEKO/KO vs . 0 . 95 ± 0 . 07 , S . D . = 0 . 49 for RBEWT/WT , p=0 . 23; calretinin intensity: 0 . 91 ± 0 . 04 , S . D . = 0 . 26 for RBEKO/KO vs . 0 . 82 ± 0 . 04 , S . D . = 0 . 28 for RBEWT/WT , p=0 . 09; n = 49 cells and N = 4 for both conditions , Mann-Whitney-Wilcoxon test for all; Figure 6—figure supplement 1 ) . The ribbon has been proposed to play a crucial role in the exocytosis of SVs at the IHC AZ ( Khimich et al . , 2005 ) . Therefore , we monitored stimulated exocytosis of SVs with perforated-patch whole-cell recordings of exocytic membrane capacitance changes ( ΔCm ) . Using IHCs from 2/3-week-old RBEWT/WT and RBEKO/KO mice , we found that ΔCm in response to step-depolarizations to the potential that elicits maximal Ca2+-influx ( −14 mV ) were not different between IHCs with or without ribbons . Both , fast exocytosis elicited by depolarizations of up to 20 ms , attributed to the fusion of the readily releasable pool of SVs ( RRP , Moser and Beutner , 2000 ) , and longer stimuli , thought to reflect sustained exocytosis , ongoing SV replenishment and fusion , were unaltered in RBEKO/KO IHCs ( Figure 7A , B , C ) . On average , ΔCm induced by 20 ms long maximal Ca2+-influx was 16 . 70 ± 1 . 67 fF ( S . D . = 5 . 80 fF; n = 12 cells , N = 7 ) for RBEKO/KO compared to 15 . 22 ± 0 . 98 fF ( S . D . = 3 . 26 fF; n = 11 cells , N = 8 ) for RBEWT/WT . Exocytic ΔCm elicited by 200 ms long maximal Ca2+-influx ( same IHCs as for 20 ms ) , on average , amounted to 62 . 09 ± 5 . 40 fF ( S . D . = 18 . 70 fF ) for RBEKO/KO versus 63 . 28 ± 6 . 64 fF ( S . D . = 22 . 04 fF ) for RBEWT/WT . Moreover , trains of 20 step-depolarizations to −17 mV of 20 ms pulse duration did not reveal impaired exocytosis in RBEKO/KO IHCs , even when the inter-stimulus interval time was as short as 160 ms ( Figure 7D; n = 11 cells , N = 5 for RBEWT/WT and n = 13 cells , N = 8 for RBEKO/KO ) . We further explored RRP recovery from partial depletion using a paired-pulse protocol ( two strong 20 ms depolarizations to −14 mV separated by 50 , 110 , 260 and 510 ms inter-pulse intervals; Figure 7E , F ) . RRP recovery , estimated as the ΔCm ratio of the second and the first pulse , was not altered in RBEKO/KO IHCs at least when probing RRP exocytosis with maximal Ca2+-influx from a hyperpolarized resting potential ( Figure 7F ) . These data are in strong contrast to our previous findings in IHCs of bassoon mutant mice , which we had equivalently analyzed . There , the loss of synaptic ribbons , combined with a loss of functional bassoon resulted in profound deficits in exocytosis ( Khimich et al . , 2005; Frank et al . , 2010; Jing et al . , 2013 ) . Given the finding of a small depolarized shift in the operating range of Ca2+-channels in RBEKO/KO IHCs ( Figure 6D ) , we also probed the voltage-dependence of ΔCm elicited by 100 ms step-depolarizations ( Figure 7G , H ) . In agreement with the results obtained at maximal Ca2+-influx , we did not find significant differences in ΔCm for stronger depolarizations ( e . g . pulses to −39 mV elicited an average ΔCm of 20 . 67 ± 7 . 46 fF , S . D . = 23 . 58 fF , nmin = 10 IHCs , N = 9 for RBEKO/KO vs . 24 . 12 ± 4 . 04 fF , S . D . = 13 . 98 fF , nmin = 10 IHCs , N = 9 for RBEWT/WT; p=0 . 20; Mann-Whitney-Wilcoxon test ) . However , for weaker depolarizations in the range of physiological receptor potentials ( Russell and Sellick , 1983 ) , we observed a subtle but significant reduction in exocytosis for RBEKO/KO IHCs ( Figure 7H , p=0 . 0115 , p=0 . 0295 and p=0 . 1321 for −45 , –43 and −41 mV; without definitive outliers as determined by Graphpad Prism: p=0 . 0017 , p=0 . 0042 and p=0 . 0489 , respectively; Mann-Whitney-Wilcoxon test for all ) . For instance , depolarization to −45 mV elicited a ΔCm of 4 . 79 ± 2 . 26 fF for RBEKO/KO ( S . D . = 7 . 14 fF; nmin = 10 cells , N = 9 ) compared to 9 . 85 ± 1 . 60 fF for RBEWT/WT ( S . D . = 5 . 05 fF; nmin = 10 cells , N = 8 ) . The Ca2+-current integral ( Ca2+-charge , QCa ) , as well , tended to be reduced for RBEKO/KO IHC at these mild depolarizations , which , however , did not reach statistical significance ( e . g . QCa for −45 mV: 3 . 90 ± 0 . 49 pC , S . D . = 1 . 54 pC for RBEKO/KO vs . 5 . 15 ± 0 . 54 pC , S . D . = 1 . 72 pC for RBEWT/WT; p=0 . 1053; t-test ) . In summary , we found exocytosis to be unaltered for strong depolarizations but mildly decreased for more physiological stimuli in RBEKO/KO IHCs , which is in line with the findings of the companion paper by Becker et al . . Next , we studied sound encoding in RBEKO/KO mice in vivo . First , we recorded auditory brainstem responses ( ABR ) and found a significant reduction in the amplitude of wave I that reflects the SGN compound action potential ( 1 . 14 ± 0 . 13 µV , S . D . = 0 . 38 µV , N = 10 for RBEKO/KO vs . 3 . 30 ± 0 . 51 µV , S . D . = 1 . 54 µV , N = 10 for RBEWT/WT , p=0 . 0007 , NPMC test ) . This indicates less synchronous SGN activation in the absence of synaptic ribbons ( Figure 8A , B ) . The subsequent ABR waves ( Figure 8—figure supplement 1 ) were normal in amplitude ( waves II , IV and V , while wave III was reduced ) indicating a degree of central auditory compensation for the sound encoding deficit , for example via coincidence detection of converging SGN input in the cochlear nucleus ( Joris et al . , 1994; Strenzke et al . , 2009 ) . We found a non-significant trend of ABR threshold to be increased across all frequencies in RBEKO/KO mice ( approximately 10 dB across all frequencies , Figure 8C; refer to the companion paper Becker et al . showing significantly increased ABR-thresholds based on a larger sample , N = 28 RBEKO/KO mice vs . 22 RBEWT/WT mice ) . Cochlear amplification , probed by recordings of distortion product otoacoustic emissions ( DPOAE , Figure 8D ) , was intact in RBEKO/KO mice . Additionally , RBEWT/KO mice showed no significant changes in ABR wave I amplitudes and ABR thresholds ( Figure 8 ) , suggesting that the subtle morphological differences observed for afferent synapses of RBEWT/KO IHCs by electron and confocal-immunofluorescence microscopy did not turn into a deficit of sound coding measurable by ABR recordings . The wave I amplitude reduction and ABR threshold elevation were much less pronounced than in bassoon mutant mice ( Khimich et al . , 2005; Buran et al . , 2010; Jing et al . , 2013 ) . We then turned to in vivo extracellular recordings from single auditory neurons by targeting glass microelectrodes to where the auditory nerve enters the anteroventral cochlear nucleus ( AVCN ) in the brainstem ( Taberner and Liberman , 2005; Jing et al . , 2013 ) . ‘Putative’ SGNs ( hereafter dubbed SGN for simplicity ) were identified based on the depth of electrode position and their firing response to pure-tone stimulation ( primary-like peristimulus time histogram and latency , Figure 9 ) and analyzed in separation from ‘putative’ cochlear nucleus neurons ( Figure 10 ) . Since all firing of the individual SGN is thought to be driven by transmitter release from a single IHC AZ ( Heil et al . , 2007; Liberman , 1978; Robertson and Paki , 2002 ) , these recordings provide insight into single AZ function . We first assessed the spontaneous firing activity and found an increased abundance of SGNs with low spontaneous firing rates in RBEKO/KO mice ( 72% with rates < 10 Hz , n = 43 SGNs , N = 9 vs . 50% in RBEWT/WT , n = 40 SGNs , N = 8; p=0 . 0267 , Kolmogorov-Smirnov test; Figure 9A ) . Frequency tuning was intact in RBEKO/KO SGNs ( Figure 9B ) : the sharpness of tuning expressed by the Q10dB ( width of tuning curve 10 dB above threshold at the characteristic frequency ( Cf ) normalized by Cf ) was comparable ( mean: 9 . 28 ± 1 . 01 , S . D . = 6 . 32 and median: 7 . 41 for RBEKO/KO SGNs , n = 39 SGNs , N = 9 vs . mean: 12 . 50 ± 1 . 98 , S . D . = 11 . 91 and median: 8 . 36 for RBEWT/WT SGNs , n = 36 SGNs , N = 9; p=0 . 28 , Mann-Whitney-Wilcoxon test ) . However , the sound threshold at Cf was significantly elevated by almost 20 dB in RBEKO/KO mice ( 35 . 60 ± 3 . 45 dB SPL , S . D . = 22 . 66 dB SPL for RBEKO/KO SGNs , n = 43 SGNs , N = 9 vs . 16 . 05 ± 2 . 47 dB SPL , S . D . = 15 . 42 dB SPL for RBEWT/WT SGNs , n = 39 SGNs , N = 9 , p<0 . 0001 , Mann-Whitney-Wilcoxon test; Figure 9C ) . Given the normal frequency tuning and DPOAE , this threshold increase seems unlikely to result from a putative functional cochlear deficit upstream of the IHCs . Next , we studied the firing response of SGNs to 50 ms tone bursts ( at Cf and 30 dB above sound threshold , 200 ms inter-stimulus interval ) , which is governed by the presynaptic glutamate release and postsynaptic spike generation . The peak firing rate at sound onset is thought to reflect the initial rate of release from the SV-occupied release sites of the RRP ( ‘standing RRP’ , [Oesch and Diamond , 2011; Pangršič et al . , 2012] ) . Refractoriness and the decline of release rate due to partial depletion of the standing RRP likely dominate the subsequent spike rate adaptation . Finally , the adapted firing rate reports SV replenishment and subsequent fusion ( reviewed in Pangršič et al . , 2012; Rutherford and Moser , 2016 ) . We observed reduced spike rates of SGNs from RBEKO/KO mice ( Figure 9D , E ) both at sound onset ( p=0 . 0001 , n = 39 SGNs , N = 8 in RBEKO/KO and n = 38 SGNs , N = 9 in RBEWT/WT , t-test ) and after short-term adaptation ( p=0 . 0023 , Mann-Whitney-Wilcoxon test ) . Both , peak and adapted rates were similarly affected by the RIBEYE-disruption , indicated by the scatter plot of peak vs . adapted rates ( Figure 9E ) . A significant peak rate reduction was also observed at higher stimulation frequencies ( 10 Hz , Figure 9G–H ) . The spike rates were better preserved in RBEKO/KO SGNs than in SGNs of bassoon mutant mice ( BsnΔex4/5 data of Jing et al . ( 2013 ) , purple data in Figure 9G–I ) . We approximated adaptation within the 50 ms response by single-exponential fitting since double exponential fitting did not regularly report two temporally discernible components in RBEKO/KO SGNs . The mean apparent adaptation time constant reported by single-exponential fitting were significantly slowed in RBEKO/KO SGNs ( 9 . 83 ± 0 . 50 ms , S . D . = 2 . 85 ms , median: 10 . 46 ms , n = 32 SGNs , N = 8 ) as compared to RBEWT/WT SGNs ( 8 . 71 ± 0 . 50 ms , S . D . = 3 . 05 ms , median: 8 . 73 ms , n = 37 SGNs , N = 9 , p=0 . 033 , Mann-Whitney-Wilcoxon test ) . The results of double-exponential fitting of RBEWT/WT and RBEKO/KO SGNs support the slowed adaptation kinetics and are presented in Table 1 . As expected for the reduced peak firing rate , we found prolonged first spike latency which also showed greater temporal jitter ( Figure 9F ) . The reduced peak firing rate together with increased first spike latency jitter likely explain the reduction in ABR wave I amplitude . The firing of putative AVCN neurons was better preserved: putative bushy cells showed normal sound driven rates and chopper cells only a mild reduction in peak rate ( Figure 10 ) . Next , we explored the encoding of sound intensity by estimating the mean firing rate during 50 ms tone bursts at different sound pressure levels . These ‘rate-level functions’ ( Figure 11A ) indicated that the spike rate increase with the sound pressure level ( p=0 . 068 , n = 24 SGNs , N = 8 in RBEKO/KO and n = 19 SGNs , N = 7 in RBEWT/WT , Mann-Whitney-Wilcoxon test , Figure 11—figure supplement 1A ) and the dynamic range of sound coding ( sound pressure level for which the spike rate changes from 10–90% , Figure 11—figure supplement 1B , p=0 . 3044 , t-test ) were not significantly altered . We then used transposed tones ( Cf at 500 Hz modulation frequency ) in order to probe for the temporal fidelity and reliability of firing in RBEKO/KO SGNs in the steady state ( Figure 11D ) . These experiments corroborated the reduced maximal firing rate of RBEKO/KO SGNs ( n = 22 SGNs , N = 7 in RBEKO/KO and n = 15 SGNs , N = 6 in RBEWT/WT , p<0 . 0001 , t-test ) and indicated that the temporal precision of sound coding is impaired also in the steady state ( reduced Synchronization Index: p=0 . 0043 , t-test ) . In order to further scrutinize the potential role of the synaptic ribbon in vesicle replenishment , we studied the response to prolonged tone-stimulation ( Figure 11B , 500 ms at Cf and 30 dB above threshold , 2 s inter-stimulus interval ) . The peak rate was better preserved in RBEKO/KO SGNs than seen with shorter inter-stimulus interval ( e . g . 200 ms , Figure 9C ) , likely reflecting more complete SV-replenishment ( i . e . larger standing RRP ) owing to the longer recovery interval ( 2 s vs . 200 ms ) . However , the adapted spike rate of RBEKO/KO SGNs was even more reduced than found with 50 ms tone bursts ( to about half of that for RBEWT/WT SGNs , Figure 11B , C ) highlighting the impaired SV replenishment during prolonged stimulation . Finally , we evaluated presynaptic vesicle pool dynamics by recording and modeling responses to forward masking protocols ( Harris and Dallos , 1979 ) , that are thought to reflect depletion and recovery of the RRP ( Figure 11E , F; Figure 11—figure supplement 1C ) . We approximated the recovery from forward masking by single exponential fitting ( Figure 11—figure supplement 1C ) to provide an estimate of the kinetics of vesicle pool replenishment . The time constant of recovery was prolonged in RBEKO/KO SGNs ( 90 . 80 ± 8 . 66 ms , S . D . = 45 . 00 ms , n = 27 SGNs , N = 8 vs . 33 . 53 ± 5 . 74 ms , S . D . = 28 . 11 ms , in RBEWT/WT SGNs , n = 24 SGNs , N = 7 , p<0 . 0001 , Mann-Whitney-Wilcoxon test ) indicating slowed RRP replenishment in the absence of the ribbon . We noted that SGNs showed considerably lower spontaneous and evoked rates during the forward masking paradigm compared to other stimulus protocols ( compare Figure 11B E ) , likely due to enhanced RRP depletion with the more extended stimulation in this protocol . Amplitude and waveform of the forward masking responses were consistent with a two-fold reduction of the number of contributing vesicular release sites of the RRP ( N’slot: contributing release sites during forward masking , Nslot: the contributing release sites for the same AZs during tone bursts at 5 Hz stimulation ) . The ratio Nslots/N’slots was estimated to be 2 . 2 for RBEKO/KO and 2 . 3 for RBEWT/WT from the drop in spontaneous and evoked SGN spiking rates . We used a previously developed biophysical model of RRP dynamics and spike generation ( Frank et al . , 2010; Jung et al . , 2015b ) to extract information on fusion and replenishment rate constants as well as the Nslot by fitting the responses to tone bursts , that is PSTHs with 100 , 200 and 2000 ms inter-stimulus interval , and also the forward masking spiking data across all recovery intervals ( Figure 11E ) . Importantly , only the first 50 ms of the 500 ms stimulus response during the PSTH 2000 were included in the fit; any later adaptation processes were disregarded , as they were not accounted by the model’s equations . The results of model fitting suggested that during the forward masking only about half of all release sites ( Nslots ) were engaged in the response ( N’slots ) . Throughout , RBEKO/KO SGNs showed a lower fusion rate than the RBEWT/WT SGNs , reflecting the reduced onset response in RBEKO/KO SGNs . When more recovery time was allowed , that is in the 0 . 5 Hz tone burst and the forward masking , where recovery times from around 250 to 500 ms occured between the probe and subsequent masker , the onset response improved in RBEKO/KO . Consequently , the estimated fusion rate almost reached the level of RBEWT/WT in the forward masking fits . With the scaling factors of approximately 2 , the estimates for the number of release sites were consistent between tone bursts and forward masking data , and in both cases only slightly smaller for the RBEKO/KO ( Table 2 ) . Taken together the fits from forward masking and tone bursts suggest that the total number of release sites ( RRP ) was only slightly reduced at ribbonless synapses of RBEKO/KO IHCs . However , a strong firing response at sound onset , that is release of a large standing RRP , required longer recovery indicating more efficient SV replenishment in the presence of ribbons , which is reflected in the larger refilling rate constants estimated by the model of RBEWT/WT synapses .
Our work confirms the central role of RIBEYE for forming synaptic ribbons ( Schmitz et al . , 2000; Magupalli et al . , 2008; Maxeiner et al . , 2016 ) . We did not observe structures reminiscent of ‘ghost ribbons’ reported for ribeye mutants in zebrafish neuromast hair cells ( Lv et al . , 2016 ) in IHCs of RBEKO/KO mice . These ghost ribbons were characterized as a halo of synaptic vesicles around a non-electron-dense area that resembled in size , though smaller , and shape to a synaptic ribbon . In zebrafish , two gene copies of ribeye ( ribeye a and b ) exist , making it harder to achieve a complete knock-out ( Lv et al . , 2016; Van Epps et al . , 2004 ) . In keeping with this notion , Lv et al . found residual immunofluorescence of ribeye a in the double mutants . Hence , we speculate that residual RIBEYE , possibly together with other scaffold proteins such as piccolo , might have formed the observed electron-translucent SV-framed structures ( Lv et al . , 2016 ) . In contrast , immunofluorescence , as well as electron microscopy , revealed the complete absence of RIBEYE and ribbons in IHCs of RBEKO/KO mice in our work and the companion study ( Becker et al . ) , which is in agreement with findings in the RBEKO/KO mouse retina ( Maxeiner et al . , 2016 ) . IHC synapses normally employ a single ribbon-type AZ . But in the absence of RIBEYE , there were typically two or more ribbonless AZs , akin to multiple conventional AZs ( Figure 2 ) . These ribbonless ‘conventional’ AZs at RBEKO/KO IHC synapses consist mostly of roundish PDs , each with a cluster of SVs , of which approximately one third were directly adjacent to the plasma membrane ( membrane-proximal: MP-SVs ) . Using electron tomography we found that about two-thirds of the MP-SVs were tethered to the AZ membrane , which was comparable to RBEWT/WT AZs ( Figure 3 ) . We speculate that SVs associated with the PD , but not facing the membrane ( PDA-SVs ) , serve to replenish the release sites once tethered MP-SVs fused , and that the ribbonless PD more likely acts in long-range SV tethering to the AZ in analogy to what is considered for conventional AZs ( Cole et al . , 2016; Fernández-Busnadiego et al . , 2013; Siksou et al . , 2007 ) . We assume that absence of RIBEYE does not alter SV size since electron tomography , which provides the most reliable estimation of SV size , did not reveal differences in SV diameter between RBEKO/KO and RBEWT/WT AZs , at least when considering all SVs . The RBEKO/KO PDs , like in RBEWT/WT , contained bassoon , CaV1 . 3 , and RIM2 , but lacked piccolino which is likely part of the ribbon in RBEWT/WT ( Figure 1 ) ( Dick et al . , 2001; Khimich et al . , 2005; Limbach et al . , 2011; Regus-Leidig et al . , 2013 ) . To some extent , the multi-AZ morphology is reminiscent of IHC synapses prior to synaptic maturation ( Huang et al . , 2012; Sendin et al . , 2007; Sobkowicz et al . , 1982; Wong et al . , 2014 ) . While we cannot rule out some sort of developmental delay of RIBEYE-deficient IHCs , we suspect that the morphological transformation into a multi-AZ morphology reflects a compensatory effort . Reasons for our interpretation include ( i ) the same morphological phenotype of RBEKO/KO IHCs at 8 months of age ( Figure 2 ) , ( ii ) the finding of highly regular PDs at RBEKO/KO IHCs synapses , which differs from less well-defined PDs at immature AZs ( Wong et al . , 2014 ) , ( iii ) the typical continuous and large PSD of RBEKO/KO IHCs synapses ( Figure 2 , see also the accompanying paper by Becker et al . ) as a characteristic of a mature synapse , rather than the several smaller PSD patches at developing IHC synapses ( Wong et al . , 2014 ) , and ( iv ) the synaptically confined CaV1 . 3 Ca2+-channel clusters , normal amplitude of IHC ICa and mature amplitude of synaptic Ca2+-signals , rather than massive extrasynaptic CaV1 . 3 abundance and larger whole-cell ICa but smaller synaptic Ca2+-signals in immature IHCs ( Wong et al . , 2014; Zampini et al . , 2010 ) . The multi-AZ morphology of the RBEKO/KO IHC synapses was also corroborated by high- and super-resolution microscopy of bassoon and CaV1 . 3 immunofluorescence ( Figure 4 ) . The organization in several smaller Ca2+-channel clusters likely explains the broader spread of the presynaptic Ca2+-signal at RBEKO/KO synapses ( Figure 6 ) . In contrast to bassoon mutant mice ( Frank et al . , 2010; Jing et al . , 2013 ) , the number of synaptic Ca2+-channels was not reduced in RBEKO/KO mice as shown by normal amplitudes of whole-cell ICa and synaptic Ca2+-signals . Therefore , the loss of synaptic Ca2+-channels from the bassoon-deficient ribbonless IHC synapses , indicates a role of bassoon in promoting Ca2+-channel tethering at the AZ likely via interaction with RIM-binding protein ( Davydova et al . , 2014 ) , which was previously shown to interact with CaV1 . 3 Ca2+-channels ( Hibino et al . , 2002 ) and is required for establishing a normal Ca2+-channel complement of the IHC AZ ( Krinner et al . , 2017 ) . Interestingly , we observed changes in Ca2+-channel function in RBEKO/KO IHCs: the voltage-dependence of Ca2+-channel activation was slightly , but significantly , shifted to more depolarized potentials both at the levels of whole-cell Ca2+-current ( Vh +2 mV ) and synaptic Ca2+-influx at individual synapses ( Vh +5 mV ) ( at 5 mM [Ca2+]e , Figure 5 and 6 ) . Similar as in this study , an enhanced inactivation ( Figure 5 ) of ICa was also found in bassoon-deficient IHCs , while their Vh was actually mildly shifted in the opposite direction ( −3 mV for imaging of synaptic Ca2+ ) and unaltered at the level of the whole-cell lCa ( Frank et al . , 2010 ) . One potential reason for why the depolarized Vh-shift of the synaptic Ca2+-influx was greater than that of the whole-cell Ca2+-influx is the contribution of extrasynaptic Ca2+-channels to the whole-cell Ca2+-influx . They are thought to contribute approximately 30% of the Ca2+-influx ( Brandt et al . , 2005 ) and are not regulated by RIBEYE/ribbon . In order to test whether the depolarized Vh-shift of synaptic Ca2+ translates into changes in transmitter release , we recorded exocytic ΔCm for different depolarization potentials . A small , but significant reduction of exocytosis for weak depolarizations in RBEKO/KO IHCs ( Figure 7 , seen also in the accompanying paper by Becker et al . ) suggests that the Vh shift is relevant for hair cell transmission ( see also below ) . How enhanced ICa inactivation might affect sound encoding is addressed by work on Ca2+-binding proteins ( CaBPs ) that are thought to antagonize calmodulin’s role in mediating ICa inactivation ( Lee et al . , 1999; Peterson et al . , 1999 ) . Among the several CaBPs expressed in IHCs ( Cui et al . , 2007; Picher et al . , 2017; Schrauwen et al . , 2012; Yang et al . , 2006 ) , CaBP2 is defective in human genetic hearing loss DFNB93 ( Picher et al . , 2017; Schrauwen et al . , 2012 ) and required for hearing likely via inhibition of IHC ICa inactivation ( Picher et al . , 2017 ) . However , deletion of CaBP4 in mice caused only a very mild increase of ICa inactivation similar to the one found here and did not alter auditory brainstem responses ( Cui et al . , 2007 ) . Future studies need to address how RIBEYE/ribbons mechanistically regulate the function and spatial organization of Ca2+-channels . Over some decades , research on retinal photoreceptors and bipolar cells , on hair cells of the inner ear and the lateral lines , on electroreceptors as well as pineal cells , has aimed to elucidate the function ( s ) of the synaptic ribbon . Current hypotheses state that the ribbon functions in ( i ) replenishing release sites ( [Bunt , 1971; Frank et al . , 2010; von Gersdorff et al . , 1996; Lenzi et al . , 2002; Maxeiner et al . , 2016; Snellman et al . , 2011; Vaithianathan and Matthews , 2014] for a deviating view see {Jackman et al . , 2009] ) , potentially by facilitated diffusion of SVs on the ribbon surface towards the site of consumption ( Graydon et al . , 2014 ) and SV priming ( Grabner and Zenisek , 2013; Snellman et al . , 2011 ) , ( ii ) establishing a large complement of vesicular release sites and Ca2+-channels at the active zone ( Frank et al . , 2010; Khimich et al . , 2005 ) , which remained hard to disentangle from potential function of bassoon ( Frank et al . , 2010; Jing et al . , 2013 ) , ( iii ) ensuring close spatial coupling of Ca2+-channels and vesicular release sites ( Maxeiner et al . , 2016 ) or enhancing presynaptic Ca2+-signals by limiting diffusional Ca2+-spread ( Graydon et al . , 2011 ) , ( iv ) contributing to multivesicular release ( Graydon et al . , 2011; Jing et al . , 2013; Mehta et al . , 2013 ) , and ( v ) contributing to SV reformation from endocytosed membranes ( Jung et al . , 2015b; Khimich et al . , 2005; Schwarz et al . , 2011 ) . Clearly , SV-replenishment was impaired at the ribbon-less IHC synapses of RBEKO/KO mice . This is shown by slowed recovery from forward-masking and the use-dependent reduction of peak and adapted firing rates , which we further scrutinized by modeling . Therefore , our study supports a role of the ribbon in vesicle replenishment , which is also found in the accompanying paper by Becker et al . . Why RRP-recovery was not significantly altered when probed with pairs 20 ms long maximal Ca2+-influx by membrane capacitance measurements in IHCs ( Figure 7 ) will need to be addressed in future studies , ideally using paired pre-and postsynaptic recordings of synaptic transmission with depolarizations of varying strength . Each of the ribbon-manipulations employed to analyze its role has strengths , but also weaknesses , such as changes in other AZ proteins and long-term compensatory processes ( e . g . bassoon deletion ) , complex manipulations ( e . g . diurnal changes or hibernation: [Hull et al . , 2006; Mehta et al . , 2013; Spiwoks-Becker et al . , 2004] ) and photoablation ( Mehta et al . , 2013; Snellman et al . , 2011 ) . Genetic RIBEYE manipulations ( Lv et al . , 2016; Maxeiner et al . , 2016; Sheets et al . , 2011; Van Epps et al . , 2004 ) have the greatest molecular specificity , but in some cases , were incomplete , and to some extent masked by compensation . In fact , our study of IHCs , unlike the situation for bipolar cell retinal ribbons ( Maxeiner et al . , 2016 ) , suggests that some features of the IHC ribbon-type AZ can be very well replaced by a ribbonless multi-AZ morphology: the synaptic complement of Ca2+-channels and SVs , as well as exocytic ΔCm elicited by strong depolarizations , were similar . Therefore , we likely underestimated the role of the ribbon in sound encoding in our present study . For sound encoding at the afferent synapses between IHCs and SGNs , we observed some commonalities and differences with the bassoon mutants and RBEKO/KO mice , whereby the stronger phenotype of bassoon mutants suggests additive effects of bassoon and ribbon loss . Recordings from single SGNs indicate reduced peak and adapted release rates at the IHC synapses of ribbonless synapses , as well as impaired temporal precision of coding . High temporal precision is a hallmark of synaptic sound encoding ( e . g . ( Köppl , 1997 ) . Reduced release rates or smaller EPSC sizes would increase the temporal jitter ( Buran et al . , 2010; Li et al . , 2014; Rutherford et al . , 2012; Wittig and Parsons , 2008 ) . Reduced spike rates and increased jitter of release likely explain the reduced ABR wave I amplitude in both mutants . A striking difference from bassoon mutants , however , is that sound encoding in RBEKO/KO mice was impaired substantially , despite unaltered exocytic ΔCm upon strong stimulation . We propose two mechanisms with likely additive effects to explain this surprising finding: i ) the small depolarized shift of synaptic Ca2+-channels might contribute the lower spontaneous and evoked firing rates as well as higher sound thresholds of SGNs and ii ) the reduced SV-replenishment might not suffice to balance the rate of consumption leading to a smaller standing RRP in vivo , while the arrest of exocytosis in the voltage-clamped IHCs for tens of seconds likely enables complete filling of the release sites ( max . standing RRP ) . The changes in Ca2+-channel gating observed in IHCs of RBEKO/KO mice were unexpected , as so far , a direct or indirect interaction of RIBEYE and Ca2+-channels have not been described . Clearly , future studies , including studies on the potential regulation of CaV1 . 3 Ca2+-channels by RIBEYE and piccolino in heterologous expression systems , paired pre- and postsynaptic recordings , as well as further computational modeling , are required .
Constitutive RIBEYE knockout mice ( RBEKO/KO derived from Ctbp2tm1 . 2Sud by Cre-recombination ) were generated by Maxeiner and colleagues ( Maxeiner et al . , 2016 ) and were back-crossed to C57BL/6 for five generations ( corresponding to a C57BL/6 background contribution of >95% ) . All experiments complied with national animal care guidelines and were approved by the University of Göttingen Board for Animal Welfare and the Animal Welfare Office of the State of Lower Saxony ( permit number: 14–1391 ) . The colony was maintained by mating heterozygous mice . Whenever possible , experiments were performed in parallel on mutant mice and their wildtype littermates . However , the experimental schedule did not always permit this and we occasionally used individual mice from the same colony but without littermate controls . Moreover , for some experiments giving rise to Figure 6—figure supplement 1 and Figure 7E , F , G , H we also used C57Bl/6 wild-type mice and combined their results with those of wildtype littermate controls . The apical 2/3 turns of organs of Corti from P14 to P28 aged mice were freshly dissected in HEPES Hank's solution containing ( in mM ) : 5 . 36 KCl , 141 . 7 NaCl , 10 HEPES , 0 . 5 MgSO4-7H2O , 1 MgCl2-6H2O , 5 . 6 D-glucose , and 3 . 4 L-glutamine ( pH 7 . 2 , ~300 mOsm ) . The basolateral membranes of the IHCs were exposed by carefully removing the surrounding cells with a suction pipette . All experiments were conducted at room temperature ( 20–25°C ) . Perforated-patch-clamp recordings were performed as described previously ( Moser and Beutner , 2000 ) . The pipette solution contained ( in mM ) : 130 Cs-gluconate , 10 tetraethylammonium ( TEA ) -Cl , 10 4-AP , 10 HEPES , 1 MgCl2 , as well as 300 μg/ml amphotericin B ( pH 7 . 2 , ~280 mOsm ) . The extracellular solution contained the following ( in mM ) : 112 NaCl , 35 TEA-Cl , 2 . 8 KCl , 1 MgCl2 , 1 CsCl , 10 HEPES , 2 CaCl2 , and 11 . 1 D-glucose ( pH 7 . 2 , ~305 mOsm ) . External solution change was done by bath exchange through a perfusion system . For live-imaging , the patch pipette solution contained ( in mM ) : 111 Cs-glutamate , 1 MgCl2 , 1 CaCl2 , 10 EGTA , 13 TEA-Cl , 20 HEPES , 4 Mg-ATP , 0 . 3 Na-GTP and 1 L-Glutathione ( pH 7 . 3 , ~290 mOsm ) . To visualize the Ca2+-hotspots and the ribbons , the Ca2+-indicator Fluo-4FF penta-K+ salt ( 0 . 8 mM , Life Technologies , Germany ) and the TAMRA-conjugated CtBP2/RIBEYE-binding dimer peptide ( 10 µM , Biosynthan , Germany ) were added to the intracellular solution . The extracellular solution contained the following ( in mM ) : 2 . 8 KCl , 102 NaCl , 10 HEPES , 1 CsCl2 , 1 MgCl2 , 5 CaCl2 , 35 TEA-Cl , and 11 . 1 D-Glucose ( pH 7 . 2 , ~300 mOsm ) . EPC-10 amplifiers controlled by Patchmaster or Pulse software ( HEKA Elektronik , Germany ) were used for the measurements . IHCs were held at −87 mV or −69 mV . All voltages were corrected for liquid junction potential offline ( 17 mV or 14 mV , depending on intra- and extracellular solutions used ) and voltage-drops across the series resistance ( Rs ) . Currents were leak corrected using a p/10 protocol in exocytosis experiments . Recordings were discarded when the leak current exceeded −55 pA , Rs exceeded 30 MΩ ( for perforated-patch ) or 15 MΩ within 4 min after break-in ( for ruptured-patch ) , or Ca2+-current rundown exceeded 25% . All passive electrical properties of the patch-clamp recording experiments are detailed in Supplementary file 2 . Exocytosis was studied by measuring the membrane capacitance increments ( ΔCm ) using the Lindau-Neher technique ( Lindau and Neher , 1988 ) . Cells were stimulated by step depolarizations of different durations to −17 or −14 mV , or by 100 ms pulses to voltages ranging from −53 to −37 mV . A resting interval of 10–100 s between the stimuli was used . Each protocol was applied two to three times and only IHCs with reproducible exocytosis during the rounds were included . For display , traces were subjected to 1 , 5 or 10 pass Binomial Smoothing using Igor Pro . Current-voltage relationships ( ‘IVs’ ) displayed in Figure 5A ( ruptured-patch , 3-week-old mice ) were obtained by applying 20 ms depolarizing step pulses of increasing voltage from −82 to 63 mV in 5 mV steps . Ca2+-imaging was performed with a spinning disk confocal scanner ( CSU22 , Yokogawa , Germany ) mounted on an upright microscope ( Axio Examiner , Zeiss , Germany ) with 63x , 1 . 0 NA objective ( W Plan-Apochromat , Zeiss ) . Images were acquired by a scientific CMOS camera ( Neo , Andor , Germany ) . Ca2+-indicator F4FF and TAMRA-conjugated peptide were excited by diode-pumped solid-state lasers with 491 nm and 561 nm wavelength , respectively ( Cobolt AB ) . The spinning disk was set to 2000 rpm to synchronize with the 10 ms acquisition time of the camera . Using a piezo positioner for the objective ( Piezosystem , Germany ) , a scan of the entire cell was performed 4 min after breaking into the cell , taking sections each 0 . 5 µm at an exposure time of 0 . 5 s in the red ( TAMRA-peptide ) channel from the bottom to the top of the cell . In order to study the voltage-dependence of Ca2+-indicator fluorescence increments at the synapses , the confocal scans were acquired every 0 . 5 µm from the bottom to the top ribbon in the RBEWT/WT mice . For the RBEKO/KO mice , the scanning was done from the bottom of the cell to +12 µm , which on average corresponds to the bottom of the nucleus . Ca2+-currents were evoked by applying a voltage ramp stimulus from −87 to +63 mV during 150 ms ( 1 mV/ms ) in each focal plane . Simultaneously , fluorescence measurements were made in the green channel ( Fluo-4FF ) with a frame rate of 100 Hz . In order to overcome the limitations of the frame rate and increase the voltage resolution of the fluorescent signal acquired , the voltage ramp protocol was applied twice , once shifted by 5 ms such that for any given frame during the second ramp the voltage was shifted by 5 mV compared to the first stimulus . Alternating planes were acquired to avoid photobleaching encountered with the consecutive plane acquisition . Apical turns of organs of Corti from 3-week-old mice were prepared for ‘whole-mount imaging’ as described in ( Ohn et al . , 2016 ) . In brief , the samples were fixed either in formaldehyde ( 4% , 10 min on ice ) , or methanol ( 20 min at −20°C ) . Afterwards , the following primary antibodies were used: mouse anti-CtBP2 ( 1:200 , BD Biosciences , Germany 612044 ) , mouse anti-PSD-95 ( 1:200 , Sigma Aldrich , Germany P246-100ul ) , mouse anti-bassoon SAP7f407 ( 1:200 , Abcam , Germany , ab82958 ) , guinea pig anti-bassoon ( 1:500 , Synaptic Systems , Germany , 141 004 ) , rabbit anti-RIM2 ( 1:100 , Synaptic Systems 140 103 ) , rabbit anti-Cav1 . 3 ( 1:75 or 1:100 , Alomone Labs , Germany , ACC 005 ) , rabbit anti-piccolino ( 1:500 , kind gift of JH Brandstätter; see Regus-Leidig et al . , 2013 ) , guinea pig anti-parvalbumin α ( 1:1000 , Synaptic Systems , 195 004 ) , mouse anti-calbindin 28 k ( 1:500 , Swant , Germany , 07 ( F ) ) , and rabbit anti-calretinin ( 1:1000 , Swant 1893–0114 ) . Secondary antibodies used were Alexa Fluor 488 conjugated anti-rabbit , Alexa Fluor 488 conjugated anti-guinea-pig , Alexa Fluor 568 conjugated anti-mouse , and Alexa Fluor 647 conjugated anti-rabbit ( 1:200 , Invitrogen , Germany , A11008 , A11004 , A11073 , and A31573 respectively ) . For high resolution STED microscopy , STAR580 and STAR635p conjugated anti-rabbit and anti-mouse ( 1:200 , Abberior , Germany , 2-0002-005-1 , 2-0012-005-8 , 2-0002-007-5 , and 2-0012-007-2 ) have been used as secondary antibodies . Images were acquired using either a Leica SP5 with a 1 . 4 NA 63x oil immersion objective or an Abberior Instruments Expert Line STED microscope , with excitation lasers at 488 , 561 , and 633 nm and STED lasers at 595 nm , 1 W , and 775 nm , 1 . 2 W , using a 1 . 4 NA 100x oil immersion objective , either in confocal or in 2D-STED mode . Images were adjusted for brightness and contrast using ImageJ . ABR , DPOAE and extracellular recordings from single SGNs were performed essentially as described before ( Jing et al . , 2013; Strenzke et al . , 2016 ) . ABR and DPOAE recordings were performed on 6-week-old mice . For extracellular recordings from individual SGNs , 6 to 10 week-old mice were anesthetized by i . p . injection of urethane ( 1 . 32 mg/kg ) , xylazine ( 5 mg/kg ) and buprenorphine ( 0 . 1 mg/kg ) , a tracheostomy was performed and the mice were then placed in a stereotactic system . After partial removal of the occipital bone and cerebellum to expose the anteroventral cochlear nucleus ( AVCN ) , a glass microelectrode was advanced through the posterior AVCN portion to reach the auditory nerve . Acoustic stimulation was provided by an open field Avisoft ScanSpeak Ultrasonic Speaker ( Avisoft Bioacoustics , Germany ) , and ‘putative’ SGNs ( auditory nerve fibers formed by the central SGN axons ) were identified and distinguished from cochlear nucleus neurons based on their stereotactic position ( >1 . 1 mm from the surface of the cochlear nucleus ) , spontaneous and noise-burst induced firing , peristimulus time histogram ( PSTH ) , regularity of firing , and first spike latency . Recordings were performed using TDT system III hardware and an ELC-03XS amplifier ( NPI Electronics , Germany ) , offline analysis using waveform-based spike detection using custom-written MATLAB software ( Source code 1 ) . High-pressure freezing , freeze-substitution followed by electron tomography were essentially performed as described previously ( Vogl et al . , 2015; Jung et al . , 2015a ) . After freeze-substitution and embedding in epoxy resin ( Agar 100 kit , Plano , Germany ) , 250 nm semithin sections for electron tomography were obtained on an Ultracut E ultramicrotome ( Leica Microsystems , Germany ) with a 35° diamond knife ( Diatome , Switzerland ) . Sections were placed on 1% formvar-coated ( w/v in water-free chloroform ) copper 100 mesh grids ( ATHENE , Plano , Germany , 3 . 05 mm Ø ) and post-stained with UAR-EMS ( Science Services , Germany ) and Reynold’s lead citrate . For electron tomography , 10 nm gold particles ( British Bio Cell/Plano , Germany ) were applied to both sides of the stained grids . Single tilt series at 12 , 000-x magnification , mainly from −60 to +60° ( if only fewer angles were possible , the tomograms were only accepted for quantification if the quality was sufficient ) were acquired with an 1° increment at a JEM2100 ( JEOL , Germany ) ) transmission electron microscope at 200 kV using the Serial-EM software ( Mastronarde , 2005 ) . The tomograms were generated using the IMOD package etomo and models were generated using 3dmod ( Kremer et al . , 1996 ) . Live-imaging and IHC-patch-clamp data were analyzed using custom programs in Igor Pro 6 . 3 ( Wavemetrics , Portland , OR , USA; Source Code 2 ) . For analysis of IV-curves , the evoked Ca2+-current was averaged from 5 to 10 ms after the start of the depolarization . The total Ca2+-charge was estimated by taking the integral of the leak-subtracted current during the depolarization step . For most protocols , ΔCm was estimated as the difference between the mean of Cm 400 ms after and before the depolarization ( the initial 60 ms after the end of depolarization were skipped ) . For paired pulse experiments , the calculation of the mean of Cm before and after the depolarization was limited to the time remaining in the inter pulse interval after skipping ( the initial 30 ms after the end of depolarization were skipped ) . ΔF images were generated by subtracting the fluorescence intensity inside the cell at the resting state ( F0 , an average of 10 frames ) from the one at the depolarized state ( an average of 6 frames during voltage ramp protocol ) . ΔF was calculated as the average of a 3 × 3 pixel square placed in the region showing the greatest intensity increase within the fluorescence hotspot . Maximal ΔF ( ΔFmax ) was the average of 5 ΔF values obtained between −17 and +8 mV during the voltage ramp ( around the peak Ca2+-influx ) . Only AZs presenting a ΔFmax greater than the mean of the fluorescence intensity plus two standard deviations at rest were considered for further analysis . For analysis of the voltage dependence of synaptic Ca2+-signals , raw traces were fitted to the following ( 1 ) F ( V ) =F0+fv⋅ ( Vr−V ) 1+e ( Vh−V ) kwhere fv is the fluorescence-voltage-relationship ΔF/ΔV obtained by linear fitting to the FV-curve in the range of 3 to 23 mV , Vr the reversal potential of 65 . 6 mV , and V the command voltage , in order to obtain Vh , the voltage of half-maximal activation , and k , the slope factor . The spatial extent of the synaptic Ca2+-signals was estimated by fitting of a 2D Gaussian function to the fluorescent hotspot using a genetic fit algorithm ( Sanchez del Rio and Pareschi , 2001 ) to obtain the full width at half maximum in the long and short axis . For each spot , the calculations were made at those confocal sections where the intensity of the spot was strongest . Activation time constants of Ca2+-currents at differing potentials were obtained by fitting to the first 5 ms of the current traces the following equation: ( 2 ) f ( t ) =y0+A× ( 1−e ( −xτ ) ) 2 Confocal and STED immunofluorescence images were analyzed and z-projected with Fiji software and further analyzed using Igor Pro . For synapse counting , co-localized pre- and postsynaptic immunofluorescent spots were counted manually . The spatial extent of the line-shaped Ca2+-channel clusters was estimated by fitting a 2D Gaussian function to the individual clusters in 2D STED images to obtain the full width at half maximum in the long and short axis . The areas of the PSD were calculated by the following formula: area = π x ( Long Axis/2 ) ( Short Axis/2 ) . The semi-quantitative immunofluorescence analysis of the proteinaceous Ca2+-buffers was performed by calculating the mean immunofluorescence intensity of a volume ( 40 ( X ) x 40 ( Y ) x 4 ( Z ) voxels or 2 . 8 × 2 . 8 × 2 µm ) below the nucleus and above the synapses . This and the count and intensity of the CtBP2 immunofluorescent spots have been analyzed in Imaris 7 . 6 . 5 with custom Matlab routines ( Source Code 4 ) . For extracellular SGN recordings , PSTHs were calculated as average firing rates across 200 presentations of 50 ms or 500 ms tone bursts presented at 0 . 1 s/0 . 2 s or 2 s intervals , resp . ( PSTH at 10/5 Hz and 0 . 5 Hz ) at Cf , 30 dB above the threshold and binned at a width of 2 ms . Peak rate was determined as the largest bin of the PSTH in a time window 3–11 ms after stimulus onset . Adapted rate was averaged in a window spanning 35–45 ms or 405–415 ms after stimulus onset ( for PSTH at 5 Hz and 0 . 5 Hz , respectively ) . Rate level functions were acquired using 50 ms tone bursts presented at Cf and at 5 Hz . 25 repetitions for each stimulus intensity ( 5 dB steps ) were recorded . Maximal steepness was calculated as the maximal increase in spike rate between two consecutive 5 dB increment steps . Dynamic range was calculated by using sigmoidal fits in the rate level functions as described in and measuring the range of sound pressure between 10% and 90% of maximal firing rate . For amplitude modulation analysis , synchronization index was calculated as described by ( Goldberg and Brown , 1969 ) . Synchronization index estimation was only considered valid when at least 15 spikes occurred in a 3 s time window and the Rayleigh statistic was below 13 . 8 . For analysis of forward masking experiments , spike counts in a 10 ms interval starting from responses of both masker and probe onset were determined and presented as the ratio of probe and masker responses for at least 25 repetitions for every masker-probe interval from each unit . Exponential fitting to the plots of each individual SGN approximated the recovery kinetics . Quantitative analysis of electron microscopy data was performed with ImageJ for conventional embedded samples and with IMOD for HPF/FS tomograms . According to the presence of ribbon-occupied and ribbonless synapses , we considered the following analysis criteria: For ribbon-occupied synapses , membrane-proximal synaptic vesicles ( MP-SVs , within a distance of ≤25 nm from the AZ membrane and ≤80 nm from the presynaptic density ) and ribbon-associated synaptic vesicles ( RA-SVs , first layer around the ribbon with a maximum distance of 80 nm from the vesicle membrane to the ribbon ) were counted ( Figure 2J , random sections analysis according to Strenzke et al . , 2016 and Figure 3G for tomograms according to 2D-random section analysis criteria ) . The tomogram analysis parameters were further modified , as used in Jung et al . , 2015a . Here , the MP-SVs were defined as vesicles with ≤50 nm from the AZ membrane and with the shortest distance from the vesicle membrane to the presynaptic density of ≤100 nm , excluding RA-SVs ( Figure 3—figure supplement 1A ) . For random sections , SV diameters were calculated by the averaged measurements of the horizontal and vertical axis . The ribbon size was measured in height and width , taking the longest axis of the ribbon excluding the PD , and the edges of the synaptic ribbon were traced manually using ImageJ . The length of the PD was measured along the AZ membrane ( Figure 2J ) . For ribbonless synapses , a presynaptic density-associated synaptic vesicle ( PDA-SVs ) pool was defined considering all clustered vesicles ≤80 nm around the PD that did not fulfill the criteria of a MP-SV ( see above , also Figure 2J for random sections ) . The MP-SV pool , as well as the SV diameter and PD length , were analyzed as for the ribbon-occupied synapses . For tomograms , the PDA-SV pool was defined as the SVs in the first layer ≤ 80 nm to the PD , excluding the MP-SVs . The MP-SV pool criteria are the same as described in the previous paragraph ( Figure 3G , according to 2D-random section and Figure 3—figure supplement 1A , according to Jung et al . , 2015a ) . For tomograms , the according pools were further distinguished into tethered and non-tethered vesicles ( Figure 3G and Figure 3—figure supplement 1A ) . All vesicles were annotated using a spherical ‘scattered object’ at its maximum projection in the tomogram , encompassing the outer leaflet of the vesicles . The vesicle radii were determined automatically ( Helmprobst et al . , 2015 ) with the program ‘imodinfo option -p’ of the IMOD software package ( Kremer et al . , 1996 ) . The coding of sound onset differs among the various SGNs in time due to different durations of the traveling wave , synaptic delays and conduction times . To obtain an average PSTH for modeling that is not smeared out due to such differences between units , the individual PSTHs were aligned beforehand by shifting their timing relative to each other . Onset detection was based on a change in spike statistics . For spontaneous activity , the 99 . 5 percentile of spike counts was determined . Next , the time at which response rises to twice this percentile was found . This is certainly a point within the sound response . Finally going back from this point , a drop back baseline activity , that is below the percentile was detected and used as onset time . Aligned PSTH from all units were averaged . This averaged PSTH from the forward masking data were fit with a model waveform using a genetic fit algorithm implemented in IGOR Pro ( Wavemetrics , Lake Oswego , OR , USA ) . The purpose of the model is to give insight into the dynamics of SV cycling at the average IHC AZ . More specifically , the notion of Ca2+-nanodomain-like control of RRP exocytosis ( Brandt et al . , 2005; Graydon et al . , 2011; Pangršič et al . , 2015; Wong et al . , 2014 ) , as well as the limited MP-SVs at the AZ ( see Figures 2 and 3 ) motivates the notion of a limited , quasi-fixed number of available vesicular release sites or slots , Nslot , ( Frank et al . , 2010; Wong et al . , 2014 ) that constitute the RRP . Each of these sites can be either empty or occupied by a release ready SV ( whereby all filled slots constitute the ‘standing’ RRP ) and at each time point , a release ready SV will fuse with a certain probability described by the fusion rate constant kfus . Its value depends on the sound pressure level in a relation we assume to be linear within the dynamic range of the synapse/fiber . While the sound pressure level rises from silence to saturation kfus increases from kfus , spont to kfus , stim . The refilling of empty sites is described by a refill rate constant krefill , which also depends on the sound intensity ( krefill , spont to krefill , stim ) . The state of the release site was described by: ( 3 ) dNslotfilled ( t ) dt=krefill ( t ) Nslot−Nslotfilled ( t ) ( kfus ( t ) +krefill ( t ) ) Although this equation is formulated for SV fusion rates , a scaling factor f can be used to account for the fraction of fusion events that cannot successfully trigger an action potential ( AP ) despite sufficient neural excitability for example because of the too small size of the elicited excitatory post synaptic current . This factor effectively operates as if the number of release sites was scaled down . The scaled equation then gives a rate R of potentially supra-threshold EPSCs as the product of the number of occupied release sites , the fusion rate constant and the scaling factor f: ( 4 ) R ( t ) =kfus ( t ) . f . Nslotfilled ( t ) The stationary solutions of Equation 3 together with Equation 4 determine steady state occupancy and steady state event rates: ( 5 ) Nsoltfilled|stradystatecondition=KrefillconditionKfusioncondition+Krefillcondition . Nsolt ( 6 ) R|stradystatecondition=Kfusioncondition . f . KrefillconditionKfusioncondition+Krefillcondition . Nsolt In this equation , ‘condition’ is either silence or saturating sound pressure level . In order to connect the postsynaptic event rate of potentially supra-threshold EPSCs to the actual AP rate , refractoriness is considered as a combination of an absolute refractory period tabs , during which the probability of an EPSC to trigger an AP is zero , with a relative refractory period during which this trigger probability returns to one with an exponential time course characterized by τrel ( Berry and Meister , 1998 ) . This description of refractoriness can be applied to the ‘driving’ EPSC rate R by means of a delayed differential equation . The equation is motivated by the concept of three possible states of the SGN: ‘absolute refractory’ , ‘relative refractory’ or ‘available’ ( fully excitable ) . At any point , the probability that the SGN turns from ‘available’ to ‘refractory’ is proportional to the rate R ( t ) . The return back to ‘available’ happens ‘delayed’ by tabs and with a probability that is proportional to 1/τrel . ( 7 ) dfavail ( t ) dt=frelref ( t ) τrel−favail ( t ) . f . R ( t ) dfrelref ( t ) dt=favail ( t−tabs ) . f . R ( t−tabs ) −frelref ( t ) τrel Together with Equation 6 the stationary solution of this description of refractoriness connects the observable steady state rates during silence and stimulation to the rate constants krefill und kfusion: ( 8 ) AP_Rate|stradystatecondition=Kfusioncondition . f . KrefillconditionKfusioncondition+Krefillcondition . Nsolt . 11+R|stradystatecondition ( tabs+τrel ) To go beyond the description of steady state event rates and to use the model for a parameterized description of the actual time course of experimentally observed PSTHs acquired during forward masking ( Figure 11E ) , it is necessary to define the relation between the applied stimulus and the fusion and refill rate constants . For the experimental data presented here , the stimulus level was increased from silence to 30 dB above fiber threshold within a 4 ms ramp having a quarter of a sin² shape . It was assumed that kfusion and krefill follow the stimulus increase simultaneously . The ordinary differential and delayed differential equations above were combined into a fit function . PSTHs ( one masker followed by one probe ) were averaged per genotype for each masker probe interval ( 4 , 16 , 64 and 256 ms ) and were fitted in parallel with one parameter set . During experiments , trials were acquired in immediate succession without gaps . Therefore , the model implements cyclic boundary conditions for the occupancy of the slots . This model only captures the short term processes , assuming that a set of experiments , for example forward masking trials quickly lead to a steady state . Slow adapting processes were not explicitly modeled . The observed drop of the apparent number of available slots in the forward masking experiments was described here as a change in the number of slots from Nslots to a reduced capacity N’slots and for a given spiral ganglion neuron that was tested with tone bursts and forward masking , the ratio Nslots/N’slots could be estimated from the change in rates ( see Results ) . The data were analyzed using Matlab ( Mathworks ) , Excel , Igor Pro 6 ( Wavemetrics ) , Origin 9 . 0 ( Microcal Software ) , and GraphPad Prism ( GraphPad Software ) . Averages were expressed as mean ± standard error of the mean ( S . E . M . ) . For every dataset , the standard deviation ( S . D . ) , number of replicates ( n ) and animals ( N ) were indicated . For Figure 7 , nmin corresponds to the minimum number of cells included in the analysis of each depolarization potential given that the number of cells for each potential differs . In order to compare two samples , data sets were tested for normal distribution ( Jarque-Bera test , D’Agostino and Pearson omnibus normality test or the Shapiro-Wilk test ) and equality of variances ( F-test ) , followed by two-tailed unpaired Student’s t-test , or , when data were not normally distributed and/or variance was unequal between samples , the unpaired two-tailed Mann-Whitney-Wilcoxon test was used . Cumulative distributions in Figure 9A were statistically compared using the Kolmogorov-Smirnov test . The ROUT method ( Q = 0 . 1% ) from GraphPad Prism was used to identify definitive outliers for Figure 7H . For multiple comparisons , statistical significance was calculated by using one-way ANOVA test ( two-way ANOVA in the case of ABR thresholds ) followed by Tukey’s test for normally distributed data or Kruskal-Wallis ( K-W ) test followed by non-parametric multiple comparisons test ( NPMC ) for non-normally distributed data . For SV diameter quantifications in random sections , a custom-written routine using Java Statistical Classes library ( JSC ) ( Bertie , 2002 ) was utilized for statistical analysis ( Source code 5 ) . Due to the tied ranks of SV diameter measurements obtained for random sections , their S . E . M . was used as a tolerance value for the usage of Kruskal-Wallis test as suggested by Bertie et al . in JSC library ( Bertie , 2002 ) , where two values were treated as equal if their difference was ≤ S . E . M . . The non-significant difference between samples is reported as n . s . , significant differences are reported as *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . | Our sense of hearing relies on our ears quickly and tirelessly processing information in a precise manner . Sounds cause vibrations in a part of the inner ear called the cochlea . Inside the cochlea , the vibrations move hair-like structures on sensory cells that translate these movements into electrical signals . These hair cells are connected to specialized nerve cells that relay the signals to the brain , which then interprets them as sounds . Hair cells communicate with the specialized nerve cells via connections known as chemical synapses . This means that the electrical signals in the hair cell activate channel proteins that allow calcium ions to flow in . This in turn triggers membrane-bound packages called vesicles inside the hair cell to fuse with its surface membrane and release their contents to the outside . The contents , namely chemicals called neurotransmitters , then travels across the space between the cells , relaying the signal to the nerve cell . The junctions between the hair cells and the nerve cells are more specifically known as ribbon synapses . This is because they have a ribbon-like structure that appears to tether a halo of vesicles close to the active zone where neurotransmitters are released . However , the exact role of this synaptic ribbon has remained mysterious despite decades of study . The ribbon is mainly composed of a protein called Ribeye , and now Jean , Lopez de la Morena , Michanski , Jaime Tobón et al . show that mutant mice that lack this protein do not have any ribbons at their “ribbon synapses” . Hair cells without synaptic ribbons are less able to timely and reliably send signals to the nerve cells , most likely because they cannot replenish the vesicles at the synapse quickly enough . Further analysis showed that the synaptic ribbon also helps to regulate the calcium channels at the synapse , which is important for linking the electrical signals in the hair cell to the release of the neurotransmitters . Jean et al . also saw that hair cells without ribbons reorganize their synapses to form multiple active zones that could transfer neurotransmitter to the nerve cells . This could partially compensate for the loss of the ribbons , meaning the impact of their loss may have been underestimated . Future studies could explore this by eliminating the Ribeye protein only after the ribbon synapses are fully formed . These findings may help scientists to better understand deafness and other hearing disorders in humans . They will also be of interest to neuroscientists who research synapses , hearing and other sensory processes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"neuroscience"
] | 2018 | The synaptic ribbon is critical for sound encoding at high rates and with temporal precision |
We discovered that optical stimulation of the mystacial pad in Emx1-Cre;Ai27D transgenic mice induces whisker movements due to activation of ChR2 expressed in muscles controlling retraction and protraction . Using high-speed videography in anesthetized mice , we characterize the amplitude of whisker protractions evoked by varying the intensity , duration , and frequency of optogenetic stimulation . Recordings from primary somatosensory cortex ( S1 ) in anesthetized mice indicated that optogenetic whisker pad stimulation evokes robust yet longer latency responses than mechanical whisker stimulation . In head-fixed mice trained to report optogenetic whisker pad stimulation , psychometric curves showed similar dependence on stimulus duration as evoked whisker movements and S1 activity . Furthermore , optogenetic stimulation of S1 in expert mice was sufficient to substitute for peripheral stimulation . We conclude that whisker protractions evoked by optogenetic activation of whisker pad muscles results in cortical activity and sensory perception , consistent with the coding of evoked whisker movements by reafferent sensory input .
Active sensing involves the integration of internally generated motor commands with sensation of the external world . In the rodent whisker system , which has been used extensively as an experimental model of active sensing , animals use their mystacial vibrissae ( whiskers ) to sample the immediate environment in rhythmic bouts of active self-generated whisker movements , called whisking ( Kleinfeld et al . , 2006; Brecht , 2007; Diamond et al . , 2008; Prescott et al . , 2011 ) . A temporal sequence of extrinsic and intrinsic whisker pad muscle activation drives exploratory whisking: the extrinsic muscle M . nasolabialis profundis initiates the forward pad translation , and then the intrinsic 'sling' muscles that surround the base of each whisker follicle drive further protraction ( Dörfl , 1982; Berg and Kleinfeld , 2003; Hill et al . , 2008; Bosman et al . , 2011; Haidarliu et al . , 2015 ) . Sensory signals arising from whisker-object contact are transmitted through the infraorbital nerve to trigeminal ganglion , ventral-posterior medial ( VPM ) thalamus , and S1 ( Petersen , 2007 ) . Because whisker pad muscles almost completely lack spindles , proprioception is thought to play a minor role , if any , in determining whisker position in space . Instead , sensations that arise from whisker self-motion – 'reafferent' signaling – is thought to play an important role in determining whisker position and object localization ( Kleinfeld and Deschênes , 2011 ) . Although earlier studies suggested that reafferent signaling is encoded in parallel to afferent ( touch-related ) signals via posterior medial ( POm ) thalamus ( Yu et al . , 2006 ) , recent evidence suggests that reafferent signaling is also processed along the same lemniscal ( VPM-S1 ) pathway to cortex as afferent input from whisker-object contact ( Moore et al . , 2015 ) . Many studies of reafferent sensory signaling have used an 'artificial whisking' paradigm to elicit whisker movements in anesthetized rodents by electrical stimulation of the buccal branch of the facial motor nerve ( Zucker and Welker , 1969; Brown and Waite , 1974; Szwed et al . , 2003 ) . Artificial whisking produces whisker protractions with amplitude and frequency that can be well controlled experimentally . This paradigm has drawbacks , however , including the necessity to perform experiments in anesthetized subjects , which makes it difficult to relate reafferent signaling to behavior; and the inability to stimulate certain muscle groups , which means that only whisker protractions , not retractions , can be evoked . Recently , optogenetic studies of motor nerves and muscles have used the hindlimb as a model system ( Liske et al . , 2013; Towne et al . , 2013; Bryson et al . , 2014; Magown et al . , 2015 ) . While various central elements of the whisker system have been targeted for optical control in behaving mice ( Poulet et al . , 2012; O'Connor et al . , 2013; Sachidhanandam et al . , 2013; Matyas et al . , 2010 ) , peripheral optogenetic stimulation have not been used to investigate control of whisker movements . In this study , we report that optogenetic stimulation of the whisker pad in Emx1-Cre;Ai27D transgenic mice evokes whisker movements due to channelrhodopsin-2 ( ChR2 ) expression in select intrinsic and extrinsic muscles . We first characterize the amplitude and frequency of whisker protractions evoked by anterior whisker pad stimulation in anesthetized mice . We then compare the electrophysiological responses in S1 to optogenetic and mechanical whisker stimulation . Finally , we show that awake , head-fixed mice are able to perceive optical whisker pad stimulation in a behavioral detection task . The results suggest that optogenetic stimulation of whisker pad muscles leads to sensory perception through reafferent signaling .
In initial screens of adult Emx1-Cre;Ai27D mice ( offspring of crossing Emx1-Cre and Ai27D lines ) , we discovered that whisker movements were evoked by blue light directed toward the whisker pad . While cortical expression of ChR2 in Emx1-Cre;Ai27D ( or the similar Emx1-Cre;Ai32 ) mice is well known ( Madisen et al . , 2012; Zagha et al . , 2013; McAlinden et al . , 2015 ) , the functional properties of incidental peripheral expression have not been characterized . Therefore , our goals were to determine 1 ) the localization of ChR2 expression in the whisker pad in Emx1-Cre;Ai27D mice and the functional properties of whisker movements evoked by peripheral optogenetic stimulation; 2 ) whether peripheral optogenetic stimulation activates S1 in a fashion similar to mechanical whisker stimulation; and 3 ) whether peripheral optogenetic stimulation induces behaviorally reported sensory detection . We first characterized the whisker movements evoked by a 2–3 mm diameter , 460 nm spot of light aimed at different regions of the whisker pad in anesthetized Emx1-Cre;Ai27D mice ( isoflurane 0 . 8–1 . 5% ) ( Figure 1A ) . The direction of movement depended on the location of the spot , such that illumination of the rostral pad resulted in whisker protraction , while illumination of the caudal-inferior pad resulted in whisker retraction ( Figure 1B ) . Stimulation at some locations elicited more complex combinations of protractions and retractions from individual whiskers ( Figure 1—figure supplement 1; see also Video 1 , Video 2 , Video 3 ) . These regional variations in light-evoked protraction and retraction were qualitatively similar in 5 of 5 mice tested . In the rest of this study , we focused on whisker protractions evoked by optogenetic activation of the rostral whisker pad . 10 . 7554/eLife . 14140 . 003Figure 1 . ChR2/tdTomato expression in whisker pad muscles and optical activation of whisker movements in Emx1-Cre;Ai27D mice . ( A ) Illustration of experimental setup showing the position of 460 nm light spot on the whisker pad used for optogenetic stimulation . ( B ) Example color map from one mouse ( isoflurane anesthesia , 0 . 8–1 . 5% ) showing the direction of movement ( rostral or caudal ) of whisker B2 evoked by optogenetic whisker pad stimulation at different locations on the pad . The maximum amplitude of whisker movement in degrees is color coded for each position tested . White pixel indicates location with no measurement . Whisker protractions evoked by rostral optical stimulation were the focus of the present study . ( C1 ) Intrinsic and extrinsic muscles of the whisker pad exhibit tdTomato fluorescence , as seen in histological sections . Photomicrographic montage of tdTomato fluorescence in a coronal section of the mystacial pad . Scale bar 200 µm . ( C2 ) Photomicrographic montage of tdTomato fluorescence in a whisker follicle in a transverse section . Scale bar 100 µm . ( C3 ) Representative photomicrographs of tdTomato fluorescence used for quantification ( as in D ) . Arrows point to regions of quantification for the external extrinsic protractor muscles ( top; pars maxillaris superficialis and pars maxillaris profunda of M . nasolabialis profundus ) , intrinsic follicular muscle ( middle ) , and internal extrinsic retractor muscles ( bottom; pars media superior and pars media inferior of M . nasolabialis profundus ) following the terminology of Haidarliu et al . , 2015 . Scale bar 100 µm . ( D ) Summary of ChR2/tdTomato fluorescence intensity . Fluorescence intensity was lowest in external extrinsic muscle and highest in intrinsic muscle . Bars are mean ± SEM and individual data points are plotted . ***p<0 . 001 , **p<0 . 01 , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 00310 . 7554/eLife . 14140 . 004Figure 1—figure supplement 1 . Analysis of retraction and protraction movements for individual whiskers . ( A ) Schematic of experiment as in Figure 1 , with video image of tracked whiskers . ( B ) Color maps showing the peak whisker movement evoked by a 20 ms , 460 nm light spot located at various positions on the whisker pad . For each colormap , the identity of the tracked whisker is indicated above . The origin ( 0 , 0 ) was defined a rostral area near the C4/D4 whiskers that evoked reliable protractions; this site was used in most additional experiments in this study . Stimulation of caudal-inferior sites , on the other hand , evoked mixed movement types in different whiskers , evident as diverse ( yellow/red ) colors on the color maps . ( C ) Time courses of movements evoked for 6 tracked whiskers ( B1 , B2 , B3 , C3 , C4 , D2 ) at each of three stimulation sites on the whisker pad , including the rostral protraction area ( 0 , 0 ) and two caudal-inferior areas [ ( -2 , -2 ) and ( -3 , -2 ) ] . Note that movements measured as near-zero for some whiskers at some sites ( e . g . , C4 at -2 , -2 ) might result from co-contraction of different muscle types that cause opposing retractions and protractions . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 00410 . 7554/eLife . 14140 . 005Figure 1—figure supplement 2 . Absence of ChR2/tdTomato expression in vibrissal nerve fibers of EMX-cre;Ai27D mice . ( A ) Example images of two whisker follicles in an Emx1-Cre;Ai27D mouse . Expression of ChR2/tdTomato was present in intrinsic muscle ( m . ) surrounding the follicle , but was not observed in nerve fibers innervating the follicle . ( B ) . Example images of two whisker follicles in a parvalbumin ( PV ) -Cre;Ai27D mouse . In contrast to the Emx1 results , PV mice showed a profusion of ChR2/tdTomato expression in follicular nerves , consistent with previous results ( Sakurai et al . , 2013 ) . Dotted curved lines indicate the interior or the cavernous sinus . Arrowheads indicate ChR2/tdTomato expression in nerve fibers and/or nerve endings at the whisker shaft ( s ) and base ( b ) . Images below are from single focal planes ( not z-projections ) at boxed regions . Each image in A and B is a maximum z-projection of 20 focal planes at 1 µm spacing in coronal sections . Scale bar , 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 005 We performed histological analysis to determine if these results could be explained by expression of ChR2 in muscles that control different types of whisker movements ( Dörfl , 1982; Hill et al . , 2008; Haidarliu et al . , 2015 ) . Indeed , analysis of the native fluorescence of the ChR2/tdTomato fusion protein in sections of the whisker pad revealed tdTomato expression in intrinsic and extrinsic whisker pad muscles ( Figure 1C ) . Intrinsic muscles appeared on both superior and inferior sides of the follicle in coronal sections , and on the rostral side in transverse sections ( Figure 1C1 , C2 ) , consistent with their sling-like morphology ( Dörfl , 1982; Haidarliu et al . , 2010 ) . No fluorescence was evident in the infraorbital ( sensory ) nerve ( Figure 1—figure supplement 2 ) . Comparison of fluorescence intensity indicated the highest intensity in intrinsic sling muscles , followed by the deep extrinsic retractor muscle ( pars maxillaris superficialis and pars maxillaris profunda of M . nasolabialis profundus ) , and lastly the superficial extrinsic protractor muscle ( the pars media superior and pars media inferior of M . nasolabialis profundus , ) ( F ( 2 , 6 ) = 57 . 66 , p=0 . 0001 , repeated measures ANOVA followed by paired contrasts; p=0 . 0004 comparing external extrinsic and intrinsic , p=0 . 0086 comparing external extrinsic and internal extrinsic , p=0 . 0341 comparing intrinsic and internal extrinsic; n=4 follicles from one mouse ) ( Figure 1D ) . These results indicate that light-evoked whisker movements in Emx1-Cre;Ai27D mice arise from activation of ChR2 expressed in extrinsic and intrinsic whisker pad muscles . To quantitatively characterize the whisker movements evoked by peripheral optogenetic stimulation , we recorded high-speed video ( 500 frames/s ) in anesthetized mice ( isoflurane 0 . 8–1 . 5% ) ( Figure 2A ) in response to 460 nm light stimulation of varying intensity , duration , and frequency , with illumination centered at the rostral protraction area . The amplitude of whisker protraction in response to a 50 ms light pulse of increasing intensity ( range , 1 . 3–10 . 3 mW ) increased approximately linearly to a maximum amplitude of 11 . 4 ± 1 . 2 degrees ( mean ± SEM , n = 4 mice; maximum 14 . 8 degrees in one mouse; Figure 2B ) . In 2 of 4 mice , the angle change appeared to saturate at less than maximal power ( 8 . 37 and 9 . 94 mW , respectively ) . The average latency of optogenetically evoked whisker movement was 13 . 5 ± 0 . 3 ms ( mean ± SEM , n = 4 mice; threshold defined as 10% of the maximum peak ) and was not affected by stimulus duration . We used 9 . 94 mW to define the relationships between whisker protraction , duration , and frequency ( below ) . We next measured the relationship between whisker protraction and optical stimuli of varying duration from 5 to 1200 ms ( at 9 . 94 mW intensity ) . Whisker protraction angle increased with the duration of the optical stimulus , saturating with durations longer than approximately 60 ms ( Figure 2C ) . 10 . 7554/eLife . 14140 . 006Figure 2 . Characterization of whisker movements evoked by optogenetic whisker pad stimulation . ( A ) Top: Illustration of experiment setup ( isoflurane anesthesia , 0 . 8–1 . 5% ) . Rostral whisker pad illumination ( 460 nm ) was used to evoke whisker protractions ( positive angle values ) . Bottom: Image of whiskers under infrared illumination as used for whisker tracking . Angle changes of individual whiskers were measured relative to the initial position ( green lines ) . ( B ) Relationship between light intensity and evoked whisker protractions . Top: Example traces from one mouse ( mean ± SD of single trial for n = 4 whiskers ) . Blue triangle indicates the onset of the light stimulation . Intensities: 1 . 3 , 3 . 1 , 4 . 1 , 6 . 5 , 8 . 4 , 10 . 3 mW . Duration of stimuli , 50 ms . Bottom: amplitude of evoked angle change ( left axis ) and movement latency ( right axis ) vs . light intensity ( bin size , 1 mW; mean ± SEM; n = 4 mice ) . ( C ) Relationship between light duration and evoked whisker protractions . Top: Example traces from one mouse ( mean ± SD of 10 trials ) . Blue triangles indicates the onset of the light stimulation . Durations: 10–80 ms at 9 . 94 mW intensity . Bottom: amplitude of evoked angle change vs . light duration ( mean ± SEM; n = 4 mice; note gap in axis between 60 and 250 ms and difference in x-axis scaling for 5–60 ms and 250–1200 ms ) . ( D ) Adaptation of evoked whisker protractions to optical pulse frequency . Top: two example traces from one mouse at 6 Hz and 22 Hz stimulation ( 9 . 94 mW ) . Blue triangle indicates the onset of the light stimulation . Bottom: Adaptation indexes ( black: an/a1 , ratio of last to first response amplitude; gray , a2/a1 , ratio of second to first response amplitude ) plotted versus stimulus frequency ( mean ± SEM; n=3 mice ) . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 00610 . 7554/eLife . 14140 . 007Figure 2—source data 1 . Data for Figure 2B . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 00710 . 7554/eLife . 14140 . 008Figure 2—source data 2 . Data for Figure 2C . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 00810 . 7554/eLife . 14140 . 009Figure 2—source data 3 . Data for Figure 2D . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 009 Finally , we tested whether whisker protractions could follow 1 s long trains of optogenetic stimulation of varying frequencies , from 1 to 45 Hz ( Figure 2D ) , covering the frequency range of natural exploratory whisking ( Welker , 1964; Carvell and Simons , 1990; Harvey et al . , 2001 ) . The duration of each pulse in the train was 2 ms . Two alternative adaptation indexes were calculated as either the ratio of amplitudes of the last response to the first response in the pulse train ( an/a1 ) or the amplitudes of the second response to the first response in the pulse train ( a2/a1 ) . A smaller adaptation index indicates a larger difference between first and second or last peaks during the optical pulse train , and therefore greater adaptation . The adaptation index of an/a1 decreased to lower values than the adaptation index of a2/a1 at the same frequency ( e . g . , at f = 28 Hz , an/a1 = 0 . 41 ± 0 . 04 , a2/a1 = 0 . 66 ± 0 . 04 ) , indicating further adaptation with increasing number of pulses ( Figure 2D ) . At frequencies greater than 30 Hz , individual evoked movements were no longer discernible , although the envelope of the angle change continued to show adaptation up to 45 Hz and became similar to movements elicited by constant prolonged light steps ( data not shown ) . These data indicate that optogenetically evoked whisker protractions show activation and adaptation over a behaviorally relevant range of frequencies . Together , the results of Figure 2 define fundamental stimulus parameters for optogenetic activation of whisker protractions in Emx1-Cre;Ai27D mice . We next investigated whether peripheral optogenetic stimulation evoked neural activity in S1 by implanting 8-channel microwire arrays in S1 of Emx1-Cre;Ai27D mice . After one week of recovery , we recorded local field potentials ( LFPs ) and multiunit spiking activity in three anesthetized mice ( isoflurane 0 . 8–1 . 5% ) in response to peripheral optogenetic stimulation ( Figure 3A ) . To account for potential differences in the locations of the arrays in S1 , we analyzed signals from the channel with the shortest latency in each mouse . Spiking activity and LFP amplitude increased with the duration of peripheral optogenetic stimulation ( Figure 3B , C ) . Plotting the responses normalized to the maximum response in each mouse indicated that spike count and LFP amplitude increase steeply with light pulse duration from 1–20 ms , and moderately between 20–100 ms ( Figure 3D ) . We used a brief mechanical deflection of the whisker to compare S1 response timing . While the active whisker protraction evoked by optogenetic stimulation provides qualitatively distinct activation of sensory input compared with passive mechanical deflection , this experiment allowed us to determine the relative latencies of S1 responses . The spike number and LFP amplitude evoked by peripheral optogenetic stimulation were on average similar to those evoked by mechanical whisker deflection ( peak spike number per stimulus in 10 ms bin: 3 . 5 ± 0 . 2 mechanical , 3 . 2 ± 1 . 1 optical; LFP peak amplitude: −168 . 5 ± 24 . 7 µV mechanical , 159 . 66 ± 45 . 1 µV optical ) . In one mouse , the largest responses observed to peripheral optogenetic stimulation were 5 . 8 spikes/stimulus and −267 . 5 µV peak LFP amplitude ( 129 . 3 ± 16 . 3% of the LFP amplitude evoked by mechanical whisker stimulation; values were 54 . 0 ± 7 . 6% , and 68 . 4 ± 2 . 6% in two other mice ) . Notably , the response latency determined from LFP recordings was 10 . 8 ± 0 . 1 ms longer for peripheral optogenetic stimulation compared to mechanical whisker stimulation ( p<1x10-5 in n = 3/3 mice; paired t-tests; 13–39 trials per mouse ) ( Figure 3C , inset; Figure 3E ) . These results suggest that the longer latency in S1 for peripheral optogenetic stimulation can likely be attributed to the 11 . 9 ± 0 . 8 ms delay associated with the initiation of evoked whisker movement ( Figure 2B; value from 9 . 3 mW intensity ) , and that sensory signals arrive rapidly in S1once whisker movement is initiated . These results are consistent with our histological data suggesting that ChR2 is expressed in muscle and absent from sensory nerve ( Figure 1C; Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 14140 . 010Figure 3 . Extracellular recordings of S1 activity in response to optogenetic whisker pad stimulation . ( A ) Illustration of experiment setup ( isoflurane anesthesia , 0 . 8–1 . 5% ) , including chronically implanted microwire array . ( B ) Example peri-stimulus time histograms ( PSTHs; bin size , 10 ms ) for one mouse displayed ± 0 . 5 s relative to stimulation onset . Blue triangles and lines denote onset of 460 nm light stimulation; black triangle and line denotes onset of mechanical whisker stimulation . ( C ) Example local field potentials ( LFPs ) from one channel in response to optical whisker pad stimulation of various durations ( 1–100 ms ) and mechanical stimulation of whisker C3 . Each trace is the mean of 30 trials . ( D ) Peak LFP and maximum spike count ( mean ± SEM , n = 3 mice ) , normalized to the maximum response for each channel . The channel that showed the largest response was selected from each mouse . ( E ) Comparison of LFP response latency for peripheral optical stimulation and mechanical whisker stimulation ( shortest latency channel selected for each mouse ) . Bar graphs show mean ( n = 3 mice ) and lines connect individual subjects . Mean latency was 17 . 3 ± 1 . 0 ms ( mean ± SEM ) for 20–100 ms optical stimuli and 6 . 5 ± 0 . 1 ms ( mean ± SEM ) for mechanical whisker stimulation . ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 01010 . 7554/eLife . 14140 . 011Figure 3—source data 1 . Data for Figure 3D . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 011 In order to determine whether Emx1-Cre;Ai27D mice can perceive peripheral optogenetic stimulation , we established a modified head-fixed sensory detection task inspired by recent studies ( O'Connor et al . , 2013; Sachidhanandam et al . , 2013 ) ( Figure 4A ) . We trained mice to report the presence of 100 ms ( 9 . 94 mW ) peripheral optogenetic stimulation by licking for water reward . In each Hit trial , mice received a water drop for licking within a 2 s time window after a stimulus . False Alarm ( FA ) trials occurred if the mouse licked when no light was delivered during the stimulus time window and resulted in presentation of a 2 s , 5 kHz tone and 5–10 s time out before the next trial . Inter-trial time randomly varied from 5–10 s . Two of four mice learned the task ( d’>1 ) within 4 sessions ( 2 sessions per day , 125 trials per session ) , showing maintained Hit rate with relatively low FA rate ( FA rate < 0 . 3 ) . The other two mice learned only after introducing 2 M salt water solution as additional punishment for licking during FA trials . We aligned learning curves for all mice relative to the start of learning ( d’>1 ) ( Figure 4B ) , which was after the introduction of salt water punishment in 2/4 mice . Overall , behavioral performance improved via a maintained high Hit rate and a decrease in FA rate , resulting in an increase in d’ from 0 . 7 ± 0 . 1 to 2 . 3 ± 0 . 4 ( mean ± SEM; p = 0 . 017 , paired t-test; n = 4 mice ) over the course of training ( Figure 4B ) . 10 . 7554/eLife . 14140 . 012Figure 4 . Behavioral performance in mice trained to detect optogenetic stimulation of the whisker pad . ( A ) Illustration of the behavioral task . Water deprived , head-fixed mice were rewarded with water for licking within a 2 s response window ( gray boxes ) after optical stimulation ( 460 nm ) of the rostral whisker pad ( Hit trials ) . Licking in the absence of stimulation resulted in a False Alarm ( FA ) and punishment ( tone and/or 2M salt water ) . ( B ) Changes in behavioral performance with training . A maintained Hit rate ( blue ) and reduced FA rate ( red ) accounted for the increase in performance ( d’; black ) over sessions ( mean ± SEM , n = 4 mice ) . Note that learning curves for 2 of 4 mice are shown from the time of introduction of salt water punishment for FA and are aligned to first learning session ( d’>1 ) . ( C ) Dependence of Hit rate ( left axis ) and reaction time ( right axis ) on stimulation duration . In expert mice , optical whisker pad stimuli of various durations were included randomly on 10% of trials during behavioral sessions ( mean ± SEM; n = 4 mice ) . ( D ) A second optical fiber delivered either visual stimulation ( left ) or S1 optogenetic stimulation ( right ) on a random 10% of trials ( 100 ms duration for all stimuli ) . Lick probability was reduced for visual stimuli ( left ) , but not S1 stimulation ( right ) . *p<0 . 05; N . S . , not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 01210 . 7554/eLife . 14140 . 013Figure 4—source data 1 . Data for Figure 4B . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 01310 . 7554/eLife . 14140 . 014Figure 4—source data 2 . Data for Figure 4C . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 014 In order to determine the psychometric curve for peripheral optogenetic stimulation , we varied the duration of optical stimulation ( stimuli from 1–100 ms presented randomly with equal probability ) in expert mice ( d’>1 . 5 ) and tested the effects on task performance . We found that Hit rate fell to chance levels with stimuli shorter than 5 ms ( Figure 4C ) , defining a lower limit of optical stimulation necessary for behavioral detection . Note that this behaviorally measured detection threshold is similar to the threshold for evoked whisker movements ( Figure 2C ) and S1 activity ( Figure 3D ) measured in anesthetized mice . Although the fiber tip was shielded , we performed additional controls to rule out the possibility that mice were responding to visual stimulation arising from the optogenetic excitation light . In expert mice ( d’ > 1 . 5 ) , we used a second optical fiber placed in front of the mouse’s head to deliver diffuse blue light to the eye on the same side of the face . Visual catch trials were added to the training regime with 10% probability ( probability of whisker pad stimulation remained at 50% ) . Lick rate for visual catch trials was 0 . 37 ± 0 . 07 compared to 0 . 87 ± 0 . 04 for peripheral optogenetic stimulation ( 328 peripheral stimulation trials , 75 visual catch trials in 6 sessions from n = 2 mice ) ( Figure 4D ) , suggesting that visual stimulation was not a salient cue involved in performance of the peripheral optogenetic detection task . We next tested whether S1 neural activity , which is elicited by peripheral optogenetic stimulation ( Figure 3 ) , is sufficient for task performance in mice trained to detect peripheral stimulation . We delivered 460 nm optogenetic stimulation to S1 through a cranial window ( implanted in the initial surgery; see Materials and methods ) in 10% of trials using a second optical fiber . Additional introduction of ChR2 was unnecessary because Emx1-Cre;Ai27D mice express ChR2 in cortical pyramidal neurons . In expert mice , lick rate in response to S1 stimulation ( 100 ms ) was not significantly different compared to peripheral optogenetic stimulation ( 100 ms ) , but was significantly greater than FA lick rate ( F ( 2 , 6 ) = 17 . 62 , p=0 . 0031 , repeated measures ANOVA followed by paired contrasts; p=0 . 43072 comparing peripheral and S1 stimulation , p=0 . 0238 comparing S1 stimulation and FA , p=0 . 016 comparing peripheral stimulation and FA; 160 peripheral stimulation trials , 156 S1 simulation trials , 160 FA trials in 2 sessions from n = 4 mice ) ( Figure 4D ) . These results indicate that optogenetic stimulation of S1 is sufficient to drive sensory detection in mice that were trained to detect peripheral optogenetic stimulation .
We initially discovered that Emx1-Cre;Ai27D mice express ChR2/tdTomato in peripheral tissue by examining pups using fluorescence goggles . A practical benefit of peripheral expression worth mentioning is that transgene transmission to offspring can be inferred simply by visual inspection under fluorescence instead of traditional DNA genotyping . Peripheral Cre expression is known in other Cre driver lines commonly used for neurobiological studies of the central nervous system . For example , Chat-Cre , often used to target cholinergic neurons of basal forebrain ( Eggermann et al . , 2014; Hangya et al . , 2015 ) also drives expression in motoneurons ( Gong et al . , 2007; Takatoh et al . , 2013 ) ; PV-Cre , often used to target a class of central GABAergic interneurons ( Cardin et al . , 2009; Sohal et al . , 2009; Gentet et al . , 2010 ) , also expresses in proprioceptive neurons of the dorsal root ganglion ( Hippenmeyer et al . , 2005 ) , sensory neurons of trigeminal ganglion ( Sakurai et al . , 2013 ) , and fast-twitch skeletal muscle fibers ( Chakkalakal et al . , 2012 ) . It should be noted that , in addition to such incidental central/peripheral expression , many other Cre driver lines have been developed exclusively for investigation of the peripheral nervous system ( da Silva et al . , 2011; Rutlin et al . , 2014 ) and muscle ( Chen et al . , 2005; Li et al . , 2005; Chakkalakal et al . , 2012; Magown et al . , 2015 ) . In our experiments , we took advantage of the central and peripheral Cre transgenic expression in Emx1-Cre;Ai27D mice to achieve optogenetic activation of facial muscles and central neurons in the same subjects ( Figure 4 ) . Our study builds upon classic studies that used electrical stimulation of the buccal motor branch of the facial nerve to induce artificial whisking ( Zucker and Welker , 1969; Brown and Waite , 1974; Szwed et al . , 2003 ) . We summarize here some of the key features of optically induced whisker movements compared to those induced by artificial whisking . In addition to the issues discussed above , the major benefits of peripheral optogenetic stimulation are the non-invasive activation of whisker movements using light and the ability to perform experiments without the use of anesthesia . We used these features to design a behavioral task in head-fixed mice in order to investigate whether mice can perceive whisker movements that result from peripheral optogenetic stimulation . Several recent studies have used similar behavioral paradigms to investigate afferent sensory perception using tasks designed to assess , for example , stimulus detection , object localization , texture or frequency discrimination ( Arabzadeh et al . , 2005; O'Connor et al . , 2010; Morita et al . , 2011; Sachidhanandam et al . , 2013; Musall et al . , 2014; Chen et al . , 2015 ) . All of these tasks were designed to test the detection of afferent input , that is , aspects of sensory input arising from an external stimulus . The goal of our task was to test the detection of reafferent input , that is , sensory input arising from self-generated movement . During natural whisking , reafferent input arises from the whiskers moving through space and , because rodents mostly lack proprioceptors in whisker pad muscles ( Moore et al . , 2015 ) , reafferent signaling is considered important for encoding whisker position and locating objects in space ( Kleinfeld and Deschênes , 2011 ) . However , reafferent signaling has been difficult to study: either the subjects are anesthetized and well controlled whisker movements are elicited by artificial whisking , or the subjects are awake are freely whisking , where whisker movements are not under experimental control . Thus , our behavior task provided a unique opportunity to investigate the detection of reafferent signaling in awake animals with well controlled stimuli . We found that mice could readily learn to detect peripheral optogenetic stimulation ( Figure 4 ) . It should be noted that two mice required introduction of salt water punishment to reduce impulsive responding , as has been used in other types of Go/NoGo tasks ( Rebello et al . , 2014 ) , and that d’ remained below that observed in other studies due to a sustained FA rate of approximately 0 . 2 ( Huber et al . , 2012; Chen et al . , 2015 ) . The reason for the sustained FA rate is not clear but could relate to motivation or hydration levels ( Guo et al . , 2014 ) that could be further optimized in future studies . Similar to afferent sensory detection tasks , behavioral performance improved via maintained Hit rate and reduced FA rate over days . Four lines of evidence suggest that mice indeed used reafferent sensory input to perform the behavioral task . ( 1 ) Electrophysiological recordings from S1 showed that the latency was approximately 10 ms longer for optogenetically compared to mechanically evoked responses ( Figure 3C , E ) , suggesting that the whisker must move before sensory signals arrive in cortex . ( 2 ) Changes in evoked whisker movements ( Figure 2C ) , neural signals in S1 ( Figure 3D ) , and behavioral responses ( Figure 4C ) showed similar relationship with the duration of optogenetic whisker pad stimulation . ( 3 ) Visual stimulation was not sufficient to substitute for peripheral stimulation in expert mice performing the detection task , suggesting that mice were not responding to visual aspects of optogenetic stimulation . ( 4 ) Optogenetic stimulation of S1 was sufficient to substitute for peripheral optogenetic stimulation in the detection task ( whereas naïve mice did not respond to S1 stimulation; data not shown ) , suggesting that S1 is involved in perception of reafferent sensory signals . We conclude that the whisker movements elicited by optogenetic activation of muscles in the whisker pad lead to sensory perception through reafferent sensory signaling . Optogenetic whisker pad stimulation provides new opportunities for studies of sensorimotor integration in behaving mice . In the future , the non-invasive nature of peripheral optogenetic stimulation could be used to further investigate reafference and whisker-object contact during evoked protractions and retractions . Furthermore , the ability to stimulate muscle directly could have therapeutic benefits in preclinical studies of motor recovery after peripheral nerve injury or motoneuron degenerative disorders .
Homozygous Emx1-Cre mice ( Stock no . 005628 , Jackson Laboratory , Bar Harbor , ME ) were crossed with homozygous Ai27D mice ( Stock no . 012567 , Jackson Laboratory ) , and the resulting Emx1-Cre;Ai27D offspring ( heterozygous for both transgenes ) were used for experiments . PV-Cre;Ai27D mice ( Stock no . 008069 , Jackson Laboratory ) in Figure 1—figure supplement 2 were generated similarly . Ai27D mice express a ChR2 ( H134R ) /tdTomato fusion protein in a Cre-dependent manner ( Madisen et al . , 2012 ) . Emx1-Cre mice , which express Cre recombinase from the Emx1 locus ( Gorski et al . , 2002 ) , have been found to express Cre in limited peripheral tissues ( http://www . informatics . jax . org/ ) in addition to the better known Cre expression in forebrain glutamatergic neurons ( Madisen et al . , 2012; Zagha et al . , 2013; McAlinden et al . , 2015 ) . All procedures were approved by Rutgers University Institutional Animal Care and Use Committee ( IACUC; protocol 13–033 ) . Male Emx1-Cre;Ai27D mice were implanted with a glass cranial window and metal head post , as in previous work ( Holtmaat et al . , 2009; Margolis et al . , 2012 ) . Briefly , 4–9 week old mice were anesthetized with isoflurane ( 4% induction , 0 . 8–1 . 5% maintenance ) and placed on a feedback controlled heating blanket maintained at 36˚C ( FHC , Bowdoin , ME ) mounted on a stereotaxic frame ( Stoelting , Wood Dale , IL ) . After cleaning the surface of the skull , bonding agent ( iBond , Heraeus Kulzer , Hanau , Germany ) and a thin layer of dental cement ( Tetric Evoflow , Ivoclar Vivadent , Schaan , Lichtenstein ) were applied covering the right side skull and the anterior and posterior left side skull . A 4 mm craniotomy was made with a dental drill ( Osada EXL-M40 , Los Angeles , CA ) , leaving the dura mater intact , centered approximately over S1 barrel cortex ( -1 mm posterior , -3 mm lateral from Bregma ) . A 4 mm diameter #1 thickness circular cover glass ( Menzel Glaser , Braunschweig , Germany ) was implanted directly on the dura . The edges of the glass window were covered with dental cement , and the junction between glass and cement was sealed with cyanoacrylate glue . A custom metal head post was cemented to the right side skull . After surgery , mice were housed under a reversed light cycle ( lights off 08:00–20:00 ) and had free access to food and water . All subsequent experiments were conducted during the dark phase of the light cycle . Beginning 3–4 days after surgery , mice were handled daily by the experimenter . Adaptation to head restraint began after at least one week of recovery . For behavioral training ( below ) , mice were water restricted to 1 ml/day . Five mice underwent peripheral optogenetic stimulation under isoflurane anesthesia ( 4% induction , 0 . 8–1 . 5% maintenance ) . In each session , whisker movements were recorded in response to one stimulus parameter ( intensity , duration , frequency; three total sessions per mouse ) . One mouse was excluded from the study because of anesthesia-related whisker motion artifacts in the first session; another mouse was excluded from the varying frequency experiment because of changing baseline whisker position . Mice were stabilized by bolting the head post to a cross bar and placed on a feedback controlled heating pad maintained at 36°C . Lack of reflex to tail/foot pinch was used to assess adequate anesthesia levels . Isoflurane levels were adjusted during recordings to maintain an approximately 1 Hz respiration rate . This was a useful benchmark for adequate but light anesthesia and helped to avoid respiration-associated whisker movements that interfered with measurements of evoked whisker movements . Optogenetic stimulation was provided by a high-powered 460 nm LED ( Prizmatix , Givat-Shmuel , Israel ) coupled to a multi-mode optical fiber ( 200 µm core , 0 . 22 NA; Thorlabs , Newton , NJ ) . The duration and frequency of light pulses were controlled by trains of TTL pulses from an Arduino Uno board to the LED current driver . The fiber tip was mounted on a micromanipulator ( Narishige , Tokyo , Japan ) and placed within 2 mm of the anterior right side the whisker pad , resulting in a 2–3 mm diameter spot covering 3–5 whiskers . Maximum power was 10 . 3 mW measured at the fiber tip with a power meter ( Thorlabs ) . For whisker tracking , the right side whiskers were illuminated from below with an infrared LED array ( 850 nm , Advanced Illuminations , Rochester , VT; Edmund Optics #66–802 ) . Evoked whisker movements were imaged through a telecentric lens ( 0 . 36x , f/6-f/18; Edmund Optics #58–257 , Barrington , NJ ) with a CMOS camera at 500 Hz frame rate ( DR1 , Photofocus AG , Lachen , Switzerland ) and acquired using Streampix software ( NorPix , Montreal , Canada ) . Each movie was 5 s in duration , including a 1 s pre-stimulus baseline . 10 trials were recorded for each stimulus parameter with an 8 s inter-trial interval . Movie files were converted to AVI format offline , and MATLAB-based ( Mathworks , Natick , MA ) whisker tracking software ( Knutsen et al . , 2005 ) was used to extract the frame-wise whisker angle for each trial . Further analysis was carried out using custom routines in MATLAB . Well-isolated individual whiskers were manually selected for whisker tracking , and in some cases traces were averaged from 3–4 tracked whiskers . Stimulus-response curves were determined by measuring changes in whisker angle relative to a manually determined resting position . In intensity plots , data were binned in 1 mW bins to account for slight intensity differences used in different subjects . Data are shown as mean ± SEM for the four mice included in the analysis . Results from one of the four mice use for intensity and duration measurements was excluded from the analysis of adaptation because of unstable data . Custom microwire arrays were implanted in S1 ( 4 x 2 array of 50 µm diameter , 1 mm length stainless steel microwires; 500 µm between-channel , 300 µm between-row spacing; Micro Probes , Inc . , Gaithersburg , MD ) . Implants were performed in four of the same mice used for whisker tracking and behavior experiments ( below ) ; one of four mice was excluded due to poor signal quality . The glass window was removed ( Goldey et al . , 2014 ) by carefully drilling the edge of the dental cement and lifting the cover glass with blunt forceps . To allow access for the array , the dura mater was punctured with a 34 gauge needle tip . The most posterior electrodes were targeted to C2-C3/D2-D3 barrel columns , as mapped by intrinsic optical signal imaging through the cranial window before removal . The microwires were inserted by stereotaxic manipulator to 500–600 µm depth . The reference electrode was located at the anterior part of the array , 3–4 mm from the most posterior part of the array . The ground wire was inserted near the olfactory bulb through a small craniotomy and fixed with a stainless steel microscrew . The array and ground wire were stabilized with dental cement , leaving the Omnetics connector exposed . Electrophysiological measurements were made with a 32 channel amplifier ( ME32 , Multi Channel Systems , Reutlingen , Germany ) sampled at 25 kHz . Raw data was analyzed with custom routines in MATLAB . Local field potential ( LFP ) data was bandpass filtered from 0 . 1 to 300 Hz , and spiking data from 300 Hz to 10 kHz . Multiunit spikes were detected using a threshold of 3 . 5 * SD of the entire recorded voltage per trial . Mechanical whisker stimulation was delivered by inserting a single whisker into a 23 gauge metal tube that was glued to a piezoelectric bending element ( Physik Instrumente PL140 , Karlsruhe , Germany ) . A 1 ms TTL pulse to the piezo current driver ( Physik Instrumente E650 ) triggered a brief rostrocaudal whisker deflection . 30 trials were averaged for each stimulus in stimulus-response curves . Optogenetic stimulus delivery was provided by the same optical fiber as in whisker tracking experiments ( above ) . The tip of the fiber was shielded with blackout tape and placed approximately 2 mm from the right side whisker pad without touching the skin or whiskers . To preclude visual detection of the 460 nm light , blackout tape was attached to a probe and installed in front of the right eye , and ambient green light ( 530 nm ) was used to flood the behavior setup . Custom Arduino routines controlled the timing and structure of trials , including triggering of the LED current driver , detection of lick timing from a capacitive touch sensor coupled to the lick spout , and triggering of the water valve . Water deprived Emx1-Ai27D mice were trained to lick for water reward in response to 100 ms peripheral optogenetic stimulation delivered to the rostral whisker pad . There was no cue for trial initiation; inter-trial interval was randomized from 5 to 10 s . 'Go' trials , when a stimulus was present , could result in either a 'Hit' or a 'Miss' behavioral response . If licks occurred within a 2 s response window after optogenetic stimulation , the trial was recorded as a Hit and mice received a 5 µl water reward . Miss trials occurred when the mouse failed to lick within the post-stimulus response window . 'NoGo' trials , when no stimulus was delivered , could result in either a 'False Alarm' ( FA ) or 'Correct Rejection' ( CR ) behavioral response . A FA response occurred when the mouse licked during the response window of a NoGo trial . Mice were punished for FAs by 2 s presentation of a 5 kHz tone followed by 5–10 s timeout ( in addition to the inter-trial interval ) before the next trial . A CR was recorded when mice did not lick when no stimulus was delivered . Mice were trained using 125 trials per session , two sessions per day . The probability of Go trials was lowered from 60% to 50% upon reaching a FA rate of <0 . 3 . Mice with high FA rate after one week of training received salt water ( 2M NaCl ) dispensed from a second spout as additional punishment ( Rebello et al . , 2014 ) . We used d’ ( d prime ) to measure behavioral performance , defined as z-score ( Hit rate ) – z-score ( FA rate ) , where Hit rate and FA rate are probabilities and z-scores are calculated from a standard normal distribution ( mean = 0 ) with unit variance . To calculate learning curves , the group mean changes in d’ , Hit rate , and FA rate were aligned in each mouse to the first session where d’ remained >1 in subsequent training sessions . In four sessions in each mouse after learning had occurred ( d’>1 ) , we introduced trials with various durations of whisker pad stimulation ( 1 , 5 , 10 , 20 , 40 , or 100 ms ) with the same total stimulus probability . Mice were rewarded with water for licking within the 2 s response window , as in standard training sessions . Trials with stimuli 20–100 ms duration were included in learning curves . In separate sessions , visual catch trials were introduced by delivering 460 nm light through a second optical fiber either in front of the head ( so that light reached the eye ) . Catch trials were randomly interleaved ( 10% stimulus probability ) with whisker pad optical stimulation trials; if mice licked during a catch trial , FA punishment was delivered . The second optical fiber was also used in separate behavior sessions to stimulate S1 with 460 nm light ( 10% stimulus probability ) . In this case , the fiber was placed <1 mm above the cranial window in a location corresponding to previously mapped S1 . Licking responses within the 2 s response window after S1 stimulation were rewarded with water , as for peripheral stimulation . Two Emx1-Cre;Ai27D mice were deeply anesthetized and transcardially perfused with phosphate buffered saline ( PBS ) followed by 4% paraformaldehyde . A commercially available depilatory was applied to the anterior facial region to remove the overlying fur and most of the whiskers protruding from the skin . Whiskers not removed by the depilatory were trimmed close to the skin . The mystacial pad and underlying tissues were dissected and stored overnight in 4% paraformaldehyde at 4°C and then in 30% sucrose in PBS at 4°C until the tissue block had sunk . The tissue was then sectioned at 50 µm with a Leica CM1520 cryostat in either the coronal or transverse plane . Sections were mounted on slides and coverslipped with a glycerol-based mounting medium ( KPL Inc . , Gaithersburg , MD ) . Fluorescent micrographs in Figure 1 were obtained using an Olympus IX51 microscope . Images used for montages were obtained with a 4x objective . Tissue from PV-Cre;Ai27 mice in Figure 1—figure supplement 2 was processes similarly , except images were acquired along with Emx1-Cre;Ai27D tissue using an Olympus FluoView FV1000 confocal microscope and 20x objective . Images for quantification were obtained using the same focus and illumination conditions for each set of measurements of the external extrinsic muscle , intrinsic muscle , and internal extrinsic muscle within a single follicle using a 10x objective . Intensities were measured from four follicles in transverse sections of the mystacial pad . Between follicles the focus and illumination conditions were adjusted to obtain optimal images . Only the field of view was adjusted to obtain images of the three muscle types from within each follicle . To quantify ChR2/tdTomato expression in whisker musculature , fluorescence intensity was measured from three 50 x 50 pixel regions of interest ( 32 . 5 x 32 . 5 µm ) placed over the muscle contained in 1392 x 1040 pixel 16-bit images using ImageJ ( http://imagej . nih . gov/ij/ ) . The intensity measured from each region of interest was then averaged to obtain a single intensity value for each image corresponding to a single intensity value for each muscle within the external to internal span of a follicle . A red look up table was applied to the images and levels were adjusted for display . Levels were adjusted equally for each panel in Figure 1C3 . Values are presented as mean ± SEM , unless otherwise noted . The number of subjects was chosen based on similarity to other in vivo studies and was not predetermined during design of the study . Statistical tests were carried out using Origin Pro or SAS . Student's two-tailed paired tests and one-way repeated measures ANOVA followed by paired contrasts were used for parametric data . Significance was measured at the level of p<0 . 05 . 10 . 7554/eLife . 14140 . 015Video 1 . Whisker movements in response to a light pulse at whisker pad position ( 0 , 0 ) , as in Figure 1—figure supplement 1 . The blue square at the upper right indicates the timing of optogenetic stimulation ( 20 ms , 460 nm ) . The original sampling rate of 500 frames per second was slowed to 25 frames per second for display . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 01510 . 7554/eLife . 14140 . 016Video 2 . Whisker movements in response to a light pulse at whisker pad position ( -3 , -2 ) , as in Figure 1—figure supplement 1 . The blue square at the upper right indicates the timing of optogenetic stimulation ( 20 ms , 460 nm ) . The original sampling rate of 500 frames per second was slowed to 25 frames per second for display . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 01610 . 7554/eLife . 14140 . 017Video 3 . Whisker movements in response to a light pulse at whisker pad position ( -2 , -2 ) , as in Figure 1—figure supplement 1 . The blue square at the upper right indicates the timing of optogenetic stimulation ( 20 ms , 460 nm ) . The original sampling rate of 500 frames per second was slowed to 25 frames per second for display . DOI: http://dx . doi . org/10 . 7554/eLife . 14140 . 017 | Mice use their whiskers to sense their environment and to detect nearby objects . Rather than simply allowing their whiskers to brush passively against objects , mice move them in rhythmic bursts in a process called whisking . Whisking enables neuroscientists to study how the brain gathers and processes actively acquired sensory information . However , controlling active whisker movements in the laboratory has proven challenging . Park et al . now offer a solution based on a technique called optogenetics . The new procedure involves introducing the gene for a light-sensitive ion channel into the facial muscles of the mouse . Shining blue light onto the area of skin where the whiskers grow – the whisker pad – activates these ion channels . Park et al . were able to use this technique to trigger the contraction of the facial muscles and the movement of the whiskers . Furthermore , stimulating different muscles in different areas of the whisker pad produced either forward or backward whisker movements . The strength of the whisker movements varied with the intensity of the light , and with how often and for how long light was applied . Recordings of neural activity showed that sensory signals from light-induced whisker movements reach the same region of the brain as signals from natural whisker movements . Behavioral experiments showed that mice could perceive these whisker movements , despite the fact that they did not generate them . By establishing a method for triggering whisker movements on demand , Park et al . have provided a convenient way of investigating active sensory processing . In addition , this method opens up new possibilities for using optogenetics after injury or degeneration of the nerves that control movement . Ultimately , by using light to trigger muscle contraction directly , it may be possible to restore movement in individuals who have sustained nerve damage through injury or disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"tools",
"and",
"resources",
"neuroscience"
] | 2016 | Peripheral optogenetic stimulation induces whisker movement and sensory perception in head-fixed mice |
The extinct ‘New World stilt-legged’ , or NWSL , equids constitute a perplexing group of Pleistocene horses endemic to North America . Their slender distal limb bones resemble those of Asiatic asses , such as the Persian onager . Previous palaeogenetic studies , however , have suggested a closer relationship to caballine horses than to Asiatic asses . Here , we report complete mitochondrial and partial nuclear genomes from NWSL equids from across their geographic range . Although multiple NWSL equid species have been named , our palaeogenomic and morphometric analyses support the idea that there was only a single species of middle to late Pleistocene NWSL equid , and demonstrate that it falls outside of crown group Equus . We therefore propose a new genus , Haringtonhippus , for the sole species H . francisci . Our combined genomic and phenomic approach to resolving the systematics of extinct megafauna will allow for an improved understanding of the full extent of the terminal Pleistocene extinction event .
The family that includes modern horses , asses , and zebras , the Equidae , is a classic model of macroevolution . The excellent fossil record of this family clearly documents its ~55 million year evolution from dog-sized hyracotheres through many intermediate forms and extinct offshoots to present-day Equus , which comprises all living equid species ( MacFadden , 1992 ) . The downside of this excellent fossil record is that many dubious fossil equid taxa have been erected , a problem especially acute within Pleistocene Equus of North America ( Macdonald et al . , 1992 ) . While numerous species are described from the fossil record , molecular data suggest that most belonged to , or were closely related to , a single , highly variable stout-legged caballine species that includes the domestic horse , E . caballus ( Weinstock et al . , 2005 ) . The enigmatic and extinct ‘New World stilt-legged’ ( NWSL ) forms , however , exhibit a perplexing mix of morphological characters , including slender , stilt-like distal limb bones with narrow hooves reminiscent of extant Eurasian hemionines , the Asiatic wild asses ( E . hemionus , E . kiang ) ( Eisenmann , 1992; Eisenmann et al . , 2008; Harington and Clulow , 1973; Lundelius and Stevens , 1970; Scott , 2004 ) , and dentitions that have been interpreted as more consistent with either caballine horses ( Lundelius and Stevens , 1970 ) or hemionines ( MacFadden , 1992 ) . On the basis of their slender distal limb bones , the NWSL equids have traditionally been considered as allied to hemionines ( e . g . Eisenmann et al . , 2008; Guthrie , 2003; Scott , 2004; Skinner and Hibbard , 1972 ) . Palaeogenetic analyses based on mitochondrial DNA ( mtDNA ) have , however , consistently placed NWSL equids closer to caballine horses ( Barrón-Ortiz et al . , 2017; Der Sarkissian et al . , 2015; Orlando et al . , 2008 , 2009; Vilstrup et al . , 2013; Weinstock et al . , 2005 ) . The current mtDNA-based phylogenetic model therefore suggests that the stilt-legged morphology arose independently in the New and Old Worlds ( Weinstock et al . , 2005 ) and may represent convergent adaptations to arid climates and habitats ( Eisenmann , 1985 ) . However , these models have been based on two questionable sources . The first is based on 15 short control region sequences ( <1000 base pairs , bp; Barrón-Ortiz et al . , 2017; Weinstock et al . , 2005 ) , a data type that can be unreliable for resolving the placement of major equid groups ( Der Sarkissian et al . , 2015; Orlando et al . , 2009 ) . The second consist of two mitochondrial genome sequences ( Vilstrup et al . , 2013 ) that are either incomplete or otherwise problematic ( see Results ) . Given continuing uncertainty regarding the phylogenetic placement of NWSL equids—which impedes our understanding of Pleistocene equid evolution in general—we therefore sought to resolve their position using multiple mitochondrial and partial nuclear genomes from specimens representing as many parts of late Pleistocene North America as possible . The earliest recognized NWSL equid fossils date to the late Pliocene/early Pleistocene ( ~2–3 million years ago , Ma ) of New Mexico ( Azzaroli and Voorhies , 1993; Eisenmann , 2003; Eisenmann et al . , 2008 ) . Middle and late Pleistocene forms tended to be smaller in stature than their early Pleistocene kin , and ranged across southern and extreme northwestern North America ( i . e . eastern Beringia , which includes Alaska , USA and Yukon Territory , Canada ) . NWSL equids have been assigned to several named species , such as E . conversidens Owen 1869 , E . tau Owen 1869 , E . francisci Hay ( 1915 ) , E . calobatus Troxell 1915 , and E . ( Asinus ) cf . kiang , but there is considerable confusion and disagreement regarding their taxonomy . Consequently , some researchers have chosen to refer to them collectively as Equus ( Hemionus ) spp . ( Guthrie , 2003; Scott , 2004 ) , or avoid a formal taxonomic designation altogether ( Der Sarkissian et al . , 2015; Vilstrup et al . , 2013; Weinstock et al . , 2005 ) . Using our phylogenetic framework and comparisons between specimens identified by palaeogenomics and/or morphology , we attempted to determine the taxonomy of middle-late Pleistocene NWSL equids . Radiocarbon ( 14C ) dates from Gypsum Cave , Nevada , confirm that NWSL equids persisted in areas south of the continental ice sheets during the last glacial maximum ( LGM; ~26–19 thousand years before present ( ka BP ) ; Clark et al . , 2009 ) until near the terminal Pleistocene , ~13 thousand radiocarbon years before present ( 14C ka BP ) ( Weinstock et al . , 2005 ) , soon after which they became extinct , along with their caballine counterparts and most other coeval species of megafauna ( Koch and Barnosky , 2006 ) . This contrasts with dates from unglaciated eastern Beringia , where NWSL equids were seemingly extirpated locally during a relatively mild interstadial interval centered on ~31 14C ka BP ( Guthrie , 2003 ) , thus prior to the LGM ( Clark et al . , 2009 ) , final loss of caballine horses ( Guthrie , 2003; 2006 ) , and arrival of humans in the region ( Guthrie , 2006 ) . The apparently discrepant extirpation chronology between NWSL equids south and north of the continental ice sheets implies that their populations responded variably to demographic pressures in different parts of their range , which is consistent with results from some other megafauna ( Guthrie , 2006; Zazula et al . , 2014; Zazula et al . , 2017 ) . To further test this extinction chronology , we generated new radiocarbon dates from eastern Beringian NWSL equids . We analyzed 26 full mitochondrial genomes and 17 partial nuclear genomes from late Pleistocene NWSL equids , which revealed that individuals from both eastern Beringia and southern North America form a single well-supported clade that falls outside the diversity of Equus and diverged from the lineage leading to Equus during the latest Miocene or early Pliocene . This novel and robust phylogenetic placement warrants the recognition of NWSL equids as a distinct genus , which we here name Haringtonhippus . After reviewing potential species names and conducting morphometric and anatomical comparisons , we determined that , based on the earliest-described specimen bearing diagnosable features , francisci Hay is the most well-supported species name . We therefore refer the analyzed NWSL equid specimens to H . francisci . New radiocarbon dates revealed that H . francisci was extirpated in eastern Beringia ~14 14C ka BP . In light of our analyses , we review the Plio-Pleistocene evolutionary history of equids , and the implications for the systematics of equids and other Pleistocene megafauna .
We reconstructed whole mitochondrial genomes from 26 NWSL equids and four New World caballine Equus ( two E . lambei , two E . cf . scotti ) . Using these and mitochondrial genomes of representatives from all extant and several late Pleistocene equids , we estimated a mitochondrial phylogeny , using a variety of outgroups ( Appendix 1 , Appendix 2—tables 1–2 , and Supplementary file 1 ) . The resulting phylogeny is mostly consistent with previous studies ( Der Sarkissian et al . , 2015; Vilstrup et al . , 2013 ) , including confirmation of NWSL equid monophyly ( Weinstock et al . , 2005 ) . However , we recover a strongly supported placement of the NWSL equid clade outside of crown group diversity ( Equus ) , but closer to Equus than to Hippidion ( Figure 1 , Figure 1—figure supplement 1a , Figure 1—source data 1 , and Appendix 2—tables 1–2 ) . In contrast , previous palaeogenetic studies placed the NWSL equids within crown group Equus , closer to caballine horses than to non-caballine asses and zebras ( Barrón-Ortiz et al . , 2017; Der Sarkissian et al . , 2015; Orlando et al . , 2008 , 2009; Vilstrup et al . , 2013; Weinstock et al . , 2005 ) . To explore possible causes for this discrepancy , we reconstructed mitochondrial genomes from previously sequenced NWSL equid specimens and used a maximum likelihood evolutionary placement algorithm ( Berger et al . , 2011 ) to place these published sequences in our phylogeny a posteriori . These analyses suggested that previous results were likely due to a combination of outgroup choice and the use of short , incomplete , or problematic mtDNA sequences ( Appendix 2 and Appendix 2—table 3 ) . To confirm the mtDNA result that NWSL equids fall outside of crown group equid diversity , we sequenced and compared partial nuclear genomes from 17 NWSL equids to a caballine ( horse ) and a non-caballine ( donkey ) reference genome . After controlling for reference genome and ancient DNA fragment length artifacts ( Appendices 1–2 ) , we examined differences in relative private transversion frequency between these genomes ( Appendix 1—figure 1 ) . We found that the relative private transversion frequency for NWSL equids was ~1 . 4–1 . 5 times greater than that for horse or donkey ( Appendix 2 , Figure 1—source data 3 , Figure 1—figure supplement 2 , and Figure 1—source data 2 ) . This result supports the placement of NWSL equids as sister to the horse-donkey clade ( Figure 1—figure supplement 3 ) , the latter of which is representative of living Equus diversity ( e . g . [Der Sarkissian et al . , 2015; Jónsson et al . , 2014] ) and is therefore congruent with the mitochondrial genomic analyses . We estimated the divergence times between the lineages leading to Hippidion , the NWSL equids , and Equus . We first applied a Bayesian time-tree approach to the whole mitochondrial genome data . This gave divergence estimates for the Hippidion-NWSL/Equus split ( node 1 ) at 5 . 15–7 . 66 Ma , consistent with ( Der Sarkissian et al . , 2015 ) , the NWSL-Equus split ( node 2 ) at 4 . 09–5 . 13 Ma , and the caballine/non-caballine Equus split ( node 3 ) at 3 . 77–4 . 40 Ma ( Figure 1 and Figure 1—source data 1 ) . These estimates suggest that the NWSL-Equus mitochondrial split occurred only ~500 thousand years ( ka ) prior to the caballine/non-caballine Equus split . We then estimated the NWSL-Equus divergence time using relative private transversion frequency ratios between the nuclear genomes , assuming a caballine/non-caballine Equus divergence estimate of 4–4 . 5 Ma ( Orlando et al . , 2013 ) and a genome-wide strict molecular clock ( following [Heintzman et al . , 2015] ) . This analysis yielded a divergence estimate of 4 . 87–5 . 69 Ma ( Figure 1—figure supplement 3 ) , which overlaps with that obtained from the relaxed clock analysis of whole mitochondrial genome data ( Figure 1 ) . These analyses suggest that the NWSL equid and Equus clades diverged during the latest Miocene or early Pliocene ( 4 . 1–5 . 7 Ma; late Hemphillian or earliest Blancan ) . The genus Equus ( Linnaeus , 1758 ) was named to include three living equid groups – horses ( E . caballus ) , donkeys ( E . asinus ) , and zebras ( E . zebra ) – whose diversity comprises all extant , or crown group , equids . Previous palaeontological and palaeogenetic studies have uniformly placed NWSL equids within the diversity of extant equids and therefore this genus ( Barrón-Ortiz et al . , 2017; Bennett , 1980; Der Sarkissian et al . , 2015; Harington and Clulow , 1973; Orlando et al . , 2008; 2009; Scott , 2004; Vilstrup et al . , 2013; Weinstock et al . , 2005 ) . This , however , conflicts with the phylogenetic signal provided by palaeogenomic data , which strongly suggest that NWSL equids fall outside the confines of the equid crown group ( Equus ) . Nor is there any morphological or genetic evidence warranting the assignment of NWSL equids to an existing extinct taxon such as Hippidion . We therefore erect a new genus for NWSL equids , Haringtonhippus , as defined and delimited below: Order: Perissodactyla , Owen 1848 Family: Equidae , Linnaeus 1758 Subfamily: Equinae , Steinmann & Döderlein 1890 Tribe: Equini , Gray 1821 Genus: Haringtonhippus , gen . nov . urn:lsid:zoobank . org:act:35D901A7-65F8-4615-9E13-52A263412F67 Type species . Haringtonhippus francisci Hay 1915 .
The suggested placement of NWSL equids within a taxon ( Haringtonhippus ) sister to Equus is a departure from previous interpretations , which variably place the former within Equus , as sister to hemionines or caballine horses ( Figure 1 ) . According to broadly accepted palaeontological interpretations , the earliest equids exhibiting morphologies consistent with NWSL and caballine attribution appear in the fossil record only ~2–3 and ~1 . 9–0 . 7 Ma ago ( Eisenmann et al . , 2008; Forsten , 1992 ) , respectively , whereas our divergence estimates suggest that these lineages to have diverged between 4 . 1–5 . 8 and 3 . 8–4 . 5 Ma , most likely in North America . Dating incongruence might be attributed to an incomplete fossil record , but this seems unlikely given the density of the record for late Neogene and Pleistocene horses . Conversely , incongruence might be attributed to problems with estimating divergence using genomic evidence . However , we emphasize that the NWSL-Equus split is robustly calibrated to the caballine/non-caballine Equus divergence at 4 . 0–4 . 5 Ma , which is in turn derived from a direct molecular clock calibration using a middle Pleistocene horse genome ( Orlando et al . , 2013 ) . Other possibilities to explain the incongruence include discordance between the timing of species divergence and the evolution of diagnostic anatomical characteristics , or failure to detect or account for homoplasy ( Forsten , 1992 ) . For example , Pliocene Equus generally exhibits a primitive ( ‘plesippine’ in North America , ‘stenonid’ in the Old World ) morphology that presages living zebras and asses ( Forsten , 1988 , 1992 ) , with more derived caballine ( stout-legged ) and hemionine ( stilt-legged ) forms evolving in the early Pleistocene . The stilt-legged morphology appears to have evolved independently at least once in each of the Old and New Worlds , yielding the Asiatic wild asses and Haringtonhippus , respectively . We include the middle-late Pleistocene Eurasian E . hydruntinus within the Asiatic wild asses ( following [Bennett et al . , 2017; Burke et al . , 2003; Orlando et al . , 2006] ) , and note that the Old World sussemione E . ovodovi may represent another instance of independent stilt-legged origin , but its relation to Asiatic wild asses and other non-caballine Equus is currently unresolved ( as depicted in Der Sarkissian et al . , 2015; Orlando et al . , 2009; Vilstrup et al . , 2013; and Figure 1 ) . It is plausible that features at the plesiomorphous end of the spectrum , such as those associated with Hippidion , survived after the early to middle Pleistocene at lower latitudes ( South America , Africa; Figure 1 ) . By contrast , the more derived hemionine and caballine morphologies evolved from , and replaced , their antecedents in higher latitude North America and Eurasia , perhaps as adaptations to the extreme ecological pressures perpetuated by the advance and retreat of continental ice sheets and correlated climate oscillations during the Pleistocene ( Forsten , 1992 , Forsten , 1996Forsten , 1996 ) . We note that this high-latitude replacement model is consistent with the turnover observed in regional fossil records for Pleistocene equids in North America ( Azzaroli , 1992; Azzaroli and Voorhies , 1993 ) and Eurasia ( Forsten , 1988 , 1992 , Forsten , 1996 ) . By contrast , in South America Hippidion co-existed with caballine horses until they both succumbed to extinction , together with much of the New World megafauna near the end of the Pleistocene ( Forsten , 1996; Koch and Barnosky , 2006; O'Dea et al . , 2016 ) . This model helps to explain the discordance between the timings of the appearance of the caballine and hemionine morphologies in the fossil record and the divergence of lineages leading to these forms as estimated from palaeogenomic data . Although we can offer no solution to the general problem of mismatches between molecular and morphological divergence estimators–an issue scarcely unique to equid systematics–this model predicts that some previously described North American Pliocene and early Pleistocene Equus species ( e . g . E . simplicidens , E . idahoensis; [Azzaroli and Voorhies , 1993] ) , or specimens thereof , may be ancestral to extant Equus and/or late Pleistocene Haringtonhippus . Three new radiocarbon dates of ~14 . 4 14C ka BP from a Yukon Haringtonhippus fossil greatly extends the known temporal range of this genus in eastern Beringia . This result demonstrates , contrary to its previous LAD of 31 , 400 ± 1200 14C years ago ( AA 26780; [Guthrie , 2003] ) , that Haringtonhippus survived throughout the last glacial maximum in eastern Beringia ( Clark et al . , 2009 ) and may have come into contact with humans near the end of the Pleistocene ( Goebel et al . , 2008; Guthrie , 2006 ) . These data suggest that populations of stilt-legged Haringtonhippus and stout-legged caballine Equus were sympatric , both north and south of the continental ice sheets , through the late Pleistocene and became extinct at roughly the same time . The near synchronous extinction of both horse groups across their entire range in North America suggests that similar causal mechanisms may have led each to their demise . The sympatric nature of these equids raises questions of whether they managed to live within the same community without hybridizing or competing for resources . Extant members of the genus Equus vary considerably in the sequence of Prdm9 , a gene involved in the speciation process , and chromosome number ( karyotype ) ( Ryder et al . , 1978; Steiner and Ryder , 2013 ) , and extant caballine and non-caballine Equus rarely produce fertile offspring ( Allen and Short , 1997; Steiner and Ryder , 2013 ) . It is unlikely , therefore , that the more deeply diverged Haringtonhippus and caballine Equus would have been able to hybridize . Future analysis of high coverage nuclear genomes , ideally including an outgroup such as Hippidion , will make it possible to test for admixture that may have occurred soon after the lineages leading to Haringtonhippus and Equus diverged , as occurred between the early caballine and non-caballine Equus lineages ( Jónsson et al . , 2014 ) . It may also be possible to use isotopic and/or tooth mesowear analyses to assess the potential of resource partitioning between Haringtonhippus and caballine Equus in the New World . Fossils of NWSL equids have been known for more than a century , but until the present study their systematic position within Plio-Pleistocene Equidae was poorly characterized . This was not because of a lack of interest on the part of earlier workers , whose detailed anatomical studies strongly indicated that what we now call Haringtonhippus was related to Asiatic wild asses , such as Tibetan khulan and Persian onagers , rather than to caballine horses ( Eisenmann et al . , 2008; Guthrie , 2003; Scott , 2004; Skinner and Hibbard , 1972 ) . That the cues of morphology have turned out to be misleading in this case underlines a recurrent problem in systematic biology , which is how best to discriminate authentic relationships within groups , such as Neogene equids , that were prone to rampant convergence . The solution we adopted here was to utilize both palaeogenomic and morphometric information in reframing the position of Haringtonhippus , which now clearly emerges as the closest known outgroup to all living Equus . Our success in this regard demonstrates that an approach which incorporates phenomics with molecular methods ( palaeogenomic as well as palaeoproteomic , e . g . [Welker et al . , 2015] ) is likely to offer a means for securely detecting relationships within speciose groups that are highly diverse ecomorphologically . All methods have their limits , with taphonomic degradation being the critical one for molecular approaches . However , proteins may persist significantly longer than ancient DNA ( e . g . [Rybczynski et al . , 2013] ) , and collagen proteomics may come to play a key role in characterizing affinities , as well as the reality , of several proposed Neogene equine taxa ( e . g . Dinohippus , Pliohippus , Protohippus , Calippus , and Astrohippus; [MacFadden , 1998] ) whose distinctiveness and relationships are far from settled ( Azzaroli and Voorhies , 1993; Forsten , 1992 ) . A reciprocally informative approach like the one taken here holds much promise for lessening the amount of systematic noise , due to oversplitting , that hampers our understanding of the evolutionary biology of other major late Pleistocene megafaunal groups such as bison and mammoths ( Enk et al . , 2016; Froese et al . , 2017 ) . This approach is clearly capable of providing new insights into just how extensive megafaunal losses were at the end of the Pleistocene , in what might be justifiably called the opening act of the Sixth Mass Extinction in North America .
We recovered Yukon fossil material ( 17 Haringtonhippus francisci , two Equus cf . scotti , and two E . lambei; Supplementary file 1 ) from active placer mines in the Klondike goldfields near Dawson City . We further sampled seven H . francisci fossils from the contiguous USA that are housed in collections at the University of Kansas Biodiversity Institute ( KU; n = 4 ) , Los Angeles County Museum of Natural History ( LACM ( CIT ) ; n = 2 ) , and the Texas Vertebrate Paleontology Collections at The University of Texas ( TMM; n = 1 ) . We radiocarbon dated the Klondike fossils and the H . francisci cranium from the LACM ( CIT ) ( Supplementary file 1 ) . For morphometric analysis , we took measurements of third metatarsals ( MTIII ) and other elements . We used a reduced data set of four MTIII variables for principal components analysis and performed logistic regression on the first three principal components , computed in R ( R Development Core Team , 2008 ) ( Source code 1 ) . We conducted all molecular biology methods prior to indexing PCR in the dedicated palaeogenomics laboratory facilities at either the UC Santa Cruz or Pennsylvania State University . We extracted DNA from between 100 and 250 mg of bone powder following either Rohland et al . ( 2010 ) or Dabney et al . ( 2013a ) . We then converted DNA extracts to libraries following the Meyer and Kircher protocol ( Meyer and Kircher , 2010 ) , as modified by ( Heintzman et al . , 2015 ) or the PSU method of ( Vilstrup et al . , 2013 ) . We enriched libraries for equid mitochondrial DNA . We then sequenced all enriched libraries and unenriched libraries from 17 samples using Illumina platforms . We prepared raw sequence data for alignment and mapped the filtered reads to the horse reference mitochondrial genome ( Genbank: NC_001640 . 1 ) and a H . francisci reference mtDNA genome ( Genbank: KT168321 ) , resulting in mitogenomic coverage ranging from 5 . 8× to 110 . 7× ( Supplementary file 1 ) . We were unable to recover equid mtDNA from TMM 34–2518 ( the francisci holotype ) using this approach ( Appendix 2 ) . We supplemented our mtDNA genome sequences with 38 previously published complete equid mtDNA genomes . We constructed six alignment data sets and selected models of molecular evolution for the analyses described below ( Appendix 1—table 1 , and Supplementary file 1; Heintzman et al . , 2017 ) . We tested the phylogenetic position of the NWSL equids ( =H . francisci ) using mtDNA data sets 1–3 and applying Bayesian ( Ronquist et al . , 2012 ) and maximum likelihood ( ML; [Stamatakis , 2014] ) analyses . We varied the outgroup , the inclusion or exclusion of the fast-evolving partitions , and the inclusion or exclusion of Hippidion sequences . Due to the lack of a globally supported topology across the Bayesian and ML phylogenetic analyses , we used an Evolutionary Placement Algorithm ( EPA; [Berger et al . , 2011] ) to determine the a posteriori likelihood of phylogenetic placements for candidate equid outgroups using mtDNA data set four . We also used the same approach to assess the placement of previously published equid sequences ( Appendix 2 ) . To infer divergence times between the four major equid groups ( Hippidion , NWSL equids , caballine Equus , and non-caballine Equus ) , we ran Bayesian timetree analyses ( Drummond et al . , 2012 ) using mtDNA data set five . We varied these analyses by including or excluding fast-evolving partitions , constrained the root height or not , and including or excluding the E . ovodovi sequence . To facilitate future identification of equid mtDNA sequences , we constructed , using data set six , a list of putative synapomorphic base states , including indels and substitutions , that define the genera Hippidion , Haringtonhippus , and Equus at sites across the mtDNA genome . To test whether our mtDNA genome-based phylogenetic hypothesis truly reflects the species tree , we compared the nuclear genomes of a horse ( EquCab2 ) , donkey ( Orlando et al . , 2013 ) , and the shotgun sequence data from 17 of our NWSL equid samples ( Figure 1—source data 2 , Appendix 1 , Appendix 1—figure 1 , and Supplementary file 1 ) . We applied four successive approaches , which controlled for reference genome and DNA fragment length biases ( Appendix 1 ) . We estimated the divergence between the NWSL equids and Equus ( horse and donkey ) by fitting the branch length , or relative private transversion frequency , ratio between horse/donkey and NWSL equids into a simple phylogenetic scenario ( Figure 1—figure supplement 3 ) . We then multiplied the NWSL equid branch length by a previous horse-donkey divergence estimate ( 4 . 0–4 . 5 Ma; [Orlando et al . , 2013] ) to give the estimated NWSL equid-Equus divergence date , following ( Heintzman et al . , 2015 ) and assuming a strict genome-wide molecular clock ( Figure 1—figure supplement 3 ) . We determined the sex of the 17 NWSL equid samples by comparing the relative mapping frequency of the autosomes to the X chromosome . We assessed the prevalence of mitochondrial and nuclear DNA damage in a subset of the equid samples using mapDamage ( Jónsson et al . , 2013 ) . Repository details and associated metadata for curated samples can be found in Supplementary file 1 . MTIII and other element measurement data are in Figure 2—source data 1 , and the Rscript used for morphometric analysis is in the DRYAD database ( Heintzman et al . , 2017 ) . MtDNA genome sequences have been deposited in Genbank under accessions KT168317-KT168336 , MF134655-MF134663 , and an updated version of JX312727 . All mtDNA genome alignments ( in NEXUS format ) and associated XML and TREE files are in the DRYAD database ( Heintzman et al . , 2017 ) . Raw shotgun sequence data used for the nuclear genomic analyses and raw shotgun and target enrichment sequence data for TMM 34–2518 ( francisci holotype ) have been deposited in the Short Read Archive ( BioProject: PRJNA384940 ) . The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature , and hence the new name contained herein is available under that Code from the electronic edition of this article . This published work and the nomenclatural act it contains have been registered in ZooBank , the online registration system for the ICZN . The ZooBank LSIDs ( Life Science Identifiers ) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix ‘http://zoobank . org/' . The LSID for this publication is: urn:lsid:zoobank . org:pub:8D270E0A-9148-4089-920C-724F07D8DC0B . The electronic edition of this work was published in a journal with an ISSN , and has been archived and is available from the following digital repositories: PubMed Central and LOCKSS . | The horse family – which also includes zebras , donkeys and asses – is often featured on the pages of textbooks about evolution . All living horses belong to a group , or genus , called Equus . The fossil record shows how the ancestors of these animals evolved from dog-sized , three-toed browsers to larger , one-toed grazers . This process took around 55 million years , and many members of the horse family tree went extinct along the way . Nevertheless , the details of the horse family tree over the past 2 . 5 million years remain poorly understood . In North America , horses from this period – which is referred to as the Pleistocene – have been classed into two major groups: stout-legged horses and stilt-legged horses . Both groups became extinct near the end of the Pleistocene in North America , and it was not clear how they relate to one another . Based on their anatomy , many scientists suggested that stilt-legged horses were most closely related to modern-day asses living in Asia . Yet , other studies using ancient DNA placed the stilt-legged horses closer to the stout-legged horses . Heintzman et al . set out to resolve where the stilt-legged horses sit within the horse family tree by examining more ancient DNA than the previous studies . The analyses showed that the stilt-legged horses were much more distinct than previously thought . In fact , contrary to all previous findings , these animals actually belonged outside of the genus Equus . Heintzman et al . named the new genus for the stilt-legged horses Haringtonhippus , and showed that all stilt-legged horses belonged to a single species within this genus , Haringtonhippus francisci . Together these new findings provide a benchmark for reclassifying problematic fossil groups across the tree of life . A similar approach could be used to resolve the relationships in other problematic groups of Pleistocene animals , such as mammoths and bison . This would give scientists a more nuanced understanding of evolution and extinction during this period . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"genetics",
"and",
"genomics"
] | 2017 | A new genus of horse from Pleistocene North America |
The P2X7 channel is involved in the pathogenesis of various CNS diseases . An increasing number of studies suggest its presence in neurons where its putative functions remain controversial for more than a decade . To resolve this issue and to provide a model for analysis of P2X7 functions , we generated P2X7 BAC transgenic mice that allow visualization of functional EGFP-tagged P2X7 receptors in vivo . Extensive characterization of these mice revealed dominant P2X7-EGFP protein expression in microglia , Bergmann glia , and oligodendrocytes , but not in neurons . These findings were further validated by microglia- and oligodendrocyte-specific P2X7 deletion and a novel P2X7-specific nanobody . In addition to the first quantitative analysis of P2X7 protein expression in the CNS , we show potential consequences of its overexpression in ischemic retina and post-traumatic cerebral cortex grey matter . This novel mouse model overcomes previous limitations in P2X7 research and will help to determine its physiological roles and contribution to diseases .
The P2X7 receptor differs from all other P2X family members by its low sensitivity to ATP , a particularly long intracellular C-terminus , and its ability to trigger various short and long-term cellular events like the induction of dye uptake and cell death ( Saul et al . , 2013; Surprenant et al . , 1996; Li et al . , 2015; Harkat et al . , 2017; Di Virgilio et al . , 2018 ) , which are incompletely understood and have mostly been described in cell culture systems . Probably best studied is its central role in NLRP3 ( NOD-like receptor family , pyrin domain containing 3 ) inflammasome activation , cytokine maturation , and inflammation that was originally established in P2X7 knockout ( P2rx7-/- ) mouse models ( Chessell et al . , 2005; Solle et al . , 2001; Di Virgilio et al . , 2017 ) . A growing body of evidence suggests that an increased P2X7 receptor function plays a role in various diseases of the central nervous system ( CNS ) and peripheral nervous system ( PNS ) and further supports its importance as a drug target ( Bhattacharya and Biber , 2016; Rassendren and Audinat , 2016; Sperlágh and Illes , 2014; Sociali et al . , 2016 ) . P2X7 receptors are expressed in cells of hematopoietic origin as well as different types of glial , epithelial , and endothelial cells . The presence and function of P2X7 receptors in neurons , however , remains a matter of debate ( Illes et al . , 2017; Miras-Portugal et al . , 2017 ) , although an increasing amount of literature describing neuronal P2X7 functions imply a wide acceptance of this view ( e . g . in [Sperlágh and Illes , 2014; Brown et al . , 2016; Engel et al . , 2012; Gulbransen et al . , 2012; Jimenez-Pacheco et al . , 2013] ) . In support of a neuronal expression , P2rx7 mRNA was detected in neurons ( Yu et al . , 2008 ) , P2X7 protein was identified in neuronal cell culture ( Ohishi et al . , 2016 ) , and P2X7 receptors were pharmacologically shown to facilitate postsynaptic efficacy and affect neurotransmitter release ( reviewed in [Sperlágh and Illes , 2014] ) . However , mRNA expression might not necessarily correlate with synthesis of the respective protein ( Carpenter et al . , 2014 ) , selectivity of the available P2X7-specific antibodies has been questioned ( Anderson and Nedergaard , 2006; Sim et al . , 2004 ) , and pharmacology of purinergic receptors is rather complex ( Anderson and Nedergaard , 2006; Compan et al . , 2012; Nörenberg et al . , 2016 ) . Also , it has been difficult to differentiate between direct effects of neuronal P2X7 activation and indirect effects of ATP-activated neurotransmitter release from glia cells ( Sperlágh and Illes , 2014; Illes et al . , 2017; Miras-Portugal et al . , 2017 ) . Taken together , the scarcity of information regarding the localization and the molecular and physiological functions of P2X7 receptors in the nervous system stands in sharp contrast to its proposed role as a drug target . To conclusively resolve these important questions , we generated transgenic mouse lines that overexpress EGFP-tagged P2X7 under the control of a BAC-derived mouse P2X7 gene ( P2rx7 ) promoter . These mice allow the direct and indirect visualization of P2X7 , its purification , and determination of functional consequences of its overexpression . Using this model , we provide the first comprehensive and quantitative analysis of the distribution of P2X7 protein within the CNS .
The P2rx7 cDNA was obtained from C57BL/6 mouse brain and C-terminally fused to the EGFP-sequence via a Strep-tagII-Gly-7xHis-Gly linker sequence ( Figure 1—figure supplement 1A ) to provide additional labeling/purification options and minimize interference with the receptor function . As two allelic P2X7 variants , 451P ( ‘wt’ ) and 451L ( SNP , present in C57BL/6 ) , with different functionality have been described ( Adriouch et al . , 2002; Sorge et al . , 2012 ) , the ‘wt’ L451P-variant was also generated by site directed mutagenesis . Efficient expression and functionality of the full-length proteins were confirmed by SDS-PAGE , patch-clamp analysis , and ATP-induced ethidium uptake in HEK cells ( Figure 1—figure supplement 1B–E ) . Both variants and the non-tagged receptors revealed similar EC50 values , indicating that the dye uptake properties of the P2X7 receptor were not influenced by the EGFP-tag . Also , current kinetics were virtually identical . Next , BAC clone RP24-114E20 , containing the full length P2rx7 and more than 100 kb of the 5´region was modified accordingly by insertion of the Strep-His-EGFP sequence in exon 13 to preserve the exon-intron structure of the gene ( Figure 1A ) . Upon verification by Southern blotting ( Figure 1B , Figure 1—figure supplement 1F ) and sequencing , the linearized BAC was injected into pronuclei of FVB/N mouse oocytes ( 451L background ) . In total , 4 ( 451L ) and 10 ( 451P ) germline transmitters were obtained and five lines ( 451L: lines 46 , 59 and 61; 451P: lines 15 and 17 ) were selected for initial characterization as described below ( Figure 1C and Figure 2—figure supplement 1 ) . Subsequent experiments were performed with the highest expressing line 17 . Southern blot analysis revealed integration of 4–15 BAC copies in the different lines . The copy numbers correlated well with the respective P2X7-EGFP protein expression levels ( Figure 1C ) , suggesting the functionality of most if not all integrated P2rx7 BAC transgenes . Endogenous P2X7 protein synthesis was unaffected by the P2X7-EGFP overexpression ( Figure 1—figure supplement 1G ) . Purification of P2X7-EGFP protein via Ni-NTA agarose demonstrated co-purification of endogenous P2X7 subunits confirming efficient co-assembly of tagged and non-tagged subunits ( Figure 1D ) . In agreement with correct plasma membrane targeting , deglycosylation with endoglycosidase H and PNGase F revealed efficient complex glycosylation , indicating that the EGFP-tag did not disturb folding and ER-exit of the transgenic P2X7-EGFP protein ( Figure 1E ) . To demonstrate functionality of the overexpressed P2X7-EGFP protein , the transgenic mice were , upon backcrossing into C57BL/6 for 8–10 generations , mated to P2rx7-/- mice ( in C57BL/6 ) to obtain ‘rescue’ mice ( Table 1 ) that express only the transgenic but not the endogenous P2X7 ( Figure 1F ) . FACS analysis of microglia from these mice confirmed that the transgene is able to fully rescue the ATP-induced DAPI uptake , which is absent in these cells from P2rx7-/- mice ( Figure 1G ) . Comparison of the kinetic and efficiency of DAPI uptake by simultaneous analysis of pooled and differentially labeled microglia revealed a stronger increase in the rescue mice compared to wt mice , most likely due to a higher number of functional P2X7 receptors at the cell surface . The specificity of the DAPI uptake was demonstrated using the P2X7 antagonist A438079 ( Figure 1—figure supplement 2 ) . To determine the overall pattern of P2X7 localization in the brain , 3 , 3'-Diaminobenzidine ( DAB ) staining with EGFP-specific antibodies was performed on brain slices from all mouse lines . As shown for five selected lines ( Figure 2A , Figure 2—figure supplement 1 ) , specific labeling with identical patterns ( Table 2 ) was obtained . A particularly high P2X7-EGFP density was found in the molecular layers of the cerebellar cortex . In addition , strong labeling was detected in the molecular layer of the dentate gyrus ( DG ) , the cerebral cortex and olfactory bulb as well as the thalamus , hypothalamus , substantia nigra , and ventral pons . Comparison of endogenous and transgenic P2X7 levels in different brain regions ( Figure 2B ) showed similar protein ratios and tissue-specific intensities , demonstrating that expression of the transgene mirrored both the expression pattern and expression level of endogenous P2X7 and thus implying that important regulatory elements governing P2X7 expression are preserved and functional in the chosen BAC construct . Due to the dense but diffuse signal and a higher background fluorescence ( probably due to structural organization and/or a high content of endogenous fluorophores ) in the cerebellar cortex , identification of cellular structures and P2X7-EGFP-expressing cell types proved difficult in adult cerebellum ( Figure 2—figure supplement 2A–C ) : Using confocal microscopy , no conclusive co-localization was seen with Purkinje cell ( calbindin D28k ) and synaptic ( vGlut2 ) marker proteins nor with astrocytes/Bergmann glia ( GFAP or S100β ) . However , analysis of the cerebellum from animals at postnatal day 7 ( P7 ) , before Bergmann glia microdomains are formed ( Grosche et al . , 2002 ) , revealed a more structured and clearer GFP signal that aligned with the cell bodies and radial extensions of S100β-immunopositive cells with typical Bergman glia morphology ( Figure 2C ) . Thus , we conclude that P2X7 is expressed in Bergmann glia , in agreement with previous findings ( Habbas et al . , 2011 ) . However , due to the localization in microdomains that give a very diffuse pattern , co-localization with the intracellular Bergmann glia marker GFAP , which visualizes mainly their radial extensions , could not be detected and co-localization with S100β was only dissolved during postnatal development . In addition to Bergman glia , P2X7 is present in microglia of the cerebellum , which was confirmed in acutely dissociated cells from adult tissue ( Figure 2-figure supplement 2D ) . Based on our data ( Figure 2E , Figure 2—figure supplement 2B ) , we exclude the expression in Purkinje cells in both adult and P7 mice . Specificity of the GFP labeling and congruency with the endogenous P2X7 expression pattern was further confirmed using a novel mouse P2X7-specific heavy chain antibody ( nanobody 7E2 fused to the hinge , CH2 and CH3 domains of rabbit-IgG ( 7E2-rbIgG ) , see ( Danquah et al . , 2016 ) and Materials and methods ) ( Figure 2D and E ) . Co-immunolabeling with cell type-specific markers ( Figure 3A–F ) and quantification of GFP-positive cells in the CA1 region of the hippocampus ( Figure 3G ) demonstrated that P2X7-EGFP is predominantly ( 57 ± 14% ) expressed in microglial cells ( 93% of all Iba1-positive cells ) , while the majority , if not all , of the remaining GFP-positive cells ( 47 ± 10% ) belong to the oligodendroglial lineage and co-express Olig2 ( 87% of Olig2-positive and 95% of NG2-positive oligodendrocyte precursor cells ) . In addition , 7 ± 4% of P2X7-EGFP-expressing cells represent S100β-positive but GFAP-negative cells . These cells comprised 8% of all S100β-positive cells and may represent either GFAP-negative astroglial cells or oligodendrocyte precursor cells . This distribution is in agreement with functional findings ( Jabs et al . , 2007 ) and cell type-specific RNA sequencing data ( P2rx7 mRNA in microglia/oligodendrocytes/astrocytes ≈ 28/26/5 fragments per kilobase of transcript sequence per million mapped fragments ) obtained from cerebral cortex ( Zhang et al . , 2014 ) ( http://web . stanford . edu/group/barres_lab/brain_rnaseq . html ) . Likewise , co-staining with Sox9 ( for astrocytes ) or neuronal ( NeuN , MAP2 ) and synaptic ( VGlut1 , PSD95 , synaptophysin , VGAT ) markers did not reveal any overlap in the CA1 , CA3 , and dentate gyrus ( Figure 3A , Figure 3—figure supplements 1 , 2 and 3B ) . A clear band of more intense GFP signal is regularly detected in the molecular layer of the dentate gyrus ( e . g . Figure 2A , Figure 3—figure supplement 1B , bottom-right panel ) and was attributed to a higher number and/or more ramified morphology of microglia that align at the border of the granular layer . In support of this explanation , the thickness of the band in this region equals the radius of microglia with their extensions ( Figure 3—figure supplement 3A ) . As P2X7 protein expression has been described in nestin-positive neuronal/glia precursor cells in the subgranular zone ( Rozmer et al . , 2017 ) , we also performed co-labeling of EGFP with nestin in this region but did not detect any co-localization ( Figure 3—figure supplement 4 ) . Likewise , no co-staining of EGFP with neurons was seen in other regions with a strong EGFP signal like basal ganglia , hypothalamus , and pons ( Figure 3—figure supplement 5 ) . In addition , we performed co-stainings of brain sections with the commercially available P2X7-specific antibodies and the nanobody-rbIgG fusion construct 7E2-rbIgG ( Danquah et al . , 2016 ) . However , the commercially available antibodies yielded unspecific or insufficient staining ( either in comparison to P2rx7-/- mice or in the P2X7-EGFP overexpressing line 17 ) ( Figure 3—figure supplement 6 ) . In contrast , 7E2-rbIgG showed specific staining of endogenous P2X7 protein in wild-type ( wt ) but not P2rx7-/- mice ( Figure 2E ) and clear overlap with the transgenic P2X7-EGFP ( Figure 2E , Figure 3—figure supplement 6 ) . To further verify that the observed transgene expression pattern correlates with the endogenous P2X7 expression , mice deficient in microglial or oligodendroglial P2X7 were generated by mating P2rx7fl/fl mice with Cx3cr1tm1 . 1 ( cre ) Jung ( Yona et al . , 2013 ) and Cnptm1 ( cre ) Kan lines ( Lappe-Siefke et al . , 2003 ) , respectively ( specificity of Cre expression is shown in Figure 3—figure supplement 7 ) . In comparison to P2rx7fl/fl mice , Cx3cr1-Cre-positive mice showed 51 . 5% ( ±4 . 5% ) and Cnp-Cre-positive mice showed 60 . 4% ( ±2 . 9% ) reduction of P2X7 protein in the brain , which correlates well with the percentage of P2X7 expressing cells determined in the brains of our transgenic mice ( Figure 3H ) . Since neuronal P2X7 expression and function has been described in amacrine cells ( interneurons ) as well as in ganglion cells , photoreceptors , and pigment epithelial cells of the retina ( Sanderson et al . , 2014 ) , we further probed if neuronal P2X7 expression was detectable in this tissue with histologically clear architecture . In contrast to previous reports , however , P2X7-EGFP was exclusively expressed in microglia ( Figure 4A ) . Likewise , P2X7-EGFP expression was not found in neurons of the spinal cord , DRG , or of teased sciatic nerve fibers ( Figure 4B–D , Figure 4—figure supplement 1 ) . In Schwann cells of the sciatic nerve fibers , however , P2X7-EGFP was localized to nodes of Ranvier and Schmidt-Lantermann incisures , in perfect agreement with the subcellular distribution pattern of endogenous P2X7 ( Figure 4C ) . At the neuromuscular junction , P2X7-EGFP was found in close association with terminal Schwann cells ( S100β-positive ) , but did not co-localize with them or with the post- ( α-bungarotoxin-positive ) or presynaptic ( synaptophysin-positive ) membrane , in contrast to previous findings ( Deuchars et al . , 2001 ) . Based on the localization and morphology , we suggest its presence on kranocytes , a fibroblast-like cell type ( Court et al . , 2008 ) . In agreement with previous reports on P2X7 localization in DRGs ( Zhang et al . , 2005; Jager and Vaegter , 2016 ) , we identified P2X7-EGFP in cells that show the localization and typical morphology of satellite glia cells which ensheath large sensory neurons . This was confirmed by co-labeling experiments using the satellite cell marker glutamine synthetase . Unlike in sciatic nerves , however , it was not found in myelin protein zero ( MPZ ) -positive Schwann cells of the DRG ( Figure 4—figure supplement 1A ) . Finally , P2X7-EGFP localization was investigated in the myenteric plexus of the colon , a part of the enteric nervous system , but was also not detected in neurons or GFAP-positive glia ( Figure 4—figure supplement 1B ) . Detailed analyses of brain parenchyma and other types of nervous tissues indicates that the BAC transgenic P2X7-EGFP is correctly regulated in our mouse model and that P2X7 protein is either below detection limit or not synthesized in neurons , at least under physiological conditions in adult mice . Despite the well-documented role of P2X7 in inflammation and cell death , P2X7 overexpression did not result in an obvious pathology or behavioral changes under physiological conditions ( Figure 5—figure supplement 3A–D ) . To test if tissue damage , that is a condition that is associated with neuroinflammation , could induce neuronal P2X7-EGFP synthesis , we proceeded our analysis with three experimental models of acute and/or invasive CNS injury: ischemic retina , stab wound , and kainic acid-induced status epilepticus . In preliminary experiments with a small number of animals ( n = 3–4 ) , an increased microglia reaction ( Figure 5A ) and microglia number ( Figure 5B ) as well as other Iba1/P2X7-positive cells ( possibly invading macrophages , Figure 5—figure supplement 1 ) were observed upon transient retinal ischemia in wt animals . Interestingly , this effect appeared to be enhanced in P2X7-EGFP transgenic mice ( Figure 5A , B ) . Importantly , however , P2X7 was not upregulated in other cell types than microglia , at least 3 days post injury ( Figure 5—figure supplement 1 ) . In this context , it should be emphasized , that a similar trend was observed in mice subjected to the stab wound injury of the somatosensory grey matter ( GM ) . Compared to the situation in wt mice ( n = 2 ) at 5 days after injury , post-traumatic GM of P2X7-EGFP transgenic mice ( n = 3 ) showed a trend toward increased reactivity of microglial cells at the injury site and increased lesion area ( Figure 5C ) . These data support the functionality and correct transcriptional regulation of the construct and suggest a deleterious effect of P2X7 overactivation on neurons ( Sperlágh and Illes , 2014 ) . The low number of animals , however , requires confirmation in future studies . Nevertheless , no induction of neuronal or astroglial P2X7-EGFP synthesis was found in the affected lesion area ( Figure 5D ) , although we cannot exclude a potential obfuscation of the EGFP signal due to autofluorescence in the direct vicinity to the injury . Finally , no induction of P2X7 protein expression was observed in neurons of the dentate gyrus , CA1 , and CA3 regions 24 hr after induction of status epilepticus by a unilateral intra-amygdala kainic acid injection , although a change of microglia morphology clearly indicated their activation ( Figure 5—figure supplement 2 ) . In conclusion , we suggest that P2X7-dependent neurodegeneration that has been observed in various studies is caused by an indirect mechanism , most likely involving P2X7 activation in microglia or oligodendrocytes .
Numerous examples demonstrate that BAC transgenics are valuable tools to investigate endogenous protein expression patterns ( Gerfen et al . , 2013; Yang and Gong , 2005 ) . In comparison to knock-in approaches , they provide the advantages of a stronger signal due to the moderate overexpression which might boost physiological functions and thus make them accessible for an in depth analysis . Together with classical and/or conditional knock-out strategies , this provides a powerful combination for the in vivo analysis of protein functions . In contrast to small transgenes , in which the expression patterns are often affected by the integration site , BAC transgenes show in most cases an expression pattern that reflects the endogenous promoter within the BAC . However , position effects such as gene deletion or aberrant expression due to integration of ( truncated ) BAC transgenes in other genes or regulatory elements cannot be excluded . Thus , five transgenic lines were analyzed in detail in this study , and all showed identical expression patterns . A correct expression pattern was further corroborated by 1 ) a comparative analysis of P2X7-EGFP transgene expression and endogenous P2X7 expression in wt mice using a novel P2X7-specific nanobody-rbIgG fusion construct and 2 ) cell type-specific P2X7 deletion in oligodendrocytes and microglia using conditional P2X7 knock-out mouse models . Together , these findings all argue against an ectopic P2X7-EGFP expression pattern and indicate a predominant expression of P2X7 in non-neuronal cells within the brain parenchyma . The absence of a neuronal localization of P2X7-EGFP contrasts with findings in a BAC transgenic reporter mouse line ( Tg ( P2rx7 EGFP ) FY174Gsat; GenSat ) in which soluble EGFP is expressed under the control of a BAC transgenic P2X7 promoter and a recently described humanized P2X7 knock-in mouse model ( Metzger et al . , 2017 ) . Both models show neuronal P2X7 expression but differ remarkably in the observed expression patterns ( Engel et al . , 2012; Jimenez-Pacheco et al . , 2013; Metzger et al . , 2017 ) . Neuronal expression of the soluble EGFP reporter in the GenSat mouse is seen in multiple brain regions whereas P2X7 transcripts in the humanized P2X7 mouse model seem to predominate in CA3 neurons ( Metzger et al . , 2017 ) ( Table 2 ) . In contrast to our findings at the protein level , the knock-in model showed only a minor reduction in P2X7 expression in microglia-specific knock-out animals . A possible explanation for these discrepancies could be that alterations in gene structure introduced into the GenSat P2X7 BAC EGFP mice influence post-transcriptional and translational regulatory mechanisms . For example , intron 1 , the importance of which is evidenced by the P2X7 k splice variant ( Nicke et al . , 2009 ) , is not fully preserved in the EGFP reporter mouse and soluble EGFP is translated from a truncated mRNA which might lack regulatory elements . In case of the humanized P2X7 model , only RNA transcripts were analyzed which might not correlate with protein synthesis . Moreover , different sensitivities of the detection methods ( in situ hybridization versus immunohistochemistry ) could also account for some of the observed differences . As even minor gene modifications , such as the flanking of the exons with the comparatively small loxP sequences , are able to influence gene expression ( Requardt et al . , 2009; Zhang et al . , 2013 ) ( Figure 3H ) , we kept the P2rx7 gene structure almost untouched and retained as much of the UTRs as possible to avoid such unpredictable influences . Although we cannot completely rule out an effect of the introduced gene modification within the BAC or a loss of possible modulatory elements that lie outside of the chosen BAC , comparison of the BAC transgenic P2X7-EGFP with endogenous P2X7 expression as analyzed with the novel nanobody-based heavy chain antibody , provide strong evidence that the expression pattern observed in our mouse line matches that of the endogenous protein . This , however , does not preclude species differences . For example , in human but not in murine Müller cells , functional and immunohistological evidence for P2X7 expression has been shown ( Franke et al . , 2005; Innocenti et al . , 2004; Pannicke et al . , 2000 ) . Finally , we cannot exclude neuronal expression that is below the detection limit . As the P2X7 receptor is known to be a highly regulated protein and has been shown to be deleterious to cultured neuronal cells ( Ohishi et al . , 2016 ) , its expression and localization should be tightly regulated in post-mitotic cells like neurons . If present in neurons , its presence would likely be limited to subcellular regions were synapse formation and selection takes place and/or to areas where damaged cells need to be removed by apoptosis . Possible extrasynaptic or growth cones localizations ( Díaz-Hernandez et al . , 2008 ) would be difficult or impossible to resolve in situ using the described antibodies and conventional microscopy . However , even upon induction of tissue damage , virtually no P2X7 expression in cerebral cell types other than microglia or oligodendrocytes was observed . Based on our data , we therefore suggest that P2X7-induced neurodegeneration is due to an indirect effect ( i . e . extended glial reaction within the acute post-traumatic period ) , which requires further investigation . Five murine P2X7 splice variants ( Masin et al . , 2012; Nicke et al . , 2009 ) have been identified . Two of these ( variants a and k ) contain the C-terminal sequence that was fused to EGFP and should be detectable in our mouse model . The other three variants ( b , c , and d ) are C-terminally truncated or altered and could therefore escape the detection by both our EGFP-tag and the most commonly used antibodies against the P2X7 C-terminus . However , the nanobody used in our study was selected on intact P2X7 expressing cells and therefore binds at the extracellular P2X7 domain which is present in splice variants b and c . Together with the fact that in the original study ( Masin et al . , 2012 ) , protein of the deleted variants was not detected in the brain , this strongly argues against the presence of the mouse P2X7 variants b and c in neurons . Splice variant d contains only one TM domain and cannot be expected to form functional receptors . We therefore consider it highly unlikely that one of the known mouse variants of P2X7 accounts for neuronal P2X7 function or detection . A 65 kD protein band is frequently detected by us and others with antibodies against the P2X7 C-terminus but , unlike the P2X7 protein ( variant a , about 75 kD ) , does not show a size shift upon glycosidase treatment and therefore , most likely does not represent a P2X7 variant . In summary , we generated and thoroughly characterized a novel transgenic P2X7-EGFP mouse model that should help to overcome previous limitations in P2X7 research . Using functional assays , we could show that transgenic expressed P2X7-EGFP rescues the phenotype of P2X7 deficiency thus confirming the functionality of the transgene . Our initial characterization , however , indicates that P2X7 overexpression does not per se induce any overt pathologies , but rather represents a natural response observed after tissue damage . This is in line with observations in Schwann cells , where overexpression of P2X7 alone does not alter the basal Ca2+ concentration: overexpression of the peripheral myelin protein 22 ( as it occurs in Schwann cells from patients suffering from Charcot Marie Tooth type 1A ) , causes instead a secondary overexpression of P2X7 and consequent Schwann cells functional derangement ( Nobbio et al . , 2009 ) , suggesting that in addition to P2X7 overexpression , other factors are required to induce P2X7-associated pathologies . Most importantly for our study , neuronal P2X7 protein expression was not induced under pathological conditions . An unresolved question is , whether the high ATP concentrations required to activate the receptor occur physiologically or if the receptor is silent under physiological conditions and mainly expressed and/or activated under pathophysiological conditions . The presented mouse model provides a new tool for addressing this question .
HEK293 cells were cultured and transiently transfected with 1 . 5-2 mg DNA/well of a 6-well-plate ( Lipofectamin , Thermo Fisher Scientific ) . After 48 h , cells were washed and collected in PBSs ( 2 wells/experiment ) , pelleted in a desktop centrifuge ( 2`at 13 , 000 rpm ) and extracted as described ( Nicke et al . , 2008 ) for 15 minutes on ice in 100 ml 0 . 1% sodium phosphate buffer ( pH 8 . 0 ) containing 1% digitonin ( Fluka , Buchs , Switzerland ) and 0 . 4 mM Pefabloc SC ( Fluka ) . 10ml of extract were separated by SDS-PAGE with with or without endoglycosidase treatment ( 30 min at 37 °C in the presence of reducing loading buffer ( 1x ) and 5 IUB miliunits EndoH or 10 IUB miliunits PNGaseF ( New England Biolabs ) ) . For ethidium uptake measurements , cells were seeded after 27 h in 96-well plates ( 5x104 cells/well ) and incubated in the presence of 20 mM ethidium bromide in PBS for 15 min . Dye influx was evaluated with a fluorescence plate reader ( Fluostar Galaxy , BMG ) upon addition of the indicated ATP concentrations , as described ( Bruzzone et al . , 2010 ) . Patch-clamp recordings were performed as described ( Nicke et al . , 2009 ) in normal ( 147 mM NaCl , 2 mM KCl , 2 mM CaCl2 , 1 mM MgCl2 , 10 mM HEPES , and 13 mM glucose ) or low divalent cation ( 0 MgCl2 , 0 . 1 mM CaCl2 ) containing extracellular solution . The Strep-tagII-Gly-7xHis-Gly-EGFP-sequence was inserted into the P2rx7 BAC clone RP24-114E20 ( Children’s Hospital Oakland Research Institute , Oakland , CA ) , immediately upstream of the P2rx7 stop codon by homologous recombination ( Warming et al . , 2005 ) using locus-specific homology arms of 50–60 bp length ( Expand High Fidelity PCR , Roche Applied Science ) . The 451P variant was generated from the obtained BAC by the same strategy . BAC DNAs were verified after each recombination step by PCR , Southern blot and DNA fingerprinting . Upon sequencing of the coding regions ( Eurofins Genomics , Germany ) , final BAC constructs were linearized with SacII ( thereby destroying the unwanted Ift81 gene ) , purified ( Sepharose CL-4B chromatography , Sigma-Aldrich ) , analyzed by pulsed field gel electrophoresis , and microinjected into pronuclei of FVB/N mouse zygotes ( 451L background ) ( for primers and probes see Supplementary files 1 and 2 ) . Genomic DNA was isolated from tail biopsies , digested with BglII , separated on an 0 . 8% agarose gel , and blotted onto Nylon membrane ( Hybond N+ , GE Healthcare ) by capilliary transfer . After immobilization by UV irradiation ( 1500 µJ/cm2 ) , DNA was hybridized to a 32P labeled probe ( Random primed labeling Kit , Roche ) corresponding to a 645 bp fragment 2 . 6 kb downstream of the P2rx7 stop codon . Autoradiographic analysis ( Phosphoimager plates , Molecular Dynamics ) specifically detected the expected hybridization signals at 5277 bp ( BAC transgene ) and 4561 bp ( endogenous P2rx7 ) . The intensity ratios were used to determine the number of inserted BAC copies ( for probes see Supplementary file 1 ) . Protein extracts were prepared as described ( Nicke , 2008 ) using a Precellys homogenizer ( Peqlab ) with CK28 beads ( 15 s , 5 . 000 rpm ) and NP40 ( Sigma ) as detergent . Protein concentrations were determined by BCA assay ( Pierce ) . 30–75 μg of total protein per lane were loaded on 8% SDS-PAGE gels . Protein was either directly visualized by EGFP fluorescence scanning ( Typhoon , GE Healthcare ) or blotted onto Immobilon-FL PVDF membranes ( Merck Millipore ) and detected with an Odyssey infrared imaging system ( LI-COR Biosciences ) using the indicated antibodies ( S1 Material and methods ) . Endoglycosidase ( New England Biolabs ) treatment was performed for 30 min at 37°C in 20 μl sample aliquots with loading buffer ( IUB miliunits: EndoH 10 , PNGaseF 20 ) . Mice ( 8–12 weeks , male and female ) were sacrificed and single-cell suspensions prepared from brains by 30 min collagenase digestion at 37°C in a shaking water bath . Cell suspensions were filtered through a 70 µm cell strainer and centrifuged for 5 min at 300x g . Microglia were separated from debris by resuspending the pellet in 5 ml of a 33% Percoll solution ( GE Healthcare ) and centrifugation ( 20 min , 300x g ) . The pellet was resuspended in 1 ml ammonium-chloride-potassium erythrocyte lysis buffer and incubated for 1 min on ice to remove erythrocytes . Cells were subsequently washed with 10 ml FACS buffer ( PBS + 0 . 2% BSA/1 mM EDTA ) and resuspended in 100 µl FACS buffer . Microglia were stained ( 30 min on ice ) with anti-CD11b-perCP ( Biolegend ) and anti-CD45-PE-Cy7 ( Biolegend ) in the presence of Fc-blocking anti-CD16/CD32 ( BioXcell ) and normal rat serum . After washing 2x with FACS buffer cells were resuspended in 200 µl RPMI medium ( Gibco ) . DAPI was added to a final concentration of 1 . 5 µM and cells were incubated in the presence or absence of 1 mM ATP at 37°C for 15 min . The DAPI uptake into CD11b+CD45low microglia was subsequently measured using a BectonDickinson Celesta flow cytometer . For monitoring of time-dependent DAPI uptake by real time flow cytometry , isolated microglia were differentiated by transgenic P2X7-EGFP expression and exogenous eFluor670-labeling ( WT or P2X7 knockout ) and pooled in one FACS tube in 500 µl RPMI medium ( Gibco ) in the presence of DAPI ( 1 . 5 µM ) in order to have identical stimulation conditions . The baseline DAPI signal was measured for 2 min at 37°C , then 1 mM ATP was added and measuring continued for 4 to 5 min . DAPI uptake over time was compared among the differentially labeled microglia . The coding region for the P2X7-specific nanobody 7E2 was cloned into the pCSE2 . 5 vector ( provided by T . Schirrmann , Technical University Braunschweig , Germany ) ( Schirrmann and Büssow , 2010 ) upstream of coding regions for the hinge , CH2 and CH3 domains of rabbit IgG ( Danquah et al . , 2016 ) . Six days after transfection of this construct into HEK-6E cells ( Zhang et al . , 2009 ) , 7E2-rbIgG was purified from the cell supernatant by affinity chromatography on a protein-G sepharose column . Buffer was exchanged by gel filtration on a PD-10 column . A panel of nanobody-rbIgG heavy chain antibodies was originally screened for binding to P2X7 transfected HEK cells before and after fixation with 4% PFA . 7E2-rbIgG was chosen because it retained the strongest staining after fixation in both immunofluorescence staining and a FACS-based dissociation assay analogous to that described in ( Fumey et al . , 2017 ) . It only weakly antagonizes gating of P2X7 by ATP and by ADP-ribosylation but its potency was not further determined ( Danquah et al . , 2016 ) . Mice were sacrificed by CO2/cervical dislocation or anesthetized ( Ketamin/Xylazin ) and transcardially perfused with 4% PFA . Brains were fixed in 4% PFA for 72 hr or 24 hr , respectively , cryoprotected in 30% sucrose , and embedded in 5% LM Agarose ( Roth , Germany ) . 30 μm sagittal brain sections were prepared ( VT1200s Leica Microsystems ) and either blocked with 4% skim milk and 10% FCS in PBS ( 1–1 . 5 hr , RT ) or , after peroxidase block ( 3% H2O2 in 0 . 01 M PBS , 30 min RT ) , with 10% Normal Goat Serum and 0 . 1% Triton X-100 in PBS ( 1 hr , RT ) , prior to primary antibody incubation overnight at 4°C . Incubation with biotinylated secondary antibodies was at 37°C for 1 hr , or at RT for 1 . 5 hr . Staining was visualized using the ABC method with the Vectastain abc kit and the DAB substrate kit for peroxidase ( Vectorlabs , USA ) or SIGMA_FAST_ DAB Tablets ( Sigma-Aldrich , Germany ) . Counterstaining was carried out with hematoxylin ( Sigma-Aldrich ) , followed by dehydration and embedding . Images were taken with an Axio Observer 7 ( Zeiss ) . Immunostaining was performed as described ( Zhang et al . , 2013 ) . In brief , mice ( P60-P90 ) were transcardially perfused with PBS and then 4% PFA in PBS . Brains or spinal cords were post-fixed over night in 4% PFA/PBS . P7 pups were decapitated and brains post-fixed in 4% PFA . After cryoprotection ( 30% sucrose in PBS ( pH 7 . 4 ) , 40 µm cryostat sections ( Microm HM560 , Walldorf , Germany ) were washed ( 3 × 10 min , PBS ) , blocked ( 5% Normal Goat Serum ( Dako Germany ) , 0 . 3% Triton X-100 ( Sigma , Munich , Germany ) in PBS , 2 hr at RT ) and incubated with primary antibodies ( 16–24 hr , 4°C ) with gentle shaking . After washing as above , sections were incubated for 2 hr with fluorescence conjugated secondary antibodies . Slices were mounted on object slides with PermaFluor Mounting Medium ( Thermo Scientific ) . Rectus femoris muscle was incubated in 30% sucrose ( 48 hr ) , embedded in OTC ( Tissue TEK ) , and frozen in liquid nitrogen . 20 μm cryostat sections were collected on object slides , fixed with 4% PFA ( 10 min ) , blocked and permeabilized ( 30 min , 1% BSA , 0 . 2% Triton X-100 in PBS ) . Incubation with Alexa 594 Fluor-conjugated α-bungarotoxin ( 1:1000 , Thermo Fisher Scientific ) and primary antibodies was overnight at 4°C , and with secondary antibodies for 1 hr at RT . Thoracic and lumbar DRGs were embedded in Tissue-Tek , sectioned in 10 μm slices , mounted on slides and frozen at −80° . Before staining as described above , sections were post-fixed for 10 min in 4% PFA in PBS , incubated for 30 min in 0 . 1 M Glycine in PBS , and blocked for 1 hr ( 5% normal goat serum , 0 . 1% Triton X-100 in PBS ) . In case of glutamine synthetase and MPZ staining , tissue was treated with 10 mM sodium citrate ( pH 6 . 0 , >95° C ) for 1 min just before blocking . Images were obtained by confocal laser scanning microscopy ( Leica SP5 or Zeiss LSM 880 ) . EGFP-positive cells were quantified in every fifth slice in a series of 25–30 sections throughout the whole rostrocaudal extension of the hippocampus . DAPI-positive cells , EGFP-positive cells , and marker protein-positive cells in the hippocampus CA1 region were counted in z-stacks . To define the counting box ( 250 × 250 × 25 µm3 ) , confocal laser micrographs of the CA1 region were obtained ( 63 x/0 , 75 NA objective ) at 1 µm intervals to a final depth of 25 µm . Cell nuclei located completely inside the counting frame and at the upper and right borders were counted . Data analysis was performed using Excel and Graphpad Prism 7 software . Data are given as mean ±SEM from N = 3 mice per group . Mice were killed by cervical dislocation and brains were rapidly removed , washed with ice cold PBS , and kept on ice . Cerebelli were homogenized using the GentleMacs neuronal tissue dissociation kit ( T ) ( Miltenyi Biotech ) according to the manufactures instruction . Dissociated cells were centrifuged ( 1000 g/ 15 min/4°C ) , and washed twice with PBS ( followed by centrifugation as above ) to remove residual trypsin . Supernatant was carefully removed and cells fixed ( 4% PFA in PBS , 10 min , 4°C ) under gentle agitation , washed three times with PBS as described above , permeabilized with blocking solution ( 2% BSA , 2% normal goat serum , 0 . 2% Triton X-100 in PBS ) , and incubated with the indicated antibodies diluted in blocking solution overnight at 4°C . After washing , cells were incubated with secondary antibodies for 1 hr , washed , incubated with DAPI ( 200 nM in PBS , 10 min ) , washed , and embedded in Aquamount ( Polyscience ) . Imaging was performed using a Zeiss Confocal microscope ( LSM 800 ) and the ZEN imaging software . In co-processed wild-type animals , no GFP ( immuno ) -fluorescence was detected . Sciatic nerves of adult mice were dissected and transferred into cold PBS . Under a stereomicroscope , the epineurium was carefully removed , nerves separated longitudinally into individual or small bundles of fascicles , transferred to a droplet of cold PBS on a superfrost slide , and gently teased apart . Samples were air-dried and stored at −20°C if not processed immediately for immunocytochemistry . Preparations were post-fixed ( 5 min ) in 4% PFA , permeabilized with ice-cold methanol ( 5 min ) , washed with PBS ( 3 × 5 min ) , and blocked ( 10% horse serum , 0 . 1% Tween 20 in PBS , 2 hr at RT ) . Slides were incubated overnight at 4°C with the primary antibodies and after washing with PBS ( 3 × 5 min ) , secondary antibodies were applied ( 2 hr at RT ) . After final washing , fibers were mounted with Vectashield Mounting Medium containing DAPI ( Vector Laboratories ) . Retinae were immersion-fixed ( 4% PFA for 2 hr ) , washed with PBS , sucrose cryoprotected and cut in 20 µm thick sections . Retinal sections were permeabilized ( 0 . 3% Triton X-100 plus 1 . 0% DMSO in PBS ) and blocked ( 5% normal goat serum with 0 . 3% Triton X-100 and 1 . 0% DMSO in PBS , 2 hr at RT ) . Primary antibodies were incubated overnight at 4°C . Sections were washed ( 1% BSA in PBS ) and incubated with secondary antibodies ( 2 hr at RT ) . Cell nuclei were labeled with DAPI ( 1:1000; Life Technologies ) . Control experiments without primary antibodies showed no unspecific labeling . Images were acquired using confocal microscopy ( Visiscope , Visitron Systems ) . For quantification of cell numbers or microglia only central retinal slices were used . Cells were quantified in a defined area of 100 µm in width ( DAPI staining ) or the whole scan field ( ~460 µm in width; microglia ) approximately 200–300 µm in distance from the optic nerve head . For each animal , 2–3 central slices were analyzed . For whole mount myenteric plexus preparation , mice were sacrificed by cervical dislocation and 1 cm segments were taken 1 cm distal from the proximal colon and transferred to ice-cold Krebs solution ( containing in mM: 117 NaCl , 4 . 7 KCl , 1 . 2 MgCl2 , 1 . 2 NaH2PO4 , 25 NaHCO3 , 2 . 5 CaCl2 , 11 glucose and aerated with carbogen to pH 7 . 4 ) in Sylgard ( Dow Corning ) -filled dissecting dishes . After flushing with Krebs buffer , segments were opened along the mesenteric border , pinned out , and fixed for 4 hr at 4°C ( 4% PFA and 0 . 2% picric acid in 0 . 1 M phosphate buffer ( pH 7 . 4 ) . Tissue was rinsed ( 3 × 10 min ) with phosphate buffer and dissected in PBS ( pH 7 . 4 ) . After carefully removing mucosa , submucosa and circular musculature using forceps and a binocular , myenteric plexus preparations were blocked ( 0 . 5% , Triton X-100 , 0 . 1% NaN3 , 4% goat serum ( Sigma-Aldrich ) in PBS , 1 hr , RT ) , incubated with primary antibodies in the above solution ( 12 hr at RT ) , rinsed 3 × 10 min with PBS , and incubated with secondary antibodies ( 2 hr at RT ) . After washing ( 3x in PBS ) preparations were mounted in PermaFluor ( Thermo Fisher Scientific ) on slides and images were acquired with a Zeiss 880 Airyscan confocal microscope and processed with ImageJ . For enrichment of distinct retinal cell types a previously described protocol was used with minor modifications ( Grosche et al . , 2016 ) . Briefly , retinae were treated with papain ( 0 . 2 mg/ml; Roche Molecular Biochemicals ) for 30 min at 37°C in the dark in Ca2+- and Mg2+-free extracellular solution ( 140 mM NaCl , 3 mM KCl , 10 mM HEPES , 11 mM glucose , pH 7 . 4 ) , washed several times in extracellular solution and incubated with DNase I ( 200 U/ml ) . Afterwards the tissue was triturated in extracellular solution containing ( mM ) 135 NaCl , 3 KCl , 2 CaCl2 , 1 MgCl2 , 1 Na2HPO4 , 10 HEPES , and 11 glucose adjusted to pH 7 . 4 with Tris . After centrifugation , cells were resuspended and incubated in extracellular solution containing biotinylated hamster anti-CD29 ( clone Ha2/5 , 0 . 1 mg/ml , BD Biosciences , Heidelberg , Germany ) for 15 min at 4°C . Cells were washed in extracellular solution , spun down , resuspended in the presence of anti-biotin MicroBeads ( 1:5; Miltenyi Biotec , Bergisch Gladbach , Germany ) and incubated for 10 min at 4°C . After washing , CD29 +Müller cells were separated using large cell ( LS ) columns according to the manufacturer’s instructions ( Miltenyi Biotec ) . To purify microglial and vascular cells in addition to Müller cells , the retinal cell suspension was subsequently incubated with CD11b- and CD31 microbeads ( Miltenyi Biotec ) and depleted from the retinal suspension using LS-columns prior to Müller cell enrichment . Total RNA was isolated from whole brain tissue using the PureLink RNA Micro Scale Kit ( Thermo Fisher Scientific , Germany ) . Upon DNase-treatment to remove genomic DNA , first-strand cDNAs were synthesized from 50 ng of total RNA ( RevertAid H Minus First-Strand cDNA Synthesis Kit , Fermentas by Thermo Fisher Scientific , Germany ) . Primers ( see Supplementary file 2 ) were designed using the Universal Probe Library Assay Design Center ( Roche ) . Transcript levels of candidate genes were measured by qRT-PCR using the QuantStudio 5 Real-Time PCR system ( 384 well , Life Technologies ) according to the company’s guidelines . All data are expressed as mean ± standard error ( SEM ) unless stated otherwise . Statistical analyses were performed using Graphpad Prism 7 Software ( San Diego , CA ) . Unless stated otherwise , the significance was determined by the non-parametric Mann-Whitney U test . Transient retinal ischemia was induced in one eye by the HIOP ( high intraocular pressure ) method as previously described ( Pannicke et al . , 2014 ) . The other eye remained untreated as internal control . Anesthesia was induced with ketamine ( 100 mg/kg body weight , intraperitoneal ( ip ) ; Ratiopharm , Ulm , Germany ) , xylazine ( 5 mg/kg , ip; Bayer Vital , Leverkusen , Germany ) , and atropine sulfate ( 100 mg/kg , ip; Braun , Melsungen , Germany ) . The anterior chamber of the test eye was cannulated from the pars plana with a 30-gauge infusion needle , connected to a saline bottle . The intraocular pressure was increased to 160 mmHg for 60 min by elevating the bottle . After removing the needle , the animals survived for 3 days and , subsequently , were sacrificed with carbon dioxide . Mice were in the C57BL/6N background . Stab wound injury was performed in the somatosensory cortex , as previously described ( Heinrich et al . , 2014; Heimann et al . , 2017 ) . Briefly , anesthetized animals received a stab wound of the somatosensory cortical GM with a lancet-shaped knife ( Alcon ) Coordinates from bregma: AP −0 . 8 to −2 . 0 , ML 1 . 6 to 2 . 0 mm and DV −0 . 6 . Animals were allocated to experimental groups regarding their genotype and kept under standard conditions with access to water and food ad libitum . Five days post injury ( dpi ) , animals were transcardially perfused and brains processed for immunohistochemistry as described above . For analysis , seven corresponding slices were prepared from each animal and triple staining of GFAP , NeuN , and Iba1 were performed sequentially ( starting with NeuN and over night fixation , and followed by GFAP and Iba1 labeling ) . Confocal images were taken at identical exposure settings with single channel maximum intensity projections set to automatic threshold . Iba1- Neu- , and DAPI-positive areas were measured using NIH ImageJ software ( Image > adjust > threshold; Analyse > measure ) . Iba1-positive areas and NeuN-negative/lesioned areas were normalized to DAPI-positive areas . Data were analyzed using Graphpad Prism 7 . Procedures were undertaken as described previously ( Jimenez-Pacheco et al . , 2013 ) in 8–12 week-old mice ( line 17/FVB/N ) bred at the Biomedical Research Facility at RCSI . Mice were anesthetized with isoflurane ( 5% induction , 1–2% maintenance ) and maintained normothermic by means of a feedback-controlled heat blanket ( Harvard Apparatus Ltd , Kent , UK ) . Fully anesthetized , mice were placed in a stereotaxic frame and a midline scalp incision was performed to expose the skull . A guide cannula ( coordinates from Bregma; AP = −0 . 94 mm , L = −2 . 85 mm ) for intra-amygdala kainic acid ( Sigma Aldrich , Dublin , Ireland ) injection was fixed in place with dental cement and status epilepticus induced in fully awake mice via microinjection of 0 . 3 µg KA ( in 0 . 2 µl phosphate-buffered saline ) into the basolateral amygdala . Control animals received 0 . 2 µl PBS . 40 min after injections , the anticonvulsive lorazepam ( 6 mg/kg , Wyetch , Taplow , UK ) was delivered i . p . to curtail seizures and reduce morbidity and mortality . Mice were killed 24 hr after lorazepam injection and perfused ( 4% PFA in PBS ) . Brains were post-fixed overnight in 4% PFA , embedded in 2% agarose , and cut by vibratome in 30 µm sections . Sections were stored in glycol at −20° C until use . Experiments were performed with 10–13 weeks old mice on 3 consecutive days in the following order . Mice were housed in standard conditions ( 22°C , 12 hr light–dark cycle , water/food ad libitum ) . Animal handling and experimental procedures were performed in accordance with German and European Union guidelines and were approved by the State of Upper Bavaria ( stab wound injury ( 55 . 2 . 1 . 54-2532-171-11 ) , retinal ischemia ( TVV 54/12; 55 . 2 DMS-2532-2-182 ) , transcardial perfusion ( 55 . 2-1-54-2532-59-2016 ) ) and Lower Saxony ( generation of BAC transgenic mice , transcardial perfusion ( 33 . 9-42502-04-12/0863 ) , behavioral experiments ( 3392 42502-04-13/1123 ) ) . Status epilepticus was induced in accordance with the principles of the European Communities Council Directive ( 86/609/EEC ) and procedures reviewed and approved by the Research Ethics Committee of the Royal College of Surgeons in Ireland ( REC 205 and 1322 ) and performed under license from the Department of Health and Children , Ireland . All efforts were made to minimize suffering and number of animals used . | The human body relies on a molecule called ATP as an energy source and as a messenger . When cells die , for example if they are damaged or because of inflammation , they release large amounts of ATP into their environment . Their neighbors can detect the outpouring of ATP through specific receptors , the proteins that sit at the cell’s surface and can bind external agents . Scientists believe that one of these ATP-binding receptors , P2X7 , responds to high levels of ATP by triggering a cascade of reactions that results in inflammation and cell death . P2X7 also seems to play a role in several brain diseases such as epilepsia and Alzheimer’s , but the exact mechanisms are not known . In particular , how this receptor is involved in the death of neurons is unclear , and researchers still debate whether P2X7 is present in neurons and in other types of brain cells . To answer this , Kaczmarek-Hájek , Zhang , Kopp et al . created genetically modified mice in which the P2X7 receptors carry a fluorescent dye . Powerful microscopes can pick up the light signal from the dye and help to reveal which cells have the receptors . These experiments show that neurons do not carry the protein; instead , P2X7 is present in certain brain cells that keep the neurons healthy . For example , it is found in the immune cells that ‘clean up’ the organ , and the cells that support and insulate neurons . Kaczmarek-Hájek et al . further provide preliminary data suggesting that , under certain conditions , if too many P2X7 receptors are present in these cells neuronal damage might be increased . It is therefore possible that the brain cells that carry P2X7 indirectly contribute to the death of neurons when large amounts of ATP are released . The genetically engineered mouse designed for the experiments could be used in further studies to dissect the role that P2X7 plays in diseases of the nervous system . In particular , this mouse model might help to understand whether the receptor could become a drug target for neurodegenerative conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"immunology",
"and",
"inflammation"
] | 2018 | Re-evaluation of neuronal P2X7 expression using novel mouse models and a P2X7-specific nanobody |
Biomedical and clinical sciences are experiencing a renewed interest in the fact that males and females differ in many anatomic , physiological , and behavioural traits . Sex differences in trait variability , however , are yet to receive similar recognition . In medical science , mammalian females are assumed to have higher trait variability due to estrous cycles ( the ‘estrus-mediated variability hypothesis’ ) ; historically in biomedical research , females have been excluded for this reason . Contrastingly , evolutionary theory and associated data support the ‘greater male variability hypothesis’ . Here , we test these competing hypotheses in 218 traits measured in >26 , 900 mice , using meta-analysis methods . Neither hypothesis could universally explain patterns in trait variability . Sex bias in variability was trait-dependent . While greater male variability was found in morphological traits , females were much more variable in immunological traits . Sex-specific variability has eco-evolutionary ramifications , including sex-dependent responses to climate change , as well as statistical implications including power analysis considering sex difference in variance .
Sex differences arise because selection acts on the two sexes differently , especially on traits associated with mating and reproduction ( Darwin , 1871 ) . Therefore , sex differences are widespread , a fact which is unsurprising to any evolutionary biologist . However , scientists in many ( bio- ) medical fields have not necessarily regarded sex as a biological factor of intrinsic interest ( Clayton , 2016; Flanagan , 2014; Karp et al . , 2017; Klein et al . , 2015; Prendergast et al . , 2014; Shansky and Woolley , 2016 ) . Therefore , many ( bio- ) medical studies have only been conducted with male subjects . Consequently , our knowledge is biased . For example , we know far more about drug efficacy in male compared to female subjects , contributing to a poor understanding of how the sexes respond differently to medical interventions ( Nowogrodzki , 2017 ) . This gap in knowledge is predicted to lead to overmedication and adverse drug reactions in women ( Zucker and Prendergast , 2020 ) . Only recently have ( bio- ) medical scientists started considering sex differences in their research ( Dorris et al . , 2015; Ingvorsen et al . , 2017; Robinson et al . , 2017; Smarr et al . , 2017; Ahmad et al . , 2017; Foltin and Evans , 2018; Thompson et al . , 2018 ) . Indeed , the National Institutes of Health ( NIH ) have now implemented new guidelines for animal and human research study designs , requiring that sex be included as a biological variable ( Clayton , 2016; Clayton and Collins , 2014; NIH , 2015a ) . When comparing the sexes , biologists generally focus on mean differences in trait values , placing little or no emphasis on sex differences in trait variability ( see Figure 1 for a diagram explaining differences in means and variances ) . Despite this , two hypotheses exist that explain why trait variability might be expected to differ between the sexes . Interestingly , these two hypotheses make opposing predictions . First , the ‘estrus-mediated variability hypothesis’ ( Figure 2 ) , which emerged in the ( bio- ) medical research field , assumes that the female estrous cycle ( see e . g . Prendergast et al . , 2014; Beery and Zucker , 2011 ) causes higher variability across traits in female subjects . A wide range of labile traits are presumed to co-vary with physiological changes that are induced by reproductive hormones . High variability is , therefore , expected to be particularly prominent when the stage of the estrous cycle is unknown and unaccounted for . This higher trait variability , resulting from females being at different stages of their estrous cycle , is the main reason for why female research subjects are often excluded from biomedical research trials , especially in the fields of neuroscience , physiology and pharmacology ( NIH , 2015a ) . Female exclusion has traditionally been justified based on the grounds that including females in empirical research leads to a loss of statistical power , or that animals must be sampled across the estrous cycle for one to make valid conclusions , requiring more time and resources . Second , the ‘greater male variability hypothesis’ suggests males exhibit higher trait variability because of two different mechanisms . The first mechanism is based on males being the heterogametic sex in mammals . Mammalian females possess two X chromosomes , leading to an ‘averaging’ of trait expression across the genes on each chromosome . In contrast , males exhibit greater variance because expression of genes on a single X chromosome is likely to lead to more extreme trait values ( Reinhold and Engqvist , 2013 ) . The second mechanism is based on males being under stronger sexual selection ( Pomiankowski and Moller , 1995; Cuervo and Møller , 1999; Cuervo and Møller , 2001 ) . Empirical evidence supports higher variability of traits that are sexually selected , often harbouring high genetic variance and being condition-dependent , which makes sense as ‘condition’ as a trait is likely to be based on numerous loci ( Rowe and Houle , 1996; Tomkins et al . , 2004 ) . Thus , higher genetic and , thus , phenotypic variance resulting from sexual selection is expected to characterise sexually selected traits . In mammals , it is likely that both mechanisms are operating concomitantly . So far , the ‘greater male variability hypothesis’ has gained some support in the evolutionary and psychological literature ( Reinhold and Engqvist , 2013; Lehre et al . , 2009 ) . Here , we conduct the first comprehensive test of the greater male variability and estrus-mediated variability hypotheses in mice ( Figure 2; Reinhold and Engqvist , 2013; Johnson et al . , 2008; Hedges and Nowell , 1995; Itoh and Arnold , 2015; Becker et al . , 2016; Beery , 2018 ) , examining sex differences in variance across 218 traits in 26 , 916 animals . To this end , we carry out a series of meta-analyses in two steps ( Figure 3 ) . First , we quantify the natural logarithm of the male to female coefficients of variation , CV , or relative variance ( lnCVR ) for each cohort ( population ) of mice , for different traits , along with the variability ratio of male to female standard deviations , SD , on the log scale ( lnVR , following Nakagawa et al . , 2015 , see Figure 1 ) . Then , we analyse these effect sizes to quantify sex bias in variance for each trait using meta-analytic methods . To better understand our results , and match them to previously reported sex differences in trait means ( Karp et al . , 2017 ) , we also quantify and analyse the log response ratio ( lnRR ) . Next , we statistically amalgamate the trait-level results to test our hypotheses and to quantify the degree of sex bias in and across nine functional trait groups ( for details on the grouping , see below ) . Our meta-analytic approach allows easy interpretation and comparison with earlier and future studies . Further , the proposed method using lnCVR ( and lnVR ) is probably the only practical method to compare variability between two sexes within and across studies ( Nakagawa et al . , 2015; Senior et al . , 2020 ) , as far as we are aware . Also , the use of a ratio ( i . e . lnRR , lnVR , lnCVR ) between two groups ( males and females ) naturally controls for different units ( e . g . cm , g , ml ) as well as for changes in traits over time and space .
We used a dataset compiled by the International Mouse Phenotyping Consortium ( Dickinson et al . , 2016 ) ( IMPC , dataset acquired 6/2018 ) . To gain insight into systematic sex differences , we only included data of wildtype-strain adult mice , between 100 and 500 days of age . We removed cases with missing data , and selected measurements that were closest to 100 days of age ( young adult ) when multiple measurements of the same trait were available . To obtain robust estimates of sex differences , we only used data on traits that were measured in at least two different institutions ( see workflow diagram , Figure 3 ) . Our dataset comprised 218 continuous traits ( after initial data cleaning and pre-processing; Figure 3 ) . It contains information from 26 , 916 mice from nine wildtype strains that were studied across 11 institutions . We combined mouse strain/institution information to create a biological grouping variable ( referred to as ‘population’ in Figure 3B; see also Supplementary file 1 , Table 1 for details ) , and the mean and variance of a trait for each population was quantified . We assigned traits according to related procedures into functionally and/or procedurally related trait groups to enhance interpretability ( referred to as ‘functional groups’ hereafter; see also Figure 3G ) . Our nine functional trait groups were: behaviour , morphology , metabolism , physiology , immunology , hematology , heart , hearing and eye ( for the rationale of these functional groups and related details , see Methods and Supplementary file 1 , Table 3 ) . We found that some means and variabilities of traits were biased towards males ( i . e . ‘male-biased’ , hereafter; turquoise shaded traits , Figure 4 ) , but others towards females ( i . e . ‘female-biased’ , hereafter; orange shading , Figure 4 ) within all functional groups . These sex-specific biases occur in mean trait sizes and also in our measures of trait variability . There were strong positive relationships between mean and variance across traits ( r > 0 . 94 on the log scale; Figure 1—figure supplement 1 ) , and therefore , we report the results of lnCVR , which controls for differences in means , in the main text . Results on lnVR are presented as figure supplements ( Figure 4—figure supplements 1 and 2 ) . There was no consistent pattern in which sex has more variability ( lnCVR ) in the examined traits ( left panel in Figure 4A ) . Our meta-analytic results also did not support a consistent pattern of either higher male variability or higher female variability ( see Figure 4B , left panel: ‘All’ indicates that across all traits and functional groups , there was no significant sex bias in variances; lnCVR = 0 . 005 , 95% confidence interval , 95% CI = [−0 . 009 to 0 . 018] ) . However , there was high heterogeneity among traits ( I2 = 76 . 5% , Supplementary file 1 , Table 4 and see also Table 5 ) , indicating sex differences in variability are trait-dependent , corroborating our general observation that variability in some traits was male-biased but others female-biased ( Figure 4A ) . As expected , specific functional trait groups showed significant sex-specific bias in variability ( Figure 4B ) . The variability among traits within a functional group was lower than that of all the traits combined ( Supplementary file 1 , Table 4 ) . For example , males exhibited an 8 . 05% increase in CV relative to females for morphological traits ( lnCVR = 0 . 077; CI = [0 . 041 to 0 . 113] , I2 = 67 . 3% ) , but CV was female-biased for immunological traits ( 6 . 59% higher in females , lnCVR = −0 . 068 , CI = [−0 . 098 to 0 . 038] , I2 = 40 . 8% ) and eye morphology ( 7 . 85% higher in females , lnCVR = −0 . 081 , CI = [−0 . 147 to ( −0 . 016 ) ] , I2 = 49 . 8% ) . The pattern was similar for overall sexual dimorphism in mean trait values ( here , a slight male bias is indicated by larger ‘turquoise’ than ‘orange’ areas; Figure 4B , right and Figure 4B , lnRR: ‘All’ , lnRR = 0 . 012 , CI = [−0 . 006 to 0 . 31] ) . Trait means ( lnRR ) were 7% larger for males ( lnRR = 0 . 067; CI = [0 . 007 to 0 . 128] ) in morphological traits and 15 . 3% larger in males for metabolic traits ( lnRR = 0 . 142; CI = [0 . 036 to 0 . 248] ) . In contrast , females had 5 . 59% ( lnRR = 0 . 057 , CI = [−0 . 107 to ( −0 . 007 ) ] ) larger means than those of males for immunological traits . We note that these meta-analytic estimates were accompanied by very large between-trait heterogeneity values ( morphology I2 = 99 . 7% , metabolism I2 = 99 . 4% , immunology I2 = 96 . 2; see Supplementary file 1 , Table 4 ) , indicating that even within the same functional groups , the degree and direction of sex bias in the mean was not consistent among traits .
Evolutionary biologists commonly expect greater variability in the heterogametic sex than the homogametic sex . In mammals , males are heterogametic , and hence are expected to exhibit higher trait variability compared to females , which is also consistent with an expectation from sexual selection theory ( Reinhold and Engqvist , 2013 ) . Our results provide only partial support for the greater male variability hypothesis , because the expected pattern only manifested for morphological traits ( see Figures 4 and 5 ) . This result corroborates a previous analysis across animals , which found that the heterogametic sex was more variable in body size ( Reinhold and Engqvist , 2013 ) . However , our data do not support the conclusion that higher variability in males occurs across all traits , including for many other morphological traits . The estrus-mediated variability hypothesis was , at least until recently ( Prendergast et al . , 2014; Smarr et al . , 2017 ) , regularly used as a rationale for including only male subjects in many biomedical studies . So far , we know very little about the relationship between hormonal fluctuations and general trait variability within and among female subjects . Our results are consistent with the estrus-mediated variability hypothesis for immunological traits only . Immune responses can strongly depend on sex hormones ( Zuk and McKean , 1996; Grossman , 1989 ) , which may explain higher female variability in these traits . However , if estrus status affects traits through variation in hormone levels , we would expect to also find higher female variability in physiological and hematological traits . This was not the case in our dataset . Interestingly , however , eye morphology ( structural traits , which should fluctuate little across the estrous cycle ) also appeared to be more variable in females than males , but little is known about sex differences in ocular traits in general ( Wagner et al . , 2008; Shaqiri et al . , 2018 ) . Overall , we find no consistent support for the female estrus-mediated variability hypothesis . In line with our findings , recent studies have refuted the prediction of higher female variability ( Prendergast et al . , 2014; Smarr et al . , 2017; Beery and Zucker , 2011; Becker et al . , 2016; Beery , 2018 ) . For example , several rodent studies have found that males are more variable than females ( Prendergast et al . , 2014; Smarr et al . , 2017; Becker et al . , 2016; Beery , 2018; Fritz et al . , 2017; Mogil and Chanda , 2005 ) . Further studies should investigate whether higher female variability in immunological traits is indeed due to the estrous cycle , or generally because of greater between-individual variation ( Figure 2 ) . In general , we found many traits to be sexually dimorphic ( Figure 5 ) in accordance with the previous study , which used the same database ( Karp et al . , 2017 ) . Although the original study also provided estimates for sex differences in traits both with and without controlling for weight ( we did not control for weight; Nakagawa et al . , 2017 ) . More specifically , males are larger than females , while females have higher immunological parameters ( see Figure 5 ) . Notably , the most sexually dimorphic trait means also show the greatest differences in trait variance ( Figures 4 and 5 ) . Indeed , theory predicts that sexually selected traits ( e . g . larger body size for males due to male-male competition ) are likely more variable , as these traits are often condition-dependent ( Rowe and Houle , 1996 ) . Therefore , this sex difference in variability could be more pronounced under natural conditions compared to laboratory settings . This relationship may explain why male-biased morphological traits are larger and more variable . We have used lnCVR values to compare phenotypic variability ( CV ) between the sexes . When lnCVR is used for fitness-related traits , it can signify sex differences in the ‘opportunity for selection’ between females and males ( Rowe and Houle , 1996 ) . If we assume that phenotypic variation ( i . e . variability in traits ) has a heritable basis , then large ratios of lnCVR may indicate differences in the evolutionary potential of each sex to respond to selection , at least in the short term ( Hansen and Houle , 2008 ) . For example , more variable morphological traits of males could potentially provide them with better capacity than females to adapt morphologically to a changing climate . We note , however , that in our study , lnCVR reflects sex differences in trait variability within strains , such that the variability differences we observe between the sexes may be partially the result of phenotypic plasticity . Demographic parameters , such as age-dependent mortality rate ( Lemaître et al . , 2020 ) can often be different for each sex . For example , a study on European sparrowhawks found that variability in mortality was higher in females compared to males ( Colchero et al . , 2017 ) . In this species , sex-specific variation affects age-dependent mortality and results in higher average female life expectancy . Therefore , population dynamic models , which make predictions about how populations change in their size over time , should take sex differences in variability into account to produce more accurate predictions ( Caswell and Weeks , 1986; Lindström and Kokko , 1998 ) . In our rapidly changing world , better predictions on population dynamics are vital for understanding whether climate change is likely to result in population extinction and lead to further biodiversity loss . It is now mandatory to include both sexes in biomedical experiments and clinical trials funded by the NIH , unless there exists strong justification against the inclusion of both sexes ( NIH , 2015a; NIH , 2015b ) . In order to conduct meaningful research and make sound clinical recommendations for both male and female patients , it is necessary to understand both how trait means and variances differ between the sexes . If one sex is systematically more variable in a trait of interest than the other , then experiments should be designed to accommodate relative differences in statistical power between the sexes ( which has not been considered before , see Flanagan , 2014; Klein et al . , 2015; Prendergast et al . , 2014; Shansky and Woolley , 2016 ) . For example , female immunological traits are generally more variable ( i . e . having higher CV and SD ) . Therefore , in an experiment measuring immunological traits , we would need to include a larger sample ( N ) of females than males ( N[female] > N[male]; N[total] = N[female] + N[male] ) to achieve the same power as when the experiment only includes males ( N[total*] = 2N[male] ) . In other words , in an experiment with both sexes we would need a larger sample size than the same experiment with males only ( N[total] > N[total*] ) . To help researchers adjust their sex-specific sample size to achieve optimal statistical power , we provide an online tool ( ShinyApp; https://bit . ly/sex-difference ) . This tool may serve as a starting point for checking baseline variability for each sex in mice . The sex bias ( indicated by the % difference between the sexes ) is provided for separate traits , procedures , and functional groups . These meta-analytic results are based on our analyses of more than 2 million rodent data points , from 26 , 916 individual mice . We note , however , that variability in a trait measured in untreated individuals maintained under carefully standardized environmental conditions , as reported here , may not directly translate into the same variability when measured in experimentally treated individuals , or individuals exposed to a range of environments ( i . e . natural populations or human cohorts ) . Further , these estimates are overall mean differences across strains and locations . Therefore , these may not be particularly informative if one’s experiment only includes one specific strain . Nonetheless , we point out that our estimates may be useful in the light of a recent recommendation of using ‘heterogenization’ where many different strains are systematically included ( i . e . randomized complete block design ) to increase the robustness of experimental results ( Voelkl et al . , 2020 ) . However , note that an experiment with heterogenization might only include a few strains with several animals per strain . Even in such a case , using just a few strains , our tool could provide potentially useful benchmarks . Incidentally , heterogenization would be key to making one’s experimental outcome more generalizable ( Webster and Rutz , 2020 ) . Importantly , when two groups ( e . g . males and females ) show differences in variability , we violate homogeneity of variance or homoscedasticity assumptions . Such a violation is detrimental because it leads to a higher Type I error rate . Therefore , we should consider incorporating heteroscedasticity ( different variances ) explicitly or using robust estimators of variance ( also known as ‘the sandwich variance estimator’ ) to prevent an inflated Type I error rate ( Cleasby and Nakagawa , 2011 ) , especially when we compare traits between the sexes . We have shown that sex biases in variability occur in many mouse traits , but that the directions of those biases differ between traits . Neither the ‘greater male variability’ nor the ‘estrus-mediated variability’ hypothesis provides a general explanation for sex differences in trait variability . Instead , we have found that the direction of the sex bias varies across traits and among trait types ( Figures 4 and 5 ) . Our findings have important ecological and evolutionary ramifications . If the differences in variability correspond to the potential of each sex to respond to changes in specific environments , this sex difference needs to be incorporated into demographic and population dynamic modelling . Moreover , in the ( bio- ) medical field , our results should inform decisions during study design by providing more rigorous power analyses that allow researchers to incorporate sex-specific differences for sample size . We believe that taking sex differences in trait variability into account will help avoid misleading conclusions and provide new insights into sex differences across many areas of biological and bio-medical research . Ultimately , such considerations will not only better our knowledge , but also close the current gaps in our biased knowledge ( Tannenbaum et al . , 2019 ) .
The IMPC ( International Mouse Phenotyping Consortium ) provides a comprehensive catalogue of mammalian gene function for investigating the genetics of health and disease , by systematically collecting phenotypes of knock-out and wildtype mice . To investigate differences in trait variability between the sexes , we only considered the data for wildtype control mice . We retrieved the dataset from the IMPC server in June 2018 and filtered it to contain non-categorical traits for wildtype mice . The initial dataset comprised over 2 , 500 , 000 data points for 340 traits . In cases where multiple measurements were taken over time , data cleaning started with selecting single measurements for each individual and trait . In these cases , we selected the measurement closest to ‘100 days of age’ . All data are from unstaged females ( with no information about the stage of their estrous cycle ) . We excluded data for juvenile and unsexed mice ( Figure 3A; this dataset and scripts can be found on https://rpubs . com/SusZaj/ESF; https://bit . ly/code-mice-sex-diff; raw data: https://doi . org/10 . 5281/zenodo . 3759701 ) . We created a grouping variable called ‘population’ ( Figure 3B ) . A population comprised a group of individuals belonging to a distinct wildtype strain maintained at one particular location ( institution ) ; populations were identified for every trait of interest . Our data were derived from 11 different locations/institutions , and a given location/institution could provide data on multiple populations ( see Supplementary file 1 , Table 1 for details on numbers of strains and institutions ) . We included only populations that contained data points for at least six individuals , and which had information for members of both sexes; further , populations for a particular trait had to come from at least two institutions to be eligible for inclusion . After this selection process , the dataset contained 2 , 300 , 000 data points across 232 traits . Overall , we meta-analysed traits with between 2–18 effect sizes ( mean = 9 . 09 effects , SD = 4 . 47 ) . However , each meta-analysis contained a total number of individual mice that ranged from 83/91 to 13467/13449 ( males/females ) . While a minimum of N = 6 mice were used to create effect sizes for any given group ( male or female ) , in reality samples sizes of male/female groups were much larger ( males: mean = 396 . 66 ( SD = 238 . 23 ) , median = 465 . 56; females: mean = 407 . 35 ( SD = 240 . 31 ) , median = 543 . 89 ) . We used the function escalc in the R package , metafor ( Viechtbauer , 2010 ) to obtain lnCVR , lnVR and lnRR and their corresponding sampling variance for each trait for each population; we worked in the R environment for data cleaning , processing and analyses ( R Development Core Team , 2017 , version 3 . 6 . 0; for the versions of all the software packages used for this article and all the details and code for the statistical analyses , see Source code 1 and repositories ) . As mentioned above , the use of ratio-based effect sizes , such as lnCVR , lnVR and lnRR , controls for baseline changes over time and space , assuming that these changes affect males and females similarly . However , we acknowledge that we could not test this assumption . We conducted meta-analyses at two different levels ( Figure 3C–J ) . First , we conducted a meta-analysis for each trait for all three effect-size types ( lnRR , lnVR and lnCVR ) , calculated at the ‘population’ level ( i . e . using population as a unit of analysis ) . Second , we statistically amalgamated overall effect sizes estimated at each trait ( i . e . overall trait means as a unit of analysis ) after accounting for dependence among traits . In other words , we conducted second-order meta-analyses ( Nakagawa et al . , 2019 ) . We used the second-order meta-analyses for three different purposes: ( A ) estimating overall sex biases in variance ( lnCVR and lnVR ) and mean ( lnRR ) in the nine functional groups ( for details , see below ) and in all these groups combined ( the overall estimates ) ; ( B ) visualizing heterogeneities across populations for the three types of effect size in the nine functional trait groups , which complemented the first set of analyses ( Figure 3I , Table 6 in Supplementary file 1 ) ; and ( C ) when traits were found to be significantly sex-biased , grouping such traits into either male-biased and female-biased traits , and then , estimating overall magnitudes of sex bias for both sexes again for the nine functional trait groups . Only the first second-order meta-analysis ( A ) directly related to the testing of our hypotheses , results of B and C are found in Supplementary file 1 and figures and reported in our freely accessible code . To obtain degree of sex bias for each trait mean and variance ( Figure 3C ) , we used the function rma . mv in the R package metafor ( Viechtbauer , 2010 ) by fitting the following multilevel meta-analytic model , an extension of random-effects models ( sensu Nakagawa and Santos , 2012 ) : ESi ~ 1 + ( 1 | Strainj ) + ( 1 | Locationk ) + ( 1 | Uniti ) + Errori , where ‘ESi’ is the ith effect size ( i . e . lnCVR , lnVR and lnRR ) for each of 232 traits , the ‘1’ is the overall intercept ( other ‘1’s are random intercepts for the following random effects ) , ‘Strainj’ is a random effect for the jth strain of mice ( among nine strains ) , ‘Locationk’ is a random effect for the kth location ( among 11 institutions ) , ‘Uniti’ is a residual ( or effect-size level or ‘population-level’ random effect ) for the ith effect size , ‘Errori’ is a random effect of the known sampling error for the ith effect size . Given the model above , meta-analytic results had two components: ( 1 ) overall means with standard errors ( 95% confidence intervals ) , and ( 2 ) total heterogeneity ( the sum of the three variance components , which is estimated for the random effects ) . Note that overall means indicate average ( marginalised ) effect sizes over different strains and locations , and total heterogeneities reflect variation around overall means due to different strains and locations . We excluded traits which did not carry useful information for this study ( i . e . fixed traits , such as number of vertebrae , digits , ribs and other traits that were not variable across wildtype mice; note that this may be different for knock-down mutant strains ) or where the meta-analytic model for the trait of interest did not converge , most likely due to small sample size from the dataset ( 14 traits , see SI Appendix , for details: Meta-analyses; 1 . Population as analysis unit ) . We therefore obtained a dataset containing meta-analytic results for 218 traits , at this stage , to use for our second-order meta-analyses ( Figure 3D ) . Our dataset of meta-analytic results included a large number of non-independent traits . To account for dependence , we identified 90 out of 218 traits , and organized them into 19 trait sub-groups ( containing 2–10 correlated traits , see Figure 3E ) . For example , many measurements ( i . e . traits ) from hematological and immunological assays were hierarchically clustered or overlapped with each other ( e . g . cell type A , B and A+B ) . We combined the meta-analytic results from 90 traits into 19 meta-analytic results ( Figure 3F ) using the function robu in the R package robumeta with the assumption of sampling errors being correlated with the default value of r = 0 . 8 ( Fisher et al . , 2017 ) . Consequently , our final dataset for secondary meta-analyses contained 147 traits ( i . e . the newly condensed 19 plus the remaining 128 independent traits , see Figure 3 , Supplementary file 1 , Table 2 ) , which we assume to be independent of each other . We created our nine overarching functional groups of traits ( Figure 3G ) by condensing the IMPC’s 26 procedural categories ( ‘procedures’ ) into related clusters . The categories were based on procedures that were biologically related , in conjunction with measurement techniques and the number of available traits in each category ( see Supplementary file 1 , Table 3 for a list of clustered traits , procedures and grouping terms ) . To test our two hypotheses about how trait variability changes in relation to sex , we estimated overall effect sizes for nine functional groups by aggregating meta-analytic results via ‘classical’ random-effect models using the function rma . uni in the R package metafor ( Viechtbauer , 2010 ) . In other words , we conducted three sets of 10 second-order meta-analyses ( i . e . meta-analyzing 3 types of effect size: lnRR , lnVR and lnCVR for nine functional groups and one for all the groups combined , Figure 3H ) . Although we present the frequencies of male- and female-biased traits in Figure 4A , we did not run inferential statistical tests on these counts because such tests would be considered as vote-counting , which has been severely criticised in the meta-analytic literature ( Higgins , 2019 ) . | Males and females differ in appearance , physiology and behavior . But we do not fully understand the health and evolutionary consequences of these differences . One reason for this is that , until recently , females were often excluded from medical studies . This made it difficult to know if a treatment would perform as well in females as males . To correct this , organizations that fund research now require scientists to include both sexes in studies . This has led to some questions about how to account for sex differences in studies . One reason females have historically been excluded from medical studies is that some scientists assumed that they would have more variable responses to a particular treatment based on their estrous cycles . Other scientists , however , believe that males of a given species might be more variable because of the evolutionary pressures they face in competing for mates . Better understanding how males and females vary would help scientists better design studies to ensure they provide accurate answers . Now , Zajitschek et al . debunk both the idea that males are more variable and the idea that females are more variable . To do this , Zajitschek et al . analyzed differences in 218 traits , like body size or certain behaviors , among nearly 27 , 000 male and female mice . This showed that neither male mice nor female mice were universally more different from other mice of their sex across all features . Instead , sex differences in how much variation existed in male or female mice depended on the individual trait . For example , males varied more in physical features like size , while females showed more differences in their immune systems . The results suggest it is particularly important to consider sex-specific variability in both medical and other types of studies . To help other researchers better design experiments to factor in such variability , Zajitschek et al . created an interactive tool that will allow scientists to look at sex-based differences in individual features among male or female mice . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology",
"evolutionary",
"biology"
] | 2020 | Sexual dimorphism in trait variability and its eco-evolutionary and statistical implications |
The factors that govern assembly of the gut microbiota are insufficiently understood . Here , we test the hypothesis that inter-individual microbiota variation can arise solely from differences in the order and timing by which the gut is colonized early in life . Experiments in which mice were inoculated in sequence either with two complex seed communities or a cocktail of four bacterial strains and a seed community revealed that colonization order influenced both the outcome of community assembly and the ecological success of individual colonizers . Historical contingency and priority effects also occurred in Rag1-/- mice , suggesting that the adaptive immune system is not a major contributor to these processes . In conclusion , this study established a measurable effect of colonization history on gut microbiota assembly in a model in which host and environmental factors were strictly controlled , illuminating a potential cause for the high levels of unexplained individuality in host-associated microbial communities .
The human gastrointestinal microbiota makes essential contributions to host metabolic , physiological , and immune functions ( Fujimura et al . , 2010 ) , and aberrations in its structure contribute to pathologies and chronic disease states ( Houghteling and Walker , 2015 ) . Microbiome assembly begins at birth and undergoes dynamic processes influenced by factors such as birth method , feeding method , and antibiotic treatment ( Bokulich et al . , 2016 ) . Once established , the community is dominated by a core set of species characteristic to the host species and is resilient to perturbations ( Dethlefsen and Relman , 2011; Martínez et al . , 2013; Schloissnig et al . , 2013; David et al . , 2014a ) . Despite the profound importance of the host-microbiome symbiosis for human health , the gut microbiota is characterized by a high degree of variation in composition and relative abundance of community members , which is referred to as ß-diversity ( Huttenhower , 2012 ) . This individuality can relate to functions important for health ( e . g . the bio-conversion and activation of dietary compounds and drugs ) , and it is likely that misconfigurations play a role in disease predisposition ( Vatanen et al . , 2016; Kostic et al . , 2015; Koh et al . , 2016; Spanogiannopoulos et al . , 2016 ) . Given these associations , a critical need exists to define the fundamental ecological principles that govern and regulate gut microbiome structure and the mechanisms that drive microbiome variation across hosts . To date , research aimed at identifying the factors that cause microbiome variation has focused largely on host genetics and diet . Host genetic variation has an established role in driving β-diversity in both mice ( Benson et al . , 2010 ) and humans ( Turpin et al . , 2016; Wang et al . , 2016 ) and a subset of fecal bacterial taxa has been shown to be partly heritable and/or associated with host single-nucleotide polymorphisms or quantitative trait loci ( Turpin et al . , 2016; Wang et al . , 2016; Goodrich et al . , 2016; Benson , 2016 ) . However , most taxa do not show any significant heritability , and for those that do , heritability estimates are generally low compared to other heritable traits ( Goodrich et al . , 2016; Hall et al . , 2017; Goodrich et al . , 2017 ) . In addition , genetic studies in humans show a low degree of repeatability , with only a small number of microbiome genome-wide associations being reproducible ( Turpin et al . , 2016; Hall et al . , 2017; Goodrich et al . , 2017 ) . Even the combined effects of host genotype and dietary variables often do not account for the overall variation in a given taxonomic ‘trait' ( Benson , 2016; Leamy et al . , 2014 ) . The rather low contribution of host genetics to microbiome variation and the lack of heritability across many taxa is reflected in studies on monozygotic twins , which although slightly more similar than dizygotic twins , exhibit a substantial degree of individuality ( Goodrich et al . , 2016; Goodrich et al . , 2014 ) . In mice , the effect of diet on microbiome structure has been shown to exceed the contribution of host genetics ( Carmody et al . , 2015 ) . However , the contribution of diet to human microbiome variation was estimated to account for only around 6% ( Wang et al . , 2016 ) , and standardizing the diet did not reduce β-diversity across study participants ( Wu et al . , 2011; David et al . , 2014b ) . Although additional factors such as lifestyle ( e . g . smoking ) , host physiology ( e . g . age ) , and medication contribute to microbiome variation , non-genetic and genetic factors each account for approximately 10% of the variation in gut microbiota and thus fail to explain the majority of the inter-individual diversity that is observed in human cohorts ( Wang et al . , 2016; Falony et al . , 2016; Rothschild et al . , 2018 ) . When attempting to understand compositional features of the gut microbiota , one must consider that the studies performed to date have not accounted for the full range of ecological processes predicted to shape community diversity ( Walter , 2015 ) . Factors such as genotype , diet , smoking , medication , and host physiology all contribute to a deterministic process ( or ‘selection’ ) that impacts diversity through niche-based mechanisms , thus selecting for specific taxa based on fitness differences ( Vellend , 2010 ) . However , according to ecological theory , variation among local communities is not only shaped by selection but also by historical and neutral processes ( Cavender-Bares et al . , 2009 ) , which can only be described in probabilistic terms and include stochastic events such as ecological drift that occur randomly with respect to species fitness differences ( Hubbell , 2001 ) . With respect to historical processes , ecological theory holds that the timing and order of species immigration during community assembly can cause variation in the structure and function of communities , or historical contingency , through priority effects and monopolization ( Fukami , 2015; De Meester et al . , 2016 ) . These effects could be paramount in shaping microbiomes as community assembly is initiated for each newborn since mammals are born germ-free ( Perez-Muñoz et al . , 2017 ) . After birth , the offspring gradually acquires a microbial population , and community assembly occurs in parallel to host immunological , physiological , and metabolic maturation , with reciprocal interactions between microbes and host postnatal development ( Arrieta et al . , 2014 ) . It is increasingly recognized that an early-life ‘window of opportunity' is a critical period for both microbiome assembly and microbiome-driven host development ( Hansen et al . , 2013; Lafores-Lapointe and Arrieta , 2017 ) . Priority effects have been suggested to influence gut microbiota assembly ( Sprockett et al . , 2018 ) , and circumstantial evidence as well as the characteristics of the adult gut microbiota ( stability and resilience despite pronounced inter-individual variation ) are consistent with a combined effect of niche-related ( deterministic ) and early-life historical processes on community assembly ( Walter and Ley , 2011 ) . However , experimental evidence for the importance of colonization history in the gut microbiota assembly is lacking . In this study , we employed a carefully controlled experimental model that allowed us to elucidate the importance of colonization history in early-life gut microbiota assembly by varying arrival order and timing of complex seed microbial communities in previously germ-free mice . The effect of arrival timing on the colonization success of specific gut bacteria and its impact on assembly trajectory of the resident community was also assessed . In both approaches , the influence of the host’s adaptive immune system in causing priority effects and historical contingency was determined .
To determine whether colonization history affects early-life gut microbiota assembly , experiments were first conducted to test the importance of arrival order when colonizing germ-free C57BL/6 mice with two compositionally distinct complex cecal microbial communities obtained from adult mice ( Mus musculus domesticus ) . Experiments consisted of three different treatment groups . In two of the treatments , donor communities were sequentially inoculated in alternating order ( Figure 1 ) , with one treatment group being inoculated first with community A and second with B ( group A/B ) and the second treatment group being first inoculated with community B followed by community A ( group B/A ) . In a third treatment group , both donor communities were inoculated at each time point ( group AB/AB ) . The age of 10 days was chosen for the first inoculation as it was the earliest age at which pups could be gavaged without jeopardizing their health or parental care . At weaning ( 21 days ) , parent mice were sacrificed and their ceca collected ( samples Parents A , Parents B , and Parents AB ) , while offspring mice were separated by sex . The second inoculation was done at day 36 ± 2 after birth via gavage ( Figure 1 ) . The cecal microbiome of the recipient mice was characterized at day 78 ± 2 , by MiSeq Illumina sequencing of 16S rRNA tags using Minimum Entropy Decomposition ( MED ) analysis , which allows the differentiation of sequences with only one nucleotide dissimilarity ( Eren et al . , 2015 ) , that are referred to as ‘types’ . The rationale to use whole complex microbiomes in contrast to defined simplified communities of strains ( which are commonly used in the field of microbial ecology research [De Roy et al . , 2014] ) was to assure that niches were for the most part filled , establishing the competitive interactions present in natural populations . To avoid microbiomes of mice that originate from commercial facilities , which are often derived from gnotobiotic mice and are frequently functionally aberrant and low in diversity ( Lagkouvardos et al . , 2016; Kreisinger et al . , 2014; Rosshart et al . , 2017 ) , selected donor mice were caught in the wild or derived from wild mice . One donor mouse was selected from a colony derived from mice caught in the Massif Central region of France and subsequently housed in a laboratory for three generations ( Wang et al . , 2014 ) ( Donor A; referred to as ‘wild-derived’ ) . The second donor was a wild mouse caught in Scotland ( Weldon et al . , 2015 ) ( Donor B; referred to as ‘wild-caught’ ) . These allopatric mouse populations differed markedly in gut microbiota composition ( Lagkouvardos et al . , 2016 ) . Prior to the actual experiments , the microbial communities from donors A and B were established in germ-free C57BL/6 mice in our facility to standardize the inocula . Cecal bacterial diversity became only marginally reduced in recipient mice 4 weeks after colonization during the standardization when compared to the wild donors , and the majority ( 85% for donor A and 82% of donor B ) of the bacterial types could be detected ( Figure 1—figure supplement 1A ) . In addition , an analysis of Bray-Curtis dissimilarities revealed that recipient microbiota clustered tightly with that of the donor ( Figure 1—figure supplement 1B ) . Overall , these analyses showed that a large proportion of the microbiota from wild mice can be established in germ-free mice under laboratory conditions with similar taxonomic distributions and proportions . The experimental model has several unique advantages in that it recapitulates ‘real life’ conditions during early-life microbiota assembly while keeping ecological variables strictly controlled . The first colonization was done in 10-day-old germ-free mice , which resemble newborn mice as they are devoid of microbes and not yet fully developed , allowing the study of gut microbiota assembly during a time window critical for both microbiota assembly and host postnatal development ( Knoop et al . , 2017 ) . The parent mice become hereby fully colonized , which allows vertical and horizontal transmission of bacteria ( i . e . through coprophagia ) to the offspring mice up until weaning when they become fully mature . From an ecological perspective , the model is uniquely suitable to disentangle the relative importance of historical processes in community assembly . First , community assembly at local scales ( mice ) occurs under standardized conditions that included the use of inbred mice under the same husbandry regimen , thereby avoiding variation in deterministic ( niche-related ) ecological factors . Second , although complex undefined microbiomes were used , the regional species pool , which can profoundly influence the ecological processes that determine community structure at local scales ( Chase and Myers , 2011 ) , including priority effects ( Sprockett et al . , 2018 ) , is standardized ( and to a large degree defined through sequencing ) . This was achieved by the administration of aliquots of the same inoculum to mice via gavage ( a highly efficient method of transmission ) and the maintenance of mice in isolators that prevented acquisition of additional members . Consequently , colonization order is the only experimental variable in the model . Analysis of donor and recipient microbiota showed that 99 . 5% of the bacterial types found in the donor microbiomes were also found in the recipient mice , while 93 . 4% of the types ( adding up to 99 . 3% of the total sequences ) in the recipient mice were also detected in the donors . These findings indicate an almost complete transfer of the communities from donors to recipients without an expansion of minority members of the donor communities in the recipient mice . In addition , the abundance distribution of bacterial types in the recipient mice reflected to a large degree those of the donors , showing that transfer of the microbiota did not result in major rearrangements in community structure ( Figure 2A ) . Alpha-diversity of the cecal bacterial community at day 78 was not influenced by colonization order ( Figure 2B ) . Interestingly , α-diversity analyses revealed that the microbiomes of recipient mice had a significantly higher number of observed bacterial types compared to the donor communities ( Figure 2B ) , irrespective of the order in which the donor microbiomes were introduced ( p=2×10−16 ) . Higher diversity was observed when species abundance and evenness ( Shannon Index ) were considered ( p=1 . 13×10−4 ) ( Figure 2C ) . These results suggest that the niche space in the cecal bacterial community , even in wild mice , is not filled , which agrees with the propositions that most ecosystems in nature are unsaturated ( Cornell and Lawton , 1992; Pinto-Sánchez et al . , 2014 ) . This finding is interesting as it provides a potential explanation for why donor-specific bacterial types can colonize the gut of a recipient host with a ‘largely’ undisrupted microbiota ( i . e . patients with metabolic syndrome ) after fecal microbiota transplantation ( Li et al . , 2016 ) . Hierarchical clustering performed with all bacterial types detected in the donor and recipient mice revealed clustering of the communities dependent on colonization order ( Figure 2A ) , indicative of historical contingency . Since it was not possible to analyze priority effects for bacterial types shared among Donors A and B , downstream analyses were performed with bacterial types specific to either donor ( Figure 2D and E ) . A total of 395 bacterial types were detected across the two donor communities , of which 101 were uniquely detected in community A ( donor and parents A ) , 83 were exclusively detected in community B ( donor and parents B ) , and 211 were common to both . Arrival order did not influence the number of successful colonizers unique to donor A ( 88 ± 6 in A/B , 86 ± 8 in B/A , and 84 ± 9 in AB/AB , F = 2 . 75; p=0 . 0521 ) , but influenced colonization success of bacterial types unique to donor B ( 67 ± 8 in B/A , compared to 48 ± 10 in A/B , and 60 ± 7 in AB/AB , F = 26 . 5; p=8 . 66×10−11 ) . These findings suggest that although a subset of members of the wild-caught B community are less fit than those of the wild-derived A community ( perhaps because the latter are pre-adapted to facility conditions ) , early arrival can partly offset this fitness disadvantage . Communities at local scales assemble through a combination of niche-related , neutral , and historical processes ( Cavender-Bares et al . , 2009 ) . Since the analysis of β-diversity can provide insight into the importance of ecological processes ( Chase and Myers , 2011 ) , analyses of Bray Curtis and binary Jaccard dissimilarities were done to determine whether arrival order affected community structure , and how microbiome assembly was impacted by dispersal limitation ( as described above , inter-individual niche-related differences should be marginal in our model ) . The overall cecal microbiomes at day 78 clustered closer to donor A when compared to donor B ( Figure 3A ) , supporting the conclusion above that members of the donor A population are on average fitter . This analysis further revealed that communities from A/B mice clustered closer to Donor A when compared to AB/AB and B/A mice , while those of B/A mice clustered closer to Donor B as compared to AB/AB and A/B mice ( Figure 3B ) , suggesting that the communities of the recipients are more similar to the donor community that arrived first . Analysis of variance of the bacterial communities of just the recipient mice using Adonis revealed that bacterial profiles ( clustered on the basis of Bray-Curtis ) were significantly affected by colonization order ( all treatments p<0 . 001 , R2 = 0 . 21; A/B vs B/A p<0 . 001 , R2 = 0 . 23; A/B vs AB/AB p<0 . 001 , R2 = 0 . 16; and B/A vs AB/AB p=0 . 002 , R2 = 0 . 11 ( Figure 3C ) . Binary Jaccard-based analysis confirmed these differences ( Figure 3—figure supplement 1 ) , indicating a measurable influence of colonization order on gut microbiota β-diversity . The experiment further demonstrated a clear importance of dispersal limitation in gut microbiota assembly . Cecal microbiomes of mice housed within the same cage were more similar ( less β-diversity ) than mice housed in different cages within the same isolator ( p=0 . 092 within A/B; p=0 . 019 within B/A; and p<1 . 00×10−7 within AB/AB; Figure 3D ) , while mice housed in different isolators ( no opportunity for dispersal ) had the largest compositional differences ( p<1 . 00×10−7 within A/B and B/A; p=6 . 1×10−3 for AB/AB; Figure 3D ) . β-diversity analysis based on Bray Curtis dissimilarities revealed significant clustering by both isolator ( Figure 3E ) and cage ( Figure 3F ) . Adonis analysis performed using ‘cage’ , ‘isolator’ , and ‘colonization order’ as explanatory variables revealed that the data was best explained by the factor cage ( p<0 . 001 , R2 = 0 . 66 ) , followed by isolator ( p<0 . 001 , R2 = 0 . 46 ) and colonization order ( p<0 . 001 , R2 = 0 . 21 ) . Analogous results were obtained when using binary ( presence/absence ) measurements ( Figure 3—figure supplement 2 ) . These results demonstrate that dispersal limitation can cause cage effects even if individual animals acquire exactly the same species members in a completely standardized environment . This finding has substantial implications for the design and interpretation of mouse experiments , as it suggests that a difference among mouse groups can arise not only from familial transmission ( Ubeda et al . , 2012 ) but also solely from ecological drift . However , although dispersal limitation does explain the largest proportion of β-diversity in the cecal bacterial population of recipient mice , colonization order makes a clear measurable contribution . The impact of arrival order on microbiota assembly prompted the evaluation of specific taxonomic groups affected , using linear mixed models . No significant differences in phyla , families , or genera abundances were observed across experimental groups ( data not shown ) . However , 20 bacterial types were significantly influenced by colonization order ( Supplementary file 1 ) , indicative of priority effects . The taxonomic distribution of these bacterial types mirrored the overall community profile , showing that no particular phylum was specifically more affected by colonization history . The fact that colonization order impacts bacterial types but not higher taxonomic levels points to the importance of competitive interactions as a major mechanism driving the observations . To gain insight into the mechanisms by which priority effects emerge , it was assessed whether bacterial types affected by colonization order were overrepresented when inoculated first or overrepresented when introduced second . If a bacterial type is overrepresented when inoculated first , it would occur either through niche-preemption , monopolization ( the evolutionary process by which early arrivals to a new patch adapt to the local conditions , gaining an advantage over later colonists ) or niche-modification that provides an advantage to the earlier colonizer , all of which are inhibitory ( Fukami , 2015 ) . In contrast , an overrepresentation of a bacterial type that that arrives second can occur through facilitative niche-modification ( Fukami , 2015 ) . This analysis revealed that priority effects were mostly inhibitory ( Supplementary file 1 ) , as in 90% of the cases , bacterial types were more successful when introduced in the first inoculation ( two examples , Types 4996 and 0857 , are shown in Figure 4A ) . However , instances of facilitative effects were also observed as two bacterial types ( Firmicutes ) had a higher abundance when introduced second ( Type_4501 is shown as an example in Figure 4A; Supplementary file 1 ) . Inhibitory effects can occur through niche pre-emption , where early arriving species inhibit the colonization of those arriving later through competitive exclusion and limiting similarity ( Fukami , 2015 ) . As closely related organisms often display overlapping niches , niche pre-emption might manifest by related organisms to exclude one another ( Cavender-Bares et al . , 2009; Violle et al . , 2011; Fukami et al . , 2016 ) . However , colonization success of bacterial types ( evaluated as the fraction of mice colonized or the average abundance of the bacterial type ) was not a function of phylogenetic distance to members of the opposite donor community ( Figure 4—figure supplement 1 ) . This finding suggests that bacterial functions among gut microbiota members are not necessarily related to phylogeny . Although an overall phylogenetic signal could not be detected , there were four instances in which closely related bacterial types introduced early prevented the establishment of related bacterial types that arrived later ( 1 to 7 mismatches in 16S rRNA gene tag , >97% – >99% similarity; Figure 4B ) . In two of the cases ( Type_5446 vs . Type_1231 , and Type_5450 vs . Type_5446 ) the two types co-existed when introduced together in group AB/AB , and earlier colonization makes either of them more competitive ( higher relative abundance ) ( Figure 4B ) . In the other two cases ( Type_4996 vs . Type_4997 , and Type_4297 vs . Type_4299 ) , one type excludes the other type when introduced together , but earlier colonization can overcome this fitness difference and give an advantage to the type that displays a lower degree of fitness ( Figure 4B ) . Overall , these patterns provide evidence for the importance of priority effects that are inhibitory through niche pre-emption . The findings described above established the importance of colonization order in community assembly when whole microbiomes were acquired in different succession . These experiments raise the question on the exact contribution of specific members within these communities toward historical contingency , and how they themselves are impacted by arrival timing . Thus , in a separate set of experiments , four bacterial strains of mouse origin ( Bacteroides vulgatus RJ2H1 , Lactobacillus reuteri Lpuph1 , Lactobacillus johnsonii DPPM , and Clostridium cocleatum ATCC 29902T ) were gavaged together into germ-free mice either before ( at 5 days of age ) , soon after ( at 14 days of age ) , or more than 3 weeks after ( at 36 days of age ) the mice received a complex cecal microbiota ( obtained from pooled cecal samples of five C3H/HeN mice raised in conventional housing at the University of Nebraska Gnotobiotic Mouse Facility ) at 10 days of age ( Figure 5A ) . Using strain-specific quantitative PCR ( qPCR ) , abundance of each strain was quantified in fecal samples collectedly weekly throughout the 36 days period after the respective inoculations , as well as in the ceca collected at day 78 . As shown in Figure 5B , arrival timing significantly impacted colonization patterns for two of the strains . Cell numbers of L . reuteri Lpuph1 were significantly higher ( p=8 . 99×10−8 time*group interaction ) at all time points when introduced on day 5 compared to days 14 and 36 ( Figure 5B ) , and the strain was only detected at the end of the experiment when it was introduced on day 5 but not at later time points ( detectable in two mice , 1 . 71 × 106 cells/g of cecal content ) . For C . cocleatum ATCC 29902T , the timing of immigration was absolutely crucial , as the organism was only detected in mice when the strain was introduced at day 5 ( p<2 . 20×10−16 time*group interaction ) ( Figure 5B ) . For B . vulgatus RJ2H1 and L . johnsonii DPPM , time of inoculation did not affect colonization success , and both strains were able to stably colonize mice at equivalent levels independently of arrival time ( Figure 5B ) . These findings indicate that specific microbiota members can differ substantially in the degree to which arrival timing relative to competitors influences their colonization success . Theory predicts that functional similarity ( niche overlap ) among colonists makes the outcome of interspecific competition sensitive to arrival order ( Fukami , 2015; Sprockett et al . , 2018 ) . This prediction was tested by determining the closest relative to the introduced species within the donor microbiome from the 16S rRNA data . The analysis showed that the two strains in the cocktail that were favored by early arrival , L . reuteri Lpuph1 and C . cocleatum ATCC 29902T , were 100% identical in their 16S rRNA gene sequence to bacteria present in the donor community , Type_3050 ( 2 . 15 ± 2 . 60% abundance ) and Type_3983 ( 0 . 42 ± 0 . 15% abundance ) ( Figure 5C ) . In contrast , the two stable colonizers not impacted to arrival order had no closely related organism in the donor community ( Figure 5C ) . Although other mechanisms cannot be excluded , our findings therefore suggest that early arrival of a bacterium is critical for its ecological success if the community contains members that overlap with its niche . If such competitors are absent , then stable colonization is independent of timing and can occur later in life , equivalent to findings in adult humans in a recent study with Bifidobacterium longum ssp . longum ( Maldonado-Gómez et al . , 2016 ) . To test if differences in arrival time of a small number of early colonizers can impact gut microbiota assembly , the cecal bacterial community of the 78-day-old mice that received the four-strain cocktail at different time points was analyzed . No differences in α-diversity were detected across treatments ( number of observed bacterial types p=0 . 464 , and Shannon Index p=0 . 841; Figure 5D ) . However , structure of the fecal bacterial communities did show significant differences based on when the cocktail of specific colonizers was introduced ( Adonis; Bray Curtis p=0 . 041 , R2 = 0 . 08; Jaccard Index p=0 . 013 , R2 = 0 . 08 ) . The fecal microbiota in mice inoculated with the four-strain mix at day 5 differed in their overall composition at day 78 compared to those that received the cocktail of strains at day 14 ( Adonis; Bray Curtis p=0 . 047 , R2 = 0 . 07; Jaccard Index p=0 . 027 , R2 = 0 . 06 ) . There was no difference between mice colonized with the cocktail at day 14 and day 36 ( Adonis; Bray Curtis p=0 . 114 , R2 = 0 . 06; Jaccard Index p=0 . 073 , R2 = 0 . 06 ) ( Figure 5E ) . Twenty-six bacterial types significantly differed in abundance across treatments ( random forest analysis; mean importance >2; Supplementary file 2 ) , none of which represented the strains from the cocktail . Overall , these findings demonstrate that differences in arrival timing of just a small number of early colonizers can bring about marked changes in the trajectory of gut microbiota assembly . From a microbial ecosystem perspective , the vertebrate gut is unique in that not only can the microbial community members adapt ( through monopolization ) during assembly ( De Meester et al . , 2016 ) , but the habitat ( host ) can also respond and adapt towards the community during postnatal development ( Atarashi et al . , 2011; Ivanov et al . , 2009 ) . Specialized features of the adaptive immune system allow the host to develop immunological tolerance to gut bacteria , for example through antigen-specific regulatory T cells ( Belkaid and Hand , 2014; Cebula et al . , 2013 ) . These immune processes differ between the neonatal and adult immune system ( Sudo et al . , 1997; Olszak et al . , 2012 ) such that early host exposure to specific bacteria may precondition the immune environment for bacterial colonization ( Knoop et al . , 2017; Gensollen et al . , 2016 ) . Differences in bacterial arrival time during community assembly could therefore cause historical contingency through adaptive immune responses that favor early colonizers during a ‘window of opportunity’ ( Knoop et al . , 2017; Torow and Hornef , 2017; Koskella et al . , 2017 ) . To test whether adaptive immune responses during postnatal immune development contribute to historical contingency in gut microbiota assembly , the same set of experiments as described above was repeated in mice of the same genetic background ( C57BL/6 ) but without adaptive immunity ( Rag1-/- ) , using the same donor microbial communities and four-strain cocktail , and compared the finding with wild-type ( WT ) mice . Equivalent to the findings in WT mice , the introduction of two whole cecal bacterial communities into Rag1-/- mice led to a significant increase in species richness ( p=6 . 63×10−14 ) when compared to the individual donor communities , regardless of colonization order ( Figure 6A ) . However , when compared to WT mice , Rag1-/- mice in two treatments ( B/A and AB/AB ) had a slight but significantly lower number of observed bacterial types ( p=7 . 3×10−4 and p=0 . 028 , respectively ) ( Figure 6B ) , and one treatment ( B/A ) had a lower Shannon diversity index ( p=0 . 043; data not shown ) . In addition , the overall cecal microbiota differed between WT and Rag1-/- mice ( Adonis based on Bray-Curtis dissimilarities , p<0 . 001 , R2 = 0 . 07 ) ( Figure 6C ) , with 21 bacterial types to significantly differ in abundance across genetic backgrounds ( Supplementary file 3 ) . These results suggest that the adaptive immune system increases diversity and affects colonization of specific bacterial strains , although the findings contrast with previous research that showed that diversity between WT and Rag1-/- mice and Zebrafish were not different ( Zhang et al . , 2015; Stagaman et al . , 2017 ) . Colonization history remained important in Rag1-/- mice , but some differences were detected . The impact of early colonization on the similarity of recipient communities to the donor communities was reduced in Rag1-/- mice ( Figure 6D ) , with only the B/A mice being significantly more similar to donor/parents B when compared to A/B mice ( Figure 6E ) . In addition , the enhanced level of colonization of donor B-specific bacteria when introduced early observed in WT mice ( Figure 2E ) , was not detected in Rag1-/- mice , and the average number of donor-B-specific bacterial types successfully colonizing the gut did not differ across treatments ( p=0 . 166 , p=0 . 279 , and p=0 . 980 ) ( data not shown ) . These results illustrate that the adaptive immune system is a contributor to the successful establishment of specific early arriving bacterial types in the gut . However , just as in WT mice ( Figure 3C ) , there was significant clustering of the ceca microbiota by colonization order in Rag1-/- mice ( Adonis; Bray Curtis and Jaccard dissimilarity , p<0 . 001 , R2 = 0 . 25; p<0 . 001 , R2 = 0 . 22 , respectively ) ( Figure 6F ) . β-diversity was impacted to a larger degree by colonization order than by the adaptive immune system ( envfit analysis of correlation between MDS axes and variables; p=0 . 001; R2 = 0 . 15 , and R2 = 0 . 07 , respectively ) , indicating a stronger effects of colonization history over host adaptive immunity . To compare which bacterial types were affected , differences in bacterial abundance were analyzed in WT and Rag1-/- mice using a linear mixed model with genetic background and colonization order as the main effects , isolator as a random effect , and an interaction term between colonization order and genetic background ( and FDR correction was applied to the P values ) . Together , forty-three bacterial types were significantly affected by colonization order , with a majority ( 81% ) showing patterns indicative of inhibitory priority effects ( Supplementary file 4 ) . Importantly , those included 19 out of the 20 bacterial types previously identified to be significantly affected in WT mice , while only two bacterial types showed a significant interaction between genetic background ( WT vs Rag1-/- ) and colonization order ( data not shown ) , illustrating similar ecological effects in the two genetic backgrounds . Overall , these findings revealed that colonization history remained an important contributor to microbiota assembly from two complex donor communities in the absence of adaptive immunity . Interestingly , experiments that tested the importance of arrival timing of specific colonizers revealed that both priority effects and historical contingency of community assembly were larger in Rag1-/- mice . The colonization patterns of L . reuteri Lpuph1 and C . cocleatum ATCC 29902T were not different in Rag1-/- and WT mice ( p=0 . 22 ) , and colonization timing remained important ( Figure 6G and Figure 6—figure supplement 1 ) . In contrast , the population sizes of L . johnsonii DPPM and B . vulgatus RJ2H1 , which were not affected by colonization timing in WT mice , were lower in Rag1-/- mice when colonization occurred later ( Figure 6G , Figure 6—figure supplement 1 ) . For both strains , later colonization ( 14 and 36 days ) resulted in significantly lower cell numbers in Rag1-/- mice compared to mice colonized at day 5 ( p=1 . 69×10−5 for L . johnsonii , and p=0 . 004 for B . vulgatus ) and to WT mice colonized at days 14 and 36 ( p=1 . 69×10−5 for L . johnsonii , and p=1 . 33×10−7 for B . vulgatus ) . The impact of the specific colonizers on the historical contingency of the bacterial community was also larger in Rag1-/- mice , as colonization of the strain cocktail at 5 and 14 days resulted in different microbiome configurations when compared to each other and when compared to mice colonized at 36 days ( Adonis test across all groups; Bray Curtis dissimilarity p<0 . 001 , R2 = 0 . 19 , Jaccard dissimilarity p<0 . 001 , R2 = 0 . 17 ) ( Figure 6H ) . Random forest analysis revealed that 46 bacterial types differed in the assembled community ( 78 days ) of Rag1-/- mice inoculated at days 5 , 14 or 36 ( Supplementary file 5 ) , showing that a greater number of types were affected by colonization order when compared to WT mice . Overall , although the comparisons of WT and Rag1-/- mice indicate that the adaptive immune system can contribute to the successful establishment of specific early colonizers , priority effects and historical contingency still occured in Rag1-/- mice , and both processes were , in several measurements , increased when compared to WT mice . More bacterial types were effected by colonization order in Rag1-/-mice in the experiments with the bacterial cocktails , and effect sizes in Adonis tests that explained the impact of colonization order on community variation were , on average , higher in both experiments . Especially the experiments using the cocktail of specific colonizers clearly showed an enhanced role of historical processes in Rag1-/- mice on both the colonization success of the individual microbes as well as the historical contingency of the entire bacterial community .
Most of the inter-individual variability observed in mammalian gut microbial communities is unaccounted for ( Benson et al . , 2010; Wang et al . , 2016; Goodrich et al . , 2016 ) . Previous work aiming to elucidate the factors that drive diversity patterns focused mostly on deterministic processes while neglecting stochastic ecological elements , that according to theory , are implicated in shaping community assembly ( Vellend , 2010; Cavender-Bares et al . , 2009; Sprockett et al . , 2018 ) . By establishing that historical processes can drive β-diversity of the gut microbiota in a setting where host genetics , diet , and the regional species pool are standardized , this study demonstrated that individuality in gut microbiota composition can arise solely from a variation of the timing by which microbes are acquired . Although the relative importance of assembly history appears small in our experiments ( and less than that of dispersal limitation ) , effects of species arrival history in real life are predicted to be amplified over time and space ( Fukami , 2015 ) . Our finding that differences in arrival timing by only four members by a few days can produce a measurable alteration in the long-term trajectory in the development of the community is therefore especially relevant . Given that there are hundreds if not thousands of species that assemble into gut microbiomes , most of whose acquisition is likely to some degree stochastic , historical events have the potential to add up to contribute substantially to individuality of gut microbiomes . Our findings therefore have important implications on our understanding of how microbiotas assemble . They further illuminate the need to develop statistical models and algorithmic approaches that could be incorporated into existing algorithms for population-driven analyses ( e . g . GWAS and MWAS ) to accurately measure the relative contributions of ecological factors , host genetic variation , and other deterministic factors in large-scale population-based studies . Although our findings clearly establish the importance of colonization history in gut microbiota assembly and suggest an importance of inhibitory priority effects , they provided limited insight into the exact mechanisms that cause historical contingency . Priority effects can arise through niche preemption or modification ( Fukami , 2015; Sprockett et al . , 2018 ) , or through in situ evolution by which earlier colonizers adapt to gain an advantage over later colonizers through monopolization ( De Meester et al . , 2016 ) . These processes might also be impacted by interactions across domains of life ( bacteria , eukaryotes , archaea ) and even bacteriophages . Future studies should therefore employ metagenomic sequencing and ‘trait-based’ approaches with careful consideration of the interactions between the functional , taxonomic , and genetic features of both the colonists and the assembled microbiomes in an attempt to determine the relative importance of niche-preemption , niche-modification ( including inter-domain interactions ) , and monopolization in historical contingency . In addition , arrival timing of microbes might also change the trajectory of gut microbiota assembly by inducing developmental , physiological , and immunological changes in the host that then in themselves shape the microbiota . In other words , the timing of species immigration might not only impact microbe-microbe interactions that would qualify as classical priority effects , but also reciprocal interactions of microbes and the host . Future studies should attempt to determine to what degree the latter contributes to historical contingency in gut microbiota assembly . If relevant , host factors and host-microbe interactions would have to be incorporated into ecological theory in order for its application to be useful for our understanding of host-associated microbial communities . In this study , the role of a key inducible host factor in historical contingency , the adaptive immune system , was systematically analyzed . We hypothesized , like others ( Sprockett et al . , 2018; Koskella et al . , 2017 ) , that ecological consequences similar to those of priority effects would arise through early colonizers gaining an advantage through the induction of tolerance by adaptive immune responses ( Knoop et al . , 2017 ) . However , our comparisons of WT and Rag1-/- mice indicated that the importance of historical processes was , with few exceptions , not reduced in the absence of adaptive immunity , and nearly all the taxa affected by colonization order in WT mice were also among those affected in Rag1-/- mice . Moreover , the overall effect of colonization history was actually amplified in most measurements in Rag1-/- mice . A potential explanation for this finding is a higher relative contribution of selection ( deterministic processes ) in Rag1-/- mice compared to their WT counterparts . Ex-germ-free Rag1-/- mice develop inflammation in the intestinal epithelial lining in response to bacterial colonization ( Peterson et al . , 2007 ) , and this inflammatory milieu could exert a higher selective pressure on the bacterial community . Stronger selection is predicted to increase competition , which in itself increases the importance of priority effects ( Fukami , 2015 ) , providing an explanation for the impact of sequence of arrival on the population of L . johnsonii and B . vulgatus in Rag1-/- but not WT mice . In summary , it appears that for some specific species , the adaptive immune system might alter niches , allowing , for example , donor-B-specific lineages to persist ( Figure 2E ) , a finding only obtained with WT but not Rag1-/- mice . However , most of our observations do not support the hypothesis that interactions between microbes and the adaptive immune system are a major contributor to historical contingency . Instead , our analysis on microbe-microbe associations ( Figure 4 ) suggests ecological interactions such as inhibitory priority effects and potentially niche-preemption as the dominant processes . Future studies are necessary to determine if additional inducible host factors ( innate immune functions , defensins , mucus , etc . ) constitute a mechanism by which arrival timing of microbes causes historical contingency in gut microbiota assembly . Although our understanding on how ecological theory can be applied to the gut microbiome is still vastly incomplete , our findings contribute to an emerging historical perspective that could explain the spatial and temporal patterns of diversity described for the gut microbiota ( Walter and Ley , 2011 ) . The seemingly paradoxical characteristics of the gut microbiota , namely resilience and high intra-individual stability despite large inter-individual variability , are consistent with a view in which stochastic historical events such as chance colonization , random extinction , ecological drift , and monopolization , in combination with niche pre-emption and modification ( Fukami , 2015 ) , drive inter-individual variability . Eco-evolutionary feedbacks in which colonizers continue to adapt to the niche opportunities that arise during the dynamic assembly process ensure that communities end up to be , despite their unpredictable configurations , composed of highly adapted members that stably occupy niches and display resilience and colonization resistance ( Walter , 2015 ) . Since some level of stochasticity during microbiota assembly would be expected to occur even among monozygotic twins , this framework provides an explanation for their surprisingly high inter-individual microbiota variation ( Goodrich et al . , 2016; Goodrich et al . , 2014 ) . In addition , low dispersal decreases rates of immigration , which is predicted to enhance priority effects as average time spans between early and later colonizers are increased ( Fukami , 2015 ) , consequently increasing β-diversity ( Hubbell , 2001; Chase and Myers , 2011; Rosindell et al . , 2012 ) . These concepts may provide one explanation for the higher microbiota individuality in people living in industrialized societies ( which likely have lower levels of dispersal due to sanitation ) when compared to non-industrialized populations ( Martínez et al . , 2015 ) . However , although these concepts are in agreement with much of the observational data on microbiomes , the findings in our study derive from experiments testing only two communities and four specific members , and it will require additional mechanistic , observational , and theoretical studies in a variety of contexts to validate the importance of historical perspective of gut microbiota assembly . Specifically , future research should be targeted at determining to what extent general concepts of ecological theory can be applied to host-associated microbial communities . Apart from contributing to our basic understanding of gut microbial ecology , this study provides a critical foundation for future experiments that test the impact of colonization order in disease predisposition and whether assembly history can be systematically influenced . Given the importance of historical contingency for gut microbiota assembly , clinical and medical interventions early in life ( e . g . antibiotics , C-sections , formula feeding ) are likely to have longer lasting consequences , driving not only inter-individual differences but potentially also aberrant patterns of colonization that could potentially be prevented by an adjustment of clinical practices to avoid priority effects ( Sprockett et al . , 2018; Dominguez-Bello et al . , 2016 ) . Our findings are also relevant for the development of strategies to modulate microbiomes . As the understanding of the health-promoting attributes of gut bacteria continues ( Olle , 2013 ) , it will be important to evaluate how they can be established more permanently . Once assembled , the gut microbiota is extremely resilient to therapeutic modulations , dietary changes and moderate doses of antibiotics ( Dethlefsen and Relman , 2011; Martínez et al . , 2013; Robinson et al . , 2010 ) , and colonization resistance constitutes a major barrier to introducing beneficial microbes ( Walter et al . , 2018 ) . Our results confirmed recent findings in humans that bacterial strains can be stably established in a climax community if closely related species are absent ( Maldonado-Gómez et al . , 2016 ) . However , if competitors are present , stable persistence cannot be achieved . More permanent persistence can be achieved if microbes are introduced early in life . In addition , the findings demonstrated that early introduction of just a few species can divert the entire trajectory of the microbiota . If such shifts can be introduced reproducibly , then early colonizers could be selected to deliberately control microbiome assembly to obtain predictable outcomes . Although there will likely always be strong stochastic elements in microbiome acquisition , priority effects will favor bacteria that are introduced first , thereby providing an opportunity to potentially prevent aberrant microbiomes , and by doing so , dysbiosis-related diseases .
Germ-free C57BL/6 wild-type ( WT ) and Rag1-/- ( which lack mature B and T lymphocytes ) mice were born and reared in flexible film isolators and maintained under gnotobiotic conditions at the University of Nebraska-Lincoln . Breeding mice and pups ( up to 21 days of age ) were fed autoclaved LabDiet 5021 ( Purina Foods , St . Louis , MO ) ad libitum . After weaning ( at 21 days of age ) , mice were fed autoclaved LabDiet 5K67 ( Purina Foods ) ad libitum . Germ-free status of the breeding colony was performed routinely as described using culture , microscopy , and PCR ( Bindels et al . , 2017 ) . The Institutional Animal Care and Use Committee of the University of Nebraska-Lincoln approved all procedures involving animals ( Project ID 731 and 817 ) . To avoid artificial niche opportunities , donor ceca communities from adult wild mice were selected . The ceca selected corresponded to i ) a mouse originating from a population of mice derived from a wild mouse population sampled in the Massif Central region of France in 2009 and was subsequently housed at the breeding facility of the Max Plank Institute for Evolutionary Biology for three generations ( Wang et al . , 2014 ) , referred to in the present manuscript as donor community A , and ii ) from a wild mouse caught in Scotland ( Weldon et al . , 2015 ) , which is referred to as donor community B . To prepare inocula , frozen ceca were introduced into an anaerobic chamber ( Bactron600 Shel Lab , Sheldon Manufacturing INC . , Cornelius , OR ) to avoid oxygen exposure , and slurries were made by suspending cecal contents in pre-reduced , sterile PBS with 10% glycerol ( pH 7 ) at a dilution of 1:10 and stored at −80°C in aliquots of 500 μl to avoid freeze-thawing . To generate standardized donor cecal inocula , the cecal communities of the wild mice was transferred into germ-free C57BL/6J mice reared at the UNL’s Animal Research Facility . This standardization was performed to i ) obtain larger volumes of inocula needed for the experiments , ii ) determine whether a wild mouse microbiota could be successfully established in C57BL/6J mice under laboratory conditions , and iii ) standardize the inocula to the same environmental conditions ( food , oxygen exposure at gavage , environmental changes due to laboratory conditions , etc . ) . To establish microbiomes in mice , 100 μl of either donor A or B were orally gavaged into 8 week old C57BL/6J mice ( N = 8 for donor A , and N = 5 for donor B ) housed in separate flexible plastic isolators . Four weeks after gavage , mice were euthanized and their ceca harvested . Ceca were excised inside a laminar flow hood with aseptic techniques , and immediately snap-frozen in liquid nitrogen . Cecal slurries of the two donor microbiomes were obtained after pooling ceca from recipient mice under anaerobic conditions and homogenizing contents with sterile , pre-reduced PBS with 10% glycerol ( 1:10 ratio ) as described above . Donor microbiomes were stored in 500 μl aliquots and kept frozen at −80°C until further use . DNA extraction was done according to Martínez and co-workers ( Martínez et al . , 2010 ) , and sequencing of the V5-V6 region of the 16S rRNA ( Krumbeck , 2015 ) was done at the University of Minnesota Sequencing Center . Analysis of the 16S rRNA data was done as explained below . The first set of experiments was designed to systematically modify colonization order of two donor cecal microbiomes ( A and B; see supplementary file for additional information on how donor inocula were selected and prepared ) , through three experimental treatment groups . In two of the treatments , donor communities were sequentially inoculated in alternating order ( Figure 1A ) , where community A was inoculated first and B second ( group A/B ) and where community B was inoculated first and A second ( group B/A ) . In a third treatment , both donor communities were inoculated at each time point ( AB/AB ) . Either germ-free C57BL/6 wild-type ( WT ) or Rag1-/- mice were inoculated at 10 ± 2 days after birth by oral gavage ( first inoculation ) , and at 36 ± 2 days after birth ( second inoculation ) . Inoculation by oral gavage consisted of introducing 50 μl of cecal contents into the mouth of each pup using a small-sized syringe . The leftover inoculum was deposited on the parents’ fur ( for 10-day-old mice ) , on the mouse’s fur ( for 36-day-old mice ) , and on the cage bedding . For the AB/AB group , separate vials containing donor A and B aliquots were introduced into the isolators and mixed right before administering to the mice . Pups were weaned at 21 ± 1 days after birth , and the parents were euthanized to collect their ceca for microbiome analysis; pups were placed into separate cages according to sex . On day 78 after birth ( ±1 day ) , mice were euthanized and their ceca harvested and stored at −80°C until DNA extraction . Mice were maintained in isolators for the entire duration of the experiment , with separate isolators for different treatments and genetic backgrounds ( WT or Rag1-/- ) . Each experiment was done in duplicate ( N ≥ 5 per replicate with mice from one or two litters ) . A second set of experiments was performed to specifically modify colonization timing of a cocktail of four bacterial strains during gut microbiota assembly relative to that of a complex microbiota . Four strains autochthonous to the mouse gut microbiota were selected: Lactobacillus reuteri Lpuph1 ( Oh et al . , 2010 ) , Lactobacillus johnsonii DPPM ( Perez-Muñoz et al . , 2014 ) , Bacteroides vulgatus RJ2H1 ( a strain isolated from laboratory mice at the University of Nebraska-Lincoln; NCBI accession number PRJNA78795 ) , and Clostridium cocleatum ATCC 29902T ( Kaneuchi et al . , 1979 ) . Growth conditions of the individual strains in the cocktail mix is detailed in the section below . The donor cecal inocula used in these experiments were prepared from ceca harvested from 8-week-old conventionally raised C3H/HeN mice ( 5 mice of the same litter ) housed at the University of Nebraska-Lincoln , as described above . Germ-free C57BL/6 wild-type ( WT ) and Rag1-/- mice were maintained inside independent , flexible-film isolators until inoculation with C3H/HeN ceca donor microbiota at 10 ± 1 days after birth . At that time , litters of pups and their parents were transferred to individually ventilated cages ( IVC ) mounted on racks with positive airflow until the end of the experiments . Inoculation of all mice with specific colonizer strains or cecal inocula was done by orally gavaging the pups 50 μl of cell suspension ( see previous section ) , except for 5-day-old mice for which a drop of the inoculum was gently placed on pups’ snouts which they leaked , thus avoiding over-handling of the pups and potentially hurting them . This technique proved successful for inoculation of the four specific strains in the mix , as seen in the consistent detection of all four strains in fecal samples . Group I received the mix of specific colonizers at day 5 ( while in the isolator and before being moved to the IVC rack and receiving the donor microbiota ) ; Group II at day 15 ( 5 days after receiving the donor microbiota ) ; and Group III at day 36 ( Figure 5A ) . Approximately 108 cells per mouse were administrated for each strain , with the exception of C . cocleatum ( around 105 cells/mouse ) . Fecal samples were collected from each mouse right before the first inoculation ( same day ) , and weekly throughout a 5-week period post-inoculation . For Group I experiments , the first fecal sample was collected from the parents instead of the pups due to their young age ( 5 days ) ; the remaining five fecal collections were obtained from each pup . For Groups II and III , all fecal pellets were collected directly from the pups/adult animals . Mice were weaned at day 21 , at which time parents were removed from the cage , euthanized , and their ceca harvested as previously described . Mice were euthanized at day 78 ( ±1 day ) , and their ceca collected for DNA extraction , quantification of the specific strains , and microbiota analysis . Each experiment was done in duplicate ( N > 5 in each replicate , from one or two litters ) . All bacterial strains were grown under anaerobic conditions at 37°C . L . reuteri and L . johnsoniii were propagated in MRS media ( BD Difco Microbiology , Houston , TX ) supplemented with 10 g/L of maltose and 5 g/L of fructose ( mMRS ) , while C . cocleatum ATCC 29902 was grown in Reinforced Clostridium media ( Oxoid Limited ) , and B . vulgatus on was grown in Tryptone Yeast Glucose broth ( TYG ) for liquid growth ( Perez-Muñoz et al . , 2014 ) and on Brain Heart Infusion Agar with 10% Sheep Blood for plating . To generate the inoculation stocks , strains were grown separately , harvested at late exponential phase , washed with reduced PBS under anaerobic conditions , cells were pelleted through centrifugation ( 3220 x g for 10 min ) , and then re-suspended in pre-reduced PBS with 10% glycerol ( pH 7 ) . Quantitative culture revealed approximately 1010 CFU/ml , with the exception of C . cocleatum , which only contained 107 CFU/ml ) . Stocks were aliquoted for single-use in the experiments ( 300 μl ) , and stored at −80°C until use in mouse experiments . Quantification of the specific colonizers was done by strain-specific qPCR . Strain-specific primers were designed ( Krumbeck , 2015 ) to target genes identified to be unique to L . reuteri Lpuph1 , L . johnsonii DPPM , C . cocleatum ATCC 29902T , and B . vulgatus R2JH1 ( JGI genome ID numbers: 2506381017 , 2606217813 , 2574179769 , and 2510065017 , respectively ) by comparing their genomes against a selection of closely related strains . The genes selected as targets encode for nicotinamide mononucleotide transporter in L . reuteri Lpuph1 ( NCBI Accession ID GCF_000179455 . 1 ) , a subtilase family protein in C . cocleatum ATCC 29902T ( NCBI GCF_900102365 . 1 ) , the ORF6C protein domain in L . johnsonii DPPM , and the TonB-linked outer membrane protein ( SusC/RagA family ) [locus tag RJ2H1_00017340] for B . vulgatus R2JH1 ( according to Joint Genome Institute annotations ) . Primer sequences and PCR fragment lengths are presented in Supplementary file 6 . Primers’ strain specificity was tested in silico by conducting a BLAST search against the NCBI database . Primers were also tested experimentally against DNA isolated from the C3H/HeN donor microbiota by qPCR , and confirmed to produce no background amplification with the donor community . Cell numbers of strains were quantified by absolute quantification using a standard curve prepared with DNA isolated from cultures for which cell numbers were determined by quantitative culture . Standard curves were prepared by using pure cultures of each microorganism harvested at late exponential phase , determined by growth curves generated for each strain under anaerobic conditions ( Maldonado-Gómez et al . , 2016; Krumbeck , 2015 ) . qPCR was performed using a Bio-Rad C1000 Thermal Cycler instrument ( Bio-Rad laboratories , CA ) , with PCR reaction volumes of 25 μl using Brilliant III Ultra-Fast SYBR Green qPCR master mix ( Agilent Technologies , Cedar Creek , TX ) , 0 . 8 μM primer concentrations , and 1 μl DNA template . Annealing temperatures of 60°C were used . Fecal and cecal samples were diluted ( 1:10 ) in sterile PBS ( pH 7 ) . 1 ml and 200 μl aliquots of cecal and fecal dilutions , respectively , were used for DNA extractions . DNA used for qPCR was extracted following a standard phenol-chloroform extraction method ( Martínez et al . , 2009 ) , while DNA used for 16S rRNA gene tag sequencing was extracted with QIAamp DNA Stool Mini kit ( QIAGEN , Hilden , Germany ) with modifications as described previously ( Martínez et al . , 2010 ) . For both methods , enzymatic and mechanical cell lysis steps were included ( Martínez et al . , 2010; Martínez et al . , 2009 ) . Cecal microbiota composition for experiments testing the impact of colonization order using whole cecal microbiota was characterized by sequencing the V5-V6 region of the 16S rRNA ( Krumbeck , 2015 ) , while the V4 region was sequenced for the experiments testing the impact of time of arrival of specific strains ( forward primer: 5’-GTGCCAGCMGCCGCGGTAA-3’ , reverse primer: 5’-GGACTACHVGGGTWTCTAAT-3’ ) . In both cases , pair-end sequencing was done in the MiSeq Illumina platform ( 2 × 300; MiSeq Reagent Kit v3 ) . Each dataset was processed independently but following the same protocol . In brief , pair-end reads ( 2 × 300; MiSeq Reagent Kit v3 ) were merged and quality controlled with the merge-pairs application of the illumina-utils toolset ( Eren et al . , 2013 ) with quality check ( Q30 ) , removal of primers , removal of sequences with at least one mismatch to the primer sequence , and P parameter = 0 . 03 ( for all other parameters default settings were used ) . Quality-controlled sequences were subjected to Minimum Entropy Decomposition ( MED ) analysis ( Eren et al . , 2015 ) , which allows the differentiation of single-nucleotide differences , therefore allowing the highest level of resolution when differentiating members of the two donor communities . Because MED is a non-cluster based algorithm , the term ‘bacterial types’ is used in the manuscript to refer to the phylogenetic bins assigned by MED , instead of operational taxonomic units that are inherently based on a clustering step . Independent MED analyses were done for the two experimental datasets ( given that the ceca microbiota used were different ) . Taxonomic assignment of the representative sequences of each bacterial type was performed using the RDP Classifier and Seqmatch tools , and verified with BLASTn by the NCBI . Quality-controlled sequences were also binned in a reference-based manner using the RDP MultiClassifier tool ( Wang et al . , 2007 ) as an independent method of sequence classification . α and β-diversity distances were calculated after rarefying the number of sequences ( 20 , 000 sequences/sample for the experiments introducing the whole ceca experiments in different sequential order , and 12 , 500 sequences/sample for the experiments testing the importance of time of arrival of specific colonizers ) . α and β-diversity distances were calculated after rarefying the number of sequences using the QIIME pipeline ( version 1 . 9 . 1 ) ( Caporaso et al . , 2010 ) . The datasets generated and analyzed during this study are available under https://figshare . com/s/a1ea177bb39717f13800 and https://figshare . com/s/1cfc381825d63780516e . Because data obtained from α-diversity measurements were normally distributed , significant differences across treatments with Analysis of Variance ( ANOVA ) and Tukey’s post-hoc tests was tested using R ( Core Team , 2014 ) . Overall community structure differences across treatments were computed with Adonis implemented in R’s vegan package ( Oksanen , 2013 ) , which performs a permutational multivariate analysis of variance based on distance matrices . The test was run with 999 permutations using Jaccard and Bray Curtis dissimilarity , and groups in the NMDS plots were displayed using the function ordiellipse . To test the effect of different factors on community structure , the envfit function of the vegan package was used with 999 permutations . Random Forest algorithm was done as previously described ( Maldonado-Gómez et al . , 2016 ) . For compositional analyses , read counts of a taxon per sample were converted to a Log10 scale , after addition of 1 count prior to transformation ( avoiding the non-defined Log10 of 0 count ) . To compare differences in taxonomic bin abundances across treatments , linear mixed models with treatment as a fixed effect , and the isolator in which the mice were housed as a random effect , were used in the lme4 package ( Bates et al . , 2015 ) , and post-hoc tests were performed with the multcomp package ( Hothorn et al . , 2008 ) in R . FDR corrections were performed to address multiple testing ( false discoveries ) . FDR adjusted p<0 . 05 were considered significant . Results are presented as mean ±standard deviation . | The microbial community living in the gastrointestinal tract of humans , also known as the gut microbiome , is essential for health . Disturbances of this community can lead to chronic diseases . Each person has a unique and stable community of gut microbes that is as personal as a ‘fingerprint’ . Studies have shown that an individual’s genetics , diet , environment , lifestyle , and physiological state all make small contributions to the variation of the gut microbiome among individuals . However , less than 30% of this variation can be explained , and even identical twins , who share the same genetics and often diets and lifestyle , have distinct gut microbiomes . This suggests that other unknown factors likely shape these microbial communities too . The microbial communities and the gut make up an ecosystem that is likely subject to many of the same ecological rules that govern ecosystems like rainforests or coral reefs . Yet many studies have overlooked the role of ecology in shaping the gut microbiota . For example , it is well known that the order in which organisms arrive in a community may influence how they interact and assemble into communities . It is possible that the order bacteria are introduced into the gastrointestinal tract of babies early in life may also change the make up of their gut microbiome , and thus introduce the variation that is currently unaccounted for . Now , Martínez et al . show that the first types of bacteria to colonize the gut of mice have a lasting impact on their microbiome . In the experiments , genetically identical mice were housed under exactly the same conditions in airtight plastic bubbles . This allowed the scientists to control when the young mice first encountered specific microbes and microbe communities . Distinct microbial communities collected from different adult mice were introduced into the gastrointestinal tract of the young mice in sequence . Martínez et al . found that the microbes they introduced into the young mice first had the strongest influence on their gut microbiome at the end of the experiments . When the experiments were repeated with a cocktail of four different bacteria the results were similar – the earlier arrivals showed enhanced colonization and had the biggest influence on the microbe community . This suggests that the timing of bacterial arrival in the gut is very important to shape the gut microbiome . Since it is highly random and unpredictable in real-life , and likely to differ even among twins , it could explain why the gut microbiome can be so unique . More studies are needed to understand how antibiotics , formula feeding , or cesarean sections affect gut microbiota early in life , and consequently health . This may help scientists develop better ways to influence the microbiota to improve health , for example , by introducing beneficial microbes early in life . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2018 | Experimental evaluation of the importance of colonization history in early-life gut microbiota assembly |
Meiotic drivers are selfish genes that bias their transmission into gametes , defying Mendelian inheritance . Despite the significant impact of these genomic parasites on evolution and infertility , few meiotic drive loci have been identified or mechanistically characterized . Here , we demonstrate a complex landscape of meiotic drive genes on chromosome 3 of the fission yeasts Schizosaccharomyces kambucha and S . pombe . We identify S . kambucha wtf4 as one of these genes that acts to kill gametes ( known as spores in yeast ) that do not inherit the gene from heterozygotes . wtf4 utilizes dual , overlapping transcripts to encode both a gamete-killing poison and an antidote to the poison . To enact drive , all gametes are poisoned , whereas only those that inherit wtf4 are rescued by the antidote . Our work suggests that the wtf multigene family proliferated due to meiotic drive and highlights the power of selfish genes to shape genomes , even while imposing tremendous costs to fertility .
Infertility can be perplexingly high within eukaryotic species . For example , more than one out of every seven human couples are infertile ( Thoma et al . , 2013 ) . This high infertility is at odds with the fundamental requirement of reproductive success for Darwinian fitness . A potential solution to this infertility paradox is the presence of selfish genes that subvert meiosis to increase their transmission into gametes ( Ségurel et al . , 2011; Presgraves , 2010; Johnson , 2010 ) ; such selfish genes might explain a subset of cases of human infertility . Gamete-killing meiotic drive alleles are one such class of selfish genes that can directly cause infertility . These genes act by killing the gametes that do not inherit them , increasing their transmission into up to 100% of the progeny of a heterozygote ( Lindholm et al . , 2016; Sandler and Novitski , 1957 ) . Meiotic drivers can also indirectly result in infertility or other disease states by interfering with natural selection’s ability to choose the most well-adapted alleles . Natural selection cannot harness the fitness benefits of alleles carried in gametes destroyed by drive . Furthermore , meiotic drivers can promote the spread of maladapted alleles that are genetically linked to the drive locus within a population ( Sandler and Novitski , 1957; Crow , 1991 ) . Because drive can be harmful to the overall fitness of a species , suppressors of drive often evolve in response ( Crow , 1991 ) . Gamete-killing meiotic drive has been observed in eukaryotes ranging from plants to mammals ( Lindholm et al . , 2016 ) . With the broadening implementation of high-throughput sequencing of both meiotic products and cross progeny to measure allele transmission , the presence of meiotic drive is being observed at an accelerated rate , and it is hypothesized that these selfish genes are common ( Lindholm et al . , 2016; Didion et al . , 2015; Ottolini et al . , 2015; Grognet et al . , 2014; Burt and Trivers , 2006 ) . However , only a handful of genes involved in meiotic drive have been identified . Lack of homology among these genes makes it nearly impossible to identify novel drive loci from genome sequences alone . Instead , rigorous genetic analyses are required to detect and map meiotic drive loci . These efforts are frequently impeded by the complexity of many drive systems; they often have multiple components and are found within chromosome rearrangements that are recalcitrant to genetic mapping ( Larracuente and Presgraves , 2012; Bauer et al . , 2012 ) . Even in the case of well-studied meiotic drive systems where one or more components have been identified , a complete understanding of the mechanistic basis of drive and its suppression has been elusive ( Grognet et al . , 2014; Larracuente and Presgraves , 2012; Bauer et al . , 2007 , 2005; Hammond et al . , 2012 ) . The prospect of characterizing meiotic drivers in a genetically tractable system spurred our study of a pair of fission yeasts , Schizosaccharomyces pombe strain 972 ( Sp ) and Schizosaccharomyces kambucha ( Sk ) . Despite being 99 . 5% identical at the nucleotide level , Sp/Sk hybrids are nearly sterile ( Rhind et al . , 2011; Zanders et al . , 2014 ) ; a reproductive barrier between these yeasts must have arisen very recently . This rapid evolution of infertility is common amongst fission yeasts that are generally categorized as isolates of the Schizosaccharomyces pombe species ( Avelar et al . , 2013 ) . In the case of Sp/Sk hybrids ( and likely other pairings ) , the infertility is caused by both chromosomal rearrangements and multiple meiotic drivers ( Zanders et al . , 2014; Avelar et al . , 2013 ) . Indeed , we previously found that genes on each of the three Sk chromosomes are capable of enacting gamete ( spore ) -killing meiotic drive against their Sp homologs ( Figure 1A ) ( Zanders et al . , 2014 ) . However , the genes responsible for the drive phenotypes were unknown . 10 . 7554/eLife . 26033 . 003Figure 1 . A complex meiotic drive landscape on Sk and Sp chromosome 3 is revealed by recombination mapping . ( A ) A cross between Sk and Sp generates a heterozygote that has low fertility and preferentially transmits Sk alleles on all three chromosomes into viable gametes ( Zanders et al . , 2014 ) . ( B ) Generation of chromosome 3 introgression diploids 1–8 . Sk-derived DNA is shown in purple while Sp-derived DNA is shown in green . The origin of the Sp/Sk mosaic chromosome is depicted in Figure 1—figure supplement 1 . ( C ) Phenotypes of rec12∆/rec12∆ introgression/Sk diploids . See Figure 1—source data 1 for breakpoints between Sk-derived DNA ( purple ) and Sp-derived DNA ( green ) . Chromosome transmission was followed using the heterozygous markers at the ade6 locus: hph is short for the hphMX4 marker gene which confers resistance to hygromycin ( HygR ) . The percentage of gametes that inherit both markers ( heterozygous disomes , likely aneuploids and diploids ) and ( after excluding the heterozygous disomes ) the percent of gametes that inherit the marker from the pure Sk chromosome are shown . Over 100 viable gametes were tested for each diploid; raw data can be found in Figure 1—source data 2 . * indicates p-value<0 . 01 ( G-test ) compared to rec12∆/rec12∆ Sk control ( from Zanders et al . ( 2014 ) ) . ( D ) Fine-scale mapping of the drive locus starting with the introgression from diploid 1 . Strains that were recombinant between the ura4 locus and an introduced kanMX4 marker gene were selected and their phenotypes were tested in crosses to Sk . The recombinant strain with the smallest amount of Sp DNA that retained the phenotype ( sensitivity to drive by an Sk chromosome ) is shown in detail . This introgression strain was mated to Sk to generate diploid 9 . These analyses identified a ~30 kb candidate region ( see blow up ) containing a drive locus . In Sp , this region contains wtf4 and the wtf3 pseudogene . The syntenic region in Sk contains only one wtf gene , wtf4 . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 00310 . 7554/eLife . 26033 . 004Figure 1—source data 1 . Breakpoints between Sp and Sk-derived DNA sequences . The introgression strains used in diploids 1–8 were sequenced and genotyped for single-nucleotide polymorphisms ( SNPs ) that reliably distinguish Sk and Sp as in ( Zanders et al . , 2014 ) . The SNPs flanking the recombination event ( left and right boundaries ) that generated each breakpoint between Sp and Sk DNA for each introgression strain are shown . The coordinates refer to the position of the SNP on Sp chromosome 3 . For the introgression used in diploid 9 , SNPs were genotyped via PCR and Sanger sequencing . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 00410 . 7554/eLife . 26033 . 005Figure 1—source data 2 . Raw data underlying Figure 1C . Diploids 1–10 ( column 1 ) were generated by crossing the indicated haploid strains ( columns 2 and 4 ) . The diploid numbers correspond to those in Figure 1 and the text . All strains are rec12∆ and transmission of chromosome 3 was followed using heterozygous markers at the ade6 locus ( columns 3 and 5 ) . hphMX4 confers resistance to hygromycin ( HygR ) . The number of viable progeny inheriting one or both ade6 markers is indicated ( columns 6–8 ) , as are the percentage of the progeny that inherited both markers ( column 10 ) . These strains have two copies of chromosome 3 , so we refer to them as disomes , although other homozygous disomes could be present in the Ade+ HygS and Ade- HygR classes as well . Amongst the progeny that inherit only one ade6 marker , we show the percent that inherit allele 2 ( column 12 ) , which is the allele from the pure Sk chromosome . For the statistical analyses ( G-tests ) , we compared the observed heterozygous disomy and allele 2 transmission to the values observed in diploid 10 , which is a pure Sk rec12∆ control ( columns 11 and 13 ) . The last column indicates the number of independent diploids that were generated and assayed of each genotype . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 00510 . 7554/eLife . 26033 . 006Figure 1—figure supplement 1 . Generation of mosaic chromosome 3 used in Figure 1B . The goal of these crosses was to generate a strain containing mostly Sp-derived DNA on chromosome 3 in an otherwise Sk background . This effort was complicated by the different karyotypes of Sp and Sk chromosomes 2 and 3 ( Zanders et al . , 2014 ) . We used rec12∆ strains to limit recombination , but rare recombinants ( e . g . SZY239 and SZY247 ) can still be obtained via selection . Markers derived from the Sk parent are shown in purple , while Sp-derived markers are green . We first isolated hybrids in which Sk and Sp markers on chromosomes 2 and 3 were uncoupled , suggesting rare recombination events had occurred between Sk and Sp chromosomes 2 and 3 . Such events have the potential to generate chromosome 3 variants with mostly Sp DNA , but with an Sk karyotype , as occurred in SZY247 . We then performed the illustrated crosses to move that chromosome into a different strain background with pure Sk chromosomes 1 and 2 . We finally sequenced SZY558 and verified the strain has Sk chromosomes 1 and 2 and Sp DNA on chromosome 3 until between SNPs at positions 1 , 804 , 477 and 1 , 810 , 659 . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 006 Here , we use genetic mapping to identify Sk wtf4 as an autonomous gamete-killing meiotic drive gene . We show that Sk wtf4 generates two transcripts from alternative start sites: a long transcript encoding an antidote and a short transcript encoding a gamete-killing poison . Whereas the poison protein is found in all the gametes , the antidote protein is enriched only in the gametes encoding Sk wtf4 , thereby ensuring that gametes that do not inherit the selfish allele are destroyed . This gene is a member of the large , rapidly evolving wtf gene family that has 25 members in Sp . We show that wtf4 is not the only driver amongst wtfs and propose a model in which meiotic drive is the ancestral function of the gene family . Our study thus identifies a novel mechanism by which meiotic drivers can act and highlights the significant role these selfish elements have played in shaping the evolution of a model eukaryote .
To study meiotic drive in fission yeast , we mate haploids to generate diploids , induce the diploids to undergo meiosis and monitor allele transmission into the gametes using genetic markers . In Sk/Sp hybrid diploids , drive of loci on all three Sk chromosomes is due to the preferential death of gametes inheriting the corresponding Sp alleles ( Zanders et al . , 2014 ) ( Figure 1A ) . In this work , we focused on chromosome 3 because it is the smallest chromosome and the drive phenotype is strong: greater than 80% of viable haploid gametes inherit an Sk marker allele from Sk/Sp hybrids ( Zanders et al . , 2014 ) . To genetically map a drive locus on chromosome 3 , we first wanted to generate a strain with Sk chromosomes 1 and 2 , but Sp chromosome 3 . Because Sp and Sk have different karyotypes on chromosomes 2 and 3 due to a translocation ( Zanders et al . , 2014 ) , we could not generate such a strain as it would lack essential genes . Instead , we generated a haploid strain with an Sk karyotype containing Sk chromosomes 1 and 2 and most , but not all , of chromosome 3 derived from Sp ( Figure 1—figure supplement 1 and Materials and methods ) . We then backcrossed this haploid strain to Sk to generate a series of haploid strains that have mosaic ( Sp and Sk-derived DNA sequences ) versions of chromosome 3 generated by recombination ( Figure 1B ) . We then crossed the recombinant haploids to Sk to generate a series of ‘introgression diploids’ ( Figure 1B and Figure 1C , diploids 1–8 ) . The introgression diploids were all homozygous null mutants for rec12 , the fission yeast ortholog of Sachharomyces cerevisiae SPO11 , which is required for inducing DNA breaks to initiate meiotic recombination ( Phadnis et al . , 2011 ) . As meiotic recombination is not induced in the introgression diploids , we could use any genetic marker on chromosome 3 to assay this chromosome for the presence of drive loci . We used the codominant markers ade6+ and ade6∆::hphMX4 to follow transmission of each chromosome into viable gametes ( Figure 1C ) . We observed three phenotypic classes amongst our introgression diploids ( diploids 1–8 , Figure 1C ) . In the first class ( diploids 1–3 ) , the allele from the pure Sk chromosome exhibited drive over the allele from the Sp/Sk mosaic chromosome . In the second class ( diploids 4–7 ) , we were surprised to observe the opposite phenotype: the allele from the Sp/Sk mosaic chromosome exhibited drive over that from the pure Sk chromosome . In the third class ( diploid 8 ) , we observed unbiased allele transmission . Our finding of three distinct phenotypic classes amongst our introgression diploids ( diploids 1–8 ) is inconsistent with the simple model of a single drive locus on Sk chromosome 3 . A single gene model predicts two phenotypic classes: ( 1 ) introgression diploids in which the pure Sk chromosome exhibits drive because the Sk/Sp mosaic chromosome lacks the Sk drive allele and ( 2 ) introgression diploids in which the chromosomes show Mendelian transmission because the Sk/Sp mosaic contains the Sk drive allele . Instead , our data are more consistent with the presence of a meiotic drive allele ( or alleles ) found on both Sk and Sp chromosome haplotypes and the existence of at least one genetically separable drive suppressor . The drive of the Sk/Sp mosaic chromosome over the pure Sk chromosome in class 2 ( diploids 4–7 ) is consistent with the presence of an Sp drive allele in these strains . The full effects of this Sp drive locus could have been missed previously in Sk/Sp hybrid crosses due to the actions of an Sp drive suppressor not found in the class 2 introgressions ( Zanders et al . , 2014 ) . Similar to what we previously observed in crosses between pure Sk/Sp hybrids ( both rec12+ and rec12∆ ) , we found that viable gametes produced by diploids of all three classes frequently inherited both alleles at the ade6 locus ( Figure 1C ) ( Zanders et al . , 2014 ) . This indicates that they are not haploid at this locus , as is expected for gametes . These gametes likely represent a mix of heterozygous diploids and heterozygous chromosome 3 aneuploids . In diploid 8 , the phenotype was extreme , with almost all the viable gametes inheriting both ade6 alleles ( Figure 1C ) . Although the frequency of meiotic chromosome missegregation is elevated in rec12∆ mutants ( Phadnis et al . , 2011 ) , we see significantly higher levels of viable gametes that inherit both alleles in diploids 1–8 than we did in a homozygous Sk rec12∆ control ( Figure 1C , diploid 10 ) . The high level of chromosome 3 aneuploidy and/or diploidy we observe in the viable progeny of Sk/Sp hybrid crosses and our introgression diploids ( 1-8 ) is also consistent with the existence of both Sk and Sp active meiotic drive loci . We previously showed in Sk/Sp hybrids that this phenotype was not due to elevated chromosome missegregation in meiosis , but rather preferential death of haploid gametes ( Zanders et al . , 2014 ) . As we proposed previously , this phenotype could result from distinct competing Sk and Sp driver loci on chromosome 3 ( Zanders et al . , 2014; Bomblies , 2014 ) . In the absence of recombination , a given haploid gamete can inherit only the Sk or Sp drive locus and is thus sensitive to being killed by the one it does not inherit . Heterozygous diploids and heterozygous aneuploids , however , would inherit both loci and be resistant to both killers . To map driver location ( s ) from the phenotypic data described above , we sequenced the haploid strains that contributed the Sk/Sp mosaic chromosomes to the introgression diploids ( diploids 1–8 ) and combined genotype information with the phenotypic data described above . We determined which regions of chromosome 3 were derived from Sk and which were from Sp in each strain ( Figure 1C and Figure 1—source data 1 ) . It was clear from our data that one or two loci are not sufficient to explain the phenotypes of all these strains . We chose to focus on the Sk/Sp mosaic chromosome found in diploid 1 . This strain has the smallest amount of Sp DNA ( ~180 kb ) , and drive of Sk in the introgression/Sk diploid suggested the strain lacks a drive allele found in Sk ( Figure 1C ) . We crossed a haploid isolate containing this chromosome to a rec12+ Sk strain to generate recombinant progeny containing smaller segments of Sp-derived DNA ( Figure 1D and Materials and methods ) . We SNP-genotyped those recombinants and tested their phenotypes by mating them to Sk to generate additional introgression diploids ( see Materials and methods ) . We selected diploid 9 for further analysis , as it contains the Sk/Sp mosaic chromosome with the smallest region of Sp-derived DNA ( ~30 kb ) that a pure Sk chromosome can drive against ( Figure 1C and Figure 1D ) . After excluding aneuploid/diploid progeny ( those that inherit both ade6 markers ) , the allele from the pure Sk chromosome shows essentially the same transmission bias in diploids 1 and 9 . These results suggest the Sk drive allele active in diploid 1 is found in this ~30 kb region . Curiously , this locus is in a region that is transmitted in a Mendelian manner ( to ~50% of progeny ) in pure Sk/Sp hybrids ( Zanders et al . , 2014 ) , suggesting that other loci can mask the effects of the driver within this ~30 kb region . In addition , it is unclear why the fraction of viable progeny that inherit both ade6 alleles drops between diploids 1 and 9 . These puzzles likely reflect the complexity of the multiple drivers and suppressor loci acting in these yeasts ( Zanders et al . , 2014 ) . We next wanted to verify the candidate drive locus using a recombination-competent ( rec12+ ) diploid . We generated introgression diploid 11 which contains the same Sk/Sp mosaic chromosome as diploid 1 , but is rec12+ . To follow the transmission of the candidate locus , we needed a closely linked marker gene , so we engineered heterozygous markers at the linked ura4 locus ( Supplementary file 1 ) . We found that the ura4 allele from the pure Sk chromosome is transmitted to 87% of the viable gametes produced by diploid 11 , which is not significantly different from the 88% transmission of the Sk allele in diploid 1 ( Figure 1C and Figure 2A ) . This result shows that ura4 is closely linked to an Sk drive locus and is consistent with that locus being within the ~30 kb candidate region . 10 . 7554/eLife . 26033 . 007Figure 2 . Sk wtf4 is a self-sufficient meiotic driver that kills gametes that do not inherit the gene . ( A ) Allele transmission and propidium iodide ( PI ) staining phenotypes of diploids 11–19 . Sk-derived DNA is purple , Sp-derived DNA is green . The cartoons depict chromosome 3 . Chromosomes 1 and 2 are derived from Sk in diploids 11–15 , but are from Sp in diploids 16–19 . For diploids 11–15 , allele transmission was monitored by following heterozygous markers at the ura4 locus , which is tightly linked to wtf4 ( estimated 7–17 cM based on physical distance [Young et al . , 2002] ) . PI dye is excluded from living spores , but not dead spores that have lost membrane integrity , such as those destroyed by drive . The percent of spores that exclude PI is shown as a proxy of fertility ( Figure 2—source data 1 ) . The PI phenotypes and ura4 locus allele transmission for diploids 11 , 12 , 14 and 15 were compared to those of the wild-type Sk control ( diploid 13 ) . * indicates p-value<0 . 01 ( G-test ) . For diploids 16–19 , allele transmission was followed using markers at the ade6 locus , which is where the empty vector or wtf gene constructs are integrated . The integrations introduced a dominant drug resistance gene and mutated ade6+ . Because these diploids all had codominant alleles at ade6 , we could detect progeny that inherited both ade6 alleles ( less than 10% of the total population ) . These progeny are excluded from the data presented above , but all the raw data are presented in Supplementary file 1 . The PI phenotypes and allele transmission for diploids 17–19 were compared to the empty vector control ( diploid 16 ) and * indicates p-value<0 . 01 ( G-test ) . See Supplementary file 1 for the markers used for each diploid and the raw data for allele transmission and Supplementary file 2 for the PI staining raw data . Over 200 viable gametes were scored for allele transmission and over 200 spores ( >50 4-spore asci ) were assayed for PI staining . ( B ) Images of PI staining and transmitted light ( TL ) in an ascus with no drive containing all alive spores ( top ) and in an ascus with drive where two of the four spores are dead ( bottom ) . Scale bar represents three microns . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 00710 . 7554/eLife . 26033 . 008Figure 2—source data 1 . PI staining correlates with viable spore yield as a measure of fertility in wild-type and wtf heterozygous crosses . The fertility of the indicated diploids was assayed using both the established viable spore yield assay and by PI staining . We avoided tetrad dissection because we found that it was complicated by disintegration of spores destroyed by drive . The viable spore yield assay is a plating assay that measures the number of viable spores produced per viable diploid induced to undergo meiosis ( Smith , 2009 ) . PI is excluded from viable spores , but enters those destroyed by wtf drivers ( Figure 2B ) . Although PI staining likely will not detect spore death by other causes that do not disrupt membrane integrity , the percent of PI-excluding cells correlates with viable spore yield in wild-type and wtf heterozygous crosses . Diploids of four genotypes are shown . Three of the diploids are used elsewhere in the paper ( diploids 17 , 22 , and 27 ) and the diploid number ( column 1 ) corresponds to the diploid numbers used in the main text and figures . The two strains that were mated to generate the diploids are shown in columns 2 and 4 . The drive-relevant genotypes of these parental strains is shown in columns 3 and 5 . The last two columns show the PI-staining and viable spore yield phenotypes of each diploid . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 008 To test whether the transmission bias we observed in diploid 11 might be caused by increased cell death amongst gametes inheriting the Sp locus , we used propidium iodide ( PI ) to stain the meiotic sacs ( asci ) that hold the spores . PI efficiently stains dead cells that have lost their membrane integrity but fails to stain viable cells ( Figure 2B and Figure 2—source data 1 ) ( Moore et al . , 1998 ) . We found that only 81% of spores generated by diploid 11 excluded PI , while wild-type strains ( e . g . diploid 13 ) have rates >90% ( Figure 2A ) . Together , our findings support the hypothesis that the Sk ~30 kb region encodes a gamete-killing meiotic driver . Near the center of the Sk 30 kb candidate region is wtf4 ( Figure 1D ) , a member of the mostly uncharacterized wtf gene family . This family contains 25 members in Sp , and its ( cheeky ) name is derived from the genes’ genomic association with Tf transposons ( Bowen et al . , 2003 ) . wtf genes are not found outside Schizosaccharomyces species ( Bowen et al . , 2003 ) . Sk wtf4 is a 1427 bp gene ( from the start to stop codon , including introns ) with six exons and encodes a protein with six predicted transmembrane domains . Sk wtf4 shares only 89% DNA sequence identity ( 82% amino acid identity ) with the gene in the orthologous locus in Sp ( Sp wtf4 ) ; this divergence is much higher than expected given the 99 . 5% average DNA sequence identity between the two genomes ( Rhind et al . , 2011; Zanders et al . , 2014 ) . We reasoned that wtf genes , in general , were good candidates for meiotic drive loci because of their rapid evolution and their transcription during meiosis ( Bowen et al . , 2003; Mata et al . , 2002; Daugherty and Malik , 2012; McLaughlin and Malik , 2017 ) . To test if Sk wtf4 is a meiotic drive gene , we deleted Sk wtf4 ( Sk wtf4Δ ) in a pure Sk background and mated that haploid to one containing the same Sk/Sp mosaic found in diploid 11 ( Figure 2A ) to produce diploid 12 . We observed a significant increase in the number of spores that could exclude PI in diploid 12 ( Sk wtf4Δ ) , compared to diploid 11 ( Sk wtf4+ ) from 81% to 96% , suggesting Sk wtf4+ promotes spore death in progeny of heterozygous diploids . In addition , Sk wtf4Δ showed more equitable allele transmission . While Sk wtf4+ is transmitted to 87% of the viable gametes produced by diploid 11 , the transmission rate of Sk wtf4Δ is reduced to 66% in diploid 12 ( Figure 2A ) . Although some residual transmission bias remains in this background , our results clearly implicate Sk wtf4 as a large contributor to gamete-killing meiotic drive . There are two known means by which gamete-killers act to eliminate competing alleles ( Lindholm et al . , 2016; McLaughlin and Malik , 2017 ) . Under one model , meiotic drivers kill gametes containing a particular target locus . For example , the Segregation Distorter ( SD ) system in Drosophila melanogaster kills sperm bearing an expansion of the Responder satellite DNA ( Larracuente and Presgraves , 2012; Wu et al . , 1988 ) . The second model is a poison-antidote model in which a gamete-killing entity ( the poison ) is encoded at a position that is closely linked to that encoding a second substance ( the antidote ) which specifically protects gametes that inherit the drive locus . For example , the unidentified rfk gene ( required for killing ) acts as a poison and the rsk gene ( resistance to spore killing ) acts as an antidote in the Spore killer-2 drive locus from Neurospora intermedia ( Hammond et al . , 2012; Harvey et al . , 2014 ) . We first tested if Sk wtf4 acts analogously to SD to kill gametes that inherit a particular Sp chromosomal locus . To test this idea , we analyzed the effect of Sk wtf4Δ/Sk wtf4+ heterozygosity in a pure Sk strain background ( diploid 14 , Figure 2A ) . As this Sk wtf4Δ/Sk wtf4+ heterozygote contains no Sp DNA , there should be no drive if wtf4 can only target and drive against Sp sequence . We did , however , observe strong drive ( 93% transmission ) of Sk wtf4+ relative to Sk wtf4Δ in diploid 14 and a concomitant decrease in the percent of spores that could exclude PI ( 59% versus 92% in wild-type; Figure 2A , diploids 14 and 13 ) . These results demonstrate that the drive of Sk wtf4 does not require an Sp target sequence . Our results are , however , consistent with a poison-antidote model of meiotic drive . The phenotype of the Sk wtf4Δ/Sk wtf4+ heterozygote ( Figure 2A , diploid 14 ) suggests that Sk wtf4 acts as the antidote because gametes lacking the gene die . If this were true and a separate gene acted as the poison , we predicted that Sk wtf4Δ homozygotes ( diploid 15 ) should have very low fertility because they would generate a poison , but no antidote . Contrary to this expectation , we found that an Sk wtf4Δ homozygote is healthy , with the same ability to exclude PI from the spores as wild-type Sk ( 92% of spores; Figure 2A , diploids 13 and 15 ) . This finding rules out the possibility that Sk wtf4 encodes a gene important for meiosis or spore development . Instead , our results suggest that Sk wtf4 acts as both poison and antidote , similar to the Spok genes of Podospora anserina ( Grognet et al . , 2014 ) . It remains unclear , however , why the phenotype of Sk wtf4 is slightly weaker in the hybrid background ( assayed in diploids 11 and 12 ) compared to the phenotypes in pure Sk ( diploids 13–15 ) or pure Sp ( diploids 16–18 ) . We speculate it could be due to the composition ( chromatin state or a sequence variant ) of the mosaic chromosome ( allele 2 in diploids 11 and 12 ) . To further test the idea that Sk wtf4 encodes an autonomous poison-antidote drive locus , we moved the gene to a naive genome and tested if it could induce drive . We integrated Sk wtf4 into the Sp genome at the ade6 locus , which is unlinked to the endogenous wtf4 locus . An Sp diploid that is hemizygous for Sk wtf4 ( Sk wtf4+/ade6+ ) produced fewer viable spores ( 54% PI-excluding spores , versus 96% in the vector-only control ) and showed a marked transmission bias ( 96% ) favoring Sk wtf4+ ( Figure 2A , diploids 16 and 17 ) . In contrast , Sp diploids homozygous for Sk wtf4+ produced viable spores that excluded PI at the same frequency as spores from wild-type diploids and showed unbiased allele transmission ( Figure 2A , diploids 18 and 16 ) . These results are consistent with Sk wtf4 acting as a complete one-gene poison-antidote drive system that causes the death of gametes that fail to inherit the locus from heterozygotes . We hypothesized that Sk wtf4 encodes two products to achieve drive ( Figure 3A ) . The first of these is a gamete-killing poison , which acts indiscriminately on all spores . The second product is an antidote that specifically rescues only the gametes encoding Sk wtf4 from the poison . To investigate how Sk wtf4 could make two products , we analyzed long-read sequence data from Sp meiotic mRNAs ( Kuang et al . , 2017 ) ( Materials and methods ) . This revealed that Sp wtf4 is transcribed during meiosis and generates two major overlapping transcripts with different start sites ( Figure 3—figure supplement 1 ) . Since the region starting 500 bp upstream of the annotated Sp wtf4 start codon until the putative second start codon is fairly well conserved ( 98% identical ) between Sp and Sk wtf4 , we hypothesized that Sk wtf4 is likely to produce similar alternate isoforms to Sp wtf4 . These alternative transcripts of Sk wtf4 could encode the two meiotic drive components – a poison and an antidote ( Figure 3B ) . 10 . 7554/eLife . 26033 . 009Figure 3 . Sk wtf4 has the capacity to make two proteins and Wtf4-GFP shows a dual localization pattern . ( A ) Model for meiotic drive of Sk wtf4 via a poison-antidote mechanism . ( B ) wtf4 creates a long and an alternative short transcript . See Figure 3—figure supplement 1 for a depiction of the long-read RNA sequencing data on which this model is based ( Kuang et al . , 2017 ) . ( C ) Sk Wtf4-GFP localization in diploids where drive does [right] or does not occur [left] . Cells were imaged prior to the first meiotic division [top] and as mature asci [bottom] . ( D ) Asci generated by diploids of the same genotypes as in ( C ) stained with PI to label dead cells ( those lacking wtf4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 00910 . 7554/eLife . 26033 . 010Figure 3—figure supplement 1 . Sp wtf4 has alternate transcriptional start sites . Our annotation of the wtf4 gene with alternate start sites predicted is shown at the top in the same format as Figures 3–5 . The PomBase annotation for Sp wtf4 is shown below that in blue . The transcript locations from one replicate of the meiotic transcript time courses sequenced by Kuang et al . ( 2017 ) are shown below in red and orange . The IsoSeq consensus reads shown should represent full-length transcripts , and each represents a number of individual sequencing reads . Only transcripts represented by 11 or more reads are displayed . Many of the transcripts vary by only a few nucleotides at the 5’ or 3’ ends and appear identical in the image . The time the samples were taken after meiotic induction are shown on the left . No transcripts with 11 or more reads were observed at earlier time points . Introns are represented by thin lines with blue arrows and the coding sequences are represented by the thick boxes . There are two major transcriptional start sites and the splice sites of intron 5 are different from those in the PomBase annotation . We did not verify two possible additional transcript types observed only at 10 hr , or explore their possible functional relevance . The data were visualized using IGV ( Thorvaldsdottir et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 010 To test the feasibility of our model , we investigated the localization of Sk Wtf4-GFP in Sp diploids induced to undergo meiosis ( Sheff and Thorn , 2004 ) . We C-terminally tagged the gene to visualize proteins generated by both the putative Sk wtf4 isoforms; this tag does not interfere with Sk wtf4’s ability to function as a drive allele ( see data for ‘GFP diploid’ in Supplementary files 1 and 2 ) . Visualizing Sk wtf4-GFP/ade6+ heterozygous diploids , we observed faint cytoplasmic Wtf4-GFP signal before the first meiotic division , which intensified throughout gamete development and filled the ascus surrounding the mature gametes ( Figure 3C ) . In mature asci , we observed a strong enrichment of Wtf4-GFP within only two of the four spores . We observed the same spore enrichment pattern in Sk wtf4-GFP/Sk wtf4+ diploids in which drive does not occur ( Figure 3C ) . We hypothesized that the diffuse Wtf4-GFP localization in the ascus corresponded to the poison , whereas the enrichment within the mature spores might reflect the localization of the antidote . If this hypothesis is correct , Wtf4-GFP should be enriched in the two spores that inherit the chromosome carrying Sk wtf4-GFP . Consistent with this idea , we stained asci from Sk wtf4-GFP/ade6+ diploids with PI and observed that the surviving PI-negative spores ( 95% of which inherit Sk wtf4-GFP ) are indeed those with the strong Wtf4-GFP signal ( Figure 3D; Supplementary file 1 ) . The localization pattern of Wtf4-GFP is consistent with our model of Sk wtf4 encoding two protein isoforms ( Figure 3A ) . To further test our poison-antidote model , we sought to generate alleles that could produce only the poison or only the antidote . We first mutated the start codon ( ATG to TAC ) that is present only in the putative short transcript . Our results ( below ) suggest that this mutant allele retains the antidote function but no longer functions as a poison: we therefore call this allele Sk wtf4antidote ( Figure 4A ) . In hemizygous diploids ( Sk wtf4antidote/ade6+ ) , Sk wtf4antidote does not cause spore death ( increased frequency of PI-stained spores ) or the transmission bias that is observed with the wild-type Sk wtf4 allele , suggesting that the mutant can no longer drive ( compare Figure 4B , diploid 20 to Figure 2A , diploid 17 ) . However , this allele still protects from meiotic drive since Sk wtf4+/Sk wtf4antidote heterozygotes produce PI-excluding spores at the same frequency as wild-type and show unbiased allele transmission ( Figure 4B , diploid 21 ) . These data assign an antidote function to the long transcript . 10 . 7554/eLife . 26033 . 011Figure 4 . Sk wtf4 creates two proteins using alternate transcripts: an antidote and a gamete-killing poison . ( A ) Separation of function wtf4 alleles . The red stars indicate start codon mutations . ( B ) Allele transmission and PI staining phenotypes of Sp diploids with the indicated Sk wtf4 alleles integrated at ade6 on chromosome 3 , as in diploids 16–19 in Figure 2A . Spores that inherited both alleles at ade6 are eliminated from the data presented above , but the complete data are found in Supplementary file 1 . * indicates p-value<0 . 01 ( G-test ) compared to empty vector ( or wild-type control ) for allele transmission and fertility as assayed by PI staining . See Supplementary file 1 for raw data and the markers used to monitor allele transmission for each diploid and Supplementary file 2 for the PI staining raw data . Over 200 viable gametes were scored for allele transmission for all diploids except diploid 24 , from which we genotyped 50 . Over 200 spores ( >50 4-spore asci ) were assayed for PI staining of each diploid . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 011 We next set out to generate a Sk wtf4poison allele by mutating the two putative start codons ( ATG to TAG ) found in exon 1 of the long transcript ( Figure 4A ) . This mutant should be able to generate only the short polypeptide . If this allele retains the ability to poison spores but has lost the antidote function , we would expect all progeny to be killed in Sk wtf4poison/ade6+ hemizygotes . Indeed , most spores generated by these diploids die ( 14% exclude PI stain , Figure 4B , diploid 22 ) . Interestingly , the Sk wtf4poison allele was modestly underrepresented ( 38% transmission ) in the few surviving spores generated by diploid 22 , indicating that the spores that inherit that allele are especially likely to be destroyed by their own poison ( Figure 4B ) . To confirm that the toxicity of the Sk wtf4poison allele was due to its lacking the Sk wtf4 antidote , we generated Sk wtf4poison/Sk wtf4+ heterozygotes . As expected , the spores that inherited the complete Sk wtf4+ gene from these diploids were immune to Sk wtf4poison toxicity , while those that inherit Sk wtf4poison die . ( Figure 4B , diploid 23 ) . These results support our model that the short Sk wtf4 transcript encodes a trans-acting gamete poison . As a final test of our model , we brought the separated poison and antidote mutant alleles back together in one diploid , but on opposite haplotypes . If they function as expected , we would predict that the Sk wtf4poison spores will die but the spores that inherit the Sk wtf4antidote will survive . This was indeed the case . Only 45% of the spores produced by Sk wtf4antidote/Sk wtf4poison heterozygotes can exclude PI stain and 88% of the surviving gametes inherit the Sk wtf4antidote allele ( Figure 4B , diploid 24 ) . We next specifically determined the localization patterns of the antidote and poison polypeptides . To visualize the antidote peptide , we generated an Sk mCherryantidote-wtf4 allele ( Figure 5A ) and found it acts similarly to the wild-type wtf4 allele ( Figure 5B , diploids 25 and 26 ) ( Hailey et al . , 2002 ) . We could not reliably use PI staining to assay fertility of mCherry-tagged strains because both signals are red , so we used viable spore yield assays ( VSY ) ( Smith , 2009 ) to confirm that the fertility of the Sk mCherryantidote-wtf4 allele was similar to untagged wtf4 in heterozygotes . Sk mCherryantidote-wtf4/ade6+ hemizygotes had a VSY of 0 . 8 ± 0 . 2 ( standard deviation ) compared to 1 . 0 ± 0 . 4 of Sk wtf4+/ade6+ , and Sk mCherryantidote-wtf4/wtf4+ diploids had a VSY of 1 . 4 ± 0 . 1 compared to 1 . 7 ± 0 . 1 of wild-type . 10 . 7554/eLife . 26033 . 012Figure 5 . Wtf4 antidote is spore-specific and Wtf4 poison spreads throughout the ascus . ( A ) Constructs tagging either the Wtf4 antidote ( top ) or poison ( bottom ) proteins . The red stars indicate start codon mutations . ( B ) Allele transmission and PI staining phenotypes for tagged alleles , as in Figure 4B . See Supplementary file 1 for raw data and the markers used to monitor allele transmission for each diploid and Supplementary file 2 for the PI staining raw data . We could not reliably use PI to assay fertility of mCherry-tagged strains because of color similarity , but in viable spore yield assays the mCherryantidote-wtf4 allele gave a similar phenotype to wtf4 . * indicates p-value<0 . 01 ( G-test ) compared to empty vector ( or wild-type control ) . Over 200 viable gametes were scored for allele transmission and over 200 spores ( >50 4-spore asci ) were assayed for PI staining . ( C ) Wtf4 poison ( cyan ) and antidote ( magenta ) protein localization prior to the first meiotic division ( left ) and in a mature ascus ( right ) . Scale bar represents three microns . TL , transmitted light . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 01210 . 7554/eLife . 26033 . 013Figure 5—figure supplement 1 . Spectral unmixing verifies true signal . Wtf4 poison ( cyan ) and antidote ( magenta ) protein localization in a mature ascus processed using linear unmixing [top] and unprocessed [bottom] . Scale bar represents three microns . DOI: http://dx . doi . org/10 . 7554/eLife . 26033 . 013 To observe the localization of the poison peptide , we generated a Sk wtf4poison-GFP allele ( Figure 5A ) ( Sheff and Thorn , 2004 ) . While this Sk wtf4poison-GFP allele is not as penetrant as the untagged Sk wtf4poison allele , it does have a poison-only phenotype ( Figure 5B , diploids 27 and 28 ) . In Sk mCherryantidote-wtf4/ Sk wtf4poison-GFP heterozygotes , we observed Sk Wtf4poison-GFP expression before the meiotic divisions and later filling mature asci . In contrast , we observe Sk mCherryantidote-Wtf4 enriched only in two of the four mature spores ( Figure 5C ) . Together , these data reconstitute the dual localization patterns we observed with Sk Wtf4-GFP and support our model of a poison-antidote system encoded by the same gene ( Figure 3A ) . We hypothesized that if Sk wtf4 is not unusual amongst the wtf genes in its ability to drive; meiotic drive could explain the ‘driving’ force behind the expansion of the wtf gene family ( Bowen et al . , 2003 ) . The large number of wtfs could also explain the complex drive landscape revealed in our recombination mapping ( Figure 1 ) . To test these ideas , we analyzed additional wtf genes from Sk . We cloned and tested six Sk wtf genes ( wtf2 , wtf5 , wtf6 , wtf28 , and wtf21 plus wtf26 [together] ) for evidence of meiotic drive . As for our tests of Sk wtf4 , we integrated the above Sk wtf genes at the ade6 locus of Sp , which disrupted the ade6+ gene . We then mated those haploids to ade6+ to generate heterozygous diploids and monitored the transmission of the Sk wtf gene ( s ) into viable progeny using the heterozygous ade6 markers . Five of the six genes had no observable drive phenotype . Sk wtf2 was transmitted to 47% ( n = 114 ) of progeny , Sk wtf5 was transmitted to 44% ( n = 454 ) , Sk wtf6 was transmitted to 51% ( n = 471 ) and the combination of Sk wtf21 and wtf26 ( cloned and integrated together ) was transmitted to 46% ( n = 111 ) . However , like Sk wtf4 , Sk wtf28 caused strong drive ( 90% transmission bias and only 57% of spores excluded PI; Figure 2A , diploid 19 ) . We also compared the sequence of each of these Sk wtf genes to the Sp wtf genes at the syntenic loci . wtf26 and wtf28 are not found in Sp , so have either been lost in Sp , or gained in Sk since divergence . While Sk wtf2 is a 1036 bp full length gene , Sp wtf2 is likely a 388 bp pseudogene ( it has a large deletion relative to other wtf genes and multiple in-frame stop codons ) . Sk wtf21 is likely a pseudogene ( multiple in frame stop codons ) , whereas Sp wtf21 is intact . The two loci share 83% DNA sequence identity . The wtf5 gene is intact in both species , and the loci share 99% DNA sequence identity and 97% amino acid identity . Sp and Sk wtf6 share 82% nucleotide identity , but only 74% amino acid identity . Altogether , the wtf loci show much greater sequence divergence than the 99 . 5% genome average identity between Sp and Sk . Such rapid evolution is a hallmark of genes involved in genetic conflicts , such as loci involved in causing or suppressing meiotic drive ( Daugherty and Malik , 2012; McLaughlin and Malik , 2017; Henikoff et al . , 2001 ) . Intriguingly , the Sk wtf28 drive gene is also the only one of the six genes we tested that also has a putative alternate start codon in exon two that could be used to make a putative short poison isoform . Additionally , Hu et al . , 2017 also identified two different wtf drivers in another Sp isolate ( CBS5557 ) and both have a potential alternate start codon in exon 2 . Of the 25 wtf loci in Sp , four ( wtf4 , wtf13 , wtf19 and wtf23 ) also appear to be capable of encoding two proteins and we predict these are active drive genes . In contrast , the intact genes we tested that did not confer drive , Sk wtf2 , wtf5 , wtf6 , and wtf26 , all encode genes similar to the antidote isoform of Sk wtf4 but appear to lack a shorter poison isoform . Together , our results and those of Hu et al . , 2017 ) are consistent with the hypothesis that the ancestral function of the wtf family is to confer meiotic drive .
Our study demonstrates that Sk wtf4 is a novel , gamete-killing meiotic drive locus . Like the Spok family of drivers found in Podospora , wtf4 is an autonomous drive system that confers both the ability to kill gametes that do not inherit the gene and the ability to protect the gametes that do ( Grognet et al . , 2014 ) . We show that wtf4 achieves these disparate functions by a previously undescribed mechanism in which the gene encodes a poison protein from one transcriptional start site and an antidote protein from an alternative transcriptional start site . We show that the poison protein is trans-acting and has the capacity to destroy all gametes , but that the antidote remains in the gametes that inherit the wtf4 locus and specifically rescues them from destruction . The poison-antidote mechanism of Sk wtf4 is comparable to the bacterial toxin-antitoxin ( TA ) systems . These systems are found in most prokaryotes and have been extensively studied . TA systems consist of a toxin that will prevent cell growth or viability and an antitoxin that neutralizes the toxin using a wide variety of mechanisms , typically being classified into six different types ( Lee and Lee , 2016 ) . Interestingly , some toxins are stable , transmembrane proteins that act by disrupting membrane integrity and are counteracted by either an unstable small RNA ( Lee and Lee , 2016; Unterholzner et al . , 2013 ) or a protein that degrades the toxin mRNA ( Wang et al . , 2012 ) . In our poison-antidote meiotic drive system , Sk wtf4 creates two putative trans-membrane proteins: a trans-acting poison and spore-specific antidote . While the exact mechanism of toxicity of Wtf4poison is unknown , we hypothesize that it could be disrupting membrane integrity in a similar manner to the membrane-lytic toxins of some TA systems ( Lee and Lee , 2016; Unterholzner et al . , 2013 ) . In contrast , we speculate that Wtf4antidote protects the spores that inherited Sk wtf4 by sequestering the poison for degradation . The spore specificity of Wtf4antidote could be due to late translation or a spore retention signal within exon 1 , because that is the only region that Wtf4poison is lacking . In addition , work by Hu et al . suggests that the C-termini of Wtf proteins may be more important for the poison than for the antidote functions , despite both proteins being generated by a single given wtf gene sharing a common C-terminus ( Hu et al . , 2017 ) . Outside of its role in meiotic drive , wtf4 has no apparent role in promoting fertility ( Figure 2A , diploid 15 ) . Instead , the gene causes about half of all gametes to be destroyed in heterozygotes . In other words , the wild-type allele of wtf4 causes infertility to promote its own fitness . This puts wtf4 into a state of genetic conflict with the rest of the genome because infertility is clearly bad for fitness of loci unlinked to wtf4 . Unlinked variants that can suppress drive would be favored by natural selection because they increase fitness ( Crow , 1991 ) . Novel wtf4 variants that can evade this suppression to reestablish drive would then be favored . This evolutionary dynamic is analogous to that observed between viruses and host immune systems and is well known to foster a ‘molecular arms race’ in which both sides must continually innovate ( Daugherty and Malik , 2012; McLaughlin and Malik , 2017 ) . Consistent with the idea that the gene is locked in such an arms race , the DNA sequence divergence between Sk and Sp at the wtf4 locus is more than 20-fold higher than the genome-wide average ( Rhind et al . , 2011; Zanders et al . , 2014 ) . The evolution of wtf4 elicits the question of how the gene can rapidly evolve while maintaining specificity between the poison and antidote it encodes . Uncoupling these components leads to sterility , an evolutionary dead end . It is possible that such variants do arise and are quickly purged from populations . We propose , however , that the coding sequence overlap between the poison and antidote could promote specificity between the two components , for example , by the antidote acting as a dominant suppressor of the poison . In this manner , the poison could diverge without losing the self-protection conferred by the antidote . Using a shared sequence to confer specificity between drive components may be a recurring theme amongst gamete-killers . In their analyses of Podospora Spok genes , Grognet and colleagues found that they could generate antidote-only alleles and a partial poison-only allele via C- and N-terminal tagging of the proteins , respectively . They were not , however , able to generate separation of function alleles of Spok1 by making N- or C- terminal deletions ( Grognet et al . , 2014 ) . Future work identifying the molecular mechanisms of both families of drivers will be required to test the idea that the fact that the poison and antidote are encoded by overlapping sequences minimizes the probability of disrupting their specificity . Combined with our previous work , the varied phenotypes of our Sp chromosome 3 introgressions reveal a complex landscape of meiotic drive loci in the Sk and Sp genomes ( Zanders et al . , 2014 ) . As Sk wtf4 is a member of the large wtf gene family , the most likely candidates underlying these other drive loci are wtf genes . For example , there are >20 wtf loci on Sp ( and likely Sk ) chromosome 3 that could contribute to the complex drive phenotypes we observe ( Figure 1 ) ( Bowen et al . , 2003 ) . Consistent with the idea that the Sk wtf4 is not unique in its ability to drive , we showed that Sk wtf28 can also cause drive and Hu and colleagues demonstrate that two additional wtf genes from another isolate of the Sp family also cause meiotic drive ( Hu et al . , 2017 ) . Although not all wtfs are capable of autonomously causing meiotic drive , their rapid evolution is still consistent with their involvement in meiotic drive ( McLaughlin and Malik , 2017 ) . We propose that different wtf genes represent distinct evolutionary stages . The putative ancestral type ( Sk wtf4 and wtf28 ) are still active as meiotic drivers and encode both poison and antidote proteins . The next stratum represents genes ( Sk wtf2 , wtf5 , wtf6 and wtf26 ) that have lost poison , but not antidote function . As we have shown for the Sk wtf4antidote allele , such alleles are unlikely to cause meiotic drive as they have lost their poison-coding capacity , but they still have protective function against the ancestral drive allele and thus may have been selectively retained as ‘domesticated parasites . ’ Over time , when the protective function is no longer beneficial and selected for ( e . g . if the ancestral drive allele is lost from the population ) , such antidote genes may also eventually degenerate . Therefore , the final stratum represents putative wtf pseudogenes such as Sk wtf21 , in which both the poison and antidote function have decayed . There are 25 wtf loci in the Sp genome and this work combined with that of Hu and colleagues implicates these genes in causing and/or modifying meiotic drive ( Bowen et al . , 2003; Hu et al . , 2017 ) . Meiotic drive has therefore played a significant role in the evolution of the Sp group of fission yeasts , despite the heavy fitness costs these selfish loci can levy . It is striking to consider how long these selfish genes went undetected in such a simple and intensely studied organism . How many more such parasites are lurking undetected in eukaryotic genomes ?
For the mapping crosses , fertility and meiotic drive assay , the crosses were carried out similar to the description in Zanders et al . ( 2014 ) . This required making stable diploids , because many of the strains used are homothallic ( h90 ) and their self-mating would generate many non-informative spores . Briefly , we mixed ~200 µL of overnight culture from each haploid parent in a microcentrifuge tube , spun down the cells and plated them on either SPA ( 1% glucose , 7 . 3 mM KH2PO4 , vitamins , agar ) or MEA ( 3% malt extract , agar ) for 12–15 hr at room temperature to allow the cells to mate . We observed no differences in meiotic drive phenotypes for diploids generated on SPA vs . MEA . We generally use SPA , but for some matings have more success isolating stable diploids from MEA . We scraped off the mated cells and spread them on a medium to select heterozygous diploids ( generally minimal yeast nitrogen base plates ) . We grew diploid colonies overnight in 5 mL of rich YEL broth ( 0 . 5% yeast extract , 3% glucose , 250 mg/L of adenine , lysine , histidine , leucine , and uracil ) . We then plated a small amount of the cultures ≤100 µL onto SPA to induce sporulation , and plated a diluted sample onto YEA ( same as YEL , but with agar ) . We screened the colonies that grew on the YEA plate via replica plating to diagnostic media to verify that the culture was comprised of heterozygous diploid cells . If not , the culture was not assayed further . After 3–7 days , we scooped up the mixture of cells , asci , and spores from the SPA plates , treated it with glusulase and ethanol to kill vegetative cells and to release spores from asci , and plated the spores on YEA . We then phenotyped the spore colonies using standard approaches . For some control loci , we could not easily verify heterozygosity in the diploid test described above . For these loci , we verified heterozygosity of the parent diploid in the progeny . If the parent diploid proved not to be heterozygous , the diploid was eliminated . For each cross , we assayed at least two independently created diploids . The number of progeny we scored varied between experiments . To map Sk wtf4 , we assayed at least 100 viable progeny per cross . To characterize Sk wtf4 , we assayed at least viable 200 gametes per cross . The one exception was the Sk wtf4poison/ Sk wtf4antidote cross in which we had to assay allele transmission by PCR and sequencing ( described below ) . For that cross , we assayed 50 viable progeny . All strains used and their genotypes can be found in Supplementary file 3 . We deposited sequencing data from all high-throughput sequencing to GenBank accession number PRJNA376152 . We chose to map first a drive allele present on Sk chromosome 3 via recombination mapping . To eliminate the effects of drivers and gross chromosomal rearrangements from chromosomes 1 and 2 , we began the mapping effort using a strain ( SZY558 ) that contains chromosomes 1 and 2 from Sk , but in which most of chromosome 3 was derived from Sp ( the mosaic chromosome illustrated in Figure 1B is from SZY558 ) . We could not use a complete Sp chromosome 3 because such a strain lacks essential genes due to a translocation between chromosomes 2 and 3 that occurred in the Sk lineage ( Zanders et al . , 2014 ) . Sequencing revealed that chromosome 3 in SZY558 was generated by a crossover event between Sp and Sk chromosomes somewhere between positions 1 , 804 , 477 and 1 , 810 , 659 on the Sp chromosome . The region to the right of this point contains Sk alleles and the strain has the Sk karyotype ( Figure 1B ) . The generation of SZY558 is described in Figure 1—figure supplement 1 . We first crossed SZY201 ( Sp ) to SZY208 ( Sk ) to generate SZY239 and SZY247 . Although no recombination was expected in this cross because the two parental strains are rec12∆ , both SZY239 and SZY247 must contain a recombinant chromosome 2 and/or a recombinant 3 because they inherited non-parental combinations of markers on chromosomes 2 and 3 , and the two species karyotypes are incompatible ( Zanders et al . , 2014; De Veaux et al . , 1992 ) . Such recombinant spores are quite rare ( Zanders et al . , 2014 ) but were obtained via selection for a nonparental combination of markers on chromosomes 2 and 3 ( e . g . His+ hygromycin resistant ) . Most such selected progeny are chromosome 3 aneuploids , so we then streaked the strains to single colonies to allow them to lose the additional copy of chromosome 3 . We crossed SZY239 to SZY247 to generate a strain ( SZY382 ) that contained the recombinant chromosome 3 , but also had the lys1-37 and his5+ markers on chromosomes 1 and 2 , respectively . The lys1-37 and his5+ markers in this strain were useful for following chromosomes 1 and 2 in a subsequent cross . We transformed SZY382 with a PCR fragment generated with oligos 255 and 256 using plasmid pAG32 as a template to generate a strain ( SZY547 ) with arg12∆::hphMX4 ( Goldstein and McCusker , 1999 ) . The arg12 locus is on chromosome 3 in Sk and in strain SZY547 , which has the Sk karyotype . We then crossed SZY547 to SZY192 ( Sk ) to generate strain SZY558 . The purpose of this cross was to move the recombinant chromosome 3 into a strain background with pure Sk chromosomes 1 and 2 ( marked with lys1+ and his5∆::natMX4 ) . For mapping Sk wtf4 , we crossed SZY558 to a differentially marked Sk strain ( SZY210 ) to generate recombinant haploid progeny ( introgression strains ) that contained a smaller fraction of chromosome 3 from Sp ( Figure 1B ) . We used genetic markers ( ura4 , ade6 and arg12 ) to select only true haploid recombinants for our introgression strains . We mated the introgression strains and Sk ( SZY196 ) to generate diploids 1–8 ( Figure 1C ) . The mapping scheme was designed such that diploids generated by these matings were homozygous rec12∆ , so recombination would infrequently separate the drive allele from the genetic markers used to distinguish the introgression and Sk chromosomes ( De Veaux et al . , 1992 ) . We sequenced at least one introgression representing each phenotype we observed amongst these strains and distinguished Sp and Sk SNPs as in ( Figure 1—source data 1 ) ( Zanders et al . , 2014 ) . SZY565 ( the haploid parent that contributed the mosaic chromosome 3 to diploid 1 in Figure 1C ) is the introgression strain that contains the smallest region of Sp-derived DNA , from position 55 , 555 to 237 , 572 ( Figure 1—source data 1 ) . The Sk chromosome drove against this introgression in test crosses . We assumed that whatever feature of the Sp genome ( either the presence of a target of killing or the absence of an antidote to killing ) that conferred the sensitive phenotype ( i . e susceptibility to being destroyed by the Sk driver ) must be within that region and , correspondingly , that the Sk drive allele must also be within or very near that region . This is because a drive allele that acts to kill gametes that do not inherit it should target the homologous locus or a closely linked site to prevent self-killing . A drive allele that killed gametes that inherit a locus not linked to the drive allele would be an evolutionary dead end because it would kill gametes bearing the drive allele as often as it would kill gametes bearing the competing allele . To narrow in on the key drive locus , we crossed SZY565 to SZY196 ( Sk ) to get a strain ( SZY649 ) with the same chromosome 3 as SZY565 , but with his5+ rather than his5∆::natMX4 on chromosome 2 . The ura4 locus , at position 116 , 726–115 , 589 on chromosome 3 , is within the Sp-derived region . We added an additional marker ( kanMX4 ) within the Sp-derived region at position 214 , 491 to generate strain SZY659 . To do this , we first generated plasmid pSZB134 which contains ~1 kb of DNA ( amplified from Sp genomic DNA with oligos 380 and 381 ) upstream of the target site ( 214 , 491 ) cloned into the BamHI and BglII sites of pFA6a , and ~1 kb of DNA ( amplified with oligos 382 and 383 ) downstream of the target site cloned into the SacI and SpeI sites of the pFA6a ( Wach et al . , 1994 ) . The transformation cassette was released from pSZB134 via NotI digest and used to make SZY659 . We then crossed SZY659 to a differentially marked Sk strain ( SZY320; rec12+ ) and screened for haploid progeny that had experienced a crossover within the Sp-derived region between the ura4∆::natMX4 allele from SZY320 and the kanMX4 allele in SZY659 ( Figure 1D ) . We tested nine such haploids by test crossing them to Sk ( SZY196 ) . Two haploids had an Sk-like phenotype in that they showed Mendelian transmission of the ura4 locus; the other six showed the sensitive phenotype . We genotyped SNPs of the haploids at a few sites within the region to roughly estimate where the recombination event ( s ) occurred ( Zanders et al . , 2014 ) . Amongst the haploids with the sensitive phenotype , SZY679 and SZY685 have the most Sk-derived DNA: they contain Sp DNA only between position ~210 , 000 ( between 207 , 954 and 210 , 312 ) to 237 , 572 ( Figure 1C , diploid 9 , Figure 1—source data 1 ) . The two strains with the Mendelian phenotype ( SZY684 and SZY686 ) contain very little Sp-derived DNA . The Sp DNA begins between positions 210 , 312 and 214 , 500 and ends before 215 , 926 . Comparing these two classes suggested that the key drive locus is located between positions ~ 210 , 000 and 237 , 572 ( but not within the small region surrounding ~214 , 000 ) . The annotated features of this region include all or part of 10 genes plus one pseudogene in Sp ( Figure 1D ) . Using oligos MESZ176 and MESZ177 , we amplified from Sk genomic DNA the region corresponding to the wtf3+wtf4 locus in Sp . The product amplified is at least 1 . 5 kb smaller than the corresponding product from Sp . We then sequenced the PCR product using oligos 557 , 560 , 565 , 566 , 567 , 568 , 569 , 570 , 595 , 597 , 598 , 599 , 601 , 602 , and 603 and assembled a 2943 bp contig . This sequence has been deposited to GenBank , accession number KY652738 . We did a BLAST search comparing our Sk sequence contig to all Sp protein sequences and got Sp wtf13 and wtf4 as top hits . The Sk region contains only one wtf-like gene , whereas the Sp region has the complete wtf4 gene and the wtf3 pseudogene . As the Sk gene appears to be orthologous to Sp wtf4 based on synteny and sequence similarity , we named the gene Sk wtf4 . We used the Sp PacBio meiotic transcriptome sequences to predict intron-exon boundaries in Sk wtf4 ( Kuang et al . , 2017 ) . The authors Kuang et al . ( 2017 ) kindly provided pre-publication access to ‘Iso-Seq’ consensus isoform sequences . wtf genes are not well-represented in the splice isoform summary tables generated for that study due to the very high nucleotide identity between wtf paralogs and stringent filtering of multiply-mapping reads . We therefore re-mapped Iso-Seq data to the Sp reference genome assembly using GMAP ( Wu et al . , 2016 ) , reporting only alignments with ≥99% identity and covering ≥99% of the length of the isoform sequence , and using the parameter ‘--suboptimal-score 20’ to reduce secondary matches ( this parameter choice successfully eliminates cross-mapping between wtf4 and wtf13 ) . We used IGV ( Thorvaldsdottir et al . , 2013 ) to visualize splice isoforms for each gene . These data reveal a coding sequence that is slightly different from that of the currently annotated Sp wtf4 gene ( http://www . pombase . org/spombe/result/SPCC548 . 03c ) . The long form of Sk wtf4 has six predicted exons and encodes a 337 amino acid protein with 82% amino acid identity to the 366 amino acid protein encoded by Sp wtf4 . The TMHMM model predicts six transmembrane helices with high probability ( >80% ) and one with lower probability ( <50% ) ( Krogh et al . , 2001 ) . To generate the Sk wtf4∆ mutant , we used the CRISPR-Cas9 system after first failing to generate the mutant via the standard homologous recombination approach ( Jacobs et al . , 2014 ) . This system requires the starting strain to be ura4- and leu1- . We generated an Sk mutant ( SZY661 ) in which leu1 was replaced with hphMX4 in strain SZY320 . We did this by first cloning a leu1∆::hphMX4 cassette ( pSZB136 ) . We made this plasmid by first cloning leu1+ ( amplified from Sk genomic DNA with oligos 413 and 414 ) into pFA6a cut with NdeI and ClaI and blunted with Klenow fragment of DNA polymerase I . This new vector was then cut with ClaI and NdeI ( within leu1 ) and blunted with Klenow: the hphMX4 cassette liberated from pAG32 with PvuII and ClaI was ligated into the gap ( Goldstein and McCusker , 1999 ) . Oligos 413 and 414 were used again to amplify the leu1∆::hphMX4 cassette for transformation . To generate plasmid pSZB184 , which encodes a guide RNA targeting the Sk wtf4 region , we annealed oligos 577 and 578 and cloned them into the CspCI site of pMZ283 ( Jacobs et al . , 2014 ) We used overlap-PCR to generate a repair cassette containing ~1 kb of homology upstream and downstream of the Sk wtf4 region flanking the kanMX4 cassette from pFA6a ( Wach et al . , 1994 ) . We stitched together the products of PCRs generated with oligos 571 and 572 , 575 and 576 , and 573 and 574 to make the repair cassette . We then transformed strain SZY661 with pMZ222 , pSZB184 , and the repair cassette . We screened through Ura+ Leu+ transformants containing both plasmids for wtf4 deletions via PCR and sequencing . We found that strain SZY862 contained a deletion of wtf4 , but unexpectedly was not resistant to G418 . Sequencing of the region revealed a truncation of the kanMX4 gene . SZY863 contains the same deletion as SZY862 , but is Ura+ due to retention of the ura4+ cassette from pSZB184 at an unknown location closely linked to the endogenous ura4 locus , although the strain retains the ura4∆::natMX4 allele at the endogenous ura4 locus . We first generated pSZB188 , a plasmid containing the kanMX4 selectable marker and a mutant ade6 allele that has 5’ , central , and 3’ deletions . This vector can be cut with KpnI within the mutant ade6 gene and then integrated into ade6+ to generate Ade- G418-resistant transformants . Other genes can be added to the vector to introduce them into the genome at the ade6 locus . To construct pSZB188 , we first made a mutant ade6 cassette via overlap PCR stitching a PCR product made from oligos 588 and 589 to one made from oligos 591 and 590 . This ade6 cassette was then digested with BamHI and XhoI and cloned into the BamHI and SalI sites of pFA6a ( Wach et al . , 1994 ) . We cloned the Sk wtf4 region into pSZB188 by first amplifying the region with oligos 619 and 620 . We digested the PCR product with SacI and cloned it into the SacI site of pSZB188 to generate pSZB189 . We introduced KpnI-digested pSZB189 into yeast and selected transformants on YEA with G418 plates . We picked red colonies , as proper integrants should harbor a mutant ade6 allele flanking the sides of the plasmid sequence . The duplicated ade6 gene makes the locus unstable and Ade+ revertants that have ‘popped out’ all plasmid-derived sequences are readily obtained . The Sk wtf4-GFP allele was made using overlap PCR . We amplified the promoter region from Sk genomic DNA using oligos 633 and 604 and the open reading frame sequence using oligos 605 and 606 . We used pKT127 as a template to amplify yEGFP using oligos 607 and 634 ( Sheff and Thorn , 2004 ) . We then stitched the three PCR products together using overlap PCR . We cut the resulting cassette with SacI and cloned it into the SacI site of pSZB188 to generate pSZB204 . This construct was integrated at ade6 as described above . For the Sk wtf4antidote allele , using overlap PCR , we stitched together two PCR products generated with oligo pairs 735 and 686 , and 620 and 736 , both using pSZB189 as a template . We cloned the stitched PCR product into the SacI site of pSZB188 to generate pSZB246 . We then cut pSZB246 and introduced it into yeast as described above . We generated the Sk wtf4poison allele using overlap PCR . Sk wtf4 has two in-frame start codons in the annotated exon 1 . Mutating the first start codon had no phenotype ( data not shown ) , so we mutated both . To mutate the first start codon , we used overlap PCR to stitch together two PCR products made by oligo pairs 701 and 686 and 620 and 702; both reactions used pSZB189 as a template . The stitched PCR product was cloned into the SacI site of pSZB188 to generate pSZB244 . We used pSZB244 as a template to mutate the second start codon via overlap PCR . We stitched together PCR fragments generated by oligo pairs 620 and 739 and 686 and 740 and cloned that product into the SacI site of pSZB188 to generate pSZB258 . We cut pSZB258 and introduced it into yeast as described above . We cloned the Sk mCherryantidote-wtf4 allele using overlap PCR . First , we purchased from IDT ( Coralville , IA ) a synthetic double-stranded DNA gene block including the Sk wtf4 promotor , the mCherry coding sequence ( Hailey et al . , 2002 ) , five glycine codons , and the first part of Sk wtf4 exon 1 . We amplified that fragment with oligos 620 and 604 and stitched that PCR product to another that contained the rest of the Sk wtf4 gene amplified with oligos 605 and 687 from plasmid pSZB189 . We then cloned that product into the SacI site of pSZB188 to generate pSZB248 , which we cut and introduced into yeast as described above . For the Sk wtf4poison-GFP allele , we amplified the 5’ end of the gene with oligos 620 and 739 using plasmid pSZB244 as a template . We amplified the 3’ end of the gene with oligos 740 and 634 using pSZB203 as a template . We then used overlap PCR to stitch those PCR fragments together and cloned the product into the SacI site of pSZB188 to generate pSZB257 , which we cut and introduced it into yeast as described above . We used the same strategy to integrate other Sk wtf genes into Sp . We used Sk genomic DNA as a template to amplify wtf21+wtf26 with oligos 643 and 644 , wtf2 with oligos 647 and 648 , wtf5 with oligos 649 and 650 , and wtf6+wtf28 with oligos 651 and 652 . We cut each cassette with SacI and cloned them into the SacI site of pSZB188 to generate: pSZB209 ( wtf21+wtf26 ) , pSZB212 ( wtf2 ) , pSZB217 ( wtf5 ) , and pSZB215 ( wtf6+wtf28 ) . We subcloned Sk wtf6 and Sk wtf28 from pSZB215 by first amplifying the individual genes using oligo pairs 732 and 652 , and 651 and 733 , respectively . We then cloned the genes into the SacI site of pSZB188 to generate pSZB252 ( wtf6 ) and pSZB254 ( wtf28 ) . All sequences of these genes have been deposited in GenBank , accession numbers KY652739-KY652742 . These constructs were all integrated at ade6 as described above . Because the alleles wtf4poison ( SZY1051 ) and wtf4antidote ( SZY1110 ) are marked with the same drug marker , to score transmission of alleles for this cross , we used sequencing . We generated diploids and spores as described above . We then plated the spores on YEA , picked the colonies to a YEA master plate and replicated to score control markers . We also prepared lysates for PCR from the master plate by scraping cells off the master plate into 20 µl of 20 mM NaOH . We boiled the cells for 5 min , froze them in liquid nitrogen , boiled again for 10 min , and spun the debris down . Using the supernatant lysate , we amplified the wtf4 region using oligos A01112 and 678 . We then sequenced the exon 1 region using oligo 861 . We analyzed the exon 1 region for the start codon mutations mentioned above ( Figure 4A ) . If the exon 1 mutations were present , we quantified this as a poison allele; if not present , as an antidote . To avoid ade6- mutant auto-fluorescence in cytology , we introduced ade6+ at the his5 locus . We amplified a region upstream of his5 to generate piece A using oligo pair 795 and 796 . We amplified a region downstream of his5 to generate piece C using oligo pair 797 and 798 , and to amplify ade6+ we generated piece B using oligo pair 799 and 800 . We stitched together pieces A , B and C using oligo pair 795 and 798 and introduced the product into yeast . For the fertility assay , we added 5–10 µl of propidium iodide ( PI , 1 mg/ml ) to 50 µl of H2O , and scraped the yeast from the SPA plate into the PI mix . We then incubated the yeast plus PI mixture at room temperature for 20 min . We took the images on a Zeiss ( Germany ) Observer . Z1 wide-field microscope with a 40x ( 1 . 2 NA ) water-immersion objective and collected the emission onto a Hamamatsu ORCA Flash 4 . 0 using µManager software . We acquired the PI images with BP 530–585 nm excitation and LP 615 emission , using an FT 600 dichroic filter . For all other fluorescence microscopy , we imaged on a LSM-700 AxioObserver microscope ( Zeiss ) , with a 40x C-Apochromat water-immersion objective ( NA 1 . 2 ) , with 488 and 555 nm excitation . We collected GFP fluorescence through a 490–55 nm bandpass filter and mCherry fluorescence through a 615-nm longpass filter . The continuously variable secondary dichroic filter was positioned at 578 nm . We also imaged using a LSM-780 ( Zeiss ) microscope , with a 40x C-Apochromat water-immersion objective and 100x alpha Plan-Apochromat oil-immersion objective ( NA 1 . 2 and 1 . 46 , respectively ) , in photon-counting channel mode with 488 and 561 nm excitation . We collected GFP fluorescence through a 481–552 bandpass filter and mCherry through a 572 longpass filter . For all images acquired on the LSM-780 ( Zeiss ) microscope , using the same objectives as described above , we also imaged in photon-counting lambda mode , with 488 and 561 nm excitation . We collected fluorescence emission over the entire visible range . After acquisition , we linear unmixed the images using an in-house custom written plugin for ImageJ ( https://imagej . nih . gov/ij/ ) . Unmixing was achieved using spectra obtained from control cells . We did unmixing to verify that there was no auto-fluorescence in the cells ( Figure 5—figure supplement 1 ) we scored . We did not score auto-fluorescent cells . Brightness and contrast is not the same for all images . We assayed at least 35 asci ( but usually >100 ) for each genotype represented in Figures 3 and 5 . | Animals , plants and fungi produce sex cells – known as gametes – when they are preparing to reproduce . These cells are made when cells containing two copies of every gene in the organism divide to produce new cells that each only have one copy of each gene . Therefore , a particular gene copy usually has a 50% chance of being carried by each gamete . There is a group of genes that selfishly increase their chances of being transmitted to the next generation by destroying the gametes that do not carry them . These “gamete killer” genes can lead to infertility and other health problems . Fission yeast is a fungus that is widely used in research . Previous studies revealed that the yeast are likely to have several gamete killers , but the identities of these genes or how they work were not clear . Nuckolls , Bravo Núñez et al . sought to identify at least one gamete killer gene and understand how it works . The experiments found that a gene called wtf4 acts as a gamete killer in fission yeast . This gene encodes two different proteins , one that acts as a poison and one that acts as an antidote . The antidote remains inside the gametes that contain the wtf4 gene , while the poison is released in the surrounding environment . The poison is capable of killing all of the gametes , but the antidote protects the gametes that contain the wtf4 gene . Further experiments show that wtf4 is just one member of a large family of genes that are also likely to play roles in selectively killing gametes . A separate study by Hu et al . found that two other members of the wtf family also act as gamete killers in fission yeast . Together , these findings expand our understanding of the nature of gamete killers and how they can contribute to infertility . This may guide the search for gamete killers in humans and other organisms . In the future , gamete killers could potentially be used to eradicate populations of pests that damage crops or spread diseases in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"genetics",
"and",
"genomics"
] | 2017 | wtf genes are prolific dual poison-antidote meiotic drivers |
Fusion of skeletal muscle stem/progenitor cells is required for proper development and regeneration , however the significance of this process during adult muscle hypertrophy has not been explored . In response to muscle overload after synergist ablation in mice , we show that myomaker , a muscle specific membrane protein essential for myoblast fusion , is activated mainly in muscle progenitors and not myofibers . We rendered muscle progenitors fusion-incompetent through genetic deletion of myomaker in muscle stem cells and observed a complete reduction of overload-induced hypertrophy . This blunted hypertrophic response was associated with a reduction in Akt and p70s6k signaling and protein synthesis , suggesting a link between myonuclear accretion and activation of pro-hypertrophic pathways . Furthermore , fusion-incompetent muscle exhibited increased fibrosis after muscle overload , indicating a protective role for normal stem cell activity in reducing myofiber strain associated with hypertrophy . These findings reveal an essential contribution of myomaker-mediated stem cell fusion during physiological adult muscle hypertrophy .
Adult skeletal muscle exhibits a dramatic capacity to regenerate after injury owing to the presence of satellite cells ( SCs ) , the resident muscle stem cells . SCs reside underneath the basal lamina in a quiescent state , and upon injury , become activated to generate a pool of myogenic progenitors ( MPs ) that differentiate and fuse to each other or existing myofibers to restore structure and function to muscle ( Relaix and Zammit , 2012; Yin et al . , 2013 ) . While SCs are clearly necessary for efficient regeneration after acute injury ( Günther et al . , 2013; Lepper et al . , 2011; Sambasivan et al . , 2011 ) , the role of SCs during adult muscle hypertrophy and maintenance is much less understood . Moreover , the regulation of myogenic fusion is relatively unknown compared to the knowledge associated with activation , proliferation , and differentiation of SCs . Elucidation of the magnitude of fusion , and the associated molecular regulation of the fusion process , for adult muscle hypertrophy may lead to strategies that augment loss of muscle mass due to chronic disease and aging . Load-induced hypertrophy is a desired adaptation of exercise resulting in an increase in skeletal muscle mass , thereby enhancing overall size and function of the affected muscles . Fundamentally , skeletal muscle hypertrophy occurs when the rates of muscle protein synthesis exceed those of protein degradation . In addition to activation of protein synthesis pathways directly in the myofiber , a myofiber could hypertrophy through fusion of SCs leading to a greater capacity for transcription of contractile machinery . The contribution of SCs for hypertrophy was initially justified through the concept of the myonuclear domain ( Cheek et al . , 1965 ) , which postulates a coordinated increase between myonuclei number and cytoplasm volume in order to maintain a constant domain size ( Snow , 1990; Tamaki et al . , 1996 ) . Moreover , myonuclear accretion precedes an increase in skeletal muscle size ( Bruusgaard et al . , 2010 ) , and loss of serum response factor ( Srf ) in the myofiber blunts hypertrophy potentially through a SC-mediated mechanism ( Guerci et al . , 2012 ) , altogether implicating a role for fusion during hypertrophy . Studies utilizing diphtheria toxin ( DTA ) -mediated ablation of SCs have revealed that effective myofiber hypertrophy occurs in the absence of satellite cells and that SCs are not required to maintain muscle mass during aging ( Fry et al . , 2015; McCarthy et al . , 2011 ) , however SC ablation leads to impaired coordination and diminished muscle performance ( Jackson et al . , 2015 ) . Additionally , SCs contribute to myofibers during aging of sedentary mice although SC activity was not globally required to maintain muscle size ( Keefe et al . , 2015; Pawlikowski et al . , 2015 ) . Recently , a report utilizing a similar DTA SC ablation model indicates that effective myofiber hypertrophy does not occur in the absence of SCs ( Egner et al . , 2016 ) . Thus , there is no consensus regarding the contributions of SCs during adult skeletal muscle hypertrophy , highlighting the need for more independent models that assess the biology of SC activity . Myomaker ( Tmem8c ) is a muscle-specific membrane protein necessary for fusion of embryonic and adult MPs ( Millay et al . , 2013 ) . In adult muscle , myomaker expression is undetectable but rapidly induced in SC descendants after cardiotoxin ( CTX ) injury . In this acute injury setting where there is complete destruction of the muscle architecture , the ensuing restoration is the sole responsibility of SCs . Here , all MPs exhibit upregulation of myomaker , and we have shown that fusion is more efficient if both fusing cells express myomaker ( Millay et al . , 2016 ) . The regulation and function of myomaker in the SC or myofiber compartment in an environment where myofibers are maintained and not destroyed , such as through increased workload and exercise , is unknown . To investigate the role of myomaker-mediated fusion in skeletal muscle hypertrophy , we explored the expression of myomaker on SCs and myofibers , and generated fusion-incompetent SCs through targeted deletion of myomaker . We show that myomaker is upregulated in SCs upon muscle overload ( MOV ) , but significant activation of myomaker within the myofiber is not detected . In the absence of myomaker induction in SCs , we report an impaired hypertrophic response to MOV , which is associated with a failure to activate Akt/mTOR-mediated protein synthesis within the myofiber . Collectively , our findings demonstrate that effective SC fusion is required for optimal muscle hypertrophy following a load-induced stimulus .
To determine the regulation of MP fusion during adult skeletal muscle hypertrophy , we first assessed the temporal kinetics of myomaker expression after MOV . In wild-type ( WT ) mice at baseline , myomaker mRNA and protein are not readily detected in muscle , however myomaker is acutely induced after MOV ( Figure 1A and B ) . Expression of myomaker peaked 7 days after MOV , and was down-regulated at later time points . To elucidate if myomaker was induced in activated MPs or by the myofiber during MOV , we analyzed Tmem8cLacZ/+ mice , which contain a LacZ cassette in intron 1 of the myomaker locus ( Millay et al . , 2014 ) . This allele results in exon 1 of myomaker splicing with LacZ , generating a transcript with exon 1 of myomaker , and an internal ribosome entry site ( IRES ) that allows independent translation of LacZ . Thus , in this system LacZ serves as a readout for myomaker transcription but not myomaker localization . X-gal staining of the plantaris muscle from Tmem8cLacZ/+ mice at multiple time points after MOV revealed detectable LacZ in discrete locations surrounding the myofiber at day 3 of MOV , and in myofibers at later stages of MOV ( Figure 1C ) . To understand the myogenic state of the different types of LacZ+ cells we stained serial sections with either x-gal or embryonic myosin ( myh3 ) , a marker of muscle differentiation . Here we observed a population of large LacZ+ cells and a population of small LacZ+ cells that were in the presumptive SC position ( Figure 1—figure supplement 1 ) . We classified the large LacZ+ cells as myofibers due to their size , and these were either myh3+ or myh3- . For myofibers , LacZ+ myh3+ cells exhibited stronger x-gal staining compared to the LacZ+ myh3- cells ( punctate staining ) suggesting that the former population are de novo fibers formed from the fusion of MPs ( Figure 1—figure supplement 1 ) . We interpret the punctate LacZ+ myh3- myofibers as existing fibers that have fused with a MP . The population of small LacZ+ cells were either myh3+ ( differentiated myocytes ) or myh3- ( MPs ) , and we classified these as the non-myofiber population ( Figure 1—figure supplement 1 ) . Quantification of these populations multiple days after MOV revealed an increase in myomaker-expressing MPs at the early stages of MOV ( days 3 and 7 ) , followed by a reduction at later stages of MOV ( days 10 and 14 ) ( Figure 1C ) . In contrast , the majority of LacZ+ myofibers were observed at later stages of MOV , with negligible occurrence at 3 days after MOV . Thus , myomaker exhibits a contrasting expression pattern in MPs compared to the myofiber compartment in response to a load-induced stimulus . 10 . 7554/eLife . 20007 . 003Figure 1 . Regulation of myomaker activation and fusion during load-induced hypertrophy . Myomaker expression at various time points after MOV was assessed by qPCR ( A ) , and western blot analysis ( B ) , showing induction of myomaker at all stages of MOV ( n = 2–4 mice ) . ( C ) Tmem8cLacZ/+ mice were subjected to MOV and plantaris sections were X-gal stained at multiple time points after surgery . LacZ , a surrogate for myomaker expression , was observed in MPs ( arrows ) during the early stages of MOV , and in myofibers ( arrowheads ) in the later stages of MOV . Quantification of the number of LacZ+ non-myofibers indicates myomaker is robustly activated in MPs at day 3 and day 7 of MOV but the number is reduced at day 10 and day 14 ( n = 3–5 mice ) . Quantification of LacZ+ myofibers demonstrates the majority of expression occurs at day 7 and day 10 after MOV ( n = 2500–4 , 100 myofibers from 3–5 independent mice ) . ( D ) Fusion of MPs with myofibers was assessed by labeling proliferating cells with BrdU and tracking their incorporation into a myofiber , identified by immunostaining with a dystrophin antibody . Mice were subjected to MOV and treated with BrdU during the initial 7 days or the last 7 days after MOV . Fusion was scored as a BrdU+ nucleus within a dystrophin+ myofiber as depicted by the arrows . Quantification of the percentage of myofibers containing a BrdU+ nucleus shows an increased labeling of fusion competent satellite cells during the first 7 days of MOV , which correlates with the highest expression of myomaker in satellite cells ( n = 380–1 , 511 myofibers from 3–9 independent mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bars: 50 μm , except inset in ( D ) which represents 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 00310 . 7554/eLife . 20007 . 004Figure 1—figure supplement 1 . Day 7 MOV serial sections were stained with x-gal or immunostained with laminin and embryonic myosin ( myh3 ) antibodies . Two populations of LacZ+ myofibers were observed . One exhibited punctate x-gal and was myh3- ( stars ) thus representing existing fibers that have fused with a MP . The second myofiber population exhibited stronger x-gal staining and was myh3+ ( blue arrows ) and these represent de novo myofibers . We also observed two different populations of small cells exhibiting x-gal staining , which we classify as non-myofiber . The first population was LacZ+ myh3+ ( green arrows ) which indicate differentiated myocytes , whereas LacZ+ myh3- ( magenta arrows ) depict myogenic progenitors . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 004 To determine if the kinetics of myomaker expression temporally coincides with fusion of MPs during MOV , we examined fusion through 5-bromo-2'-deoxyuridine ( BrdU ) labeling of proliferating cells . After synergist ablation surgery , WT mice were treated with BrdU either during the initial or final seven days of the overload stimulus , and fusion was defined as the incorporation of a BrdU+ nucleus within a dystrophin+ myofiber . Labeling of fusion-competent MPs during the first 7 days of MOV resulted in a greater percentage of myofibers containing a BrdU+ nucleus compared to labeling during the final 7 days of MOV ( Figure 1D ) . While these data indicate that fusion-competent MPs are generated throughout the MOV stimulus , there is greater labeling of fusogenic cells during the initial stages of MOV when myomaker-LacZ+ MPs are more prominent , suggesting a temporal relationship between myomaker expression and fusion . While MP expression of myomaker-LacZ indicates that myomaker is actively transcribed in progenitor cells , LacZ activity in myofibers could result from fusion with a LacZ+ MP or direct transcription of myomaker-LacZ from a myonucleus . To determine the source of LacZ in myofibers during MOV , we blocked the ability of MPs to fuse through genetic deletion of myomaker . Tmem8cLacZ/loxP mice were bred with satellite cell-specific Cre recombinase ( Pax7CreERT2/+ ) mice ( Lepper et al . , 2009 ) , and subsequently treated with tamoxifen to genetically delete myomaker in muscle progenitor cells ( myomaker scKO ) . Controls include tamoxifen-treated Tmem8c+/+; Pax7CreERT2/+ mice , or vehicle-treated Tmem8cLacZ/loxP; Pax7CreERT2/+ mice ( Figure 2—figure supplement 1 ) . Mice were injected with tamoxifen for 5 consecutive days prior to MOV and maintained on a tamoxifen dosage throughout the protocol . qPCR analysis confirmed the efficiency of this tamoxifen-based strategy , as myomaker expression was eliminated in myomaker scKO muscle compared to control muscle 7 days after MOV ( Figure 2A ) . To be certain that this model maintains the presence of SCs during MOV we crossed the Rosa26mTmG genetic lineage reporter with our Tmem8cLacZ/loxP; Pax7CreERT2/+ mice ( Figure 2—figure supplement 1 ) . This results in a ubiquitously expressed membrane-targeted tdTomato ( mTom ) , but upon Cre-mediated recombination tdTomato is excised allowing membrane-GFP expression . These mice were subjected to MOV and assayed for the presence of mGFP+ cells ( MPs ) using flow cytometry . We observed a non-significant reduction in mGFP+ cells after MOV in myomaker scKO muscle ( Figure 2—figure supplement 2 ) . We also performed immunostaining for Pax7 , which revealed no major differences in SC number during MOV when normalized to fiber number ( Figure 2—figure supplement 2 ) . These results indicate that while there could be subtle alterations in SC dynamics , the myomaker scKO model does not significantly alter the frequency of muscle stem cells . 10 . 7554/eLife . 20007 . 005Figure 2 . Myomaker is mainly expressed in MPs and not myofibers during MOV . ( A ) Tmem8cLacZ/loxP; Pax7CreERT2/+ mice ( myomaker scKO ) were treated with tamoxifen to ablate myomaker specifically in SCs . Control mice include vehicle-treated Tmem8cLacZ/loxP; Pax7CreERT2/+ mice and tamoxifen-treated Tmem8c+/+; Pax7CreERT2/+ mice . Mice were subjected to MOV and myomaker expression was assessed through qPCR analysis 7 days post-surgery showing that myomaker was efficiently deleted in myomaker scKO muscle ( n = 3–4 independent mice ) . ( B ) To determine the origin of LacZ staining observed in control myofibers ( Figure 1C ) , MOV was performed on myomaker scKO mice and sections were X-gal stained . Ablation of myomaker in SCs results in restriction of myomaker expression to MPs and lack of expression in the myofiber . Quantification of LacZ+ non-myofibers shows a significant increase at day 7 in myomaker scKO mice compared to control mice because myomaker scKO are fusion defective and thus remain outside the myofiber ( n = 3–5 mice ) . The percentage of LacZ+ myofibers are reduced in myomakerscKO mice suggesting that myofiber LacZ is acquired through fusion of MPs ( n = 2500–4 , 100 myofibers from 3–5 independent mice ) . In the quantification of ( B ) , the control bars are from Figure 1C . ( C ) To genetically assess the requirement of myofiber-derived myomaker , tamoxifen was administered to Tmem8cLacZ/loxP; ACTA1CreERT2 mice ( myomaker mKO ) prior to MOV . qPCR analysis of myomaker in control , myomaker scKO , and myomaker mKO mice after MOV shows that myomaker is only reduced in myomaker scKO mice , demonstrating that myomaker is not transcribed from a myofiber nucleus ( n = 3–4 mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 00510 . 7554/eLife . 20007 . 006Figure 2—figure supplement 1 . Schematic showing the mice used to ablate myomaker in satellite cells , and their associated control groups . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 00610 . 7554/eLife . 20007 . 007Figure 2—figure supplement 2 . Myogenic progenitors are maintained in myomaker scKO muscle during MOV . ( A ) Flow cytometry for mGFP+ progenitors in tamoxifen-treated Pax7CreERT2/+; RosamTmG ( control ) and Tmem8cLacZ/loxP; Pax7CreERT2/+; RosamTmG mice ( myomaker scKO ) reveals a distinct GFP- ( red ) and GFP+ ( green ) population of cells 7 and 10 days after MOV . Top panels reveal gating strategy derived from forward and side scatter plots . Plots shown are representative from one experiment . Quantitative analysis confirms similar levels of mGFP+ cells between control and myomaker scKO muscles at 7 and 10d of MOV ( n = 3 for each group ) , indicating that myogenic progenitors are present after satellite cell ablation of myomaker . ( B ) Representative images of Pax7+ nuclei in plantaris sections further reveal that satellite cells are present in fusion-incompetent muscles throughout two weeks of MOV . Quantification revealed a significant difference 7 days after MOV when Pax7+ cells are analyzed per section but not when normalized to fiber number ( n = 4–6 ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bar: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 00710 . 7554/eLife . 20007 . 008Figure 2—figure supplement 3 . Myomaker expression is reduced in myomaker mKO mice when tamoxifen is present concominant with satellite cell activation suggesting ACTA1CreERT2 is active in myoblasts prior to fusion ( n = 3–4 mice ) . These data indicate the ACTA1CreERT2 transgene is functional . Data are represented as mean ± SEM , *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 008 We next assessed how the loss of myomaker in MPs impacts LacZ+ activity within the muscle . X-gal staining of plantaris sections from myomaker scKO mice after MOV demonstrates that myomaker expression is restricted to non-myofibers ( Figure 2B ) . Quantification of LacZ+ non-myofibers revealed an accumulation at 7 days post-MOV in myomaker scKO mice compared to control mice , reflecting the inability of myomaker-deficient MPs to fuse to myofibers ( Figure 2B ) . We also observed a lack of LacZ activity in myofibers in myomaker scKO mice at days 7–14 of MOV , suggesting that LacZ in myofibers in control mice is contributed through fusion of MPs ( Figure 2B ) . Overall , these data indicate that myomaker is not actively transcribed from a myofiber nucleus in response to MOV . One possibility for the lack of myonuclear transcription in myomaker scKO mice is that the myofiber requires expression of myomaker on MPs to activate myomaker transcription within the myofiber . To exclude this possibility , we genetically deleted myomaker using a myofiber-specific Cre and assessed myomaker levels after MOV . Specifically , we crossed Tmem8cLacZ/loxP mice with a tamoxifen-inducible Cre under control of the human skeletal actin promoter ( ACTA1CreERT2 ) . We treated Tmem8cLacZ/loxP; ACTA1CreERT2 mice with tamoxifen ( myomaker mKO ) for 5 days prior to MOV and analyzed myomaker transcript levels through qPCR analysis . If myomaker was transcribed from a myofiber nucleus in the presence of WT MPs , we would expect a reduction of myomaker levels in myomaker mKO muscle after MOV . However , myomaker mKO MOV plantaris did not exhibit a reduction in myomaker levels compared to control MOV , whereas a reduction was achieved in myomaker scKO MOV plantaris ( Figure 2C ) . To be certain the ACTA1CreERT2 was functional , we assessed myomaker levels in myomaker mKO mice when tamoxifen delivery was maintained after MOV . Using this protocol , we did detect a reduction of myomaker expression in myomaker mKO mice , suggesting a functional ACTA1CreERT2 allele ( Figure 2—figure supplement 3 ) . Myomaker deletion in myomaker mKO mice was achieved only with maintenance of tamoxifen after MOV , likely because the human skeletal actin ( ACTA1 ) promoter is activated as MPs differentiate and prior to fusion . Taken together , our results suggest the LacZ expression observed in control myofibers is acquired primarily from LacZ+ MPs through fusion , and not from transcription by existing myofiber nuclei . Hence , myomaker is activated mainly in MPs during muscle overload . Since myomaker is mainly activated in MPs during MOV , we utilized myomaker scKO mice to determine the role of myomaker during hypertrophy . To ensure effective deletion of myomaker , control and myomaker scKO mice were subjected to five daily doses of tamoxifen prior to MOV , and kept on a tamoxifen regimen until sacrifice ( Figure 3A ) . Following 14 days of overload , evaluation of plantaris weight normalized to tibia length revealed a two-fold increase in control MOV compared to sham ( Figure 3B ) . In contrast , myomaker scKO MOV plantaris was smaller than control MOV , although still increased above myomaker scKO sham weight ( Figure 3B ) . Muscle responds to a load-induced stimulus by generating new myofibers and activating hypertrophic pathways in existing myofibers . Indeed , through histological examination of the plantaris after MOV , in control mice we observed an increase in myofiber size , a more direct evaluation of hypertrophy ( Figure 3C ) . This characteristic was not observed in myomaker scKO MOV plantaris sections , highlighting the importance of myomaker on MPs for efficient hypertrophy ( Figure 3C ) . Cross-sectional area ( CSA ) analysis of fiber size revealed an increase in control MOV compared to control sham ( Figure 3C ) . In contrast , no differences were detected between myomaker scKO sham and MOV , indicating defective myofiber hypertrophy following MOV with genetic deletion of myomaker in MPs ( Figure 3C ) . In addition , frequency distribution analysis confirms the presence of fewer large myofibers in myomaker scKO muscles after overload ( Figure 3C ) . Given that skeletal muscle is comprised of multiple different fiber types that have the capacity to dynamically adapt in response to physiological stimuli , we assessed if all fiber types were resistant to hypertrophy after loss of myomaker in SCs . We immunostained sections after 14 days of MOV with antibodies specific to Type I , Type IIa , and Type IIb myosin and quantified myofiber CSA for each fiber type . This analysis revealed no differences in fiber type frequency after MOV , and that genetic ablation of myomaker in SCs resulted in a lack of hypertrophy of Type IIa , IIb , and IIx myofibers ( Figure 3—figure supplement 1 ) . 10 . 7554/eLife . 20007 . 009Figure 3 . Genetic loss of myomaker in MPs results in an absence of muscle hypertrophy . ( A ) Schematic showing timing of tamoxifen treatment by intraperitoneal ( IP ) injections , MOV , and tissue harvest . Tamoxifen was maintained after MOV through IP injections every other day . ( B ) Plantaris weight normalized to tibia length indicates a dramatic increase in size in control mice , which is blunted in myomaker scKO mice ( n = 7–13 mice ) . ( C ) Representative laminin-stained plantaris sections shows an increase in myofiber size after muscle overload in control mice but not in myomaker scKO mice . Quantification of cross-sectional area ( CSA ) of myofibers at the plantaris mid-belly reveals a significant inhibition of myofiber hypertrophy with satellite cell ablation of myomaker ( n = 3200–5 , 400 myofibers from 7–11 independent mice ) . Relative frequency of myofiber sizes in the various groups of mice demonstrates that control muscle exhibits a greater percentage of larger fibers after MOV . ( D ) Fiber number was quantitated at the mid-belly of the plantaris , and reveals a significant inhibition of new fiber formation in plantaris muscle when myomaker was ablated in satellite cells ( n = 7–11 mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 00910 . 7554/eLife . 20007 . 010Figure 3—figure supplement 1 . All fiber types fail to hypertrophy in myomaker scKO muscle after MOV . Representative plantaris sections immunostained with antibodies for laminin , myh7 ( Type I myosin , red ) , myh2 ( Type IIa , yellow ) , and myh4 ( Type IIb , green ) . Type IIx fibers are unstained . Quantification of fiber type cross-sectional area at the plantaris mid-belly reveals a significant increase in Types IIa , IIb , and IIx myofiber size in control mice after overload , whereas myofiber hypertrophy in each of these fiber types is significantly inhibited with satellite cell ablation of myomaker ( n = 28–54 fibers ( Type I ) or 500–1 , 900 fibers ( Types IIa , IIb , IIx ) from 3–5 independent mice ) . Frequency distribution analysis reveals that fiber type composition remains relatively unaltered after 14 days of MOV ( n = 1100–1 , 700 fibers from three independent mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 010 Furthermore , the total number of fibers is elevated in control MOV compared to control sham mice , but fiber number is similar between sham and overloaded muscles in myomaker scKO mice ( Figure 3D ) . These data suggest an inability of myomaker-deficient MPs to fuse to each other to generate de novo myofibers , indicating a distinct requirement for myomaker in augmenting fiber number during the response to overload . The blunted hypertrophic phenotype observed in myomaker scKO mice can therefore be attributed to an impairment in hypertrophy of existing myofibers and a failure to generate de novo fibers following overload . Collectively , our findings demonstrate the requirement of satellite cell-derived myomaker for load-induced adult skeletal muscle hypertrophy . We utilized two independent approaches to determine whether the diminished hypertrophic response to overload in myomaker scKO mice is associated with deficits in MP fusion . First , mice were administered daily dosages of BrdU post-surgery throughout the entire duration of MOV while maintaining the tamoxifen regimen ( Figure 4A ) . As described above , BrdU labels proliferating cells ( including MPs ) and fusion was assayed by tracking the incorporation of a BrdU+ nucleus into a dystrophin+ myofiber . After 14 days of overload in control mice , BrdU+ nuclei within a dystrophin-stained myofiber were readily detected ( Figure 4B ) . In comparison , BrdU+ nuclei reside primarily in the interstitium between myofibers in overloaded myomaker scKO mice , indicating efficient abrogation of fusion ( Figure 4B ) . Indeed , quantification of the percentage of BrdU+ myofibers revealed ~30% of myofibers in control mice underwent a fusion event , however in myomaker scKO mice BrdU+ myofibers are at basal levels observed in shams ( Figure 4B ) . Given that this is one level of a muscle , if a BrdU myonucleus was above or below this plane then that fiber would not be counted as a positive fusion event . Therefore , this BrdU analysis may underestimate the extent of fusion that occurs during MOV . 10 . 7554/eLife . 20007 . 011Figure 4 . Myomaker is required for fusion during overload-induced muscle hypertrophy . ( A ) Schematic showing treatment with tamoxifen and BrdU . ( B ) Representative images of BrdU+ nuclei within a dystrophin-stained myofiber ( arrows ) indicating fusion . In myomaker scKO mice , BrdU+ nuclei were observed only in the interstitium between myofibers . Quantification of the percentage of BrdU+ myofibers shows that 30% of fibers undergo fusion in control samples after MOV while this is dramatically reduced in myomaker scKO mice ( n > 300 fibers from three independent mice ) . ( C ) Tmem8cLacZ/loxP; Pax7CreERT2/+ were crossed with RosamTmG mice to genetically assay fusion . Tamoxifen-treated Tmem8c+/+; Pax7CreERT2/+; RosamTmG mice were used as controls . Upon tamoxifen treatment , SCs will be converted from mTomato-expressing to mGFP-expressing . mGFP+ myofibers indicates fusion with MPs . ( D ) In sham mice , only minimal mGFP+ fibers were observed . Control MOV mice display numerous mGFP+ myofibers , indicating the majority of myofibers undergo a fusion event . Myomaker scKO mice exhibited marginal mGFP+ fibers ( n = 4 mice ) . ( E ) Single fibers were isolated from control and myomaker scKO mice and analyzed for mTomato and mGFP expression . Quantification of mGFP intensity in single fibers shows an increase in control MOV , which is inhibited when myomaker is ablated in satellite cells ( n = 90–120 myofibers from 3–4 independent mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bars: 50 μm , except insets in ( B ) which are 25 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 011 To complement our BrdU assay and establish greater insight into the magnitude of fusion during physiological hypertrophy , we used a fluorescent genetic reporter that offers enhanced sensitivity in detecting fusion events . Specifically , Tmem8c+/+; Pax7CreERT2/+; Rosa26mTmG mice ( controls ) and Tmem8cLacZ/loxP; Pax7CreERT2/+; Rosa26mTmG ( myomaker scKO ) mice were treated with tamoxifen to induce expression of mGFP in MPs ( Figure 4C ) . With this strategy , existing myofibers will maintain mTom expression and also be mGFP+ if they have fused with MPs , while myofibers that are solely mGFP+ would represent de novo myofibers arising from the fusion of MPs with each other . After muscle overload the majority of existing myofibers in control mice undergo fusion with MPs , evaluated by expression of both mTom and mGFP , which was not observed in overloaded myofibers of myomaker scKO mice ( Figure 4D ) . To quantify fusion in this genetic model , we isolated single myofibers and analyzed their relative mGFP intensity normalized to total fiber area using ImageJ ( Figure 4E ) . Here , we detected a significant increase in mGFP intensity of single fibers in control mice after overload , which is reduced to sham levels after genetic ablation of myomaker in MPs ( Figure 4E ) . Our overall findings substantiate that expansive MP fusion occurs during load-induced skeletal muscle hypertrophy , and that myomaker expression on MPs is essential for this myogenic process . Moreover , these results indicate that myomaker-mediated fusion is necessary for myofiber hypertrophy in response to physiological stimuli . An established mechanism for eliciting muscle hypertrophy is through the Akt1 pathway , which activates mTOR and p70s6k and ultimately leads to enhanced protein synthesis ( Egerman and Glass , 2014 ) . We explored how this pathway is perturbed with defective MP fusion during adult muscle hypertrophy . To this end , western blot analysis revealed an expected increase in phosphorylated levels of Akt ( Ser473 ) and p70s6k ( Thr389 ) in control MOV muscles at 7 and 14 days of overload ( Figure 5A ) . Activation of these pathways was significantly blunted in myomaker scKO overload muscle , highlighting a mechanism by which myomaker expression on MPs regulates hypertrophy ( Figure 5A ) . Additionally , total levels of p70s6k were elevated in control MOV muscles compared to myomaker scKO overload muscles , which could be indicative of enhanced overall protein translation in control MOV samples during hypertrophy ( Bakker et al . , 2016 ) . Quantitative analysis of normalized pAkt and p-p70s6k levels further demonstrates significant disruptions in the Akt/mTOR pathway in fusion-incompetent muscles subjected to overload ( Figure 5A ) . 10 . 7554/eLife . 20007 . 012Figure 5 . Hypertrophic signaling pathways are reduced in the absence of fusion during MOV . ( A ) Western analysis for pAkt ( Ser473 ) , total Akt , p-p70S6K ( Thr389 ) , and total p70S6K . Control plantaris muscle displays pAkt and p-p70S6K activation after MOV compared to sham controls , however the pathways are not fully activated in myomaker scKO MOV muscles . Quantification of the western signal obtained for pAkt normalized to total protein , and quantification of the western signal obtained for p-p70S6K normalized to total protein both reveal altered Akt/mTOR signaling in myomaker scKO muscles after MOV ( n = 4 mice ) . ( B ) Analysis of protein synthesis through puromycin incorporation into nascent peptides . Mice were injected with puromycin and sacrificed thirty minutes later . Sections were then immunostained with antibodies to puromycin and myh3 , a marker of newly generated muscle cells . Puromycin staining is increased in both myh3+ myofibers ( de novo , arrowhead ) and myh3- myofibers ( existing , arrow ) in control samples after MOV . Only myh3- myofibers were observed at 14 days of MOV . Quantification revealed significantly more puromycin incorporation in myh3- myofibers in control samples at both 7 and 14 days after MOV compared to myomaker scKO MOV muscles ( 2200–3400 fibers from n = 3–4 independent mice ) . Data are represented as mean ± SEM , *p<0 . 05 . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 012 The ultimate function of mTOR signaling is to activate protein synthesis in myofibers for hypertrophy , therefore we directly investigated rates of protein synthesis after MOV using a technique by which puromycin is incorporated into nascent polypeptides , and is thus a readout of active protein synthesis ( Schmidt et al . , 2009 ) . We injected puromycin into control and myomaker scKO mice thirty minutes prior to sacrifice and assayed for puromycin incorporation into myofibers through immunostaining of muscle sections with a puromycin antibody ( Goodman et al . , 2011 ) . We also co-immunostained these sections with embryonic myosin ( myh3 ) to evaluate the relative contribution of protein synthesis from de novo and existing myofibers . We observed minimal puromycin in sham samples , however in control muscle puromycin was elevated in both myh3- ( existing ) and myh3+ ( de novo ) myofibers 7 days after MOV ( Figure 5B ) . At 14 days of MOV , we detected minimal myh3+ myofibers although puromycin remained elevated in existing fibers of control mice , whereas myomaker scKO muscle did not exhibit significant puromycin incorporation at either time point analyzed ( Figure 5B ) . Of note , we did not observe myh3+ cells in myomaker scKO muscle 7 days after MOV , further demonstrating that myomaker is required to form new myofibers ( Figure 5B ) . Given that myh3+ myofibers are not present in myomaker scKO muscle we only quantified the puromycin intensity in myh3- myofibers , which revealed a significant inability for fusion-incompetent muscle to activate protein synthesis ( Figure 5B ) . Our findings indicate that SC-derived myomaker is required for full activation of Akt/mTOR signaling and protein synthesis in the myofiber , which could serve as a potential mechanism though which fusion drives the hypertrophic response . In addition to the activation of signaling pathways to achieve optimal hypertrophy , muscle hypertrophy is also associated with remodeling of extracellular matrix ( ECM ) content . While the mechanisms that govern ECM alterations during muscle hypertrophy and regeneration are relatively unknown , one aspect of this regulation is dependent on the presence of MPs ( Fry et al . , 2014 , 2015; Murphy et al . , 2011 ) . Thus , we sought to determine the consequence on ECM in fusion-defective muscle that is sensing increased load . We performed Masson’s Trichrome staining on control and myomaker scKO sham and MOV muscle , which revealed modest fibrosis in control MOV , indicating that ECM remodeling is a normal process for hypertrophy ( Figure 6 ) . In contrast , we observed an increase in fibrosis in the overloaded muscles of myomaker scKO mice ( Figure 6 ) . Quantification of the fibrotic area indeed revealed a significant increase in fibrosis in myomaker scKO samples ( Figure 6 ) . Consequently , the enhanced accumulation of ECM in myomaker scKO MOV muscle may explain the increase in muscle weight despite the lack of myofiber hypertrophy ( Figure 3 ) . This non-functional gain in muscle size illustrates that in place of de novo and larger myofibers , there is an increased formation of connective tissue in muscles with non-fused MPs , suggesting a protective role for fusion in stabilizing muscle architecture associated with increased workload . 10 . 7554/eLife . 20007 . 013Figure 6 . Loss of myomaker in satellite cells results in increased fibrosis during physiological hypertrophy . Trichrome stained sections from control and myomaker scKO mice demonstrate a significant increase in fibrosis in myomaker scKO mice after MOV . Quantification of the fibrotic area further demonstrates increased fibrosis in fusion-incompetent muscle . Values indicate mean ( n = 4–12 mice ) ± SEM , *p<0 . 05 . Scale bar: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 20007 . 013
Skeletal muscle hypertrophy is a highly desirable adaptation of various therapeutic strategies to restore strength and function in muscle wasting conditions . Functional gains in adult muscle mass are derived from elevations in protein synthesis within the myofiber , and our findings highlight an additional source for muscle hypertrophy through the contributions of muscle stem cell fusion . Specifically , we use BrdU labeling and genetic lineage tracing approaches to evaluate the extent of fusion during overload-induced hypertrophy . While BrdU analysis revealed approximately 30% of fibers to undergo a fusion event , findings from the fluorescent reporter model demonstrate that the majority of myofibers fuse ( mGFP+ ) , indicating the BrdU strategy is an underestimation of the fusion process during muscle hypertrophy . We also reveal that in response to increased muscle workload , myomaker is primarily upregulated in activated SCs , which is required for fusion with a myofiber . In the absence of SC-derived myomaker , the activation of pro-hypertrophic signaling pathways and protein synthesis , and increases in myofiber size and number are diminished . Thus , myomaker-mediated fusion is required for load-induced hypertrophy of skeletal muscle . The requirement of muscle stem cells during adult skeletal muscle hypertrophy has long been a source of debate . Indeed , blunting of hypertrophy was observed after elimination of satellite cells through gamma irradiation ( Adams et al . , 2002 ) or through genetic manipulation of IL-6-dependent effects on SC/MP activity ( Guerci et al . , 2012; Serrano et al . , 2008 ) . However , two independent groups utilized DTA-mediated ablation of SCs and produced widely contrasting results . The initial SC DTA experiments indicated that SCs are not necessary for hypertrophy , where CSA was increased 10% in control mice with normal SC activity . ( Fry et al . , 2014; McCarthy et al . , 2011 ) . In contrast , a recent report also using ablation of SCs through DTA provides evidence that muscles fail to undergo overload-induced hypertrophy in the absence of SCs , and this study reported a 26% increase of CSA in control mice ( Egner et al . , 2016 ) . While the interpretation of these two studies are vastly different , the magnitude of hypertrophy achieved , which is best assessed through CSA analysis and not gross muscle weights , is not similar and may account for the divergent outcomes . The magnitude of CSA increase in our study was 29% in control mice after MOV , which is more similar to the work by Egner et al . We also observed an approximate 10% CSA increase in myomaker scKO MOV samples compared to myomaker scKO sham . Although this increase was not significant it suggests that myofibers harbor a minor hypertrophic potential in the absence of fusion consistent with Fry et al . Through generation of fusion-defective SCs however , we provide independent evidence for the necessity of SCs for any major muscle hypertrophy . While our findings demonstrate that fusion during muscle overload is inhibited by loss of myomaker in SCs , the exact contribution of fusion for hypertrophy remains to be elucidated . During compensatory hypertrophy , newly acquired myonuclei precede muscle hypertrophy ( Bruusgaard et al . , 2010 ) . Thus , fusion may provide new nuclei to facilitate enhanced transcription of contractile proteins necessary for hypertrophy , as supported by our findings that myofibers are not able to activate protein synthesis in the absence of fusion . Addition of a new nucleus to the myofiber may also function to maintain a constant myonuclear domain ( Moss and Leblond , 1970 ) , however this concept has been challenged especially during postnatal muscle development ( White et al . , 2010 ) . Alternatively , fusion may not only provide new nuclei but also other factors necessary for hypertrophy , such as ribosomes for protein translation , or a bolus of signaling molecules to activate hypertrophic pathways similar to the activation of Akt by Wnt7a ( von Maltzahn et al . , 2012 ) . These concepts clearly require more extensive work to reveal definitive mechanisms through which SC fusion drives adult skeletal muscle hypertrophy . The Akt/mTOR pathway is controlled by insulin-like hypertrophy factor 1 ( IGF-1 ) and myostatin , a member of the transforming growth factor-β ( TGF-β ) . IGF-1 directly activates the Akt pathway ( Rommel et al . , 2001 ) while myostatin is a negative regulator of muscle mass through SMAD-dependent inhibition of Akt ( Sartori et al . , 2009; Trendelenburg et al . , 2009 ) . The inability for muscle to initiate Akt/mTOR signaling without myomaker expression in myogenic progenitors was surprising since previous studies have shown that myofibers have the ability to intrinsically activate this canonical protein synthesis pathway ( Bodine et al . , 2001; Lai et al . , 2004 ) . Indeed , SC activation is not observed during muscle hypertrophy induced through a constitutively active Akt transgene ( Blaauw et al . , 2009 ) , or inhibition of myostatin action by pharmacological blockade of its receptor ACVR2B ( Lee et al . , 2012 ) . It is conceivable that during overload-induced hypertrophy , the increased tension placed on the myofiber requires SC-mediated activity for hypertrophy , whereas there is an absence of mechanical strain on the myofiber through treatment with IGF-1 or ACVR2B antibody . The ultimate goal of muscle hypertrophy is to increase contractile capability to deal with heightened physiological demands . Hypertrophy associated with loss of myostatin may not yield long-term improvements in muscle function suggesting that addition of myonuclei are necessary for functional hypertrophic gains ( Gundersen , 2016 ) . We did not assess muscle contractile function in our fusion-defective mice after MOV , however this will be important to fully elucidate the physiological relevance for MP fusion during muscle hypertrophy . Our finding of increased fibrosis in muscles of myomaker scKO mice after overload further suggests a protective role for SC fusion on myofiber stability in response to mechanical strain , however the cellular circuitry driving fibrosis development remains unknown . DTA-mediated deletion of SCs also results in increased ECM components during overload-induced hypertrophy and aging , potentially through lack of interactions between SCs and fibroblasts ( Fry et al . , 2014; Judson et al . , 2013; Murphy et al . , 2011 ) . Indeed , recent work suggests that the main function of MPs during overload-induced hypertrophy is to release miR-206-containing exosomes that dampen collagen production in fibroblasts , and that loss of MPs facilitates development of fibrosis ( Fry et al . , 2017 ) . After overload of myomaker scKO mice , MPs are still present in muscle , suggesting the major reason for fibrosis can be attributed to an inability to maintain myofiber integrity due to loss of SC activity . It is formally possible that MPs lacking myomaker are not only fusion-defective but also lack an ability to communicate with fibroblasts through paracrine signaling . However , given that no evidence exists supporting this possibility , and that we observed significant fusion during MOV , our data are more consistent with a model where fibrosis occurs in myomaker scKO MOV muscles due to an inability to repair myofiber mechanical strain . This model of myofiber mechanical strain eliciting fibrosis is not unlike what occurs in other contexts , such as in cardiac muscle subjected to pressure overload where fibrosis functions to maintain tissue architecture ( Creemers and Pinto , 2011; Teekakirikul et al . , 2012 ) . One intriguing question to arise from this study pertains to the fate of fusion-incompetent MPs . Compared to control , we observed an increase in the number of LacZ+ cells ( MPs ) in myomaker scKO muscle at day 7 after MOV . These data could indicate an accumulation of MPs because they have not fused . However , 10 days after MOV the number of LacZ+ cells was comparable between control and myomaker scKO muscle , suggesting that accumulated LacZ+ fusion-defective cells die between day 7 and day 10 of MOV or that myomaker is down-regulated and they become LacZ- . Analysis of MPs using a genetic reporter ( mGFP ) found no dramatic difference in the percentage of mGFP+ cells between control and myomaker scKO MOV samples . These inconsistencies highlight the current lack of knowledge regarding fusion-defective MPs . It is possible that the LacZ+ cells represent a small fraction of MPs and not enough to shift the percentage of mGFP+ MPs analyzed through flow cytometry . Also , since mGFP+ MPs are expressed as a percentage of the total cell population , changes in non-MP populations could indirectly alter the mGFP+ percentage . Nonetheless , the behavior and fate of fusion-incompetent MPs requires future investigation . In conclusion , we have shown that myomaker is activated on SCs allowing fusion with a myofiber during muscle hypertrophy , thereby promoting hypertrophic signaling pathways and minimizing fibrosis associated with mechanical strain . These findings modify current theories for the role of muscle stem cells during hypertrophy and may lead to new intervention strategies for muscle wasting diseases .
The generation of myomaker mouse strains used in this study was described previously ( Millay et al . , 2014 ) . Briefly , Tmem8cLacZ/+ mice contain a LacZ cassette in intron 1 of the myomaker locus . Tmem8cloxP/+ mice contain loxP sites that flank exon 2 . Tmem8cLacZ/loxP mice were then bred with mice carrying the satellite cell-specific ( Pax7CreERT2 ) ( Lepper et al . , 2009 ) or myofiber-specific ( ACTA1CreERT2 ) ( Schuler et al . , 2005 ) Cre recombinase . To assess fusion , Tmem8c+/+; Pax7CreERT2/+ mice ( control ) and Tmem8cLacZ/loxP; Pax7CreERT2/+ mice ( myomaker scKO ) were crossed with Rosa26mTmG mice ( Jackson Laboratory # 007676 ) . Experimental mice were at least 3 months of age prior to any procedure performed . All animal procedures were approved by Cincinnati Children’s Hospital Medical Center’s Institutional Animal Care and Use Committee . Tamoxifen ( TAM; Sigma-Aldrich ) was mixed in corn oil with 10% ethanol at 25 mg/ml . Prior to muscle overload surgery , 3-month old mice were administered five daily doses of TAM at 0 . 075 mg/g/day through intraperitoneal injections . Following surgery , mice were maintained on a TAM regimen ( with the exception of mice used in Figure 2C ) throughout MOV and administered 10 μl/g BrdU ( Thermo Fisher Scientific ) as indicated . No statistical differences in body weights were observed after tamoxifen treatment in either control or myomakerscKO mice . Muscle overload of the plantaris muscles was achieved through bilateral synergistic ablation of soleus and gastrocnemius muscles as described by others ( Dearth et al . , 2013 ) . Briefly , the soleus and gastrocnemius muscles were exposed by making an incision on the posterior-lateral aspect of the lower limb . The distal and proximal tendons of the soleus were cut followed by cutting of the distal gastrocnemius tendon and excision of 75% lateral and medial gastrocnemius . Plantaris muscles were dissected , blotted dry , and weighed . Hindlimbs were dissected and dissolved overnight in lysis buffer containing 0 . 4 mg/ml proteinase K at 55°C . Tibia length was assessed by a digital caliper , and the average of three measurements was recorded . For cryosections , muscles were embedded in 1% tragacanth/PBS ( Sigma ) , and frozen in liquid nitrogen-cooled 2-methylbutane . Sections were then cut at 10 μm . For gene and protein analysis , mice were fasted for four hours , and muscles were flash-frozen in liquid nitrogen and stored at −80°C . For Tmem8cLacZ/loxp; Pax7CreERT2/+; Rosa26mTmG mice , plantaris muscles were fixed in 4% PFA/PBS for 5 hr at 4°C and sucrose protected overnight ( 30% sucrose/PBS ) followed by embedding and cryosectioning . Cryosections were stained with X-gal as described previously ( Millay et al . , 2014 ) . After staining , the samples were rinsed in PBS and fixed in 4% PFA/PBS for at least 10 min . Tissue sections were co-stained with light eosin ( 0 . 7% ) , dehydrated , and mounted with Permount ( Fisher ) . Samples were visualized on an inverted bright-field microscope ( Olympus BX51 ) , and the numbers of LacZ+ non-myofibers , LacZ+ myofibers , and all myofibers in a given field of view were manually counted . Data presented is representative of the average number of LacZ+ non-myofibers , and the percentage of LacZ-expressing myofibers , quantified from several non-overlapping fields of view . We isolated skeletal muscle stem cells using techniques described previously ( Liu et al . , 2015 ) . Briefly , plantaris muscles from either tamoxifen-treated Tmem8c+/+; Pax7CreERT2/+; Rosa26mTmG or Tmem8cLacZ/loxp; Pax7CreERT2/+; Rosa26mTmG mice were collected and minced in 10% horse serum ( HS ) ( Gibco # 26050–070 ) , and then incubated in 800 U/ml Collagenase Type 2 ( Worthington # CLS-2 ) solution at 37°C with gentle agitation for 1 hr . Following centrifugation , pellets were resuspended in 10% HS with 1000 U/ml Collagenase Type II and 4 . 8 U/ml Dispase ( Roche # 4942078001 ) , and incubated at 37°C with gentle agitation for another 30 min . Following this second round of incubation , samples were triturated with a 20-gauge needle , centrifuged and resuspended in 10% HS . Cell suspensions were subsequently filtered through a 40 μm nylon cell strainer ( Corning # 352340 ) , centrifuged , and resuspended in 2% Fetal Bovine Serum/PBS ( Hyclone # SH30071 ) . Flow cytometry analysis on cell suspensions was performed with a BD Biosciences LSR II Flow Cytometer configured with the 488 nm laser for GFP and the 561 nm laser for Tomato . Voltages were determined using cell suspensions from vehicle-treated Tmem8c+/+; Pax7CreERT2/+ and Tmem8c+/+; Pax7CreERT2/+; Rosa26mTmG mice . Analysis was performed using FACSDiva software . Cryosections were fixed in 1% PFA/PBS; permeabilized with 0 . 2% Triton X-100/PBS or denatured with 2 M HCL in 0 . 5%Triton X-100/PBS and subsequently neutralized with TBS ( pH 8 . 4 ) for BrdU labeling . After neutralization , sections were blocked with 1% BSA , 1% heat-inactivated goat serum , and 0 . 025% Tween20/PBS; and incubated with primary antibody overnight at 4°C . Following 1 hr of incubation with a secondary Alexa Fluor antibody ( 1:200 ) ( Invitrogen ) , slides were mounted with VectaShield containing DAPI ( Vector Laboratories ) . For assessment of hypertrophy , samples were stained with anti-laminin ( 1:100; Sigma-Aldrich # L9393 ) , and representative non-overlapping images from all sections were captured at 20x on a Nikon A1R confocal system . The cross-sectional area ( CSA ) of myofibers was quantified through a binary thresholding algorithm developed in Imaris ( Bitplane ) . Whole section images were also taken at 10x , and the total number of myofibers per muscle section was determined through an automated image segmentation program from ImageJ . To assay for fusion , samples were co-stained with anti-BrdU ( 1:20; Roche #11170376001 ) and anti-dystrophin ( 1:100; Abcam #ab15277 ) . The following antibodies from Developmental Studies Hybridoma Bank were also used: Pax7 ( 1:50 ) , myh7 ( 1:100; clone BA-D5 ) , myh2 ( 1:100; clone SC-71 ) , myh4 ( 1:10; clone BF-F3 ) , and myh3 ( 1:10; clone F1 . 652 ) . For Pax7 immunostaining , the above protocol was used in addition to antigen retrieval ( 1x Antigen Retrieval Citra Solution ( Biogene # HK086-9K ) and boiled for 30 min ) , followed by incubation with Pax 7antibodies overnight . The following day , 1:200 M . O . M . biotinylated anti-mouse IgG ( Vecta Laboratories ) was incubated for 30 min , washed , and then 1:200 Strepdavidin-Alexa Fluor 488 ( Invitrogen ) was incubated for 30 min . All samples were visualized at 40x with a Nikon A1R+ LUNV on a Ti-E Inverted Microscope unless otherwise noted . For mTom/mGFP analysis , sucrose-protected samples were visualized at 10x . The authenticity of mGFP signal was verified by establishing baseline levels for auto-fluorescence in vehicle-treated mice ( GFP- ) . Masson’s Trichrome staining was performed using standard protocols , and total fibrotic area within a section was quantified with a threshold intensity program from ImageJ . To quantify overall mGFP+ expression , single fibers were isolated from plantaris muscles of sham and overloaded mice , as described previously ( Kanisicak et al . , 2009 ) . Whole plantaris muscles were incubated in 0 . 2% Type I collagenase ( Sigma-Aldrich , # C-0130 ) for 1 hr at 37°C . Muscles were transferred to DMEM ( Hyclone ) containing 10% horse serum and triturated to release individual fibers . Fibers were then fixed in 4% PFA/PBS for 30 min at 4°C , washed three times in PBS , permeabilized and mounted with Vectashield containing DAPI , and visualized at 10x on a Nikon A1R confocal system . Identical settings were utilized to capture mGFP and mTomato fluorescent images from all fibers and then processed in ImageJ . Fibers isolated from vehicle-treated mice were used to establish an impartial baseline for auto-fluorescence , and the threshold for GFP intensity was set at a constant value . The absolute GFP intensity for individual experimental fibers was then determined by tabulating the integrated density of saturated pixels above threshold levels . Similarly , tomato expression was used to quantify myofiber area by calculating the total pixelated area above background . GFP intensity was then normalized to myofiber area . Changes in protein synthesis in whole muscle cross-sections were detected through utilization of antibiotic puromycin , a non-radioactive technique known as surface sensing of translation ( SUnSET ) ( Goodman et al . , 2011; Schmidt et al . , 2009 ) . Following 4 hr of fasting , mice were administered 0 . 04 μmol puromycin/gram of body weight dissolved in PBS intraperitoneally for 30 min , and plantaris muscles subsequently collected for cryosections . Sections were immunostained with anti-puromycin ( 1:1000; EMD Millipore; clone 12D10 ) , and anti-embryonic myosin ( myh3 ) . For the purpose of defining the myofiber borders we also stained the sections with anti-laminin . Representative images were captured at 20x , and a binary algorithm developed in Imaris was used to calculate the mean puromycin intensity within a laminin-stained myofiber . Samples were automatically distinguished as myh3+ ( de novo ) or myh3- ( existing ) myofibers within the same image , and puromycin intensities for individual myofibers in each group were quantified separately . Total RNA was extracted from plantaris muscle with Trizol ( Invitrogen ) , and cDNA was synthesized using MultiScribeTM reverse transcriptase with random hexamer primers ( Applied Biosystems ) . Gene expression was assessed using standard qPCR approaches with PowerUpTM SYBR Green Master Mix ( Applied Biosystems ) . Analysis was performed on a 7900HT fast real-time PCR machine ( Applied Biosystems ) with the following myomaker SYBR primers: Forward , 5’-ATCGCTACCAAGAGGCGTT-3’; and Reverse , 5’-CACAGCACAGACAAACCAGG-3’ . GAPDH primers ( Forward , 5’-TGCGACTTCAACAGCAACTC-3’; and Reverse , 5’-GCCTCTCTTGCTCAGTGTCC-3' ) were used as internal controls . A myomaker custom antibody was generated through YenZym Antibodies LLC . Rabbits were immunized with amino acids #137–152 of mouse myomaker ( MKEKKGLYPDKSIYTQ ) after conjugation to KLH . We used antigen-specific affinity purified products for western analysis . Plantaris muscles were homogenized with a bead homogenizer ( TissueLyser II; Qiagen ) in a lysis buffer [10 mM Tris ( pH 7 . 4 ) , 1 mM EDTA , 1 mM dithiothreitol , 0 . 5% Triton X-100 , 2 . 1 mg/ml NaF] containing protease and phosphatase inhibitor cocktails ( 5 μl/ml; Sigma-Aldrich ) , and following centrifugation , the amount of protein in supernatants was determined through Bradford protein assay . Muscle homogenates were heated at 65°C for 30 min , separated on 10% ( 30 μg of protein for detection of signaling proteins along the Akt/mTOR pathway ) or 12% ( 100 μg of protein for myomaker detection ) SDS-page gels , and transferred to PVDF-FL membranes ( Immobilon ) . Membranes were subsequently blocked in 5% dry milk or 5% BSA in Tris-buffered saline ( TBS ) -Tween , and incubated overnight at 4°C with an antibody against myomaker ( 1:200 ) , phosphorylated Akt ( Ser473 ) ( 1:750; Cell Signaling # 9271 ) , total Akt ( 1:750; Cell Signaling # 9272 ) , phosphorylated p70s6k ( Thr389 ) ( 1:500; Santa Cruz # SC-11759 ) , or total p70s6k ( 1:750; Cell Signaling # 2708 ) . GAPDH ( 1:5000; Millipore ) served as a control for sample loading . Membranes were then washed and incubated with an IRDye 800CW anti-rabbit secondary antibody ( 1:5000; LI-COR Biosciences ) . The relative abundance of phosphorylated and total protein levels of Akt and p70s6k was quantified using the Odyssey infrared detection system ( LI-COR Biosciences ) . Quantitative data sets are presented as means ± SEM , and analyzed with an unpaired t-test ( Figures 1D and 2A ) , a one-way analysis of variance ( ANOVA ) ( Figures 1A , C and 2C ) , or a two-way ANOVA ( all other data sets ) using GraphPad Prism 6 software . For assessments of myofiber CSA , total fiber number , BrdU incorporation , single fiber fluorescent intensity , and levels of protein synthesis in myofibers both image captures and data analyses were performed in a blinded fashion . Statistical significance was set at a p value <0 . 05 . | Skeletal muscle has a remarkable capacity to adapt to a variety of stimuli , including an ability to become larger and stronger through exercise . In embryos , new muscles develop from muscle stem cells , which either replicate themselves or “differentiate” into mature muscle cells . Adult muscles also contain stem cells , which are normally dormant , but activate when the muscle is damaged . The stem cells subsequently differentiate and fuse with one another or to existing muscle fibers to restore the muscle . What is not fully understood is whether this fusion process also helps undamaged adult muscles to increase in size ( for example , in response to exercise ) . Fusion proteins such as myomaker – which specifically acts in muscles – help the stem cells to fuse . To investigate myomaker’s role in adult muscle growth , Goh and Millay deleted the gene that produces it from the muscle stem cells of mice . The mice then experienced two weeks of increased muscle activity , after which their muscle growth was compared with that of normal mice that had been subjected to the same activity routine . Goh and Millay discovered that myomaker is important in muscle stem cells , and not in muscle fibers , for adult muscle growth . After two weeks of increased muscle activity , substantial levels of muscle stem cell fusion had occurred in normal mice , and their muscles had grown significantly . However , the muscles of mice that lacked myomaker in their muscle stem cells did not increase in size . Additional experiments showed that normal muscle stem cell fusion activates signaling pathways that create new proteins and drive muscle growth . Furthermore , scarring occurred in muscles that lacked myomaker , suggesting that stem cell fusion also protects muscle fibers from damage during increased activity . Overall , the findings presented by Goh and Millay reveal that the fusion of muscle stem cells is an important event for adult muscle growth . Further studies are now needed to determine the relevance of muscle stem cell fusion during the normal aging process , and to uncover the relationship between fusion and the activation of pro-growth signaling pathways . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"cell",
"biology"
] | 2017 | Requirement of myomaker-mediated stem cell fusion for skeletal muscle hypertrophy |
Socially-conveyed rules and instructions strongly shape expectations and emotions . Yet most neuroscientific studies of learning consider reinforcement history alone , irrespective of knowledge acquired through other means . We examined fear conditioning and reversal in humans to test whether instructed knowledge modulates the neural mechanisms of feedback-driven learning . One group was informed about contingencies and reversals . A second group learned only from reinforcement . We combined quantitative models with functional magnetic resonance imaging and found that instructions induced dissociations in the neural systems of aversive learning . Responses in striatum and orbitofrontal cortex updated with instructions and correlated with prefrontal responses to instructions . Amygdala responses were influenced by reinforcement similarly in both groups and did not update with instructions . Results extend work on instructed reward learning and reveal novel dissociations that have not been observed with punishments or rewards . Findings support theories of specialized threat-detection and may have implications for fear maintenance in anxiety .
Neuroscientists have built a rich understanding of the brain systems that govern learning from punishments and rewards , which seem to be largely conserved across species ( Balleine and O'Doherty , 2010; Haber and Knutson , 2010 ) . Yet humans possess a distinctive capacity to learn from socially-conveyed rules and instructions , which is an important example of a broader capacity , shared by many species , of learning from events other than simple reinforcers ( Tolman , 1949 ) . Most of us do not need to get burnt to avoid putting our hands on a hot stove: A verbal warning serves as a sufficient threat . How does instructed knowledge influence our subsequent responses to ongoing threats in the environment ? We tested whether dynamic feedback-driven aversive learning is modulated when individuals are informed about contingencies in the environment . It is unknown whether instructions integrate with the systems that support error-driven learning , as most reinforcement learning models fail to account for the potential influence of explicit information , despite the fact that verbal information shapes responses across nearly all domains , including aversive learning ( Wilson , 1968; Phelps et al . , 2001; Costa et al . , 2015; Mertens and De Houwer , 2016 ) . We sought to characterize computationally the contribution of instructed knowledge to dynamic aversive learning in humans . Our aim was to determine whether instructed knowledge influences the neural mechanisms of aversive learning , or whether separate neural systems process feedback-driven and instructed knowledge . Recent studies in the appetitive domain suggest that instructions about rewarding outcomes modulate learning-related responses in the striatum ( Doll et al . , 2009; 2011; Li et al . , 2011a ) and ventromedial prefrontal cortex ( Li et al . , 2011a ) and that this modulation might depend on the prefrontal cortex ( Doll et al . , 2009; 2011; Li et al . , 2011a ) . In the aversive domain , such interactions , by which instructed knowledge might help to overcome learned expectations of threat , are of particular importance due to their relevance for anxiety and post-traumatic stress disorder . However , no studies have tested whether instructions have the same effects on dynamic aversive learning , which is known to depend on the amygdala ( Maren , 2001 ) but also involves the striatum ( Seymour et al . , 2004; 2005; Delgado et al . , 2008b ) and ventromedial/orbitofrontal cortex ( VMPFC/OFC ) ( Phelps et al . , 2004; Kalisch et al . , 2006; Schiller et al . , 2008 ) . Instructions might modulate learning in the amygdala as well as the striatum and VMPFC/OFC , or amygdala responses might be insensitive to cognitive instruction ( Ohman and Mineka , 2001 ) , as suggested by theories of automatic threat detection in the amygdala ( Ohman , 2005 ) . All participants performed a Pavlovian aversive learning task in which one image , the Original conditioned stimulus ( CS+ ) , was paired with mild electric shock ( the unconditioned stimulus [US] ) on 30% of trials , and a second image , the Original CS- , was not paired with shock . Contingencies reversed three times . Participants assigned to an Instructed Group were informed about initial contingencies and instructed upon reversal ( Figure 1 ) , whereas participants in an Uninstructed Group learned through reinforcement alone . Models were fit to skin conductance responses ( SCRs ) , a traditional measure of the conditioned fear response in humans . We combined quantitative modeling of behavior with functional magnetic resonance imaging ( fMRI ) to examine how instructions influence brain responses and skin conductance responses ( SCRs ) during fear conditioning . We evaluated quantitative learning models , fit to SCR , and confirmed model conclusions with task-based fMRI analyses . We focused on responses in the amygdala , striatum , and VMPFC/OFC . We hypothesized that instructions about contingencies would modify learning-related signals and brain responses during fear conditioning , and that , as in the appetitive domain , this modulation would involve the prefrontal cortex . 10 . 7554/eLife . 15192 . 003Figure 1 . Experimental design . ( A ) Prior to the conditioning phase of the experiment , participants in the Instructed Group saw each image and were informed about initial probabilities . Participants in the Uninstructed Group also saw the images prior to the experiment , but were not told about contingencies . ( B ) Participants in both groups underwent a Pavlovian fear conditioning task with serial reversals . There were three reversals across the duration of the task , leading to four continuous blocks of twenty trials . In each block , one image ( the conditioned stimulus , or CS+ ) was paired with a shock ( the unconditioned stimulus , or US ) 30% of the time , leading to 4 reinforced trials and 8 unreinforced trials , whereas a second image ( the CS- ) was never paired with a shock . Images were presented for 4 s , followed by a 12-second inter-stimulus interval . ( C ) Upon each reversal , the Instructed Group was informed that contingencies had reversed . Button presses were included to ensure participants were paying attention to the instructions but had no effect on the task itself or task timing . Instructions were always immediately followed by at least two unreinforced presentations of each CS before the new CS+ was paired with a shock . The figure presents one of two pseudorandom trial orders used during the experiment ( see Materials and methods ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 003
Upon completion of the experiment , participants reported the number of perceived contingency reversals and retrospectively rated shock expectancy and affect in response to each image . Participants recognized that three reversals had occurred ( M = 3 . 22 , SD = 0 . 79 ) , and there was no group difference in the estimated number of reversals ( p>0 . 1 ) . Two-way ANOVAs revealed a main effect of Group on reported affect ( F ( 1 , 68 ) = 8 . 57 , p<0 . 01 ) , such that the Uninstructed Group reported less positive affect for both stimuli . There were no Group differences in shock expectancy ratings , nor were there any main effects of Stimulus ( Original CS+ vs Original CS- ) or Group x Stimulus interactions on either outcome measure . We tested whether participants showed differential SCRs to unreinforced CS presentations during fear acquisition ( i . e . larger responses to the Original CS+ than Original CS- ) , and whether responses were modulated after participants were instructed that contingencies had reversed ( Instructed Group ) or after they received a shock paired with the new CS+/ previous CS- ( Uninstructed Group ) . As reported in Table 1 , both groups showed differential responses that reversed in response to contingency changes throughout the task ( CS+ > CS-; ß = 0 . 04 , t = 5 . 82 , p<0 . 0001 ) . Differential responses were larger in the Instructed Group than the Uninstructed Group ( ß = 0 . 03 , t = 3 . 98 , p=0 . 0002 ) , and SCRs habituated over time ( ß = -0 . 07 , t = 10 . 59 , p<0 . 0001 ) . 10 . 7554/eLife . 15192 . 004Table 1 . Group differences in differential SCRa . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 004AnalysisModelInterceptStimulus ( Original CS+ > Original CS- ) Reversal effect ( Original contingencies vs . Reversed contingencies ) Stimulus x Reversal Interaction ( Current CS+ > Current CS- ) TimeAll participants ( n = 68 ) Within-subjects effects , controlling for group ( first level ) ß = 0 . 22n . s . ( p = 0 . 06 ) n . s . ( p = 0 . 445 ) ß = 0 . 04ß = - . 07t = 12 . 37t = 5 . 82t = 10 . 59p <0 . 0001p<0 . 0001p<0 . 0001Effect of group ( second level ) n . s . ( p = 0 . 11 ) n . s . ( p = 0 . 15 ) n . s . ( p = 1 . 0 ) ß = 0 . 03n . s . ( p = 0 . 09 ) t = 3 . 98p = 0 . 0002Learners only ( n = 40 ) Within-subjects effects , controlling for group ( first level ) ß = 0 . 27ß = 0 . 01n . s . ( p = 0 . 95 ) ß = 0 . 06ß = - . 08t = 11 . 62t = 2 . 79t = 6 . 5t = -11 . 18p<0 . 0001p = 0 . 0082p<0 . 0001p<0 . 0001Effect of group ( second level ) n . s . ( p = 0 . 13 ) n . s . ( p = 0 . 20 ) n . s . ( p = 0 . 35 ) ß = 0 . 04ß = - . 02t = 3 . 67t = -2 . 53p = 0 . 0007p = 0 . 0157Instructed Group learners ( n = 20 ) Entire taskß = 0 . 30n . s . ( p = 0 . 30 ) n . s . ( p = 0 . 55 ) ß = 0 . 10ß = -0 . 09t = 8 . 93t = 5 . 35t = -9 . 26p<0 . 0001p<0 . 0001p<0 . 0001Entire task , second half of each runß = 0 . 27ß = 0 . 02ß = 0 . 04ß = 0 . 12n . s . ( p = 0 . 69 ) t = 7 . 82t = 2 . 15t = 4 . 18t = 6 . 19p<0 . 0001p = 0 . 0314p<0 . 0001p<0 . 0001Trials following the first reversalß = 0 . 28n . s . ( p = 0 . 65 ) ß = -0 . 16ß = 0 . 09ß = -0 . 07t = 6 . 89t = -2 . 36t = 4 . 19t = -5 . 05p<0 . 0001p = 0 . 0184p<0 . 0001p<0 . 0001Trials following the first reversal , second half of each runß = 0 . 22n . s . ( p = 0 . 81 ) n . s . ( p = 0 . 30 ) ß = 0 . 09n . s . ( p = 0 . 09 ) t = 5 . 67t = 4 . 49p<0 . 0001p<0 . 0001Uninstructed Group learners ( n=20 ) Entire taskß = 0 . 23ß = 0 . 01n . s . ( p = 0 . 54 ) ß = 0 . 03ß = - . 06t = 7 . 30t = 2 . 09t = 4 . 58t = -6 . 10p<0 . 0001p = 0 . 0368p<0 . 0001p<0 . 0001Entire task , second half of each runß = 0 . 22ß = 0 . 02ß = 0 . 02ß = 0 . 04n . s . ( p = 0 . 678 ) t = 6 . 86t = 3 . 18t = 2 . 44t = 5 . 45p<0 . 0001p = 0 . 0015p = 0 . 0149p<0 . 0001Trials following the first reversalß = 0 . 21n . s . ( p = 0 . 58 ) n . s . ( p = 0 . 49 ) ß = 0 . 01ß = - . 04t = 5 . 28t = 1 . 67t = -2 . 49p<0 . 0001p = 0 . 0967p = 0 . 0127Trials following the first reversal , second half of each runß = 0 . 19n . s . ( p = 0 . 65 ) n . s . ( p = 0 . 21 ) ß = 0 . 02ß = 0 . 03t = 5 . 83t = 2 . 04t = 2 . 88p<0 . 0001p = 0 . 0413p = 0 . 004aThis table presents results of linear mixed models that included normalized skin conductance response ( SCR ) as a dependent measure . In the Instructed Group , contingencies and reversals are coded relative to instructed reversal . In the Uninstructed Group , contingencies and reversals are coded relative to reinforcement ( i . e . reversals occur when the previous CS- is paired with a shock ) . Within groups , we analyzed SCRs across the entire task as well as following the first reversal ( without the acquisition phase ) . We also examined responses within the second half of each run , as well as across all trials ( including trials that immediately followed reversals ) . These results reflect group-level effects and group differences across all participants . However , a subset of participants did not show differential SCRs prior to the first reversal . As our primary research question concerns the effects of instructions on the neural systems of aversive reversal learning , the strongest tests are in those individuals who learn contingencies prior to the first reversal . Given this , we restricted our analyses to 'learners . ' As described in Materials and methods , we defined learners as those individuals who showed greater SCR to the CS+ relative to the CS- in late acquisition ( the second half of the first run; 20/30 Instructed Group participants , 20/38 Uninstructed Group participants ) . In this subset of learners , the main effects and interactions on SCR reported above remained significant and increased in magnitude ( Figure 2A , C; Table 1 ) . Importantly , both the Instructed and Uninstructed Groups showed SCRs that were responsive to changing contingencies when we examined each group separately , and when we restricted analyses to trials that followed the first reversal ( Table 1 ) , suggesting that effects were not driven entirely by initial learning . The Uninstructed Group showed significant reversals when we analyzed trials from the second half of each block in this post-acquisition analysis , and marginal effects when we included all post-acquisition trials ( see Table 1 ) . This is expected given the low reinforcement rate ( 33% ) that we used in this study , and is consistent with somewhat slow learning . The quantitative models and fMRI analyses reported below focus on the 20 learners in each group . Combined fMRI analyses that include data from the full sample are entirely consistent with these findings and are reported in figure supplements and source data . 10 . 7554/eLife . 15192 . 005Figure 2 . Effects of instructions on skin conductance responses ( SCR ) and aversive learning . Mean normalized skin conductance responses ( SCRs ) as a function of group and condition . Both groups showed significant reversals of SCR responses throughout the task ( p<0 . 001; Table 1 ) , and effects were larger in the Instructed Group ( see Table 1 ) . Error bars reflect within-subjects error . ( A ) Mean SCR in the Instructed Group as a function of original contingencies . Runs are defined relative to the delivery of instructions . ( B ) Dynamics of expected value based on fits of our modified Rescorla-Wagner model , fit to SCR in the Instructed Group . Fitted model parameters were consistent with SCR reversing almost entirely in response to instructions ( ρ = 0 . 943 ) . This timecourse was used in fMRI analyses to isolate regions involved in instruction-based learning . ( C ) Mean SCR in the Uninstructed Group as a function of original contingencies . A new run is defined when the previous CS- is paired with a shock . ( D ) Dynamics of expected value based on the model fit to SCR from the Uninstructed Group . This timecourse was used in fMRI analyses to isolate regions involved in feedback-driven learning in both groups . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 005 We also tested whether instructions immediately update autonomic responses by examining instructed reversals in Instructed Group learners . Each reversal featured a delay between instruction delivery and reinforcement of instructions ( Figure 1C ) : Following instructions , each CS was presented without reinforcement at least twice before the previous CS- was paired with a US . We compared responses during these post-instruction , pre-reinforcement windows with an equivalent number of trials prior to each instruction ( e . g . the last two CS+ and CS- trials in each phase ) . We found a significant effect of instructions on differential responding ( ß = 0 . 04 , t ( 19 ) = 3 . 50 , p= 0 . 0024 ) , such that SCR responses reversed immediately after each instruction , before any actual reinforcement was delivered . We were interested in understanding how instructions shape error-driven learning and the development of expectations . To this end , we focused on the computation of expected value ( EV ) and used a simple Rescorla-Wagner model modified to capture flexible effects of instructions and test for dissociations . Our quantitative models assume that SCR at cue onset reflects EV ( see Materials and methods ) . Shocks were incorporated as reinforcements , and thus positive EV corresponds to an expectation for a shock . Consistent with standard Rescorla-Wagner models , EV updates in response to prediction error ( PE ) , and the speed of updating depends on learning rate ( α ) . We focus on correlations with EV in this manuscript , due to concerns of algebraic collinearity when EV , shock , and PE are included in the same model . Models fit to the Uninstructed Group’s SCRs isolate learning-related processes that respond to reinforcement history alone ( i . e . feedback-driven learning ) , since this group was not informed about cue contingencies or reversals . Models fit to the Instructed Group’s SCRs , however , should capture the immediate effects of instruction reported above . To acknowledge this flexible effect of instructions , we modified the standard Rescorla-Wagner model . We introduced an instructed reversal parameter , ρ , which determines the extent to which EV reverses upon instruction ( see Materials and methods ) . If ρ = 1 , the EVs of the two CSs are swapped completely when instructions are delivered , whereas if ρ = 0 , each CS maintains its current EV and the model reduces to a standard experiential Rescorla-Wagner model . The best-fitting parameters when fit across Instructed Group subjects revealed that EV reversed almost completely at the time of instructions in the Instructed Group ( ρ = 0 . 943; Table 2 ) , suggesting that instructions immediately influence EV/SCR , as illustrated in Figure 2B . When we fit the same model to SCRs from the Uninstructed Group ( for whom the additional effect should not be observed since no instructions were given ) estimates were indeed consistent with associations not reversing at the time when the instructions would have been delivered ( ρ = 0 . 0; Table 2 ) . The resulting time course , which captures slower reversals of EV based on purely feedback-driven learning , is depicted in Figure 2D . We also fit the model to individual participants in both groups ( see Materials and methods ) and found that instructed reversal parameters ( i . e . ρ ) differed significantly as a function of Group ( Instructed Group > Uninstructed Group , t ( 38 ) = 6 . 53 , p<0 . 0001; Table 2 ) . 10 . 7554/eLife . 15192 . 006Table 2 . Quantitative model of instructed learning: Rescorla-Wagner modulated with instruction parameter ( ρ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 006GroupAnalysis typeαρDevianceInstructed Group learners ( n = 20 ) Across-subjects analysis0 . 0610 . 94360 . 83Within-subjects analysisM = 0 . 07 , SD = 0 . 09M = 0 . 69 , SD = 0 . 32M = 2 . 73 , SD = 0 . 88Uninstructed Group learners ( n = 20 ) Across-subjects analysis0 . 042040 . 86Within-subjects analysisM = 0 . 11 , SD = 0 . 22M = 0 . 10 , SD = 0 . 25M = 1 . 98 , SD = 0 . 93 Our goal was to determine whether and how the neural systems that support feedback-driven aversive learning are modulated by instructions . Thus our computational neuroimaging analyses proceeded in two stages , guided by the quantitative models reported above . First , we examined each group separately using the model fit to behavior in that group . Thus we used the model depicted in Figure 2D to isolate neural correlates of feedback-driven learning in the Uninstructed Group , and the model depicted in Figure 2B to isolate neural correlates of instruction-based learning in the Instructed Group . Next , we directly compared the two sources of learning in the Instructed Group , since these participants were exposed to both forms of feedback ( i . e . instructions about contingencies and reversals , as well as experiential learning from reinforcement between each reversal ) . In each analysis , we focused on results in amygdala , striatum , and VMPFC/OFC ( see Figure 3—figure supplement 3 ) to determine whether these a priori regions of interest ( ROIs ) were sensitive to feedback-driven learning and/or whether they updated with instructions . Finally , we tested the conclusions from quantitative models with task-based analyses that relied strictly on our experimental design , thus eliminating the influence of assumptions derived from our models . We first focused on the neural correlates of experiential learning by examining responses in the Uninstructed Group . Regressors were based on the best-fitting parameters from the model fit to the Uninstructed Group , thus isolating feedback-driven EV ( Figure 2D; see Materials and methods ) . ROI-based analyses within the Uninstructed Group revealed a main effect of Region ( F ( 2 , 40 ) = 5 . 13 , p=0 . 011 ) . Post-hoc t-tests revealed that this was driven by positive correlations between feedback-driven EV and responses in the amygdala ( bilateral: t ( 1 , 19 ) = 4 . 21 , p=0 . 0005; Left: t ( 1 , 19 ) = 3 . 94 , p=0 . 001; Right: t ( 1 , 19 ) : = 3 . 72 , p=0 . 001 ) and striatum ( bilateral: t ( 1 , 19 ) = 2 . 31 , p=0 . 0017; left: t ( 1 , 19 ) = 2 . 58 , p=p =0 . 018; right: p=p =0 . 063 ) . Voxel-wise FDR-corrected results confirmed ROI-based findings , and isolated additional correlations in the VMPFC/mOFC ( see Figure 3A , Figure 3—figure supplement 1 , and Figure 3—figure supplement 1—source data 1 and 2 ) . The striatum and amygdala both showed positive correlations with EV , associated with increased activation for the stimulus currently predicting an aversive outcome . The VMPFC/mOFC showed negative correlations with EV , consistent with prior work showing increased VMPFC/OFC responses to conditioned stimuli predicting safe , relative to aversive , outcomes ( Schiller et al . , 2008 ) . Additional regions that correlated with feedback-driven EV are reported in Figure 3—figure supplement 1 and associated source data . 10 . 7554/eLife . 15192 . 007Figure 3 . Neural correlates of expected value . ( A ) Neural correlates of feedback-driven expected value ( EV ) were isolated by examining correlations between the timecourse depicted in Figure 2D and brain activation in response to cue onset in Uninstructed Group learners ( n = 20 ) . Top: ROI-based analyses ( see Figure 3—figure supplement 3 ) revealed significant correlations with feedback-driven EV in the amydala and striatum . Error bars reflect standard error of the mean; ***p<0 . 001; *p<0 . 05 . Bottom: Voxel-wise FDR-corrected analyses confirmed ROI-based results and revealed additional correlations in the VMPFC/OFC , as well as other regions ( see Figure 3—figure supplement 1 , Figure 3—figure supplement 1—source data 1 , 2 ) . ( B ) Neural correlates of instruction-based EV were isolated by examining correlations between the timecourse depicted in Figure 2B and brain activation in response to cue onset in Instructed Group learners ( n = 20 ) . Top: ROI-based analyses revealed a significant negative correlation with instruction-based EV in the VMPFC/OFC . Bottom: Voxel-wise analyses confirmed these results and revealed strong positive correlations in the bilateral striatum , as well as the dACC , insula , and other regions ( see Figure 3—figure supplement 2 and Figure 3—figure supplement 2—source data 1 and 2 ) . We did not observe any correlations between amygdala activation and instruction-based EV . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 00710 . 7554/eLife . 15192 . 008Figure 3—figure supplement 1 . Feedback-driven EV in the Uninstructed Group . Voxelwise FDR-corrected results of neural correlates of feedback-driven EV in Uninstructed Group participants , based on the modified learning model fit to learners in the Uninstructed Group ( across-subjects fits , see Materials and methods ) . Warm colors reflect positive correlations with EV , which was coded such that positive EV denotes expected shock . Cool colors reflect negative correlations . Top: Results in learners , or those individuals who showed differential SCR prior to the first reversal ( n = 20 ) . Bottom: Results across the entire Uninstructed Group ( n = 38 ) . For complete results in tabular format , please see 'Figure 3—figure supplement 1—source data 1' and 'Figure 3—figure supplement 1—source data 2' . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 00810 . 7554/eLife . 15192 . 009Figure 3—figure supplement 1—source data 1 . Neural correlates of feedback-driven expected value ( EV ) : Uninstructed Group Learners ( n = 20 ) . This table presents brain regions that correlate with feedback-driven EV ( derived from the across-subjects model fit to Uninstructed Group learners ) within Uninstructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 00910 . 7554/eLife . 15192 . 010Figure 3—figure supplement 1—source data 2 . Neural correlates of feedback-driven EV: Entire Uninstructed Group ( n = 38 ) . This table presents brain regions that correlate with feedback-driven EV ( derived from the across-subjects model fit to Uninstructed Group learners ) within the entire Uninstructed Group ( n = 38 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01010 . 7554/eLife . 15192 . 011Figure 3—figure supplement 2 . Instruction-based EV in the Instructed Group . Voxelwise FDR-corrected results of neural correlates of instruction-based EV in Instructed Group participants , based on the modified learning model fit to learners in the Instructed Group ( across-subjects fits , see Materials and methods ) . Warm colors reflect positive correlations with EV , which was coded such that positive EV denotes expected shock . Cool colors reflect negative correlations . Top: Results in learners , or those individuals who showed differential SCR prior to the first reversal ( n = 20 ) . Bottom: Results across the entire Instructed Group ( n = 30 ) . For complete results in tabular format , please see 'Figure 3—figure supplement 2—source data 1' and 'Figure 3—figure supplement 2—source data 2' . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01110 . 7554/eLife . 15192 . 012Figure 3—figure supplement 2—source data 1 . Neural correlates of instruction-based EV: Instructed Group Learners ( n = 20 ) . This table presents brain regions that correlate with instruction-based EV ( derived from the across-subjects model fit to Instructed Group learners ) within Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q<0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01210 . 7554/eLife . 15192 . 013Figure 3—figure supplement 2—source data 2 . Neural correlates of instruction-based EV: Entire Instructed Group ( n = 30 ) . This table presents brain regions that correlate with instruction-based EV ( derived from the across-subjects model fit to Instructed Group learners ) within the entire Instructed Group ( n = 30 ) . Results are whole-brain FDR-corrected ( q<0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01310 . 7554/eLife . 15192 . 014Figure 3—figure supplement 3 . Regions of interest . Regions of interest for ROI-based FMRI analyses . See Materials and methods for details of ROI selection . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 014 We used parameters from the best-fitting instructed learning model to isolate the neural correlates of aversive learning that updates expected value on the basis of instructions in the Instructed Group . In this model , EV updates immediately with instructions ( Figure 2b ) , consistent with the SCRs measured in the Instructed Group . ROI-based ANOVAs revealed a significant effect of Region ( F ( 2 , 40 ) = 6 . 49 , p=0 . 0038 ) , driven by significant negative correlations with EV in the VMPFC/OFC ROI ( t ( 1 , 19 ) = -2 . 61 , p=0 . 0173; see Figure 3B ) . Voxelwise FDR-corrected results also revealed robust activation in the bilateral caudate , which showed positive correlations with instructed EV ( see Figure 3B , Figure 3—figure supplement 2 , Figure 3—figure supplement 2—source data 1 , and Figure 3—figure supplement 2—source data 2 ) . Additional regions that tracked instructed EV in whole brain analyses and results across the entire Instructed Group are reported in Figure 3—figure supplements 2 , 3 and associated source data . The preceding results , from separate groups using EV signals driven by instructions or feedback , suggest that responses in the amygdala , striatum , and VMPFC/OFC were driven by reinforcement in the Uninstructed Group , while only VMPFC/OFC and striatal responses were sensitive to instructions the Instructed Group . To test for formal dissociations , we directly compared the neural correlates of instructed and feedback-driven aversive learning in Instructed Group participants , who were exposed to both instructions and experiential learning . This within-subjects analysis ensures that potential dissociations indicated above are driven by differences in the computational sources of neural activation , rather than differences in performance between the groups . To isolate brain responses that were sensitive only to reinforcement ( despite the presence of instructions about contingencies and reversals ) , we used the feedback-driven EV regressor generated from the model fit to the Uninstructed Group’s behavior , which takes advantage of the fact that trial sequences were identical for both groups . The instruction-driven EV regressor was generated from the model fit to the Instructed Group’s behavior and reported above . We included both EV regressors in a within-subjects voxel-wise analysis in the Instructed Group . Between reversals , experiential learning would be somewhat correlated across models . Thus , to remove shared variance , we did not orthogonalize regressors in this analysis ( see Materials and Methods ) . A contrast across the two EV regressors formally tests whether each voxel is more related to feedback-driven or instruction-driven EV . Results from this contrast are presented in Figure 4 . Voxelwise FDR-corrected analyses revealed that the bilateral amygdala was preferentially correlated with feedback-driven EV , while the right caudate and left putamen showed preferential correlations with instruction-based EV ( see Figure 4B , Figure 4—figure supplement 1 , and Figure 4—figure supplement 1—source data 1 ) . ROI-based analyses confirmed voxelwise results , with a main effect of Region ( F ( 2 , 40 ) = 3 . 76 , p=0 . 0324; Figure 4A ) , driven by amygdala correlations with feedback-driven EV but not instructed EV ( t ( 1 , 19 ) = 2 . 57 , p=0 . 0189 ) , although striatal differences between models were not significant when averaged across the entire ROI . Although these effects were somewhat weak when limited to learners , effects increased in magnitude when we examined the entire Instructed Group , including individuals who did not exhibit measurable SCRs ( see Figure 4—figure supplement 1 and Figure 4—figure supplement 1—source data 2 ) . When we included all Instructed Group participants , we also observed negative correlations with instruction-based EV in the VMPFC/OFC . ROI-wise analyses across the entire Instructed Group revealed that the main effect of Region ( F ( 2 , 60 ) = 7 . 12 , p =0 . 0017 ) was driven by both bilateral amygdala specificity for feedback-driven EV ( t ( 1 , 29 ) = 2 . 93 , p =0 . 0066 ) as well as VMPFC/OFC specificity for instructed EV ( t ( 1 , 29 ) = 2 . 29 , p =0 . 0293 ) . 10 . 7554/eLife . 15192 . 015Figure 4 . Dissociable effects of instructed and feedback-driven learning in the Instructed Group . ( A ) ROI-based effects of feedback-driven and instruction based EV signaling within Instructed Group learners from the model including both signals ( see Materials and Methods ) . Direct model comparisons within Instructed Group learners revealed a significant effect of Model in the amygdala ( p<0 . 05 ) . VMPFC differences were marginal within learners ( p = 0 . 11 ) and were significant when all Instructed Group participants were included in analyses ( p<0 . 05 ) . Error bars reflect standard error of the mean . ***p<0 . 001; *p<0 . 05; †p<0 . 10 . ( B ) Voxelwise direct comparison between feedback-driven and instruction based EV signaling within the Instructed Group . Regions in warm colors , including bilateral amygdala ( left ) , showed preferential correlations with feedback-driven EV . Regions in cool colors , including left caudate ( middle ) , dorsal anterior cingulate and medial prefrontal cortex ( right ) , showed higher correlations with instruction-based EV . Additional regions that showed significant differences as a function of model are presented in Figure 4—figure supplement 1 , Figure 4—figure supplement 1—source data 1 and 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01510 . 7554/eLife . 15192 . 016Figure 4—figure supplement 1 . Feedback-driven vs instruction-based EV in the Instructed Group . Voxelwise FDR-corrected results of direct comparison between feedback-driven and instruction based EV signaling within the Instructed Group . Feedback-driven EV is based on the model fit to the Uninstructed Group learners , while instruction-based EV is based on fit to the Instructed Group learners . Regions in warm colors showed preferential correlations with feedback-driven EV , while regions in cool colors showed higher correlations with instruction-based EV . Top: Results in learners , or those individuals who showed differential SCR prior to the first reversal ( n = 20 ) . Bottom: Results across the entire Instructed Group ( n = 30 ) . For complete results in tabular format , please see 'Figure 4—figure supplement 1—source data 1' and 'Figure 4—figure supplement 1—source data 2 . 'DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01610 . 7554/eLife . 15192 . 017Figure 4—figure supplement 1—source data 1 . Feedback-driven vs instruction-based EV: Instructed Group Learners ( n = 20 ) . This table presents brain regions that show preferential correlations with either feedback-driven or instruction-based EV , based on direct contrasts between the two signals . Analyses are restricted to Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01710 . 7554/eLife . 15192 . 018Figure 4—figure supplement 1—source data 2 . Neural correlates of instruction-based EV: Entire Instructed Group ( n = 30 ) . This table presents brain regions that show preferential correlations with either feedback-driven or instruction-based EV , based on direct contrasts between the two signals . Analyses include the entire Instructed Group ( n = 30 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 018 Quantitative models revealed that the striatum and VMPFC/OFC track aversive learning that updates with instructions , whereas the amygdala learns from aversive feedback irrespective of instruction . To further verify these dissociations with traditional contrast-based analyses that were independent of learning models and observed behavior , we employed a task-based , event-related General Linear Model ( GLM ) . This allowed us to isolate the effects of instructed reversals within the Instructed Group by focusing specifically on the trials surrounding instructions , prior to subsequent reinforcement . The analysis leverages the delay between instruction delivery and US reinforcement ( Figure 5A ) and mirrors our behavioral analysis of the immediate effects of instructed reversals on SCR . We compared brain responses pre- and post-instruction by computing a CS ( [previous CS+ > previous CS-] ) x Phase ( [Pre - Post] ) interaction analysis . This provided a within-subjects comparison between instructed reversals and feedback-driven reversals . Furthermore , this analysis is entirely independent of the Uninstructed Group , and therefore allows us to ensure that the results and conclusions reported above do not simply reflect a difference in performance or learning rate between the two groups . 10 . 7554/eLife . 15192 . 019Figure 5 . Task-based effects of instructed reversals and relationship with modeled behavior . ( A ) We examined responses in the Instructed Group surrounding the three instructed reversals to dissociate regions that are sensitive to instructions from those that learn from feedback and seem to be insensitive to instructions . Regions that are sensitive to instructions should show differential responses that reverse immediately upon instruction . The lavender timecourse depicts this pattern with greater activation on CS+ trials ( blue ) than CS- trials ( yellow ) prior to instruction , and the opposite pattern after instructions are delivered . Regions that update from aversive feedback and show no effect of instructed reversals would follow the orange timecourse , with greater activation to the previous CS+ than CS- both pre- and post-instruction . This feedback-driven pattern does not update until the new CS+ has been reinforced . ( B ) A number of regions showed differential responses that reversed upon instruction , including the right VS and VMPFC/OFC ( left; see also Figure 5—figure supplement 1 , Figure 5—figure supplement 1—source data 1 and 2 ) . The VS showed greater activation to the current CS+ relative to the current CS- , whereas the VMPFC/OFC showed deactivation to the CS+ . The right amygdala showed differential activation that did not reverse with instructions ( right ) . Additional regions that did not reverse with instructions are presented in Figure 5—figure supplement 2 , Figure 5—figure supplement 2—source data 1 and 2 . ( C ) We conducted brain-behavior correlations to explore the relationship between neural activity in the period surrounding instructions and the magnitude of each individual’s behavioral response to instructions . We tested for correlations between each individual’s ρ parameter ( based on within-subjects fits ) and the magnitude of the reversal effect using an exploratory threshold of p<0 . 001 , uncorrected . We observed significant correlations between ρ and the magnitude of instructed reversals in dACC ( left ) and the bilateral caudate tail and thalamus ( right ) , as well as bilateral DLPFC ( see Figure 5—figure supplement 3 ) , suggesting that those individuals who showed stronger reversals in SCR also showed stronger reversals in these regions . In addition , we found that the individuals who showed the least evidence for updating with instructions also showed the largest non-reversing differential responses in the right amygdala ( right ) . Full results of brain-behavior correlations are reported in Figure 5—figure supplement 3 and Figure 5—figure supplement 3—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 01910 . 7554/eLife . 15192 . 020Figure 5—figure supplement 1 . Immediate reversal with instructions ( CS [previous CS+ > previous CS-] x Phase [Pre - Post] interaction ) Voxelwise FDR-corrected results of regions that show immediate reversals with instructions in the Instructed Group , based on the window surrounding the delivery of instructions ( see Materials and Methods ) . Regions in warm colors showed greater activation to the current CS+ relative to the current CS- , while regions in cool colors show relatively greater activation to the CS- ( or deactivation to the CS+ ) . Top: Results in learners , or those individuals who showed differential SCR prior to the first reversal ( n = 20 ) . Bottom: Results across the entire Instructed Group ( n = 30 ) . For complete results in tabular format , please see 'Figure 5—figure supplement 1—source data 1' and 'Figure 5—figure supplement 1—source data 2 . 'DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02010 . 7554/eLife . 15192 . 021Figure 5—figure supplement 1—source data 1 . Immediate reversal with instructions ( CS x Phase interaction ) : Instructed Group Learners ( n = 20 ) . This table presents brain regions whose differential responses ( CS+ vs CS- ) reversed immediately upon instruction . Regions that were positive in this contrast show greater activation with the current CS+ relative to the current CS- , whereas those that are negative show deactivation to the CS+ or increases with the CS- . Analyses are restricted to Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q<0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02110 . 7554/eLife . 15192 . 022Figure 5—figure supplement 1—source data 2 . Immediate reversal with instructions ( CS x Phase interaction ) : Entire Instructed Group ( n = 30 ) . This table presents brain regions whose differential responses ( CS+ vs CS- ) reversed immediately upon instruction . Analyses include the entire Instructed Group ( n = 30 ) . Results are whole-brain FDR-corrected ( q<0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02210 . 7554/eLife . 15192 . 023Figure 5—figure supplement 2 . No reversal with instructions ( main effect of CS without interaction; i . e . [previous CS+ > previous CS-] ∪ [new CS- > new CS+] ) Voxelwise FDR-corrected results of regions that show no evidence for reversal when instructions are delivered , based on continued differential responses pre- and post-instruction ( see Materials and Methods ) . Regions in warm colors showed greater activation to the pre-instruction CS+ relative to the CS- both pre- and post-instruction , while regions in cool colors show relatively greater activation to the CS- ( or deactivation to the CS+ ) . Top: Results in learners , or those individuals who showed differential SCR prior to the first reversal ( n = 20 ) . Bottom: Results across the entire Instructed Group ( n = 30 ) . For complete results in tabular format , please see 'Figure 5—figure supplement 2—source data 1' and 'Figure 5—figure supplement 2—source data 2 . 'DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02310 . 7554/eLife . 15192 . 024Figure 5—figure supplement 2—source data 1 . No reversal with instructions ( main effect of CS without interaction ) : Instructed Group Learners ( n = 20 ) . This table presents brain regions whose differential responses ( CS+ vs CS- ) did not reverse upon instruction in Instructed Group learners ( n = 20 ) . Regions that were positive in this contrast show greater activation to the pre-instruction CS+ relative to the CS- both pre- and post-instruction , whereas those that are negative show deactivation to the CS+ or increases with the CS- . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02410 . 7554/eLife . 15192 . 025Figure 5—figure supplement 2—source data 2 . No reversal with instructions ( main effect of CS without interaction ) : Entire Instructed Group ( n = 30 ) . This table presents brain regions whose differential responses ( CS+ vs CS- ) did not reverse upon instruction in the entire Instructed Group ( n = 30 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02510 . 7554/eLife . 15192 . 026Figure 5—figure supplement 3 . Correlations with instructed reversal ( ρ ) parameters . Voxelwise FDR-corrected results of brain-behavior correlations that tested for correlations between each individual’s ρ parameter ( based on within-subjects fits ) and the magnitude of the reversal effect or absence of reversals using an exploratory threshold of p<0 . 001 , uncorrected . Regions in warm colors show positive correlations between the magnitude of the instructed reversal parameter and the strength of reversal ( Top ) or sustained differential response ( Bottom ) in the region , while regions in cool colors show negative correlations . We focused on Instructed Group learners for these analyses . Top: Correlation between the instructed reversal parameter and immediate reversals with instructions ( CS x Phase interactions ) . Bottom: Correlation between the instructed reversal parameter and the absence of instructed reversal ( main effect of CS without reversal ) . For complete results in tabular format , please see 'Figure 5—figure supplement 3—source data 1' and 'Figure 5—figure supplement 3—source data 2 . 'DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02610 . 7554/eLife . 15192 . 027Figure 5—figure supplement 3—source data 1 . Correlation between instructed reversal parameter ( ρ ) and instructed reversal effects: Instructed Group Learners ( n = 20 ) . This table presents brain regions in which the ρ parameter correlated positively ( warm ) or negatively ( cool ) with the magnitude of the CS x Phase interaction , which indicates differential responses ( CS+ vs CS- ) that reverse upon instruction in Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02710 . 7554/eLife . 15192 . 028Figure 5—figure supplement 3—source data 2 . Correlation between instructed reversal parameter ( ρ ) and continued response to previous CS+ vs CS- ( no reversal effect ) : Instructed Group Learners ( n = 20 ) . This table presents brain regions in which the ρ parameter correlated positively ( warm ) or negatively ( cool ) with the magnitude of the sustained differential response in Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 028 We first identified regions that reversed immediately upon instruction ( CS x Phase interaction ) . Such regions should show differential activation to the previous CS+ than CS- prior to instruction , and similar activation to the new CS+ relative to the new CS- after instructions , prior to reinforcement . We observed immediate instructed reversal effects in the ventral striatum and VMPFC/OFC . The VMPFC/OFC was identified in both ROI-based and voxel-wise analyses , while the striatal activation was evident in voxel-wise analyses ( see Figure 5B , Figure 5—figure supplement 1 , and Figure 5—figure supplement 1—source data 1 ) . Post-hoc analyses revealed that the VS showed greater activation to the current CS+ , relative to the corresponding CS- , while VMPFC/OFC showed greater deactivation to the current CS+ ( t ( 1 , 19 ) = -4 . 82 , p=p =0 . 0001 ) . Instructed reversal effects in the striatum and VMPFC/OFC were stronger and more widespread when the entire Instructed Group was included in the analysis ( see Figure 5—figure supplement 1 and Figure 5—figure supplement 1—source data 2 ) . Whole-brain exploratory analyses also revealed immediate reversals with greater activation to the current CS+ in the dorsal anterior cingulate cortex ( dACC ) , bilateral insula , thalamus , and midbrain surrounding the periaqueductal gray ( PAG ) , and deactivation in bilateral hippocampus ( see Figure 5B , Figure 5—figure supplement 1 , Figure 5—figure supplement 1—source data 1 , and Figure 5—figure supplement 1—source data 2 ) . We also identified regions that continued to show differential responses to the previous contingencies despite the fact that instructions had been delivered ( main effect of CS without interaction; i . e . [previous CS+ > previous CS-] ∪ [new CS- > new CS+] ) . These regions presumably update with reinforcement , as our analysis incorporates all three reversals and therefore regions must show greater activation to the new CS+ prior to each subsequent instructed reversal . Voxel-wise analyses revealed that the right amygdala showed differential responses that did not reverse upon instruction ( Figure 5B ) , although effects were not significant when averaged across the entire ROI . We note that we also observed a small cluster in the left putamen that did not reverse with instructions; however this region showed greater activation to the current CS- than CS+ , unlike the robust ventral striatal activation observed in all of our other analyses , which reveal greater activation with higher likelihood of aversive outcomes . Results of whole-brain exploratory analyses and across the entire sample are reported in Figure 5—figure supplement 2 , Figure 5—figure supplement 2—source data 1 and 2 . Finally , we tested whether these task-based effects of instructions on neural activation were related to our quantitative models and effects on behavior . We used each individual’s ρ parameter ( based on within-subjects model fitting; see Materials and Methods ) to characterize the behavioral effects of instructions on SCR and aversive learning . We tested for correlations between this quantity and the magnitude of the instructed reversal effect ( CS x Phase interaction ) as well as the main effect without reversal ( [previous CS+ > previous CS-] ∪ [new CS- > new CS+] ) , using an exploratory threshold of p <0 . 001 , uncorrected . We found that the magnitude of the behavioral reversal effect was positively correlated with instructed reversal effects in bilateral dorsolateral prefrontal cortex ( DLPFC ) , bilateral caudate tail and thalamus , dACC , and right anterior insula ( see Figure 5C , Figure 5—figure supplement 3 , and Figure 5—figure supplement 3—source data 1 ) . Conversely , differential responses in the amygdala ( which did not reverse with instructions ) were strongest in those individuals whose behavior showed the weakest influence of instructions ( see Figure 5C , Figure 5—figure supplement 3 , and Figure 5—figure supplement 3—source data 2 ) . The results reported above reveal dissociable effects of instructions on individual brain regions involved in aversive learning , and relate neural effects with observed behavior . Our final question was whether responses to instructions themselves influence subsequent learning-related neural responses . To understand how instructions influence aversive learning , we searched for brain regions that were uniquely sensitive to instructions , i . e . that showed a group difference across trials ( CS onset , collapsed across CS+ and CS- ) . Although no regions survived FDR-correction , the left dorsolateral prefrontal cortex ( DLPFC; middle frontal gyrus , peak voxel xyz = [-43 43 21] ) showed greater activation across all trials in the Instructed Group at an uncorrected threshold of p<0 . 001 ( Figure 6A ) . 10 . 7554/eLife . 15192 . 029Figure 6 . Relationship between dorsolateral prefrontal cortex response to instructions and instructed reversal effects . ( A ) The left dorsolateral prefrontal cortex ( DLPFC ) showed group differences across all trials , with greater activation in the Instructed Group than the Uninstructed Group . ( B ) We extracted the magnitude of the DLPFC response to instructions for each individual within the Instructed Group . C ) The magnitude of the DLPFC response to instructions was correlated with the magnitude of instructed reversals in VMPFC/OFC ( left ) and dorsal putamen ( right ) . High DLPFC responders showed larger reversals , with putamen activation to the new CS+ relative to the new CS-and VMPFC/OFC deactivation to the new CS+ relative to the new CS- . See also Figure Supplement 1 and Figure 6—figure supplement 1—source data 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 02910 . 7554/eLife . 15192 . 030Figure 6—figure supplement 1 . Correlations with left dorsolateral prefrontal cortex response to instructions . Voxelwise FDR-corrected results of brain-behavior correlations that tested for correlations between each Instructed individual’s DLPFC response to instructions ( see Figure 6 ) and the magnitude of the reversal effect or absence of reversals using an exploratory threshold of p<0 . 001 , uncorrected . Regions in warm colors show positive correlations between the magnitude of the DLPFC response and the strength of reversal , while regions in cool colors show negative correlations . We focused on Instructed Group learners for these analyses ( n = 20 ) . For complete results in tabular format , please see 'Figure 6—figure supplement 1—source data 1' . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 03010 . 7554/eLife . 15192 . 031Figure 6—figure supplement 1—source data 1 . Correlation between dorsolateral prefrontal response to instructions and instructed reversal effect: Instructed Group Learners ( n = 20 ) . This table presents brain regions in which the DLPFC response correlated positively ( warm ) or negatively ( cool ) with the magnitude of the CS x Phase interaction , which indicates differential responses ( CS+ vs CS- ) that reverse upon instruction in Instructed Group learners ( n = 20 ) . Results are whole-brain FDR-corrected ( q < 0 . 05 ) and clusters are defined based on contiguity with voxels at uncorrected p<0 . 001 and p<0 . 01 . DOI: http://dx . doi . org/10 . 7554/eLife . 15192 . 031 We then tested whether responses to instructions in this DLPFC region predicted the extent to which regions reversed immediately with instructions . We extracted DLPFC activation to the presentation of instructions for each Instructed Group participant ( Figure 6B ) , and correlated this quantity with the magnitude of the instructed reversal effect ( the CS x Phase interaction reported above ) throughout the brain . We observed significant positive correlations in the right putamen ( peak voxel xyz = [24 8 11] ) and negative correlations in the VMPFC/mOFC ( peak voxel xyz = [-5 40 -17]; see Figure 6C ) , although the VMPFC/mOFC cluster was slightly smaller than our cluster threshold ( 9 voxels rather than 10 ) . In both cases , individuals who showed greater DLPFC activation during instruction showed stronger reversals , with greater activation to the current CS+ than CS- in the putamen , and greater deactivation to the current CS+ in the VMPFC/mOFC . Whole brain results are reported in Figure 6—figure supplement 1 and Figure 6—figure supplement 1—source data 1 .
We found dissociable effects of instructions on the brain regions involved in aversive learning . Many regions , including the striatum and VMPFC/OFC , updated immediately with instructions . However , responses in the amygdala were not shaped by instructions in our task , as the amygdala required aversive feedback in order to update . These results indicate that the neural mechanisms of dynamic aversive learning include processes that perform the same computations across appetitive and aversive domains ( i . e . striatal learning that updates with instructions ) , as well as unique patterns that may be specific to aversive learning due to evolutionary adaptations and the biological relevance of threat processing .
Eighty right-handed English-speaking participants were enrolled in the experiment . Participants were not enrolled if they were taking any anti-depressant or anti-anxiety medication , had heart or blood pressure problems , were pregnant , or had completed a study that applied electric shocks within the previous six months . All participants provided informed consent as approved by New York University’s Institutional Review Board , the University Committee on Activities Involving Human Subjects ( UCAIHS; protocol #12–8965 and #13–9582 ) . Participants were informed that the study was voluntary , and that the goals of the study were to learn about emotional responses to images that predict shock , and to learn more about how physiological responses and the activity of different regions of the brain are related to emotion and learning . Three consented participants did not participate in the experiment because we were unable to measure skin conductance . Data from eight participants were not included in the analyses either because they did not complete the study due to discomfort during the scan ( n = 3 ) or because they slept for an extended period of time during the experiment ( n = 5 ) . Finally , intermittent signal loss prevented us from analyzing skin conductance data for one participant , leaving a final sample of sixty-eight participants ( 46 Female; mean age = 22 . 1 years ( SE = 0 . 42 ) ) . Participants received mild electric shocks ( 200 ms duration , 50 pulses per second ) to the right wrist using a magnetically shielded stimulator and electrodes ( Grass Medical Instruments , West Warwick , RI ) attached by Velcro . We measured skin conductance using shielded silver-silver chloride electrodes ( BIOPAC Systems , Inc . , Goleta , CA ) filled with standard NaCl electrolyte gel , and attached by Velcro to the middle phalanges of the second and third fingers of the left hand . Pupillometry data were also collected , but data were corrupted for the majority of participants and are therefore not included in the current analysis . Our learning models assume that SCR correlates with dynamic quantities derived from feedback-driven or instructed learning models . Below we describe the quantitative models we evaluated , followed by our general procedures for model fitting . | Around the start of the twentieth century , Pavlov discovered that dogs salivate upon hearing a bell that has previously signaled that food is available . This phenomenon , in which a neutral stimulus ( the bell ) becomes associated with a particular outcome ( such as food ) , is known as classical conditioning . The network of brain regions that supports this process – which includes the striatum , the amygdala and the prefrontal cortex – seems to work in a similar way across most animal species , including humans . However , humans don’t learn only through experience or trial-and-error . We do not need to burn our hands to learn not to touch a hot stove: a verbal warning from others is usually sufficient . Experiments have shown that giving people verbal instructions on how to obtain rewards alters the activity of the striatum and prefrontal cortex . That is , the instructions interact with the circuit that also supports learning through experience . But is this the case for learning how to avoid punishments ? That process depends largely on the amygdala , and it is possible that systems designed to detect threats may be less sensitive to verbal warnings . To address this question , Atlas et al . taught people to associate one image with a mild electric shock , and another with the absence of a shock . After a number of trials , the relationships were reversed so that the previously neutral picture now predicted a shock and vice versa . Telling the participants about the reversal in advance triggered changes in the activity of the striatum and part of the prefrontal cortex . By contrast , such warnings had no effect on the amygdala . Instead , the activity of the amygdala changed only after the volunteers had experienced for themselves the new relationship between the pictures and the shocks . A key next step is to find out whether this distinction between the two types of learning signals ( those that can be updated by instructions and those that cannot ) is specific to humans . While the current study relied upon language , there are other methods that could be used to explore this issue in animals . Furthermore , knowing that the human brain has a specialized threat detection system that is less sensitive to instructions could help us to understand and treat anxiety disorders . Atlas et al . hope to test this possibility directly in the future . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2016 | Instructed knowledge shapes feedback-driven aversive learning in striatum and orbitofrontal cortex, but not the amygdala |
Leishmania donovani causes visceral leishmaniasis ( VL ) , the second most deadly vector-borne parasitic disease . A recent epidemic in the Indian subcontinent ( ISC ) caused up to 80% of global VL and over 30 , 000 deaths per year . Resistance against antimonial drugs has probably been a contributing factor in the persistence of this epidemic . Here we use whole genome sequences from 204 clinical isolates to track the evolution and epidemiology of L . donovani from the ISC . We identify independent radiations that have emerged since a bottleneck coincident with 1960s DDT spraying campaigns . A genetically distinct population frequently resistant to antimonials has a two base-pair insertion in the aquaglyceroporin gene LdAQP1 that prevents the transport of trivalent antimonials . We find evidence of genetic exchange between ISC populations , and show that the mutation in LdAQP1 has spread by recombination . Our results reveal the complexity of L . donovani evolution in the ISC in response to drug treatment .
Parasites of the Leishmania donovani species complex cause visceral leishmaniasis ( VL ) , the most severe presentation of leishmaniasis that is usually fatal if untreated . There are probably between 200 , 000 and 300 , 000 VL cases annually ( Alvar et al . , 2012 ) , leading to as many as 50 , 000 deaths per year ( den Boer et al . , 2011; Lozano et al . , 2012 ) . VL is widespread in both the New and Old Worlds ( Pigott et al . , 2014 ) , but as much as 80% of the global VL burden occurs in the Indian sub-continent ( Alvar et al . , 2012 ) . Recent intensified control efforts have led to a notable decline in cases ( Chowdhury et al . , 2014 ) but the problem is not yet eliminated . VL is a key neglected tropical disease , affecting the poorest regions of the world and the poorest communities within these regions ( Boelaert et al . , 2009 ) . VL was first reported in the Indian sub-continent ( ISC ) in the 1820s , but initially confused with malaria until the discovery of L . donovani in 1903 ( Gibson , 1983 ) . Although VL was nearly eliminated from the ISC in the 1960s ( Thakur , 2007 ) by antimalarial spraying campaigns with DDT , it re-emerged in 1977 and has caused several subsequent major epidemics ( Dye and Wolpert , 1988 ) . Widespread chemotherapy for VL in the region has been ongoing since the 1820s , initially with quinine and other drugs , followed by extensive use of the trivalent antimonial SbIII ( 1915 ) and compounds of the less toxic pentavalent SbV ( 1922 ) such as sodium stibogluconate ( SSG ) , and since 2005 with miltefosine ( MIL ) that is freely supplied through a government-subsidized control program . The parasite developed resistance to both SbIII and SbV , and after ten years of clinical use there has been a notable decline in MIL efficacy ( Rijal et al . , 2013; 2007; Sundar et al . , 2012 ) . Leishmania parasites can re-shape their genome rapidly in vitro in response to stress ( Leprohon et al . , 2009 ) , suggesting structural variation is an important feature by which they can rapidly adapt to changing environmental conditions and drug pressure . However , there is little data on the diversity of clinical Leishmania populations or how they evolve during treatment . While an extensive literature has made use of molecular methods to study the population genetics of Leishmania ( e . g . Alam et al . , 2009; Lukes et al . , 2007; Mauricio et al . , 2006; Schonian et al . , 2008 ) , existing genetic markers have relatively poor resolution , and in particular L . donovani within the ISC show very little genetic differentiation based on these approaches ( Alam et al . , 2009; Downing et al . , 2012 ) . Whole-genome sequence data has the potential to show significant population structure within the ISC , and also allows us to identify changes in genome structure . Here we report the genome sequences of 204 L . donovani isolates ( Figure 1 , Supplementary file 1 ) , obtained from VL patients between 2002 and 2011 from regions in Nepal ( N=98 ) , India ( N=98 ) and Bangladesh ( N=8 ) that represent the epicentre of the on-going VL epidemic in the ISC ( Figure 1a ) . 10 . 7554/eLife . 12613 . 003Figure 1 . History and geography of Indian subcontinent L . donovani . ( a ) Location of the patients from which the 204 L . donovani genomes were isolated , and of historical Kala-Azar outbreaks . Genetic groups of the parasite isolates are indicated by the colour of the dots representing them , matching those in Figure 2a , c . Sampling dates and locations are summarised in Figure 1—figure supplement 1 , and detailed information about each strain including GPS coordinates are given in the source data file . Citations are to historical primary literature reviewed and cited in ( Gibson , 1983 ) . Posterior probability distributions of estimated ages for the oldest split in ( b ) the main population in Bihar and Nepal and ( c ) the ISC5 group associated with Sb resistance . Dark shading shows estimates under a strict molecular clock , light shading from relaxed molecular clock and lines show relaxed clock results with Bangladeshi and putative hybrid isolates included . ( d ) Estimated effective population size through time for ISC5 population ( green ) and the rest of the parasite population ( black/grey ) . Lines show median of posterior distributions , dark and light shading cover 50% and 95% of the posterior density respectively . Dates for all splits on this phylogeny and other results of phylogeographic analysis are shown in Figure 1—figure supplement 2 . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00310 . 7554/eLife . 12613 . 004Figure 1—figure supplement 1 . Sampling of genetic groups . Pie charts indicate the number of samples in each year ( columns ) , for each genetic group ( rows ) coming from each country ( grey shading ) . Horizontal lines connect and surround isolates of each group , with colours matching the groups shown in panels ( b ) and ( c ) . *8 samples from Bangladesh were all sampled in 2006 , and form a distinct population to Nepalese and Indian isolates ( ISC2 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00410 . 7554/eLife . 12613 . 005Figure 1—figure supplement 2 . Full results of discrete-space , constant population size molecular clock Bayesian phylogeography analysis of core population . ( a ) Maximum posterior probability phylogeny , with tips coloured by country of origin for sample ( green for India , blue for Nepal ) , and branches coloured by maximum posterior probability of country reconstructed in discrete phylogeography model . Values on nodes indicate posterior probability of assigned country/colour , with filled circles marking nodes with probability 1 . Other panels represent posterior probability distributions for rates of migration ( lineage switches per month ) from ( b ) Nepal to India and ( c ) from India to Nepal . Note the mode ( maximum posterior probability estimate ) for migration from Nepal is zero , but non-zero migration in the reverse direction is supported . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 005
Calling variants against a reference genome assembly for a Nepalese L . donovani strain ( BPK282/0cl4; Figure 2—figure supplement 1 ) , we identify three divergent genetic lineages circulating in this region ( Figure 2b ) : a core group of 191 closely related parasites found in the highly endemic lowlands of all three countries , a small population of 12 Nepalese isolates found most frequently in the highlands ( ISC1 ) and a single divergent Nepalese isolate ( BPK512/0cl9 ) ( Downing et al . , 2011 ) . These two main groups show fixed differences at 45 , 743 sites ( Supplementary file 2 , table a ) , and two previously sequenced Sri Lankan L . donovani isolates ( Zhang et al . , 2014 ) were more closely related to the core population ( 21 , 546 fixed differences ) than to ISC1 ( 45 , 743 fixed differences ) . Parasites within each group show little SNP variation with only 5 , 628 variable sites in ISC1 and just 2 , 418 sites varying within the core population ( Supplementary file 2 , table b ) and correspondingly few SNPs in protein-coding regions ( Supplementary file 2 , table c ) . Core population isolates differ at an average of 88 . 3 nucleotide sites with an average nucleotide diversity of 9 . 7 per Mb ( Supplementary file 2 , table d ) . 10 . 7554/eLife . 12613 . 006Figure 2 . Genealogical history of L . donovani from the ISC . ( a ) Maximum-likelihood tree based on SNPs called for 191 strains ( see Figure 2—figure supplement 1 ) from the core population in the Indian subcontinent . Samples are coloured by population assignment , with putative hybrid strains not clustered in the main groups in black . Further analysis confirms the hybrid ancestry of some of these isolates ( Figure 2—figure supplement 2 ) . ( b ) Unrooted phylogenetic network of the L . donovani complex based on split decomposition of maximum-likelihood distances between isolates described here , reference genome isolates and two published Sri Lankan isolates ( Zhang et al . , 2014 ) . ( c ) Model-based clustering of 191 isolates from the core population reveals six discrete monophyletic groups , and some groups and other samples of less certain ancestry . Coloured bars show the fraction of ancestry per strain assigned to a given cluster , with colours assigned to the population most closely related to each cluster . More detailed population clustering analysis shows largely congruent results ( Figure 2—figure supplements 3 and 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00610 . 7554/eLife . 12613 . 007Figure 2—figure supplement 1 . Flowchart of SNP detection using COCALL . Overview of the SNP detection method COCALL ( COnsensus of SNP CALL ) . COCALL finds genetic variants that show a concordant signal over five different SNP callers ( Cortex , Freebayes , GATK , Samtools Mpileup and Pileup ) . See supplementary methods for details . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00710 . 7554/eLife . 12613 . 008Figure 2—figure supplement 2 . Haplotype networks for core population isolates . Haplotype networks indicate putative hybrids as isolates with ancestry from multiple distinct populations . Chromosomal haplotype neighbour-joining networks of the phased data for the core population were constructed using the R ape package . Each node represents one haplotype variant for ( a ) chromosome 32 and ( b ) chromosome 35 , coloured by group . Black lines are network edges and red lines connect haplotype variants from the same isolate for selected isolates where haplotypes appear in different parts of the network ( with isolate names shown ) . Six ungrouped isolates ( BHU815/0 , BHU764/0cl1 , BHU274/0 , BHU574cl4 , BHU581cl2 , BHU572cl3 ) have mixed ancestry from ISC5 and other groups , and two ( BHU744/0 and BHU774/0 ) have a mix of ISC6/7/8/9/10 haplotypes . No mixing among ISC2/3/4 was evident . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00810 . 7554/eLife . 12613 . 009Figure 2—figure supplement 3 . Haplotype similarity for core population isolates . Heatmap showing the mean expected number of haplotypes shared between pairs of core population isolates . Samples listed on the y-axis are haplotype donors to those on the x-axis . 18 , 747 phased genotypes at 2397 SNPs sites were computed with Chromopainter v0 . 0 . 2 using recombination rates from PHASE for 79 haplotype chunks with c=0 . 00054 effective chunks . This image confirms six discrete populations ISC2-7 and illustrates complex ancestry in certain samples not belonging to these groups . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 00910 . 7554/eLife . 12613 . 010Figure 2—figure supplement 4 . Mosaic ancestry patterns in eight putative hybrid L . donovani isolates . Representative samples from three potentially parental groups ( BPK282/0cl4 , ISC6; BHU200/0 , ISC7; BPK275/0cl18 , ISC5 ) were compared to eight putative hybrid samples ( BHU815/0 , BHU764/0cl1 , BHU274/0 , BHU574cl4 , BHU581cl2 , BHU572cl3 , BHU744/0 and BHU774/0 ) . To the left is a maximum-likelihood tree constructed with RAxML showing the evolutionary history of the aligned haplotypes . The table shows a set of SNPs for which ChromoPainter assigned ancestry probability values >0 . 4 in any of these eight hybrids . Individual SNPs are coloured if the sample had an ancestry probability >0 . 4: uncoloured ones represent those observed in multiple ISC populations . All isolates have mixed ancestry from the two groups , but four isolates ( BHU574cl4 , BHU764/0cl1 , BHU815/0 and BHU274/0 ) have haplotypes that appear to have a more complex origin . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 010 While a panel of microsatellite markers shows no variation between isolates from the core population ( Downing et al . , 2012 ) , we reveal significant spatial and temporal genetic structure within this group despite this extremely low level of overall diversity ( Figure 2a , c; Supplementary file 2 , table e ) . Phylogenetic and clustering methods identify six congruent monophyletic groups ( ISC2-7 ) . Three other groups ( ISC8-10 ) and 21 ungrouped isolates had more complex and less certain evolutionary histories ( Figure 2a , c; Supplementary file 2 , table f ) . Most of the ISC groups are present throughout our sampling window ( 2002–2011 ) , and many are present in both India and Nepal ( Figure 1—figure supplement 1 ) . There are some exceptions: ISC7 represents a recent radiation ( first observed in 2006 ) with almost no diversity ( 20 unique SNPs; Supplementary file 2 , table g; π=1 . 8 per Mb ) and is restricted to India , while ISC6 is an older and more diverse group restricted to Nepal ( π=12 . 2 per Mb ) . We observe subsequent evolution within some groups: ISC5 is distinguished from other groups by just 32 SNP sites ( Supplementary file 2 , table h ) , but contains a subgroup with multiple novel SNPs and lower somy ( Supplementary file 2 , table i ) . Bayesian phylogenetic models in an explicit temporal and spatial framework revealed that the core population diverged in the mid 19th century ( Figure 1b ) , matching the dates of the earliest reports of large-scale VL outbreaks in the ISC ( Gibson , 1983 ) and thus suggesting that modern lowland parasites descend from these early epidemics . Within the core population , the Indo-Nepalese population itself appeared around 1900 ( Figure 1—figure supplement 2 ) , almost certainly in India ( 0 . 89 posterior probability ) , matching the dates of the first reported outbreaks in Bihar ( Gibson , 1983 ) , more precisely in Purnea ( Figure 2d ) . Most subsequent diversification is more recent , with many groups ( ISC2 & ISC4-6 ) radiating from the 1960s ( Figure 1b ) , coinciding with the end of the DDT spraying campaign . The estimated rate of migration from India to Nepal in the Core 191 group was significantly greater than that from Nepal to India , suggesting that India acts as a source population seeding the Nepalese epidemic ( Figure 1—figure supplement 2 ) . A lack of linkage disequilibrium decay between SNP pairs with genomic distance in the core population ( r2~0 . 33 at 5–1 , 400 kb ) reflects a lack of detectable recombination within the six main genetic groups ( ISC2-7 ) across the entire genome ( Supplementary file 2 , table j ) . While the low number of SNPs varying within the core population limits our power to detect recombination , we find compelling evidence of hybridisation among eight of the samples not assigned to any of the ISC groups ( Figure 2—figure supplements 2–4 ) . The identity of these isolates as hybrids and our assignment of other isolates to groups is supported by allele-frequency based methods ( f-statistics ) , which should be robust to gene flow between groups ( Supplementary file 2 , tables e , f ) and population structure analysis based on haplotype sharing ( Supplementary files 2 , tables k–n ) . The four-allele test also confirms that recombination is largely restricted to these hybrids ( Supplementary file 2 , table o ) . These isolates appear to result from multiple independent recent hybridizations between distinct ancestors of either ISC5 and ISC6 , ISC5 and ISC7 , or ISC6 and ISC7 ( Figure 2—figure supplement 2 ) . We detect extensive variation in the structure of these L . donovani genomes . Local copy-number variants ( CNVs ) cover ~11% of the genome . These include sporadic gene duplication , dynamic tandem gene array sizes ( Figure 3—figure supplement 1 ) and long sub-telomeric amplifications/deletions , the latter generally spanning whole transcription units . While structural variation in Leishmania is often considered a transient adaptation , particularly to culture conditions in vitro , we find striking conservation of many CNVs across all core population groups here . Two multigenic intra-chromosomal duplicated regions ( MAPK1 and H-locus; Downing et al . , 2011 ) are present in variable numbers in all core population isolates but are absent in ISC1 ( Figure 3b , c; Figure 3—figure supplement 2 ) . Conserved heterozygous SNPs in both of these structural variants confirm that these regions have duplicated once and been maintained throughout the evolution of this population . All known genes on these duplicated regions are associated with virulence ( MAPK1 , ASS , sAcP; Fernandes et al . , 2013; Lakhal-Naouar et al . , 2012; Wiese , 1998 ) or drug resistance ( Brotherton et al . , 2013 ) , indicating that extensive structural variation allows these parasites to alter local copy number in response to changing environments: both aneuploidy and CNV regulate gene expression ( Leprohon et al . , 2009 ) . Most isolates are aneuploid ( Figure 3—figure supplement 3 ) , even excluding the generally tetrasomic chromosome 31 , and almost all chromosomes show some variation in somy ( Figure 3a ) . Aneuploidy ( r2=0 . 15 , p=2 . 7x10-118 ) , CNVs ( r2=0 . 26 , p=7 . 5x10-218 ) and indels ( r2=0 . 30 , p=2 . 1x10-254 ) are significantly correlated with SNP variation in the core isolates , suggesting that these variants have appeared gradually during the evolution of the population in the field . Most strikingly , we find two cases of recent epidemic expansions associated with major changes in aneuploidy and heterozygosity ( Figure 4 ) . Variation in somy can thus lead to changes in heterozygosity , which could allow selection to eradicate recessive deleterious mutations in the absence of recombination ( Roze and Michod , 2010 ) . 10 . 7554/eLife . 12613 . 011Figure 3 . Structural variations in ISC L . donovani . ( a ) Stacked barplots per chromosome showing the proportion of ISC strains that are monosomic , disomic , trisomic , tretrasomic or pentasomic for the respective chromosome . A full breakdown of somy per strain is presented in Figure 3—figure supplement 3 , and a complete catalogue of other structural variants in Figure 3—figure supplement 1 . Violin plots showing the copy number of MAPK1 ( b ) and H-locus ( c ) per ISC group , except for ISC1 where these amplicons were absent . These amplicons are intra-chromosomal ( Figure 3—figure supplement 2 ) . ( d ) Tetrameric protein model of the transport protein aquaglyceroporin-1 . The C-terminus part that is affected by the 2-nucleotide frameshift found in all ISC5 isolates is shown in magenta . Image was created using PyMOL version 1 . 50 . 04 ( Schrödinger ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 01110 . 7554/eLife . 12613 . 012Figure 3—figure supplement 1 . Copy number variants in all 206 genomes . The position in the genome is shown on the y-axis , while individual isolates are shown on the x-axis . Colours of each copy number variant ( CNV ) represent the haploid depth variation ( D ) compared to the median depth for that chromosome ( see legend for colour key ) . When the depth of the majority of the strains is high like the episome in ch23 , this appears as a reduced depth in the strains that lack the episome . The length of each CNV is reflected by its length along the y-axis ( i . e . thickness of the line ) . Four major CNVs – gp63 , rDNA , an episome in ch23 and the MAPK amplification – are indicated with arrows . Group-specific copy number variants were highlighted with a box and numbered – detailed information about these CNVs are given in the table . The 206 samples included here are 204 ISC samples with L . infantum JPCM5 ( MCAN/ES/1998/LLM-877 ) and L . donovani LV9 ( MHOM/ET/1967/HU3 ) for reference . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 01210 . 7554/eLife . 12613 . 013Figure 3—figure supplement 2 . Copy number variation by intrachromosomal tandem duplication or extrachromosomal linear amplification in clinical isolate . ( a , c ) Chromosomes from L . donovani BPK282/0cl4 ( ISC6 ) , BPK380/0 ( ISC9 ) and BPK026/0cl5 ( ISC1 ) were separated by pulsed-field gel electrophoresis ( PFGE ) . ( b ) The MAPK1-locus and H-locus were detected by southern blot hybridization with probes specific for MAPK1 or HTBF , respectively . Hybridization was only observed in fragments of lengths equal to those of chr36 ( ~2 . 5 Mb ) and chr23 ( ~1 Mb ) and no additional smaller fragments were observed , indicating the absence of extra-chromosomal amplifications . ( d ) In contrast , linear extrachromosomal amplification ( as evidenced by a second and smaller band ) is shown for chromosome 35 in BPK380/0 by hybridization of a probe specific to the LinJ35 . 4130 gene . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 01310 . 7554/eLife . 12613 . 014Figure 3—figure supplement 3 . Chromosome number variation in L . donovani in the ISC . Average number of chromosomes found within each cell culture for each of the 36 chromosomes ( y-axis ) and each of the 204 L . donovani strains ( x-axis ) . Samples are coloured by population assignment following Figure 1c , with strains not clustered in the main populations shown in white . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 01410 . 7554/eLife . 12613 . 015Figure 4 . SNP heterozygosity and somy variation in two subclades . Two subclades show an expansion of polysomic strains from disomic ancestors ( below ) and an expansion of disomic strains from polysomic ancestors ( above ) . Somy variation per chromosome ( 1–36; above heatmap ) and the total number of heterozygote SNPs ( right to heatmap ) are shown for each individual strain . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 015 We find no statistically significant association between any individual SNP or structural genetic variant and in vitro SSG resistance , or SSG or MIL treatment outcomes ( Supplementary file 2 , tables p–r ) , but the distribution of antimony susceptibility was uneven across different ISC populations ( Supplementary file 1 ) . 9 of 11 ISC5 samples tested were highly SbV-resistant and two out of four ISC5-ISC6/7 hybrids tested have intermediate levels of resistance . One variant – a two-bp insertion introducing a frameshift and premature stop codon in the aquaglyceroporin-1 gene ( LdBPK_310030 , AQP1 ) – is homozygous in all 52 ISC5 isolates ( Table 1 ) , and heterozygous in six hybrids between ISC5 and either ISC6 or ISC7 . ISC5 isolates also share other genomic features – such as higher copy number of both the H-locus and MAPK1 amplicons ( Figure 3b , c ) . The H-locus includes MRPA , a gene involved in the efflux of SbIII and associated with drug resistance ( Leprohon et al . , 2009 ) . Other lines of evidence strongly link AQP1 with antimony resistance . While recent antimonial drugs such as SSG are compounds of pentavalent antimony ( SbV ) , SbV is thought to act mostly as a pro-drug , being reduced to SbIII in both the macrophage phagolysosome ( Frézard et al . , 2001 ) and in the parasite itself ( Denton et al . , 2004; Decuypere et al . , 2012 ) . AQP1 is known to assist with SbIII uptake , both genetic and transcriptional changes at this locus have been associated with Sb resistance ( Gourbal et al . , 2004; Monte-Neto et al . , 2015; Mukherjee et al . , 2013; Uzcategui et al . , 2008 ) and a homologous transporter is associated with drug resistance in trypanosomes ( Baker et al . , 2012 ) . Recently , an AQP1 knockout line of Leishmania major was shown to be resistant to SbIII due to reduced uptake ( Plourde et al . , 2015 ) . The truncated frameshift protein found in ISC5 is predicted to be incapable of forming a functional trans-membrane channel ( Figure 3d ) . We find three other independent frameshifts in AQP1 gene in other antimony resistant isolates , including one in BPK181/12 ( ISC6 ) , an isolate taken from a patient following failure of ten months of antimony treatment that was absent in the pre-treatment isolate from the same patient ( BPK181/ 0cl11 , Table 1 ) . 10 . 7554/eLife . 12613 . 016Table 1 . Small indels . The first half of the table summarises the numbers and types of indels detected in each group . The second half of the table shows the proportion of samples within a cluster that share each group-specific coding-region indel . DOI: http://dx . doi . org/10 . 7554/eLife . 12613 . 0161 . Number of indels ISC002 ISC003 ISC004 ISC005ISC006ISC007ISC008ISC009ISC010Total number ofindels found within each group 586073796555558460Number of group-specific Indels shared by part of the strains of that group951112708227Number of group-specific Indels shared by all the strains of that group613300002Number of group-specific Indels within coding regions 0123401312 . Indels within coding region Gene IDGene productPositionTypeLdBPK_310030AquaglyceroporinLd31_000777420 . 04LdBPK_310030AquaglyceroporinLd31_000773521LdBPK_310030AquaglyceroporinLd31_000766210 . 04LdBPK_310030AquaglyceroporinLd31_000809920 . 11LdBPK_291860Putative historie H2ALd29_0816454-20 . 25LdBPK_040410Conserved hypothetical proteinLd04_015549130 . 08LdBPK_070540Conserved hypothetical proteinLd07_0230487-30 . 12LdBPK_190080Conserved hypothetical proteinLdl9_001515110 . 04LdBPK_261790Conserved hypothetical proteinLd26_065174840 . 11LdBPK_301000Conserved hypothetical proteinLd30_0311376-10 . 02LdBPK_310690Conserved hypothetical proteinLd31_0241951-30 . 11LdBPK_332580Conserved hypothetical proteinLd33_099596010 . 04LdBPK_366590Conserved hypothetical proteinLd36_2473775-30 . 25LdBPK_110650hypothetical , unknown functionLdll_0245832-30 . 56LdBPK_292330hypothetical , unknown functionLd29_1008496-30 . 08 We propose that the AQP1 truncation is associated with antimonial resistance in the ancestor of ISC5 , and has been transmitted to a group of hybrid parasites . The ISC5 lineage emerged following the end of the DDT campaigns but then proliferated quickly in the 1970s ( Figure 1c , d ) , when antimonial dosage had to be doubled because of its declining efficacy . The persistence of this lineage beyond the era of Sb treatment perhaps reflects the increased fitness ( Vanaerschot et al . , 2013 ) of Sb-resistant parasites . These observations tally with a stronger signature of purifying selection on the ISC5 lineage , measured as a lower rate of derived allele accumulation compared to other ISC populations , which may be a consequence of higher historical exposure to drug stress ( Supplementary file 2 , table s ) . Sb resistance is also present in other genetic groups , with 4 out of 15 ISC4 lines tested in vitro being SbV-R , indicating resistance has emerged independently and recently multiple times in ISC L . donovani , and that other genetic variants responsible for SbV resistance must be present in this population . Indeed , other SbV resistance mechanisms are known in this population: previous work has shown that two resistant strains from ISC4 ( BPK087 and BPK190 ) show significantly decreased transcription of an AQP1 locus encoding a wildtype protein sequence ( Decuypere et al . , 2005 ) , and BHU764 combines a different indel mutation in AQP1 and reduced expression of MRPA , an efflux transporter of SbIII ( Mukhopadhyay et al . , 2011 ) . The failure of any single resistance locus to sweep through this population may reflect the low level of gene flow and the presence of a large reservoir of untreated asymptomatic cases ( Ostyn et al . , 2011 ) .
We have shown that genomic data can retrospectively unravel the evolution and epidemiology of this parasite population , and gain new insight into possible mechanisms of drug resistance against a background of extensive variation in genome structure . We report the first analysis of the structure and history of a Leishmania population , aligned with clinical and epidemiological records , enabled by the higher resolution of genome sequence data than other genotyping approaches . These data have allowed us to describe a mechanism of resistance to one of the most ancient drugs used in the human pharmacopeia , antimonials , not only identifying a key locus , but also showing the epidemiological dynamics of a population carrying a loss-of-function variant at this locus . Continued genetic surveillance of parasite populations is key to rapidly identify and respond to the emergence of treatment failure . In the recent emergence of artemisinin resistance in Plasmodium falciparum , genomic data has led to the identification of the major locus underlying resistance ( Ariey et al . , 2014; Cheeseman et al . , 2012 ) , revealed the genetic architecture of resistance ( Miotto et al . , 2015 ) and shed light on the population genetic context in which resistance is appearing ( Miotto et al . , 2013 ) . Genomic surveillance is playing a key role in defining the geographic boundaries of the spreading artemisinin-resistant population . Failure of anti-Leishmania chemotherapy could become a similar public health emergency: miltefosine has shown reduced efficacy in both India ( Sundar et al . , 2012 ) and Nepal ( Rijal et al . , 2013 ) . While amphotericin B is now being used against visceral leishmaniasis in ISC , few alternative treatments are available , and continued genomic surveillance will facilitate tracking the response of the Leishmania population to continued use of these drugs . Monitoring drug resistance in clinical settings is challenging: the data set we present was generated as part of a five-year collaboration between clinicians in the endemic countries , parasitologists and genome biologists . This collaboration is critical in generating data that reflects the evolution of parasite populations in close to real time and as such is directly applicable in a public health context . The data we present here provide baseline information on the diversity of Leishmania donovani in the ISC that will contribute to future studies of drug resistance and epidemiology of this population . Our results show the promise of genomic surveillance for other Leishmania populations , where patient symptoms , the parasites involved and the main treatment modalities all differ from those in the ISC ( Sundar and Chakravarty , 2015 ) .
The ethics committee of ( i ) the Nepal Health Research Council , Kathmandu , ( ii ) the Institute of Medical Sciences , Banaras Hindu University ( BHU ) , Varanasi , India and ( iii ) the corresponding bodies at the Institute of Tropical Medicine of Antwerp and the Antwerp University , Belgium , reviewed and approved the study protocol . Informed written consent was obtained from each patient or his/her guardian for those <18 years of age . All the patients and caretakers/parents had the study purpose explained to them in local language . A total of 204 parasite isolates were obtained from clinically confirmed VL patients in the high endemic regions of the Indian subcontinent ( ISC ) by the B . P . Koirala ( BPK ) Institute of Health Sciences in Dharan ( Nepal , Terai , N=98 ) , the Kala-azar Medical Research Center in Muzaffarpur ( India , Bihar , N=98 ) and the Mymensingh Medical College in Mymensingh ( Bangladesh , BD , N=8 ) . The Indian and Nepalese isolates were collected as part of a multi-center collaborative project to investigate drug resistance in ISC and were all typed as Leishmania ( Leishmania ) donovani . Complete clinical and epidemiological data were available for the Indian and Nepalese isolates ( Supplementary file 1 ) . The 204 L . donovani isolates were obtained from confirmed visceral leishmaniasis patients in previous clinical studies as described elsewhere ( Rijal et al . , 2013; 2007 ) . PCR-RFLP of the cysteine proteinase gene ( Quispe Tintaya et al . , 2004 ) typed all isolates as Leishmania donovani . Strain names consisted of 2–3 letters that indicated the location of isolation ( BD , BHU , BPK ) , 2–4 digits that indicated the patient number in that location , a forward slash followed by 1–2 digits that indicated when the sample was isolated ( 0: before treatment , 1: 1 month after treatment , etc ) and optionally the number of the parasite clone if the strain was cloned ( clone one is listed as 'cl1'; clone two is listed as 'cl2' , etc ) . Cloning was performed using the micro-drop method ( Van Meirvenne et al . , 1975 ) . Patient treatment outcome was monitored at the end of treatment and at 3 , 6 and 12 months post-treatment ) . Treatment non-response was defined as a case with positive parasitology at the end of the treatment course . Patients who were successfully cured at the end of treatment but in whom symptoms re-emerged within the 12 month follow-up period were classified as relapse cases . Patients who were cured at the end of treatment and remained cured within the 12 month follow-up period were classified as definite cures . If patients were lost to follow-up , the last known treatment outcome was recorded . Seven pre- and post-treatment samples coming from the same patients were obtained . Patient treatment outcome after treatment with miltefosine ( MIL ) and pentavalent antimonials ( SSG ) was monitored during 12 months ( at the end of treatment , month 3 , month 6 and month 12 after treatment ) . 50 strains were phenotyped for their susceptibility to SSG using a standardized in vitro susceptibility test as described elsewhere ( Downing et al . , 2011; Rijal et al . , 2007 ) . An SSG-susceptible reference strain ( BPK206/0cl10 ) was included in each assay . The classification into resistance and susceptible strains was determined by calculating the activity index ( AI ) : the ratio of the EC50 of the strain in question versus the EC50 of the susceptible reference strain . AI values clustered strongly , with most strains showing an AI≤1 ( 25; classified as SSG-sensitive ) or ≥6 ( 18; classified as SSG- resistant ) . A few strains ( 7 ) showed AI values around 3 and were considered as showing intermediate resistance . DNA isolation , sample preparation , DNA quantification and DNA library preparation were done as outlined previously ( Downing et al . , 2011 ) . 100 bp paired-end sequence reads were generated ( median coverage 44 per sample ) with the Illumina Hiseq 2000 platform according to standard protocols . Read data are available under study ERP000140 at the European Nucleotide Archive ( http://www . ebi . ac . uk/ena/data/view/ERP000140 ) . Reads were mapped to the reference L . donovani genome BPK282/0cl4 using Smalt v5 . 7 ( www . sanger . ac . uk/resources/software/smalt/ ) . Options for exhaustive searching for alignments and random assignment of repetitively mapped reads were used to properly estimate read coverage . Non-mapping read exclusion , read file merging , sorting and elimination of PCR duplicates were implemented with Samtools v0 . 1 . 18 and Picard v1 . 85 . The reference genome was masked at regions of the genome that were repetitive , duplicated , close to contig edges , structurally variable , or potentially mis-assembled . Five criteria masked a total of 6 , 358 , 203 bp out of the 32 , 444 , 998 bp reference genome sequence for L . donovani BPK282/0cl4 , resulting in SNPs being called at 26 , 086 , 795 or 80 . 4% of the nuclear genome . Criteria were: 1 . Manually identified repeats , commonly duplicated or deleted regions , regions with excessive rates of common SNPs and non-unique regions ( Downing et al . , 2011 ) identified with Gnuplot , the Artemis Comparison Tool , Artemis and Samtools tview ( 1 , 740 , 084 bp ) . 2 . Duplicated regions determined by DNA similarity as Blast v2 . 2 . 25 ( Altschul et al . , 1990 ) hits between the two reference genome sequences for L . donovani BPK282/0cl4 and L . infantum JPCM5 , with E-value less than 10e-20 ( 2 , 082 , 546 bp ) . 3 . Low complexity repeat regions determined by Tantan v0 . 13 ( www . cbrc . jp/tantan/ ) ; ( 2 , 495 , 070 bp ) . 4 . 100 bp regions adjacent to each contig edge ( 1 , 641 , 511 bp ) – initially 13 . 8% of candidate SNPs were in these regions . 5 . The first 300 bp and last 5 kb of all chromosomes , which are more likely to contain mis-assemblies . SNPs were ascertained using a consensus calling approach ( COCALL ) that is based on the framework outlined for the 1000 Genomes project ( 1000 Genomes Project Consortium , 2012 ) . COCALL applied five different variant detection approaches and combines evidence from them to calculate the support for each genotype . For complete details on the algorithm testing and development , see Appendix 1 . In short , this approach avoids bias associated with systematic errors unique to each individual SNP caller by examining their consistency and identifying discordant mutations symptomatic of false positives . The five callers used were FreeBayes v0 . 9 . 5 , GATK 2 . 0–38 , Samtools Pileup v0 . 1 . 16 and Mpileup v0 . 1 . 18 based on the DNA read mapping by Smalt , and Cortex v1 . 0 . 5 . 13 based on its own de novo assembly and mapping . In a large population of genetically homogenous strains , superior inference power was achieved by examining the population-wide genotype at each candidate SNP site ( i . e . population-based COCALL; Figure 2—figure supplement 1 ) . Candidate SNPs with genotype qualities of 40+ across all five callers were retained . SNPs with population normalized read depth ≤0 . 5 or ≥1 . 75 or with multiple derived alleles across the five callers were excluded . Candidate SNPs in soft-masked regions were accepted where the number of callers ≥3 . 5; those in non-masked regions were kept where the number of callers ≥2 . 5 . SNP sites retained in the final set of retained SNP sites were supported by a mean of 4 . 5 callers out of 5 . Chromosomal read depths were computed using a trimmed median read depth ( calculated as the median of read depths for sites with depths within one standard deviation of an initial , untrimmed , median read depth of each chromosome ) and normalized as the depth per haploid genome as outlined previously ( Downing et al . , 2011 ) . Somy levels were estimated as the median normalized chromosomal read depths ( Downing et al . , 2011 ) . Local copy number variants ( CNVs ) were detected where the local read depth was significantly different from the median depth of approximately 60 samples from ICS4 , ISC6 and ISC8 whose depth profile is similar to that of BPK282/0cl4 , and were measured with respect to the haploid depth to exclude somy variability . Two CNVs in particular , the MAPK1 and H-locus , were further investigated as they show functions potentially relevant to parasite adaptation ( Downing et al . , 2011 ) . A quantitative PCR assay in a subset of 46 samples was performed to confirm the copy number variation of the MAPK1 and H-locus amplicons . The nature of the amplification ( extra- or intra-chromosomal ) was determined by pulsed-field gel electrophoresis ( PFGE ) and southern blot hybridization comparing two strains that showed differential amplification of these loci ( ISC6 strain BPK282/0cl4: amplification; and ISC1 strain BPK026/0cl5: no amplification ) . To exclude the possibility that the amplicons are a culturing artefact , PCRs using primers that enabled amplification of circular episomes or tandem duplications was also attempted directly on five bone marrow samples from VL patients . Indels were detected using a consensus calling method based on the concordance of results across four tools: Cortex , Freebayes , GATK and Samtools Mpileup . For complete details on the Somy , CNV , indel and episome detection , see Appendix 2 . Haplotypes were inferred using PHASE v2 . 1 . 1 ( Stephens et al . , 2001 ) : 0 . 1% of genotypes in the Core 191 and 0 . 9% in ISC1 had confidence scores <0 . 95 . Haplotypes were inferred with a general recombination rate model ( Li and Stephens , 2003 ) with ten runs , each with a burn-in of 100 steps , 100 iterations and a single MC thinning step and recombination rate estimated for each chromosome . Convergence was examined for each chromosome: recombination estimates were consistent , though there was more variation between phasing runs for chromosome 16 in the core population and consequently inferred haplotypes are less certain for that chromosome . There was no correlation between the mean chromosome copy number and mean recombination rate or PHASE probability values for inferred haplotypes ( r2=0 . 011 ) . While variation in somy is not explicitly accounted for in the phasing process , the rapid flux in the somy levels of aneuploid chromosomes may mean this variation has no effect on haplotype inference . Of 3 , 567 heterozygous sites , 3 , 076 ( 86% ) had a PHASE probability of exactly 100% and 437 had PHASE probabilities < 0 . 95: these lower-confidence haplotypes were masked . Haplotypes for BHU1087/0 were inferred along with the core population . The phased core population SNP set had 2 , 401 SNPs: 17 singletons were masked . The smaller sample size meant that phasing within ISC1 was less successful: phase was successfully inferred for 2 , 308 sites using the same ( 0 . 95 ) confidence score threshold: 524 were not phased and were excluded from further analysis . No correlation between phasing confidence score and trisomy or tetrasomy was apparent . Linkage disequilibrium ( LD ) was inferred as the correlation in genotypes ( r2 values ) between SNP pairs using Bcftools v0 . 1 . 17 screened with Samtools Mpileup given SNP mapping qualities >30 and base qualities >25 . These pairwise r2 values were used to examine genome-wide LD patterns and LD decay with distance . Recombination was confirmed using the four-gamete test ( Hudson and Kaplan , 1985 ) . Mean chromosomal estimates of LD in the core population did not correlate with somy level if the tetrasomic chromosome 31 was excluded ( r2=0 . 001 ) but did if chromosome 31 was included ( r2=0 . 167 ) . Somy had little impact on the variance of LD per chromosome ( r2=0 . 017 with chr31 , r2=0 . 000 without chr31 ) . Variance in somy level across chromosomes had no association with either the mean or variance of LD per chromosome . We calculated zygosity as the probability that a SNP exists at a distance d from a SNP at a site x assuming diploidy ( Lynch , 2008 ) . No differences between homozygous and heterozygous SNP clustering measured as a product of chromosomal distance was observed . L . infantum JPCM5 ( MCAN/ES/1998/LLM-877 ) from Spain and LV9 ( MHOM/ET/1967/HU3 ) from Sudan were used ( Downing et al . , 2012 ) for comparison with the L . donovani genomes generated in this study . Variants were called for these two samples using the approaches outlined above . For two additional L . donovani isolates from Sri Lanka ( Zhang et al . , 2014 ) , we mapped Illumina GAII reads using Smalt v5 . 7 as above and called candidate SNPs at non-masked regions using Samtools Pileup v0 . 1 . 16 ( Li et al . , 2009 ) , followed by screening steps as above . Two genomes were excluded in the final analyses because sequence reads were of insufficient quality ( for MHOM/IN/10/BHU1087/0 ) or because of a suspected mixed infection ( for MHOM/IN/10/BHU790/0 ) . BHU790/0 is distantly related to the core population ( most likely ISC3 ) and appears to be a mixed infection rather than a hybrid because its average read allele frequency of heterozygous SNPs approximates 0 . 17 , whereas most detected hybrids had mean read allele frequencies of 0 . 4–0 . 5 . Remaining data were used to construct phylogenies using the 211 , 536 sites containing verified SNPs in the entire sample set ( ISC1/2/3/4/5/6/7/8/9/10 and ungrouped , LV9 , JPCM5 , BPK512/0cl9 ) . JPCM5 , LV9 , the two Sri Lankan isolates and one sample from our collection ( BPK512/0cl9 ) represented genetically distinct lineages , distinct to both the ISC1 ( n=12 ) and core populations ( n=191 ) . Seven SNPs in the core population and ten in ISC1 had multiple derived alleles compared to reference genome sample BPK282/0cl4 ( Supplementary file 2 , table t ) . These were included in all diversity analyses , but not those involving phased haplotypes . Genome-wide phylogenetic trees were constructed with RAxML v8 . 1 . 1 ( Stamatakis , 2014 ) using the GTR+G substitution model and 1000 bootstrap replicates for 10 runs for the core population ( 881 alignment patterns ) , ISC1 ( 349 alignment patterns ) , and all samples including the CL and VL samples from Sri Lanka ( Zhang et al . , 2014 ) ( 2274 alignment patterns ) . The best fitting substitution model determined using MEGA v6 ( Tamura et al . , 2011 ) for the core population was GTR+G . The final phylogenies were visualised using MEGA v6 ( Tamura et al . , 2011 ) and Splitstree v4 ( Huson and Bryant , 2006 ) . Unrooted haplotype trees for the phased SNPs for each chromosome were constructed from maximum-likelihood distances for the TN93 substitution model using the package Ape ( Paradis et al . , 2004 ) v3 . 1–4 in R version 3 . 12 . Samples in the core population of 191 isolates were classified using model-based clustering as implemented in Structure v2 . 3 . 2 . 1 ( Pritchard et al . , 2000 ) and principal component analysis ( PCA ) of the allele frequencies . Given a number of genetically distinct clusters ( K ) , samples were probabilistically assigned to a population independent of a mutation model with a prior of 1/K based on the correlation in genotypes of each sample with estimated population allele frequencies . 1≤K≤15 was examined with admixture and incomplete membership allowed to reduce overfitting . We used 105 burn-in steps before a run of 2x106 steps with three independent runs per K to confirm chain convergence . The most likely number of clusters was based on the second-order rate of change of the likelihood function ( ΔK , Evanno et al . , 2005 ) . At K=4 the groups were composed of ISC2/3/9/10 , ISC4 , ISC5 and ISC6/7/8 . Inter-population differentiation was lower for ISC2/3/9/10 ( FST=0 . 36 ) compared to the others ( 0 . 85<FST<0 . 98 ) . K=7 was the most probable K value ( ΔK=25 . 8 ) : the groups were composed of ISC2 , ISC3 , ISC4 , ISC5 , ISC6/7/8 , and ISC9/10 ( all FST>0 . 79 ) – the 21 ungrouped samples collectively had an FST=0 . Most population membership assignments were >0 . 97 with few ambiguous values ( range 0 . 80–0 . 97 ) . At K=9 , ISC6/7/8 split into ISC6 and ISC7/8 ( both FST>0 . 85 ) . At K=10 , ISC7/8 segregated into ISC7 and ISC8 . Dated phylogenies , historical population sizes and migration patterns were modelled for the 191 core clinical isolates using BEAST v1 . 8 . 1 ( Drummond et al . , 2012 ) . For molecular clock analyses , hybrid isolates not assigned to any of the ISC groups were removed from the dataset , as were the Bangladeshi outgroups for most analyses . Dates for each were fixed to the month of isolation , with sampling dates for those for which only isolation year data was available estimated during the MCMC but given a uniform prior on sampling ages within that year . Broadly consistent date estimates were obtained under three different models: with an uncorrelated lognormal relaxed clock model and a TVM substitution model and a Bayesian skyride model for population sizes , under the same model but with a strict clock model and finally under a GTR substitution model , with a simple constant population size coalescent model for data including the outgroups . Migration rate estimates were obtained by including a simple continuous-time Markov model of a discrete trait representing the country ( Nepal/India ) of isolation , so that ancestral states and rates of change in geographical location were estimated along the phylogeny . All analyses were made with a minimum of 8 independent MCMC runs , for 200 million update generations per run . Convergence was assessed by inspection in Tracer v1 . 6 , confirming that at least 5 of the 8 runs had converged to the same stationary distribution of parameters and that this had the highest likelihood . In most analyses , seven or eight chains all converged to the same posterior distribution , but the Bayesian skyride analyses converged more slowly . ESS estimates for almost all parameters across runs was over 500 , except for some skyride population size parameters . The first 20 million generations of each MCMC run were removed before combining all converged runs for inference . Historical population sizes were estimated both with the Bayesian skyride model and by transforming lineage-through-time data for all trees in the posterior probability distribution from the strict clock model above using the package Ape ( Paradis et al . , 2004 ) v3 . 1–4 in R version 3 . 12 . To compare population sizes between the drug resistant clade and others , we split ISC5 from other data and removed coalescent events between the ISC groups ( the oldest six ) to make these comparable with the ISC5 coalescence . f-statistics describe the correlation in allele frequencies between populations ( Patterson et al . , 2012; Reich et al . , 2009 ) . The simplest ( f2 ) is simply the sum-of-squares difference in allele frequency between two populations averaged across loci , and so captures the amount of divergence , or branch length between two populations . Two more complex statistics , f3 and f4 are calculated as differences between f2 statistics between groups of 3 and 4 related populations . f3 ( C;A , B ) has the property that , for a population C derived from populations A and B , it is expected to be positive if A , B and C are related by a simple history of divergence and genetic drift , but negative if admixture from A or B has contributed to the genetic composition of population C , while being robust to the details of the relationship between the three populations . In contrast , the value of f4 ( A , B , C , D ) does depend on the evolutionary history of populations A , B , C and D and so can be used to test a proposed relationship: if the four populations are related as ( ( A , B ) , ( C , D ) ) the f4 statistic is expected to be zero; for ( ( A , C ) , ( B , D ) ) it is expected to be positive and for ( ( A , D ) , ( B , C ) ) , negative . Finally , if the evolutionary history of three ancestral populations is known , the ratio of two f4 ratios is an estimate of the relative contribution of two potential parental populations to a fourth admixed population , given an outgroup . Whereas groups ISC2/3/4/5/6/7 seemed clearly defined phylogenetically and by Structure , ISC8/9/10 were not and no simple relatedness among the 21 ungrouped samples was detected . Consequently , we used inferred haplotypes to test whether these represented genetically discrete populations , or whether some of those samples were mixtures of ISC3/4/5/6/7 generated by hybridisation between these groups ( Lawson et al . , 2012 ) . Chromopainter v0 . 0 . 2 and FineStructure v0 . 0 . 2 inferred ancestral patterns of haplotype similarity among samples without a prior assumption of a given number of populations or of independence between mutations . Co-ancestry matrices for the core population were computed using Chromopainter v0 . 0 . 2 as the number of segments potentially donated to or received from individual samples , using the phased haplotypes . Recombination rates between pairs of SNPs inferred by PHASE were used for each of 36 unlinked chromosomes . Groups of SNPs on a single chromosome were expected to be exchanged as blocks of different sizes , so a higher number and longer lengths of shared blocks indicate recent common ancestry . The most likely ancestral sample or population was assigned according to its similarity to corresponding segments in a set of donor isolates . Two main datasets were generated by ChromoPainter: a co-ancestry matrix where all 191 could donate to all 191 as recipients ( 191x191 ) , and another where six representative samples were used as the only donors ( 191x6: BD09 for ISC2 , BPK067/0cl2 for ISC3 , BPK087/0cl11 for ISC4 , BPK275/0cl18 for ISC5 , BPK282/0cl4 for ISC6 , BHU200/0 for ISC7 ) . The expected number of chunks was minimised for the six representative samples , with k=80 segments and an effective number of chunks c=0 . 02 . Reducing the number of representative strains to represent distinct groups identified by Structure with smaller K parameters resulted in smaller k and larger c values , suggesting that using six representative samples was the optimal number for discrimination within the core population . Though ISC7 was a subset of ISC6 , ISC7 had a large number of fixed SNPs sufficient to differentiate it from ISC6 with Structure , so it was included . For the 191x191 comparison , k=79 segments was expected and the effective number of chunks was lower ( c=0 . 00054 ) because the total diversity of the donor set per SNP had decreased . These 191x191 and 191x6 co-ancestry matrices represented the most probable number of segments copied from each donor to each recipient , and also the relative probability of ancestry across the set of donors for each SNP for each sample . The number of donors per recipient was set to 100 . 20 expectation-maximisation algorithm iterations was sufficient to maximise the recombination-scaling coefficient ( Ne ) and copying probabilities with 10<k<1000 iterations across different number of donor samples assuming a minimum recombination rate of 10-15 Morgans/bp . For the 191x191 matrix , the Ne=523 . 3 and the mutation rate ( µ ) was 0 . 000181 . For the 191x6 matrix , the Ne=1015 . 1 and µ=0 . 000628: Ne and µ were higher because there were more mutations per sample . The 191x191 matrix was clustered for 106 MCMC ( Markov chain Monte Carlo ) steps with a burn-in of 10 , 000 and a skip of 100 steps using FineStructure v0 . 02 to obtain aggregated expected segment sharing between samples and populations with 100 trees examined per merge step . This distinguished complex ancestral patterns of segment sharing for the strains which Structure could not fully assign to single populations . To verify FineStructure and Structure results , the correlation in the SNP allele frequencies across samples was examined in the core population for six principal components with p<10-7 using PCA implemented by smartPCA in Eigensoft v4 . 2 ( Price et al . , 2006 ) . The first PC separated ISC2 ( 10 . 1% of all variation ) , the second ISC4 ( 6 . 4% ) , the third ISC5 ( 5 . 8% ) , the fourth ISC3 ( 4 . 9% ) , the fifth BPK035/0cl1 and BPK043/0cl2 ( 4 . 2% ) and the sixth a subset from ISC9/10 ( 3 . 9% ) . This was repeated for the 2353 variable sites in the core population ( ISC3/4/5/6/7/8/9/10 and ungrouped samples , n=183 ) excluding the 8 samples from Bangladesh ( ISC2 ) . This partitioned ISC5 ( PC1 , 7 . 4% ) , then ISC4 ( 6 . 8% ) , third ISC3 ( 5 . 6% ) , and fourth BPK035/0cl1-BPK043/0cl2 ( 4 . 8% ) . Eigenstrat and FineStructure PCA results were effectively the same but with some different axis labels – PC1 in the former was PC3 in the latter . FineStructure 191x191 ancestry patterns partitioned ISC4 vs ISC6 over PC1 ( 16 . 8% of variation ) , and ISC5 vs ISC6 over PC2 ( 15 . 5% ) . The next ( 12 . 3% ) differentiated ISC2 , and PCs 4 ( 6 . 4% ) and 5 ( 5 . 7% ) separated ISC3 . PC6 in FineStructure differentiated the BPK035/0cl1-BPK043/0cl2 pair . Information on in vitro SbV-resistance was available for 50/191 Core 191 isolates , from which 25 were sensitive and 25 resistant ( Supplementary file 1 ) . Links between genetic diversity ( SNP , indel , CNV and somy ) and in vitro SbV-resistance were assessed using the Fisher Exact test ( FET ) , Mann Whitney U-tests ( MWU ) and odds ratios ( ORs ) , implemented on 103 CNVs and 17 indels ( in 14 genes ) as well as 2 , 392 phased SNPs genotypes . SNPs were assigned to the 5’ and 3’ UTR if they were within 1 kb of the start or end of the gene ( respectively ) . To counter bias associated with the small sample size , FET and MWU were used initially . For the FET , variants were defined as discrete variables: SNPs as 0 , 1 or 2 non-reference alleles , and small indels as the diploid number of inserted or deleted basepairs . For the MWU , mutations were considered as a continuous variable such that the somy state was the haploid chromosome state , and CNVs were the haploid copy number times the somy state . The null hypothesis was that there were no significant genetic differences between SbV-R and SbV-S strains ( subject to p<0 . 01 ) . The FETs and MWU were limited by the partial association of different mutations with the phenotypes , so we examined ORs of the derived alleles segregating in multiple ISC populations with 6+ non-reference alleles for which the absolute difference in SbV-R and SbV-S allele frequencies >0 . 1 using the log-scaled EC50 values . We compared the log-scaled EC50 values of each allele pair using t-tests . We also examined samples for which the patient was treated with SbV and was either cured or not , samples for which the patient was treated with miltefosine ( MIL ) and was either cured or not , and also in vitro MIL resistance levels as implemented above for SbV . Evidence of historical differences in selective processes on the ancestors of the major ISC populations was assessed as the rate of accumulation of derived alleles . Stronger purifying selection should purge deleterious derived alleles more quickly , detected as an excess of nonsynonymous changes relative to synonymous ones , as previously observed for ISC isolates ( Downing et al . , 2011 ) . This signature should be most apparent for derived alleles , which should accumulate at a net rate dependent on the historical effective population sizes and selective coefficients . Using L . infantum JPCM5 as the outgroup , the relative abundance of derived alleles in one population that were absent in the other for each ISC population pair ( ISC2-7 ) were determined as the statistic R ( Do et al . , 2015 ) . The associated ratio R2 denoted the relative rate of homozygous derived allele accumulation between populations . R and R2 should approximate 1 assuming no difference in the strength of selection , and primarily depend on the derived allele frequency per population , so the main confounder was variance in historical effective population sizes among ISC populations . To calculate confidence intervals for these R values that take into account correlation between neighbouring sites , we used a Weighted Block Jackknife by splitting the SNPs according to chromosome ( Busing et al . , 1999 ) to counter the extensive linkage disequilibrium between SNPs ( Moorjani et al . , 2011 ) : discrete chromosomal blocks may still be linked . This was adjusted for the number of SNPs per block to reflect the variability in the relative selective pressure ( Kunsch , 1989 ) . A threshold of four times the standard error of these jackknife estimates was used as a criteria for identifying comparisons deviating significantly from expected values ( Do et al . , 2015 ) . A protein model of the intact Leishmania donovani AQP1 from BPK282/0cl4 was created using MODELLER 9 . 14 ( Sali and Blundell , 1993 ) . The template for homology modelling was the crystal structure of the aquaglyceroporin from Plasmodium falciparum in complex with glycerol ( PDB code: 3c02 ) published by Newby and co-workers ( Newby et al . , 2008 ) . The sequence identity between the target and the template was approximately 33% . PyMOL version 1 . 50 . 04 ( Schrödinger ) was used to generate the biological units for the aquaglyceroporin from Plasmodium falciparum ( generation of symmetry mates function in pymol ) . The C-alpha atoms of chain A , B , C and D of the tetramer template were restrained during homology modeling using MODELLER in order to reduce the number of interatomic distances that needed to be calculated . | The parasite Leishmania donovani causes a disease called visceral leishmaniasis that affects many of the world's poorest people . Around half a million new cases develop every year , but health authorities lack safe and effective drugs to treat them . Up to 80% of these cases occur in the Indian subcontinent , where devastating epidemics have occurred in the last decades . One reason these epidemics continue to occur is that the parasites develop genetic mutations allowing them to adapt to and resist the drugs used to kill them . As there are few existing drugs that can kill L . donovani , it is crucial to understand how drug resistance emerges and spreads among parasite populations . Imamura , Downing , Van den Broeck et al . have now investigated the history of visceral leishmaniasis epidemics by characterising the complete genetic sequence – or genome – of 204 L . donovani parasite samples . This revealed that the majority of parasites in the Indian subcontinent first appeared in the nineteenth century , matching the first historical records of visceral leishmaniasis epidemics . The genomes show that most of the parasites are genetically similar and can be clustered into several closely related groups . These groups first appeared in the 1960s following the end of a regional campaign to eradicate malaria . The most common parasite group is particularly resistant to drugs called antimonials , which were the main treatment for leishmaniasis until recently . These parasites have a small genetic change that scrambles most of a protein known to be involved in the uptake of antimonials . Parasites may also be able to develop resistance to drugs through additional mechanisms that allow them to produce many copies of the same gene . These mechanisms could allow the parasites to rapidly adapt to new drugs or changes in the populations it infects . The work of Imamura et al . looks only at parasites isolated from patients then grown in the laboratory , so further research is now needed to explore how variable the Leishmania genome is in both of the parasite’s hosts: humans and sandflies . Imamura et al . ’s study reveals how L . donovani has spread throughout the Indian subcontinent in fine detail . The genome data can be used to create simple molecular tools that could form an "early warning system" to track the success of disease control programs and to determine how well the current drugs are working . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2016 | Evolutionary genomics of epidemic visceral leishmaniasis in the Indian subcontinent |
Sleep plays a crucial role in the consolidation of newly acquired memories . Yet , how our brain selects the noteworthy information that will be consolidated during sleep remains largely unknown . Here we show that post-learning sleep favors the selectivity of long-term consolidation: when tested three months after initial encoding , the most important ( i . e . , rewarded , strongly encoded ) memories are better retained , and also remembered with higher subjective confidence . Our brain imaging data reveals that the functional interplay between dopaminergic reward regions , the prefrontal cortex and the hippocampus contributes to the integration of rewarded associative memories . We further show that sleep spindles strengthen memory representations based on reward values , suggesting a privileged replay of information yielding positive outcomes . These findings demonstrate that post-learning sleep determines the neural fate of motivationally-relevant memories and promotes a value-based stratification of long-term memory stores .
It is well established that sleep contributes to memory consolidation processes ( Diekelmann and Born , 2010; Maquet , 2001; Stickgold and Walker , 2013 ) . In animals and humans memory traces are reinforced through neural reactivation and trimmed through synaptic downscaling . Critically , both neural replay and synaptic downscaling entail the tagging of neural elements coding for specific memories and coincide with the occurrence of specific oscillatory elements in sleep such as sleep spindles , and slow wave activity ( Ringli and Huber , 2011; Tononi and Cirelli , 2003; Schwindel and Mcnaughton , 2011; Wilhelm et al . , 2011; Káli and Dayan , 2004 ) . For memory consolidation to be adaptive , information that is critical for survival , such as stimuli with an emotional or rewarding value , should be consolidated in priority during sleep ( Perogamvros and Schwartz , 2012; Sterpenich et al . , 2009 ) . Consistent with this hypothesis , behavioral studies in humans suggest that reward may modulate sleep-related memory consolidation ( Fischer and Born , 2009; Werchan and Gómez , 2013; Oudiette et al . , 2013 ) . At the neural level , animal studies provide evidence for spontaneous coordinated replay of neural activity in memory and reward brain structures during slow-wave sleep ( Lansink et al . , 2009; Pennartz et al . , 2004 ) , suggesting that the activation of reward circuits during sleep influences neural plasticity . Further , within the rat hippocampus , the firing patterns of CA1 and CA3 neurons encode the high or low reward outcome of items in an associative memory task ( Mckenzie et al . , 2014 ) . In humans , at wake , the activation of the mesolimbic dopaminergic system enhances associative memory through interactions with the hippocampus ( Shohamy and Wagner , 2008 ) , as well as subjective confidence and feeling of goal attainment during successful retrieval ( Shohamy and Wagner , 2008; Adcock et al . , 2006; Wolosin et al . , 2012 ) . Yet , it is unclear whether the replay of rewarded memories during sleep primarily involves the transfer of hippocampal memories to cortical sites ( Frankland and Bontempi , 2005; Gais et al . , 2007; Takashima et al . , 2009 ) , or the strengthening of memory-reward associations , implicating hippocampal-striatal interactions ( Lansink et al . , 2009 ) . Based on these findings and recent theoretical proposals suggesting that sleep may influence associative memory by facilitating the integration of multi-item sequences ( Stickgold and Walker , 2013; Ellenbogen et al . , 2007 ) , a critical and unanswered question that we ask here is whether reward values linked to recent memory items control the remodeling of associative knowledge during sleep . Specifically , we test whether reward influences sleep-dependent memory consolidation by promoting associative processes in the hippocampus , ultimately prioritizing long-term retention and subjective confidence for highly ( over lowly ) rewarded stimuli .
We designed an associative memory task in which participants learned series of pictures yielding high or low reward outcomes . To assess the influence of sleep and reward , and their potential interaction , memory for these series was tested following a nap or a rest period , and then again 3 months later during the retest session ( Figure 1A ) . The task was composed of eight series of six successive pictures each; four series were associated with high reward outcome ( HR: dollar coin ) and four with low reward outcome ( LR: cent coin ) all participants received the same total payment . The participants first saw the eight series once during the encoding session ( Figure 1B and Figure 1—figure supplement 1A ) , followed by the 2-alternative forced choice learning session with feedback during which participants successively selected the next picture in the series among two presented pictures Figure 1—figure supplement 1B ) . Then , the participants either took a 90-min nap ( Sleep group , n = 16 ) or spent an equivalent period quietly awake ( Wake group; n = 15 ) , both monitored by polysomnography EEG . After this delay and again three months later , the participants performed an associative memory task on pairs of pictures with different relational distances ( direct , inference of order 1 and order 2 ) . They were presented with one image and were then asked which one among two images was part of the same series as the first image ( while the other image belonged to a different series ) and gave confidence ratings for each answer ( Figure 1C and Figure 1—figure supplement 1C ) . Functional MRI ( fMRI ) data were acquired during all experimental sessions and analyzed using SPM8 ( see Materials and methods ) . 10 . 7554/eLife . 07903 . 003Figure 1 . Experimental design . ( A ) Overview of the experimental protocol for the Sleep group ( upper line ) and the Wake group ( lower line ) , composed of three main MRI sessions , each preceded by PVT . Learning and test sessions were performed during one afternoon and separated by a nap ( Sleep group ) or rest ( Wake group ) interval monitored by EEG . The retest session occurred three months later . ( B ) Each series of pictures started by a high ( dollar symbol ) or low ( cent symbol ) reward cue and was composed of the following series of pictures: pillow , sofa , kitchen , bedroom , house , and landscape . ( C ) Examples of direct trials ( left ) , inference of order 1 trials ( middle ) and inference of order 2 trials ( right ) . Direct trials were used during the learning , test , and retest sessions , while inference 1 and 2 trials were only used during the test and retest sessions . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 00310 . 7554/eLife . 07903 . 004Figure 1—figure supplement 1 . Experimental procedure . ( A ) Encoding: example of the presentation of the successive pictures of a high reward outcome series . Each picture was presented for 2 s followed by a 2 s fixation cross . ( B ) Learning: example of the presentation of the successive pictures of a high reward outcome series yielding three good responses out of five . Each picture was presented for 2 s followed by a 2 s fixation cross and by the choice screen ( red ) . For 2 s participants could not answer , they just examined the choice screen before the question screen ( ‘choose the next picture’ ) appeared , prompting them to press the left or right button on the button box . Participants had a maximum of 8 s to answer before the next picture appeared . We used the choice screen ( red ) as onset times for all reported fMRI analyses to minimize motion-related artifacts . ( C ) Test and retest: example of an inference of order 1 trial . The picture was presented for 2 s followed by a 2 s fixation cross and by the 2 s choice screen ( red ) . Participants answered after the ‘choose the next picture’ sentence appeared on the screen . Then , a confidence choice screen was displayed for 2 s , during which participants were invited to think about how they answered the trial . After the ‘how did you answer the trial’ sentence appeared on the screen , participants selected one of the four response options proposed by pressing the corresponding button on the button box . For all choices , participants had a maximum of 8 s to answer . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 004 First , we assessed the effect of reward on learning performance ( hit rate ) during the learning phase using an ANOVA with Reward ( high , low ) and learning Block ( 1 , 2 , 3 , see Materials and methods ) as within-subject factors and Group ( Sleep , Wake ) as between-subject factor . Performance was superior for high reward ( HR ) than for low reward ( LR ) picture pairs ( F ( 1 , 28 ) = 35 . 86 , p<0 . 001 ) , with a significant learning effect over Blocks ( F ( 2 , 56 ) = 20 . 68 , p<0 . 001 ) , and a Reward x Block interaction effect ( F ( 2 , 56 ) = 4 . 72 , p=0 . 012 ) . Importantly , there was no Group difference ( F ( 1 , 28 ) = 2 . 59 , p>0 . 05 ) and participants of the two groups reached similar levels of performance for HR and LR trials at the third block of the learning ( post-hoc test p>0 . 05; Figure 2—figure supplement 1A ) . For all three learning blocks , performance was above chance level ( one-way T-tests comparing performance to 50% , all p<0 . 05 ) . Reward-related performance improvement during the learning phase ( HR vs . LR ) was paralleled by an increase in midbrain activity , in a region compatible with the ventral tegmental area ( VTA ) [z-score = 3 . 71 ( -3x , -13y , -20z ) , p<0 . 05 , small-volume corrected ( SVC ) for familywise error ( Bunzeck and Düzel , 2006 ) ; see Materials and methods] ( Figure 2—figure supplement 1B and Supplementary file 2 ) . Next , we tested whether the consolidation of rewarded associative memory was selectively enhanced by post-encoding sleep . After a nap ( Sleep group ) or rest ( Wake group ) , participants were tested on direct pairs of pictures in the series , and on non-consecutive pairs of pictures , that is inference of order 1 and 2 measuring associative memory , namely the strength of the integration of non-consecutive images in a sequence ( Ellenbogen et al . , 2007 ) ( Figure 1C and Figure 1—figure supplement 1C ) . Both groups performed above chance level for HR and LR trials and direct , inference 1 and inference 2 trials ( one-way T-tests p<0 . 05 ) . Participants performed better for HR than LR trials ( F ( 1 , 84 ) = 15 . 88 , p<0 . 001 ) , and for close relational distance trials ( F ( 2 , 84 ) = 4 . 16 , p=0 . 018 ) . Further , the Sleep group performed better than the Wake group ( F ( 1 , 84 ) = 4 . 52 , p=0 . 036; Figure 2A ) . Notably , within HR trials , the sleep Group performed better than the wake group ( F ( 1 , 84 ) = 5 . 07 , p=0 . 027 ) , which was not the case for LR trials ( F ( 1 , 84 ) = 1 . 52 , p=0 . 22 ) . However the Group x Reward interaction was not significant ( F ( 1 , 84 ) = 0 . 23 , p=0 . 63 ) . In the Sleep group , the number of slow spindles ( 11–13 Hz ) correlated with memory improvement from learning to test specifically for HR ( R = 0 . 69 , p<0 . 001 ) and not for LR trials ( R = -0 . 27 , p>0 . 05; Figure 2B ) . The correlation for HR trials was significantly higher than that for LR trials ( Fisher’s z-score = 2 . 15 , two-tailed p=0 . 01 ) ( Lee and Preacher , 2013 ) . Additionally , the correlation effect was not linked to sleep duration as there was no correlation between performance improvement in the HR trials and sleep duration , while the correlation with performance improvement in HR persisted when considering slow spindles density ( number of slow spindles per second of sleep; R = 0 . 586 , p=0 . 028 ) . 10 . 7554/eLife . 07903 . 005Figure 2 . Test results . ( A ) Better performance for the Sleep group than the Wake group and also for High than Low reward trials . ( B ) Memory improvement for HR series correlated with the number of slow spindles . ( C ) Increased right hippocampal activity for HR than LR for the Sleep compared to the Wake group . ( D ) Increased right hippocampal response for HR inference of order 2 compared to inferences of order 1 correlated with the number of slow spindles for the Sleep group . All activation maps are displayed on the mean T1 anatomical scan of the whole population . For display purposes , hippocampal activations are thresholded at p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 00510 . 7554/eLife . 07903 . 006Figure 2—figure supplement 1 . Learning results . ( A ) Performance increased along the three blocks , better results for HR than for LR trials in Block 1 and 2 , but not 3 , leading to same baseline for future consolidation . ( B ) Midbrain activity mediated reward-related learning enhancement . Activity in a midbrain region compatible with the VTA ( -3x , -13y , -20z ) showed a selective increase for HR versus LR trials , correlating with individual increase in performance between HR and LR trials . Activation map displayed on the mean proton-density weighted scan computed over the whole population . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 00610 . 7554/eLife . 07903 . 007Figure 2—figure supplement 2 . Behavioral results at test . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 007 To investigate fMRI responses from the Sleep and Wake groups during the test session , we used a general linear model distinguishing successful trials according to Reward ( HR , LR ) , Relational Distance ( direct , inference 1 , inference 2 ) and Confidence ( ‘certain’ vs lower confidence judgments , i . e . ‘not sure’ , ‘guess’ , ‘by elimination’ ) . A direct comparison between the Sleep and Wake groups for HR versus LR ( independently of Relational Distance ) revealed increased activity in the right hippocampus [z-score = 4 . 04 ( 30x , -10y , -23z ) , p<0 . 05 SVC; Figure 2C ) ] . Within this interaction , post-hoc analysis of the extracted beta values showed that both for Sleep and Wake groups HR condition was different from LR condition ( both p=0 . 02 ) . Because the integration of distant associative memories possibly relies on sleep spindles ( Tamminen et al . , 2010 ) , we tested for the brain correlates of this functional relationship . To test specifically for the integration of associative memories ( and not the consolidation of previously seen associations , as in direct trials ) , we considered associations between non-consecutive images with distinct relational distances by comparing inference 2 to inference 1 trials ( inference 2 > inference 1 ) . We found that the number of slow spindles in the Sleep group positively correlated with activation in the right hippocampus for HR distant associations [HR inference 2 versus HR inference 1; z-score = 3 . 75 ( 27x , -16y , -11z ) , p<0 . 05 SVC; Figure 2D] , and not for LR associations . We assessed long-term memory in a retest session identical to the test session , three months later . Both Sleep and Wake groups performed above chance level for HR and LR series ( one-way T-tests , all p<0 . 05; Figure 3A ) . While we found no main effect of Group , Reward , or Relational Distance ( ANOVA all p>0 . 05 ) , we found a Group x Reward interaction ( F ( 1 , 66 ) = 4 . 21 , p=0 . 044 ) . Only the Sleep group remembered HR compared to LR series better ( T ( 32 ) = 2 . 91 , p=0 . 006 ) , while no such difference was found in the Wake group ( T ( 38 ) = -0 . 25 , p=0 . 80 ) . In fMRI , we tested for Group differences between HR and LR trials according to Relational Distance . We observed a selective increase of left hippocampus activity for HR versus LR during inference 2 trials in the Sleep group compared to the Wake group [z-score = 3 . 78 ( -36x , -28y , -8z ) , p<0 . 05 SVC; Figure 3B] , we observed no Group difference for direct and inference 1 trials . Post-hoc analyses of the hippocampal beta values showed that the interaction was due to an increased hippocampal response to HR vs . LR trials in the Sleep group only ( p=0 . 03 ) . Further , using psychophysiological interaction ( PPI ) , we asked whether sleep had some long-term effects on the functional coupling between the hippocampus and other brain regions during the processing of HR ( vs . LR ) trials ( see Materials and methods ) . Both the caudate nucleus [z-score = 3 . 41 ( 15x , 20y , 7z ) ; p<0 . 05 SVC] and the medial prefrontal cortex [z-score = 2 . 99 ( 15x , 59y , -5z ) ; p<0 . 05 SVC] showed such pattern of increased functional connectivity in the Sleep ( compared to the Wake ) group ( Figure 3C ) . Because striatal activation may be relevant for the reprocessing of rewarded associations during sleep ( Lansink et al . , 2009 ) , we tested whether the strength of the caudate-hippocampal functional coupling correlated with the number of slow sleep spindles during the nap in the Sleep group , and found that this correlation was significant ( R = 0 . 591 , p=0 . 028 , Figure 3D ) . Note that no such correlation was observed for the connectivity between the hippocampus and medial prefrontal cortex ( mPFC ) . 10 . 7554/eLife . 07903 . 008Figure 3 . Retest results . ( A ) The Sleep group performed better for HR trials than for LR trials . ( B ) Increased left hippocampus activity during the retest session for HR vs . LR for the Sleep compared to the Wake group , selectively during inference 2 trials . ( C ) PPI for the retest session , using the seed in the right hippocampus from Figure 2C . Increased functional coupling with the caudate nucleus and the medial prefrontal cortex during HR vs . LR trials , selectively for the Sleep group compared to the Wake group . ( D ) Beta values of the PPI around the caudate nucleus peak correlated with the number of slow sleep spindles . Activation map displayed on the mean T1 anatomical scan of the population . For display purposes , hippocampal activations are thresholded at p<0 . 005 . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 00810 . 7554/eLife . 07903 . 009Figure 3—figure supplement 1 . Detailed retest behavioral results . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 009 Beyond performance effects , it has been shown that remembering emotional events elicits a stronger feeling of recollection ( Sharot et al . , 2004 ) , yet higher confidence seems adaptive only if associated with accurate memory recall ( Schultz , 2006 ) . We therefore analyzed ‘certain’ confidence responses . To correct for potential performance biases , we considered the percentage of ‘certain’ responses for hits only . For the test session , we found a main effect of Group ( F ( 1 , 78 ) = 4 . 90 , p=0 . 029; more ‘certain’ responses for the Sleep group ) , of Reward ( F ( 1 , 78 ) = 59 . 06 , p<0 . 001; higher confidence for HR ( Figure 4A ) . Further , we found a Group effect ( Sleep>Wake ) within HR hits ( F ( 1 , 78 ) = 6 . 63 , p=0 . 01 ) , and not for LR hits ( F ( 1 , 78 ) = 0 . 73 , p=0 . 39 ) . Importantly , the Group effect was not due to an unspecific influence of sleep , as confidence judgments for incorrect responses did not differ between the groups , neither for HR trials ( U = 72 . 50 , p=0 . 35 ) nor for LR trials ( U = 84 , p=0 . 51 ) . Note that we performed U-Tests because of small sample size and non-normal distribution of incorrect responses . For the retest session , we found a main effect of Reward ( higher confidence for HR trials; F ( 1 , 66 ) = 18 . 92 , p<0 . 001 ) and a post-hoc difference between HR and LR confidence within the Sleep group only ( p=0 . 003 for the Sleep group and p=0 . 38 for Wake group ) ( Figure 4B ) . At retest , there were very few certain responses . To obtain correctly sized samples to analyze confidence judgments at retest , we grouped ‘certain’ , ‘by elimination’ and ‘not sure’ answers together and contrasted them to ‘guess’ answers . In line with our main imaging results for inference 2 trials ( Figure 3B ) , this analysis on confidence yielded higher activation in the parahippocampus [z-score = 3 . 49 ( 27x , -25y , 26z ) p<0 . 05 SVC] and in the putamen [z-score = 3 . 46 ( -18x , -1y , 7z ) , p<0 . 05 SVC] for Sleep > Wake selectively for HR inference 2 when comparing some confidence to guess ( Figure 4C ) . 10 . 7554/eLife . 07903 . 010Figure 4 . Confidence results . ( A ) Higher confidence on hits for the Sleep group than the Wake group and for HR vs . LR trials during the test session . ( B ) Higher confidence on HR vs . LR hits for the Sleep group during the retest session . ( C ) Greater activation of the left caudate and right hippocampus for the ‘certain & by elimination & not sure’ vs ‘guess’ confidence judgments for inference 2 HR trials in the Sleep compared to the Wake group during the retest session . Activation map displayed on the mean T1 anatomical scan of the population . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 010
Here we show that sleep favors the selectivity of memory consolidation and confidence for these memories by promoting the integration and long-term retention of the most important ( i . e . here highly rewarding ) memories . First , our results show that slow sleep spindles were critically involved in the consolidation of series with high reward outcome ( Figure 2B ) , consistent with Schmidt et al . ( Schmidt et al . , 2006 ) who reported that memory improvement for word-pairs over a nap implicated slow spindles . Sleep spindles may induce long-term synaptic plastic changes ( Marshall et al . , 2006 ) , thus consolidating newly learned information into a more stable form of long-term memory . Slow spindles ( 11–13 Hz ) are dominant over the prefrontal lobe ( Saletin and Walker , 2012 ) and are related to a cross-linking of transferred information within prefrontal circuitry ( Clemens et al . , 2007 ) . Moreover , activity in dopaminergic reward regions such as the VTA is increased during slow spindles in humans ( Schabus et al . , 2007 ) . Consistent with these observations , we found that slow spindles after learning were related to the functional interplay between the hippocampus and the caudate nucleus during the long-term recall of high rewarded series . From these connectivity findings , we conclude that slow spindles may primarily favor the long-term consolidation of rewarded stimuli across striatal and hippocampal networks . We may speculate that enhanced hippocampal recruitment for HR trials at retest could possibly foster recall processes subserved by the mPFC , a brain region reportedly involved in the retrieval of episodic memories ( Düzel et al . , 1999 ) . Second , our results confirm that sleep does not only strengthen memory for recently ( and possibly more strongly ) encoded items ( Diekelmann and Born , 2010 ) but also boosts the integration of associative memories , via a hippocampus-dependent mechanism ( Werchan and Gómez , 2013; Ellenbogen et al . , 2007; Lau et al . , 2010 ) ( Figure 2D ) . Importantly , there was no difference between reaction times for inference trials and direct trials ( Table 1 ) , supporting the idea that the consolidation of associative memory involved the integration of discrete events ( i . e . , generalization of memories ) rather than sequential inferential reasoning ( Shohamy and Wagner , 2008 ) . This mechanism could promote the conversion of implicit forms of memory into more explicit and conscious memories ( Wilhelm et al . , 2013 ) , and also facilitate the access to remote associations ( Zeithamova et al . , 2012 ) . 10 . 7554/eLife . 07903 . 011Table 1 . Reaction times ( mean ± SEM ) for the test phase . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 011HRLRDirectInference 1Inference 2DirectInference 1Inference 2SleepMean939 . 83 ± 19 . 161144 . 29 ± 21 . 511167 . 24 ± 22 . 351207 . 46 ± 21 . 361434 . 43 ± 22 . 691318 . 69 ± 22 . 61Median752 . 07 ± 18 . 591006 . 20 ± 21 . 391024 . 83 ± 21 . 861066 . 83 ± 20 . 121385 . 43 ± 23 . 821132 . 13 ± 22 . 21WakeMean1164 . 14 ± 19 . 181353 ± 27 . 811405 . 74 ± 21 . 911220 . 55 ± 20 . 871305 . 55 ± 21 . 741290 . 21 ± 26 . 50Median994 . 17 ± 20 . 831226 . 83 ± 29 . 181261 . 27 ± 22 . 991067 . 90 ± 21 . 861197 . 50 ± 23 . 131231 . 40 ± 27 . 78 Third , we show that sleep after learning reinforces the subjective feeling of recollection for correct recall ( Figure 4A ) , and at long-term retest selectively enhances performance for rewarded information ( Sharot et al . , 2004 ) ( Figure 3A ) , although this interaction effect was not present at short-term . These effects were not present for missed trials and were thus not due to a general increase in confidence judgments after sleep . Moreover , we report that medium to high confidence is characterized by parahippocampal and striatal activation at retest . Both these structures have previously been linked to confidence measures ( Daniel and Pollmann , 2012; Eichenbaum et al . , 2007 ) , with parahippocampal activation involved in remembering contextual information about a memory , which contributes to increased confidence about the information being recalled ( Eichenbaum et al . , 2007 ) , while striatum ( in particular putamen ) activation has been shown to reflect an internal error signal about probable outcome in the absence of feedback ( or prediction error of confidence measures ) ( Daniel and Pollmann , 2012 ) . At a more conceptual level , rewards may act as a ‘relevance tag’ that would prioritize the neural reprocessing of associative memories during sleep via a 2-step process: ( i ) during encoding , potential rewards activate dopaminergic midbrain regions , which would seal a ‘relevance tag’ to recently learned and rewarded information; ( ii ) during sleep ( in particular sleep oscillations ) , the activation of a striatal-hippocampal network would favor the reprocessing of recent memories with a high relevance ( Perogamvros and Schwartz , 2012 ) . Overall , by determining the fate of motivationally-relevant memory traces , post-learning sleep , even in the form of a nap , sharpens the skyline of our memories .
Thirty-one healthy young volunteers ( 16 women and 15 men , age range = 18–30 years old ) gave written informed consent and received financial compensation for their participation in this study , which was approved by the Ethics Committee of Geneva University Hospitals . All participants were right-handed , non-smokers , free from psychiatric and neurological history , and had a normal or corrected-to-normal vision . They were within the normal ranges on self-assessed questionnaires for depression , anxiety , circadian typology , had no excessive daytime sleepiness and reported taking regular naps . Before inclusion in the study , all selected participants came for a habituation nap monitored by polysomnography . They then kept a regular sleep-wake schedule during five days prior to the experimental day . Compliance was documented by actigraphy ( Actiwatch , Cambridge Neuroscience , Cambridge , UK ) and sleep diary . Moreover , they were requested to refrain from all caffeine and alcohol-containing beverages and intense physical activity for the 48h preceding the experiment . Participants were randomly assigned to either a ‘Sleep’ group ( n = 16 , 8 men ) , or to a ‘Wake’ group ( n = 15 , 7 men ) . There was no group difference for any of the questionnaires ( all p>0 . 05 ) , so both groups had in particular equal sensitivity to reward as assessed by the SPSRQ questionnaire . Participants arrived at 12:45 PM at the Brain and Behavior Laboratory of the University of Geneva . Before each fMRI session , participants got acquainted with the task on two series of pictures that were not used in the main experiment . At 1 PM , the participants were comfortably installed in the fMRI scanner , and performed the encoding session directly followed by the learning session ( Figure 1A ) . Then , electrodes were applied to all the participants to ensure similar experimental conditions for all participants: between 2 PM and 3:30 PM participants of the Sleep group took a nap and participants of the Wake group stayed quietly awake in a sound-attenuated room . At 4:30 PM , participants underwent the test fMRI session . A surprise retest session similar to the test session took place three months later at 3:30 PM . Before each fMRI session , a psychomotor vigilance task ( PVT ) was performed ( Blatter and Cajochen , 2007 ) ( see Table 2 ) . 10 . 7554/eLife . 07903 . 012Table 2 . Psychomotor vigilance task results ( mean ± SEM ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 012PVT 1PVT 2 Sleep groupPVT 2 Wake groupPVT 3Mean RT264 . 92 ± 4 . 42263 . 29 ± 4 . 45256 . 44 ± 4 . 74267 . 67 ± 4 . 37Median RT255 . 02 ±3 . 95250 . 81 ± 3 . 64249 . 38 ± 4 . 30258 . 12 ± 4 . 21False alarms0 . 42 ± 0 . 950 . 55 ± 0 . 941 . 15 ± 1 . 410 . 4 ± 0 . 89Lapses0 . 42 ± 0 . 80 . 72 ± 1 . 020 . 38 ± 0 . 700 . 4 ± 0 . 7Mean RT fastest 10%215 ± 3 . 59204 . 93 ± 4 . 48208 . 69 ± 4 . 80213 . 50 ± 4 . 63 Behavioral task . We developed an associative memory task comprising 8 series of pictures; 4 series yielded high reward outcome ( HR ) ( 1 dollar symbol for each picture correctly selected ) while 4 series yielded low reward outcome ( LR ) ( 1 cent symbol ) . Each series was composed of 6 photographs presented in the following order: pillow , sofa , kitchen , bedroom , house , and landscape ( Figure 1B ) . We told the participants that their final gain depended on their performance , namely that the dollar sign indicated a sequence that would be more rewarded while the cent indicated a sequence proportionally less rewarded . Critically , we also explicitly told them that both the dollar and the cent were ‘symbols’ and that we would convert their performance ( not the accumulated US tokens ) into Swiss Francs at the end of the learning session , so that if they performed well they would receive the maximal amount ( i . e . 120 Swiss Francs ) . This manipulation was meant to ( i ) ensure that the participants paid attention to the distinct reward levels associated with each sequence , ( ii ) increase the participants’ motivation to obtain the rewards , and ( iii ) also ensure that each participant got the same final monetary outcome ( i . e . 120 Swiss Francs ) . Participants were alsoinformed in the initial instructions that they would have to do an inference task , and were explicitly asked to memorize series as wholes rather than as isolated pair-wise associations . 10 . 7554/eLife . 07903 . 013Table 3 . Number of correct responses during the three learning blocks ( mean ± SEM ) . Only pairs that were selected correctly 2 or 3 times were considered learnt . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 013HR ( 20 direct pairs ) LR ( 20 direct pairs ) 3 correct15 . 60 ± 0 . 4812 . 93 ± 0 . 552 correct3 . 37 ± 0 . 395 . 00 ± 0 . 451 correct0 . 83 ± 0 . 171 . 57 ± 0 . 290 correct0 . 20 ± 0 . 090 . 50 ± 0 . 14 Encoding session . Participants watched each of the 8 series once , one picture at a time ( 2000 ms each ) , and were asked to encode them . At the beginning of each series , a dollar or a cent picture indicated the reward value ( high or low ) associated with the series ( Figure 1B and Figure 1—figure supplement 1A ) . Learning session . Each series is formed by 6 pictures , which that can therefore be grouped into 5 successive direct pairs ( pillow-sofa , sofa-kitchen , kitchen-bedroom , bedroom-house , house-scene ) . Participants were trained on all successive pairs of each of the 8 series in the following way : participants were shown the first picture ( pillow ) of the series . Then , the same picture was shown together with two options for the second item ( sofa ) . The onset of this display was used for the fMRI analysis ( red frame on Figure 1—figure supplement 1B ) . Two seconds later , when the sentence ‘choose next picture’ appeared in the middle of the screen , participants could give their answer by pressing on an MRI compatible response box ( Current Designs Inc . , USA ) ( Figure 1—figure supplement 1B ) . Next , the correct second item ( sofa ) was presented for two seconds , followed by this item together with the two options for the third item ( kitchen ) , and so on until the sixth picture ( landscape ) . At the end of each series , participants were shown how much they earned in dollars or cents , for high or low reward outcome series respectively . During the learning , participants got feedback about the correctness of their choice on each trial: after each ‘choice’ display , the correct picture was presented on the subsequent trial , as the next element in the sequence for which the following picture should be selected , and so on until the last trial , for which the ultimate correct choice ( landscape ) was also shown . The total amount of correct responses on a given sequence was also summarized at the end of the sequence , as a display with filled dollar ( for HR ) or cents ( for LR ) symbols ( see Figure 1—figure supplement 1 ) . All volunteers underwent 3 blocks of learning; in each block , each of the 8 series was presented once , in a randomized order . Nap time . Half of the participants took a nap ( Sleep group ) , the other half stayed awake ( Wake group ) , both for 1h30 . The Wake group was allowed to read in dim light ( 25 lux ) , on a topic not involving memorization or high cognitive load . For both groups the temperature of the room was controlled ( 21°C ) and polysomnographic data was continuously recorded ( see below EEG acquisition ) . Participants in the Sleep group were allowed to sleep for up to 90 min and were woken only from sleep stages 1 , 2 or REM ( see Supplementary file 1 for characteristics of the nap period for participants in the Sleep group ) . Before starting the test phase , the participants spent at least 40 min awake , in order to dissipate effects of sleep inertia in the Sleep group , and they all completed the St . Mary's Hospital questionnaire . One participant’s sleep data were lost due to a computer failure . This participant was excluded from the sleep data analyses . Test session . Participants were tested on all possible direct ( or immediately consecutive ) pairs of pictures in the series , and also on pairs of non-consecutive pictures , i . e . , inference of order 1 and 2 pairs . In inference trials , participants were also presented with a cue picture and had to select which of two pictures belonged to the same series as the cue picture , but for pairs of pictures that were more distant in the series ( i . e . , separated by one or two intermediate pictures; Figure 1C and Figure 1—figure supplement 1C ) . During test and retest sessions all possible combinations were presented once per session: all 5 direct pairs ( from each of the 8 series ) , inference 1 pairs ( for each of the 8 series there are 4 possible inference 1 pairs: pillow-kitchen , sofa-bedroom , kitchen-house and bedroom-scene ) and inference 2 pairs ( for each of the 8 series there are 3 possible inference 2 pairs: pillow-bedroom , sofa-house , kitchen-landscape ) . The order of the presentation of the pairs was randomized . Like in the learning session , they were first shown one picture alone , then the 2-alternative forced choice display , which was used as onset time for the fMRI analysis ( indicated by a red frame on Figure 1—figure supplement 1C ) . They could only give their answer after a 2 s delay when the ‘choose next picture’ sentence appeared . After each response , participants rated how confident they were when selecting the correct picture among 4 possible options: ‘certain’ , ‘not sure’ , ‘guess’ , or ‘by elimination’ ( Figure 1—figure supplement 1C ) . In order to prevent further learning , no feedback at all was shown during the test and retest sessions . Two participants ( one of the Sleep and one of the Wake group ) did not understand the confidence question , the data of these participants wer excluded from the confidence analysis . Retest session . Three months after the experiment , the participants were asked to come back for a retest session that was similar to the test session . Participants did not know at test that there would be a retest session . Out of the 31 participants 25 were able to come back after three months for the retest session . During the learning session , a picture in a series was considered learnt when the participant had selected the correct picture at least twice out of three times during the three learning blocks . Pictures that did not meet this learning criterion were removed from the analysis of the test and retest sessions ( 3 . 1 ± 2 . 26 pairs per participant out of 40 , see Table 3 ) . One participant had to be excluded of all analyses because of memory performance below two standard deviations of the group mean , thus the final group comprised 15 participants for both Sleep and Wake groups , which were included in fMRI analyses . All behavioral analyses were performed using Statistica ( Version 11 , www . statsoft . com , StatSoft , Inc . TULSA , OK , USA ) ; non parametric tests were used when normal distribution and equal variance criteria were not met . Post-hoc tests were performed using the Scheffe method . We first removed false alarms ( reaction times under 50ms ) and then computed the mean and median reaction times values . We did not remove slow reaction times as this was not a rapidity task , and participants were told that they could take up to 8 s to answer . We performed ANOVAs on mean and median RT values with Reward ( high , low ) and Relational Distance ( direct , inference 1 , inference 2 ) as within-subject factors , and Group ( Sleep , Wake ) as between-subject factor . These analyses revealed a main effect of Reward for both means and medians ( F ( 1 , 84 ) = 4 . 8263 , p=0 . 031 and F ( 1 , 84 ) = 6 . 8350 , p=0 . 011 , respectively ) but no main effect of Relational Distance and no interaction with Relational Distance . Participants were faster for trials belonging to HR series as compared to LR series , thus attesting an influence of the reward manipulation on behavior , see Table 1 . Similar analyses on the mean and median reaction times for ‘certain’ responses during the test phase did not show any main effect of Reward , Relational Distance , or Group , see Table 4 . 10 . 7554/eLife . 07903 . 014Table 4 . Reaction times ( mean ± SEM ) for the certain answers of the test phase . DOI: http://dx . doi . org/10 . 7554/eLife . 07903 . 014HR LRDirectInference 1Inference 2DirectInference 1Inference 2SleepMean727 . 98 ± 18 . 52876 . 96 ± 19 . 29882 . 20 ± 20 . 68609 . 91 ± 21 . 69952 . 90 ± 21 . 21721 . 28 ± 23 . 30Median665 . 43 ± 17 . 20792 . 3 ± 18 . 77770 . 67 ± 19 . 71553 . 5 ± 21 . 06861 . 167 ± 21 . 05682 . 93 ± 22 . 71WakeMean844 . 88 ± 18 . 211064 . 95 ± 24 . 451095 . 13 ± 27 . 81686 . 73 ± 24 . 63881 . 59 ± 21 . 59767 . 54 ± 22 . 88Median743 . 23 ± 18 . 111086 . 60 ± 28 . 671068 . 27 ± 27 . 28637 . 33 ± 23 . 90712 . 17 ± 20 . 62651 . 80 ± 22 . 05 PVT was administered three times: before the encoding session ( PVT 1 ) , before the test session ( after the sleep/rest period; PVT 2 ) , and before the retest session ( PVT 3; see Figure 1A ) . Analysis of the PVT data showed that reaction times were normally distributed for Sleep and Wake groups during the learning , test and retest sessions . Importantly , there was no group difference for reaction times , false alarms , or lapses at any time point ( all p>0 . 05 ) , see Table 2 . Nap-time was monitored using a V-Amp recorder ( Brain Products , Gilching , Germany ) . Standard polysomnography included 6 EEG ( Fz , Cz , Pz , Oz , C3 , C4 , reference on both mastoids ) , chin EMG , and vertical and horizontal EOG recordings ( sampling rate: 250 Hz ) . For PSG analyses , we used FASST ( fMRI Artifact rejection and Sleep Scoring Toolbox; Cyclotron Research Centre , University of Liège , Belgium ) implemented in Matlab ( MATLAB version 7 . 13 . 0 . 564 R2011b , Natick , Massachusetts: The MathWorks Inc . , 2011 ) . Fourteen naps and fifteen periods of quiet wakefulness were visually scored on a 20 s epoch basis by two independent scorers , according to standard criteria by the AASM Manual for the Scoring of Sleep ( Iber et al . , 2007 ) . Additionally , automatic detection of spindles was performed . Sleep spindles were detected based on an algorithm previously developed by Molle et al . ( Mölle et al . , 2002 ) . Sleep spindles were separated according to their main frequency ( i . e . maximum power amplitude within the 11–15 hz range ) , in line with the global standards used in sleep research ( Schabus et al . , 2007; Maquet , 2010 ) . Slow sleep spindles were defined as spindles with predominant frequency between 11 and 13 Hz and fast spindles between 13 . 1 and 15 Hz . Spindle frequency computation was done by a Matlab toolbox ( FASST , Cyclotron Liège ) ( Schabus et al . , 2007; Sterpenich et al . , 2007 ) . The distribution of fast and slow spindles has been reported to be bimodal ( Astill et al . , 2014 ) and corresponds to an approximate topographical localization , fast spindles having a predominant distribution over parietal regions and slow spindles over prefrontal regions ( Schabus et al . , 2007; Andrillon et al . , 2011 ) . Slow spindles had a predominant frequency below 13 Hz ( mean: 12 . 55; SEM: 0 . 08 ) and fast spindles had a predominant frequency above 13 . 5 Hz ( mean: 14 . 33; SEM: 0 . 06 ) . Sleep data are summarized in Supplementary file 1 . None of the participants in the Wake group fell asleep . MRI data were acquired on a 3 Tesla MRI scanner ( SIEMENS Trio System , Siemens , Erlangen , Germany ) . Multislice T2*-weighted fMRI 2D images were obtained with a gradient echo-planar sequence using axial slice orientation ( 36 slices; voxel size , 3 . 2 × 3 . 2 × 3 . 2 mm; repetition time ( TR ) = 2100 ms; echo time ( TE ) = 30 ms; flip angle ( FA ) = 80° , FOV = 205 mm ) . A whole-brain structural image was acquired at the end of the test part with a T1-weighted 3D sequence ( 192 contiguous sagittal slices; voxel size , 1 . 0 × 1 . 0 × 1 . 0 mm; TR = 1900 ms; TE = 2 . 27 ms; FA = 9° ) . An additional structural image was acquired with a proton-density weighted sequence ( 20 axial slices; voxel size , 0 . 8 × 0 . 8 × 3 . 0 mm; TR = 6000 ms; TE = 8 . 4 ms; FA = 149° ) . This acquisition served for the localization of the VTA ( D'Ardenne et al . , 2008 ) . All stimuli for fMRI were designed and delivered using a MATLAB Toolbox ( Cogent 2000 , http://www . vislab . ucl . ac . uk/cogent_2000 . php ) . Functional images were analyzed using SPM8 ( Wellcome Department of Imaging Neuroscience , London , UK ) . This analysis included standard preprocessing procedures: realignment , slice timing to correct for differences in slice acquisition time , normalization ( images were normalized to an EPI template ) , and smoothing ( with an isotropic 8-mm FWHM Gaussian kernel ) . A general linear model ( GLM ) approach was then used to compare conditions of interest at the individual level and then these contrasts from each participant entered a second-level random-effects analysis . All group comparisons were performed using ANOVAs . Correction for multiple comparisons was performed by submitting all reported activations to small-volume correction ( SVC ) for familywise error ( p<0 . 05 ) using regions of interest based on the Anatomy toolbox of SPM8 for the hippocampus ( SPM Anatomy toolbox 2 . 1 , Forschungszentrum Jülich GmbH ) , the automated anatomical labeling ( aal ) atlas for the caudate nucleus , the parahippocampus and the putamen ( Tzourio-Mazoyer et al . , 2002 ) , and using the coordinates from ( Bunzeck and Düzel , 2006 ) for VTA-compatible midbrain regions and from ( Düzel et al . , 1999 ) for the medial prefrontal cortex . Coordinates of brain regions are reported in MNI space . To investigate fMRI responses from the learning session , we first used a general linear model at the single individual level including 3 sessions for the 3 learning blocks . Each session contained 2 regressors for successful trials and 2 regressors for misses according to their reward outcome ( HR , LR ) and the 6 motion parameters derived from the spatial realignment added as covariate of no interest . We then performed a contrast between HR and LR hits for each participant and entered the resulting statistical maps into a second-level 2-sample t-test which also included the difference in performance between HR and LR series ( i . e . , HR minus LR performance ) as a covariate . This analysis allowed us to identify brain regions selectively contributing to reward-related performance improvement . This analysis revealed increased activation in a midbrain region compatible with the VTA [z-score = 3 . 71 ( -3x , -13y , -20z ) , p<0 . 05 , small-volume corrected ( SVC ) for familywise error] using the VTA coordinate from Bunzeck and Duzel ( Bunzeck and Düzel , 2006; Figure 2—figure supplement 1B ) . Psychophysiological interaction ( PPI ) analysis was computed to test the hypothesis that functional connectivity between the seed region ( the right hippocampus seed from the contrast of the test session , Figure 2C at ( 30x , -10y , -23z ) ) ( see Results ) and the rest of the brain differed for HR vs . LR hits during the retest session . Therefore , we took as psychological factor the contrast between HR and LR hits , irrespective of trial type ( direct , inference 1 and inference 2 trials ) . A new linear model was prepared for PPI analyses at the individual level , using three regressors . The first regressor represented the psychological factor , composed of HR vs . LR hits . The second regressor was the activity in the right hippocampus . The third regressor represented the interaction of interest between the first ( psychological ) and the second ( physiological ) regressor . To build this regressor , the underlying neuronal activity was first estimated by a parametric empirical Bayes formulation , combined with the psychological factor and subsequently convolved with the hemodynamic response function ( Gitelman et al . , 2003 ) . The model also included movement parameters . A significant psychophysiological interaction indicated a change in the regression coefficients between any reported brain area and the reference region , related to the correct retrieval of HR vs . LR trials . Next , individual summary statistic images obtained at the first-level ( fixed-effects ) analysis were spatially smoothed ( 6 mm FWHM Gaussian kernel ) and entered a second-level ( random-effects ) analysis using ANOVAs to compare the functional connectivity between groups . Finally , based on existing animal data suggesting a coordinated replay of rewarded associations within striatal and hippocampal regions ( Lansink et al . , 2009 ) , we tested whether reward-related regions showing increased functional connectivity with the hippocampus as a function of reward level ( selectively in the Sleep group ) correlated with sleep spindles . We thus extracted the beta values around relevant PPI result peaks using a 10 mm diameter sphere and performed a Spearman’s rank correlation analysis ( appropriate for small samples ) with the number of spindles detected during the nap for the participants in the Sleep group . | Fresh memories are strengthened while we sleep . However , we don’t remember every detail of our daily life experiences . Instead , it is essential that we retain information that promotes our survival , such as what we call "rewards" ( including food , money or sex ) and dangers that we should avoid . Igloi et al . sought to find out how the human brain picks out important memories to be consolidated during sleep , while discarding irrelevant information . Healthy participants learned series of pictures associated with either high or low rewards . After learning , some of the participants had a nap , while others remained awake . Directly after this and three months later , all the participants returned for a memory test . Igloi et al . found that the highly rewarded pictures were better remembered at both time points ( at the expense of lowly rewarded ones ) , but only for participants who had slept after learning . Further analysis revealed that distinctive bursts of brain activity occurring during sleep , so-called “sleep spindles" , favor the reorganization of memories stored in a region of the brain called the hippocampus , often considered to be the organ of memory . These findings uncover how sleep enhances long-term memory selectivity thus demonstratethat sleep does not just passively increase the retention of all memories . In the future , this work may inspire educational strategies that combine the careful use of rewards followed by an overnight period of sleep . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2015 | A nap to recap or how reward regulates hippocampal-prefrontal memory networks during daytime sleep in humans |
Combining rabies-virus tracing , optical clearing ( CLARITY ) , and whole-brain light-sheet imaging , we mapped the monosynaptic inputs to midbrain dopamine neurons projecting to different targets ( different parts of the striatum , cortex , amygdala , etc ) in mice . We found that most populations of dopamine neurons receive a similar set of inputs rather than forming strong reciprocal connections with their target areas . A common feature among most populations of dopamine neurons was the existence of dense ‘clusters’ of inputs within the ventral striatum . However , we found that dopamine neurons projecting to the posterior striatum were outliers , receiving relatively few inputs from the ventral striatum and instead receiving more inputs from the globus pallidus , subthalamic nucleus , and zona incerta . These results lay a foundation for understanding the input/output structure of the midbrain dopamine circuit and demonstrate that dopamine neurons projecting to the posterior striatum constitute a unique class of dopamine neurons regulated by different inputs .
A longstanding challenge in neuroscience is to understand how neurons compute their output by integrating information from multiple sources . Mapping precise neural connectivity is a critical step towards understating neural computation . Recent developments in viral , molecular , and imaging techniques offer unprecedented opportunities to outline various aspects of the brain's wiring diagram in systematic and quantitative manners ( Denk et al . , 2012; Watabe-Uchida et al . , 2012; Huang and Zeng , 2013; Osten and Margrie , 2013; Pollak Dorocic et al . , 2014; Hart et al . , 2014; Ogawa et al . , 2014; Oh et al . , 2014; Weissbourd et al . , 2014; Callaway and Luo , 2015 ) . Midbrain dopamine neurons play important roles in various brain functions including motivation , reinforcement learning , and motor control ( Wise , 2004; Redgrave and Gurney , 2006; Ikemoto , 2007; Schultz , 2007 ) . Recording experiments have shown that many dopamine neurons signal reward prediction error ( RPE ) : the discrepancy between predicted and actual reward value ( Schultz et al . , 1997; Bayer and Glimcher , 2005; Bromberg-Martin et al . , 2010; Clark et al . , 2012; Cohen et al . , 2012; Schultz , 2015 ) . Classically , it is thought that RPE coding is relatively uniform among dopamine neurons , and that dopamine's major function is to guide behavior toward maximizing future rewards . However , recent studies have suggested that there are at least two types of dopamine neurons , value-coding and salience-coding ( Matsumoto and Hikosaka , 2009 ) , although the extent of physiological diversity remains controversial ( Fiorillo , 2013; Fiorillo et al . , 2013a , 2013b ) . Most dopamine neurons reside in midbrain nuclei called the ventral tegmental area ( VTA ) , the substantia nigra pars compacta ( SNc ) , and the retrorubral field ( RRF ) . Clusters of dopamine neurons in and around these nuclei are designated A10 , A9 , and A8 , respectively . In the mouse brain , just ∼30 , 000 dopamine neurons reside in these nuclei ( Zaborszky and Vadasz , 2001 ) . As with other monoamine neurotransmitters in the brain such as serotonin and noradrenaline , this small population of midbrain dopamine neurons exerts its influence over much of the brain as a neuromodulator . However , compared to the other monoamine neuromodulators , the dopamine system is anatomically unique in that the collateralization of dopamine neurons is limited ( Moore and Bloom , 1978; Swanson , 1982; Sobel and Corbett , 1984 ) . In other words , unlike the other monoaminergic neurons , a single dopamine neuron tends to target just one brain area , raising the possibility that different functional populations can be defined by the region that they target . Recent studies have shown that dopamine neurons projecting to different targets have distinct properties ( Lammel et al . , 2008 , 2011 , 2012 , 2014; Kim et al . , 2014 ) . For instance , a population of dopamine neurons in the posterior medial VTA that projects to the medial prefrontal cortex ( mPFC ) in mice , is unique with respect to many features: low expression of dopamine transporter ( DAT ) , tyrosine hydroxylase ( TH ) , and the D2 dopamine receptor , narrow spike waveform , high baseline firing and low level of cocaine-induced synaptic plasticity ( Lammel et al . , 2008 , 2011 , 2012 , 2014 ) . Dopamine neurons that project to different areas are located at slightly different , but highly overlapping locations in the midbrain ( Swanson , 1982; Yetnikoff et al . , 2014 ) . These results have raised the possibility that projection target , rather than anatomical location , can better classify functional groups of dopamine neurons . To understand how dopamine neurons compute their output , it is critical to know their inputs . Early systematic studies produced a thorough list of input areas to dopaminergic nuclei ( Geisler and Zahm , 2005; Geisler et al . , 2007 ) using conventional tracers . A recent study using a cell-type specific transsynaptic tracing method employing a rabies virus confirmed that many of these areas actually project directly onto dopamine neurons ( Watabe-Uchida et al . , 2012 ) . This study also showed that VTA dopamine neurons received a distinct set of inputs compared to SNc dopamine neurons . It is possible that the differences observed in this study between inputs to VTA and SNc dopamine neurons were attributable to the uneven distribution of populations of dopamine neurons with different projection targets between these two nuclei . Both VTA and SNc contain dopamine neurons projecting to different targets , and each population of dopamine neurons projecting to a given target is distributed in a complex manner , often encompassing both VTA and SNc . Thus , it remains to be determined whether dopamine neurons projecting to different targets receive different or similar inputs . In the present study , we classified dopamine neurons according to their projection targets and defined the distribution of monosynaptic inputs of each subpopulation . We compared the inputs of dopamine neurons defined by eight major targets: different parts of the striatum ( ventral striatum [VS] , dorsal striatum [DS] , tail of the striatum [TS] ) , cortex ( mPFC , orbitofrontal cortex [OFC] ) , central amygdala ( Amy ) , globus pallidus ( GP ) , and lateral habenula ( lHb ) . To collect and analyze this large data set , we developed a data acquisition and analysis pipeline using a brain clearing method ( CLARITY ) ( Chung and Deisseroth , 2013 ) , whole-brain imaging using light-sheet microscopy ( Keller et al . , 2010 ) , and semi-automated software for analysis . We found that populations of dopamine neurons projecting to most of these targets receive a similar set of inputs , while dopamine neurons projecting to TS ( ‘tail of the striatum’ or ‘posterior striatum’ ) are a clear outlier . These results indicate that classifying dopamine neurons based on their projection target is a viable approach , and lay a foundation for studying the mechanisms by which dopamine neurons are regulated . Our approach is based on an automated imaging and analysis suite that will facilitate similar anatomical studies in other brain systems in an efficient and quantitative manner .
To visualize the monosynaptic inputs to dopamine neurons , we used a retrograde transsynaptic tracing system based on a modified rabies virus ( SADΔG-EGFP ( EnvA ) ) ( Wickersham et al . , 2007 ) . This rabies virus is pseudotyped with an avian sarcoma and leukemia virus ( ALSV-A ) envelope protein ( EnvA ) , so that the initial infection is restricted to cells that express a cognate receptor ( TVA ) in mammalian brains . In addition , this rabies virus lacks the gene encoding the rabies virus envelope glycoprotein ( RG ) , which is required for trans-synaptic spread . This allows us to restrict trans-synaptic spread to cells that exogenously express RG . In this way , only monosynaptic ( but not poly-synaptic ) inputs are trans-synaptically labeled . In our previous work , we used two helper viruses to express TVA and RG under the control of Cre recombinase ( AAV5-FLEX-TVA-mCherry and AAV8-FLEX-RG ) ( Watabe-Uchida et al . , 2012 ) . We previously injected these helper viruses into the VTA or SNc of transgenic mice expressing Cre specifically in dopamine neurons ( dopamine transporter-Cre , or DAT-Cre ) ( Backman et al . , 2006 ) to specifically label the monosynaptic inputs of these dopamine neurons throughout the brain ( Watabe-Uchida et al . , 2012 ) . In the present study , we sought to restrict the initial rabies infection to subpopulations of dopamine neurons defined by their projection sites . To do this , we injected a helper virus that expresses TVA under the control of Cre recombinase ( AAV5-FLEX-TVA ) into a dopamine projection site and co-injected a virus bearing blue fluorescent protein ( AAV1-CA-BFP ) to mark this site . Because it has been shown that minute TVA expression is sufficient for infection by ALSV-A ( Belanger et al . , 1995 ) , we reasoned that DAT-Cre-expressing dopamine neurons in the midbrain which were retrogradely infected by AAV5-FLEX-TVA would express enough TVA for initial rabies infection in a Cre-dependent manner ( Figure 1A ) . To verify this , we injected AAV5-FLEX-TVA into one of the projection sites , the DS . After 3 weeks , rabies virus was injected into both VTA and SNc . In this experiment , we did not inject AAV8-FLEX-RG , so the rabies virus did not spread trans-synaptically . We did this to visualize only neurons that were directly infected by rabies virus ( Figure 1B ) . Consistent with our intention , we observed that many VTA/SNc neurons were infected by the rabies virus and therefore expressed GFP ( Figure 1B; Figure 1—figure supplement 1A ) . Based on antibody staining , almost all of the GFP-positive neurons were also positive for TH ( Figure 1C–F ) , a marker for dopamine neurons ( 98 ± 1%; n = 600 neurons , n = 3 mice ) . Given that injecting pseudotyped rabies virus alone resulted in very few infections ( Watabe-Uchida et al . , 2012 ) , these results indicate that the injection of AAV5-FLEX-TVA into a dopamine projection site allowed us to restrict infection of rabies virus to dopamine neurons in a projection specific manner . 10 . 7554/eLife . 10032 . 003Figure 1 . Labeling projection-specific dopamine neurons and their monosynaptic inputs throughout the brain with rabies-GFP . ( A ) A schematic of the injections used to label projection-specific populations of dopamine neurons . The blue circle represents the site of infection with adeno-associated virus ( AAV ) -FLEX-TVA and green neurons represent the DAT-Cre-expressing dopamine neurons projecting to that area . ( B ) Horizontal optical section showing rabies-GFP signal following AAV-FLEX-TVA injection into the striatum of a DAT-Cre animal followed by rabies injection into the ventral tegmental area ( VTA ) and substantia nigra pars compacta ( SNc ) . The numbers of infected neurons are shown in Figure 1—figure supplement 1 . Bar indicates 2 mm . ( C ) Coronal physical section showing rabies labeled neurons from ( B ) in green and anti-TH antibody staining in red . Bar indicates 500 μm . ( D–F ) Higher magnification image of rabies labeled neurons and tyrosine hydroxylase ( TH ) staining . Bars indicate 200 μm . ( G ) A schematic of the injections used to label the inputs of projection-specific populations of dopamine neurons throughout the brain . The blue circle represents the site of infection with AAV-FLEX-TVA and the red circle represents the site of infection with AAV-FLEX-RG . Green neurons outside of VTA/SNc represent the monosynaptic inputs labeled throughout the brain . ( H ) Horizontal optical section showing rabies-GFP signal following AAV-FLEX-TVA injection into the striatum and AAV-FLEX-RG into the VTA and SNc of a DAT-Cre animal followed by rabies injection into the VTA and SNc . Number of infected neurons shown in Figure 1—figure supplement 1 . Bar indicates 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 00310 . 7554/eLife . 10032 . 004Figure 1—figure supplement 1 . Number of starter cells and inputs labeled . ( A ) The average numbers of projection-specified dopamine neurons labeled in each condition , using the injection schematic shown in Figure 1A with the indicated injection sites of AAV-FLEX-TVA on the x-axis of the graph . Mean ± s . e . m . ( B ) The average numbers of inputs ( outside of the VTA/SNc ) labeled in each condition using the injection schematic shown in Figure 1H with the indicated injection sites of AAV-FLEX-TVA on the x-axis of the graph . Mean ± s . e . m . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 004 To allow rabies virus to spread trans-synaptically , we performed the same AAV5-FLEX-TVA injection to infect DS-projecting dopamine neurons , but this time also injected AAV8-FLEX-RG into both the VTA and SNc . Because a large quantity of RG is required for robust trans-synaptic spread , we injected this virus near the cell bodies of dopamine neurons ( in the VTA and SNc ) rather than near their axons . After 3 weeks , rabies virus was injected into both the VTA and SNc ( Figure 1G ) . This resulted in a large number of trans-synaptically labeled GFP-positive neurons outside VTA/SNc ( Figure 1H; Figure 1—figure supplement 1B ) . Together , these results demonstrate that our method allows us to label subpopulations of dopamine neurons defined by their projection sites and the monosynaptic inputs to these populations . We used this method to identify the monosynaptic inputs to dopamine neurons projecting to different targets . Comparing the eight experimental groups required us to acquire and process a large data set . To achieve this goal , we developed a new data acquisition and analysis suite . In order to quantify the spread of rabies virus in each condition , we cleared brains using CLARITY to prepare them for light sheet microscopy ( Figure 2A ) . Before imaging each brain , we pre-screened to ensure sufficient optical clarity by shining a 488 nm light sheet through the ventral part of the brain and collecting light through an objective near the dorsal surface ( ∼6 mm away ) . 77 of 89 processed brains ( ∼87% ) were sufficiently clear for visualization of the ventral-most cells , and only these brains were used . We imaged each brain from the dorsal and ventral sides ( such that each image was a ‘horizontal’ optical section ) , and then merged these images to create a continuous 3D image of the brain ( Figure 2B ) . 10 . 7554/eLife . 10032 . 005Figure 2 . Automated acquisition and analysis of whole-brain tracing data . ( A ) A schematic of the brain clearing , imaging , and analysis pipeline used to acquire data from brains labeled using the injection schematics outlined in Figure 1A and Figure 1G . ( B ) A graphical explanation of the image acquisition and stitching process . Whole brains were imaged horizontally , from the dorsal ( orange arrow ) and ventral ( purple arrow ) sides and these images were stitched and combined to create a whole brain image . An example brain expressing tdTomato under the control of genetically encoded Vglut2-Cre is shown . ( C ) An example of a horizontal optical section acquired as described in A , B . Boxes indicate the locations of inset panels . Bar indicates 2 mm . ( D ) The automatically generated segmentation of C , with each labeled cell being represented by a single pixel . Bar indicates 2 mm . ( E–H ) Insets displaying raw images and automatically generated segmentations from the indicated regions ( cortex , striatum , midbrain , and cerebellum ) . Bars indicate 200 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 005 Because rabies spreads to monosynaptic input neurons , the only information we ultimately needed to collect from each brain for the present goal of comparing the distribution of inputs was the position of each labeled cell body . Therefore , we only needed to image with sufficient resolution to distinguish every cell body throughout the brain ( Figure 2C ) . This resulted in ∼2 Tb of image data from each brain , allowing us to image many brains , whereas imaging with even 2× higher magnification in each dimension would have led to prohibitively ( based on our present processing capacity ) large file sizes of ∼20 Tb per brain . We applied multiple segmentation algorithms to this data to classify each pixel as ‘cell’ or ‘non-cell’ ( Figure 2D; see ‘Materials and methods’ ) . Quality of segmentation was consistent across many regions and imaging depths , including cortex ( Figure 2E ) , striatum ( Figure 2F ) , midbrain ( Figure 2G ) , and cerebellum ( Figure 2H ) . Because each cell was present in several images ( at different depths ) , we then collapsed all neighboring ‘cell’ pixels in all three dimensions to generate a list of the centroids of the detected cells . To compare the distribution of cells across brains , we aligned the autofluorescence image captured from each brain to a common ‘reference space’ and used these transformations to align each set of centroids into this common space ( see ‘Materials and methods’ ) . By aligning brains with different genetically defined populations of neurons labeled , we were able to determine the boundaries of major brain regions in this reference space in three dimensions . This strategy for analysis mirrors the analysis pipeline called ‘CUBIC’ , proposed for comparable analyses of whole-brain images acquired using a similar clearing method ( Susaki et al . , 2014 ) . The main target structure of midbrain dopamine neurons is the striatum . However , dopamine neurons also project to other brain areas including other basal ganglia structures , amygdala , and much of the cortex . We characterized the distribution of dopamine neurons that project to cortex ( mPFC , OFC ) , striatum ( VS , DS , TS ) , GP , central amygdala ( Amy ) and lateral habenula ( lHb ) . To do this , AAV5-FLEX-TVA was injected into one of the eight projection sites in DAT-Cre animals expressing Cre specifically in dopamine neurons ( without injecting AAV8-FLEX-RG into the midbrain ) . 3 weeks later , rabies virus was injected into both VTA and SNc to infect dopamine neurons projecting to the area of the AAV5-FLEX-TVA injection ( Figure 1A ) . Brain samples were collected after 1 week . As previously reported , we found that dopamine neurons with distinct projection targets reside in different , but overlapping , areas of the midbrain ( Figure 3; Figure 3—figure supplement 1; Figure 3—figure supplement 2; Figure 3—figure supplement 3 ) ( Swanson , 1982; Bjorklund and Dunnett , 2007; Lammel et al . , 2008; Haber , 2014 ) . Interestingly , we observed an overlapping but dorso-laterally shifted distribution of dopamine neurons that project to VS , DS , and TS , in this order ( Figure 3—figure supplement 1; Figure 3—figure supplement 2 ) . Because many of the labeled populations of neurons overlapped in space , it is difficult to discriminate between populations based on location alone . In the case of Amy-projecting dopamine neurons , we observed labeling throughout VTA/SNc and also substantial labeling in supramammillary areas ( A10rv ) ( Yetnikoff et al . , 2014 ) . In short , each brain area receives projections from dopamine neurons in slightly different , but highly overlapping subareas of VTA/SNc . Furthermore , we found that the distribution of labeled neurons in the VTA/SNc was similar in cases with and without transsynaptic spread ( with and without RG , respectively ) , indicating that transsynaptic spread between primarily infected dopamine neurons and other dopamine neurons did not lead to the nonspecific infection of all dopamine neurons ( Figure 3—figure supplement 3 ) . 10 . 7554/eLife . 10032 . 006Figure 3 . Distribution of monosynaptic inputs to projection-specific populations of dopamine neurons throughout the brain . A summary of the inputs to each of the projection-specified populations of dopamine neurons assayed ( using the injection scheme from Figure 1G ) normalized such that the 1500 neurons were randomly sampled from each brain . We then combined such subsampled neurons from three randomly selected animals for each condition and plotted them in corresponding 400 μm sections . Inputs to: VS-projecting , DS-projecting , TS-projecting , GP-projecting , Amy-projecting , OFC-projecting , mPFC-projecting , and lHb-projecting dopamine neurons as well as selected coronal sections for reference ( A–H ) . In this figure , and all others , the following abbreviations were used: ‘VS’ for ventral striatum , ‘DS’ for anterior dorsal striatum , ‘TS’ for tail of the striatum ( posterior striatum ) , ‘GP’ for globus pallidus , ‘Amy’ for amygdala , ‘OFC’ for orbitofrontal cortex , ‘mPFC’ for medial prefrontal cortex , and ‘lHb’ for lateral habenula . Data collected in the same manner from experiments with no transsynaptic spread shown in Figure 3—figure supplement 1 and Figure 3—figure supplement 2 , comparison of the two conditions shown in Figure 3—figure supplement 3 , and injection sites used shown in Figure 3—figure supplement 4 and Figure 3—figure supplement 5 . Bars indicate 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 00610 . 7554/eLife . 10032 . 007Figure 3—figure supplement 1 . Distribution of projection-specific populations of dopamine neurons within the midbrain . A summary of the projection-specified populations of dopamine neurons labeled ( using the injection scheme from Figure 1A ) in cyan , compared to the distribution of all DAT-cre expressing neurons in the area of the injection in red . Distribution of: VS-projecting dopamine neurons , DS-projecting dopamine neurons , TS-projecting dopamine neurons , GP-projecting dopamine neurons , Amy-projecting dopamine neurons , OFC-projecting dopamine neurons , mPFC-projecting dopamine neurons , lHb-projecting dopamine neurons , and selected landmarks for reference ( A–D ) . Plots were prepared by projecting all neurons within 400 μm of the selected coronal plane onto a single image . Bars represent 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 00710 . 7554/eLife . 10032 . 008Figure 3—figure supplement 2 . Distribution of projection-specific populations of dopamine neurons within the midbrain: Maximum Intensity Projection . For each graph , 75 neurons were sampled at random from 3 of the ( −RG ) brains shown in Figure 3—figure supplement 1 . These 225 neurons were then plotted as a ‘maximum intensity projection’ ( i . e . , their Z coordinate was ignored ) . In the case of Lhb-projecting dopamine neurons , labeling was very sparse , so we plotted all 122 neurons . The distribution of all DAT-cre expressing neurons is shown in red . The distribution of dopamine neurons projecting to the indicated site in each case is shown in cyan . Grid lines indicate 500 microns . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 00810 . 7554/eLife . 10032 . 009Figure 3—figure supplement 3 . Comparison of labeled cell distribution between control and experimental groups . A comparison of the distribution of labeled VTA/SN neurons in experiments lacking RG ( labeling dopamine ‘starter neurons’ only ) and experiments with RG ( labeling dopamine ‘starter neurons’ as well as their monosynaptic inputs ) . Coronal planes match Figure 3—figure supplement 1B , C . VS , DS , and TS are color-coded , while the total distribution of all DAT-Cre expressing neurons in the area are shown in red . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 00910 . 7554/eLife . 10032 . 010Figure 3—figure Supplement 4 . Injection Sites . CLARITY-based brain clearing did not preserve native BFP fluorescence , so brains were physically sectioned and stained with an anti-BFP antibody to label the injection site of AAV-FLEX-TVA in each animal . All injection sites used for input tracing ( based on the schematic in Figure 1G ) are summarized in the left panels and a single raw image from each is displayed in the panels on the right . Injection sites were intended to allow TVA uptake into VS-projecting ( A–A′ ) , DS-projecting , ( B–B′ ) , TS-projecting ( C–C′ ) , GP-projecting ( D–D′ ) , Amy-projecting ( E–E′ ) , OFC-projecting ( F–F′ ) , mPFC-projecting ( G–G′ ) , and lHb-projecting ( H–H′ ) dopamine neurons . Red: injection sites . Blue: autofluorescence . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 01010 . 7554/eLife . 10032 . 011Figure 3—figure supplement 5 . Subdivisions of the striatum . ( A ) Maximum intensity projection ( horizontal ) of the outline of the brain ( black ) , the DS ( green ) , the nucleus accumbens ( orange ) , and the injections into VS ( blue ) , DS ( cyan ) , and TS ( red ) . Stereotactic coordinates are shown as grid lines with 1 mm spacing . ( B ) Maximum intensity projection ( sagittal ) of the outline of same injection sites as shown in above . ( C ) Coronal views of the approximate centers of the VS/DS and TS injections . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 011 Next , we mapped the monosynaptic inputs to these eight subpopulations of dopamine neurons ( Figure 3—figure supplement 4; Figure 3—figure supplement 5 ) defined by their projection target . Because the number of labeled neurons is different in each condition ( Figure 1—figure supplement 1 ) , comparing different conditions required a method of normalization . We first observed the distributions of labeled neurons by plotting the positions of labeled neurons for each condition . To normalize for the different numbers of total neurons labeled for visualization , we randomly sampled 1500 neurons from each brain , and plotted corresponding neurons in coronal sections ( Figure 3 ) . In our previous studies , we mapped the monosynaptic inputs to all of the dopamine neurons in VTA and SNc ( Watabe-Uchida et al . , 2012; Ogawa et al . , 2014 ) . We initially hypothesized that each of the eight populations would receive inputs from a subset of the regions identified in these previous studies . Surprisingly , we found that the distribution of inputs to each population appeared to be largely similar ( Figure 3 ) . We found that monosynaptic inputs to most dopamine neurons were concentrated in the ventral striatum ( Figure 3A ) and that several areas containing inputs are continuous around the medial forebrain bundle , along the entire anterior–posterior axis ( across the nucleus boundaries ) ( Figure 3A–E ) as was observed previously ( Geisler and Zahm , 2005; Watabe-Uchida et al . , 2012 ) , regardless of the projection targets . However , we found that TS-projecting dopamine neurons did not show a similar concentration of inputs: these neurons received many fewer inputs from anterior regions ( Figure 3A , B ) and ventromedial regions ( Figure 3C , D ) . In short , these observations suggested that the overall distribution of inputs to TS-projecting dopamine neurons is different from those of the rest of the dopamine neuron populations ( whose inputs largely overlapped with one another ) . We next examined these similarities and differences in more systematic and quantitative manners . To quantify the differences between inputs to dopamine neurons with different projection sites , we compared the percentage of input neurons observed in each brain area in each condition ( Figure 4 ) . This normalization allowed us to average between brains within a given condition ( n = 3–5 mice per condition , n = 4 for TS ) and then to compare across conditions . Although we found some differences between subpopulations , the overall patterns were surprisingly similar . Among the brain systems , we found that the basal ganglia and hypothalamus provided the majority of inputs to dopamine neurons ( Figure 4 , Figure 4—figure supplement 1 ) . In the basal ganglia , the ventral striatum ( nucleus accumbens core and shell ) , DS and ventral pallidum were the largest sources of inputs . In hypothalamus , the lateral hypothalamus ( LH ) ( including the parasubthalamic nucleus [PSTh] ) contained the largest number of inputs , and the preoptic areas also provided a substantial number . In the midbrain , the superior colliculus and dorsal raphe provided inputs as well . 10 . 7554/eLife . 10032 . 012Figure 4 . Comparison of the percentage of inputs from each region across populations of projection-specific dopamine neurons . A summary of the distribution of inputs to dopamine neurons with different projection sites , with bars representing the average % of inputs observed ( out of all labeled input neurons outside of the VTA/SNc/SNr/RRF ) per region ( mean ± s . e . m . ) . Each color represents inputs to a different population of dopamine neurons , as indicated in the inset . The 20 most prominent inputs are compared in Figure 4—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 01210 . 7554/eLife . 10032 . 013Figure 4—figure supplement 1 . Comparison of the percentage of inputs from selected regions across populations of projection-specific dopamine neurons . A summary of the distribution of inputs among the 20 most prominent regions providing input to dopamine neurons . These areas were selected based on the maximum percentage of inputs among eight conditions for each area . Plots are identical to those displayed in Figure 4 and each condition is labeled with the same color as in Figure 4 . p-Values are given based on 1-way ANOVA . Asterisks ( * ) indicate significant differences among conditions after Holm-Sidak corrections for multiple comparisons . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 013 In spite of similarities among dopamine populations , several prominent differences emerge upon focusing on TS-projecting dopamine neurons . While the ventral striatum is one of the largest sources of inputs to most of the subpopulations , it is a relatively minor source of inputs to TS-projecting dopamine neurons ( Figure 3A , Figure 4 ) . Instead , TS-projecting neurons receive an increased proportion of their inputs from the GP ( Figure 3C , Figure 4 ) , entopeduncular nucleus ( EP , Figure 3D , Figure 4 ) , PSTh , subthalamic nucleus ( STh ) and zona incerta ( ZI ) ( Figure 3E , Figure 4 ) . To statistically verify these observations , we first performed a 1-way analysis of variance ( ANOVA ) among the 20 main input areas ( Figure 4—figure supplement 1 ) . After correcting for multiple comparisons , nucleus accumbens shell , GP , STh and ZI remained significant ( p = 0 . 00045 , 0 . 0017 , 0 . 00023 , and 0 . 0028 , respectively; significant after Holm-Sidak corrections for multiple comparisons ) . These differences are mainly attributable to the effect of the TS-projecting dopamine population . If we remove TS in the analysis , none of these areas retain statistical significance with the exception of the ventromedial hypothalamus ( p = 0 . 0028 , 1-way ANOVA ) , which appears to be a slightly larger source of inputs to VS-projecting dopamine neurons . Furthermore , in all of the above cases , the percentage of inputs from TS was significantly larger than the group mean of the remainder of the populations ( p < 0 . 001; t-test ) . These results suggest that many dopamine neurons are embedded in a ‘canonical’ circuit similar to VS-projecting dopamine neurons ( which are known to encode RPE ) , while TS-projecting dopamine neurons have a unique distribution of inputs and therefore form a distinct class of dopamine neurons . By carefully examining the data , we could also discern some trends among the ‘canonical’ populations of dopamine neurons . For example , DS-projecting dopamine neurons received more inputs from the neocortex than other populations ( Figure 4 ) , while VS-projecting dopamine neurons received more inputs from the olfactory tubercle and other ventromedial structures such as the ventromedial hypothalamus ( Figure 4 ) . Furthermore , DS-projecting and OFC-projecting dopamine neurons appeared to receive more inputs from the dorsal-most part of the DS ( Figure 3B ) . In the posterior midbrain , all populations received inputs from the dorsal raphe , but lHb-projecting dopamine neurons received fewer inputs from nearby areas ( Figure 3H ) . In addition , lHb-projecting dopamine neurons received fewer inputs from posterior midbrain regions such as the periaqueductal grey and gigantocellular reticular nuclei ( Figure 4 ) . By far , though , the most distinct outlier was the set of inputs to TS-projecting dopamine neurons . In all of the cases in which our data attained statistical significance across all populations after correction for multiple comparisons ( ventral striatum , ventromedial hypothalamus , GP , ZI , and STh ) , TS-projecting neurons were the clear outliers ( Figure 4—figure supplement 1 ) . To further quantify the overall similarity of patterns of inputs between conditions , we calculated pair-wise correlations ( Pearson's correlation coefficients ) between the percentage of input neurons across anatomical areas ( Figure 5 ) . We found that most pairs had relatively high correlations ( r = 0 . 85–0 . 98 ) . Here again , TS-projecting dopamine neurons had consistently lower correlations than other populations ( Figure 5A ) : TS-projecting dopamine neurons were most distinct from VS- and DS-projecting dopamine neurons ( r = 0 . 65 and 0 . 69 , respectively ) ( Figure 5A ) , although the similarity was slightly higher compared to non-striatal regions ( r = 0 . 69–0 . 82 ) . Hierarchical clustering supported this view: TS-projecting dopamine neurons were an outlier among the populations tested ( Figure 5B ) . Together , these results show that TS-projecting dopamine neurons receive the most unique set of inputs among the populations of dopamine neurons that we examined . 10 . 7554/eLife . 10032 . 014Figure 5 . Correlation between projection-specific populations of dopamine neurons . ( A ) Scatter plots comparing the percent of inputs from each anatomical region for VS- , DS- , and TS-projecting dopamine neurons . Each circle represents the average percent of inputs for one anatomical area for each condition . r: Pearson's correlation coefficient . The axes are in log scale . ( B ) Dendrogram ( left ) and correlation matrix ( right ) summarizing all pair-wise comparisons . Color on the right indicates correlation values . The dendrogram was generated based on hierarchical clustering using the average linkage function . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 014 When parceling the data based on brain regions , we found that most dopamine neurons receive more inputs from structures along the ventral medial part of the basal ganglia and hypothalamus , while TS-projecting dopamine neurons receive more inputs from subthalamic areas . Because we noticed that labeled neurons often did not respect the boundaries between regions ( i . e . , they did not form discrete clusters corresponding to each sub-region ) and that distributions of labeled neurons were not always uniform in each area ( e . g . , the LH ) , we also examined the general topography of inputs in these areas irrespective of regional boundaries . We found that , in several coronal planes along the A-P axis , the center of mass of labeled neurons differed between conditions . Interestingly , we observed that the inputs to TS-projecting dopamine neurons were shifted dorso-laterally compared to the inputs to other subpopulations ( Figure 6 ) . In the anterior part of the brain , this shift is caused by the relatively low number of inputs from the ventral striatum to TS-projecting dopamine neurons ( Figure 6A ) . In the middle sections of the brain , the shift is caused by an increased number of inputs from GP ( Figure 6B ) and EP ( Figure 6C ) , as well as a decrease in inputs from the ventromedial hypothalamus . Finally , in a more posterior section of the brain , this shift is caused by the large number of inputs from the subthalamic and parasubthalamic nuclei to TS-projecting neurons ( Figure 6D ) , as well as the lower number of inputs from the ventromedial hypothalamus . 10 . 7554/eLife . 10032 . 015Figure 6 . Topological shift in the center of mass of input neurons to projection-specified dopamine neurons . Four coronal optical sections ( 400 μm thick ) were chosen to demonstrate the dorsolateral shift of inputs to TS-projecting dopamine neurons . ( A–D ) Coronal optical sections , with a region of interest marked in yellow for reference . Bars represent 2 mm . The distribution of inputs to each population of neurons is plotted in magenta and neurons within the ventral striatum ( A ) , GP ( B ) , entopeduncular nucleus ( EP ) ( C ) , and subthalamic nucleus ( STh ) ( D ) are indicated . A fixed number of neurons were randomly chosen from each brain and those neurons from three randomly chosen animals per condition were plotted for corresponding coronal sections . ( E–H ) Coronal optical sections , with a yellow box showing the location of the insets displayed below . The center of mass for each population is shown , with vertical and horizontal lines representing the standard error in the y-axis and x-axis , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 015 In summary , we found that the center of mass of inputs to TS-projecting dopamine neurons were dorso-laterally shifted compared to inputs to other subpopulations in several coronal planes ( Figure 6E–H ) . This suggests that the input pattern to TS-projecting dopamine neurons was unique , regardless of the method used for analysis . Although the meaning of the shift is not clear , overall , our analysis revealed that the inputs to subpopulations of dopamine neurons were topographically organized within and across nucleus boundaries . As we discussed above , the most prominent inputs to the ‘canonical’ dopamine neurons are from the nucleus accumbens , a part of the ventral striatum , though TS-projecting dopamine neurons received far fewer inputs from the nucleus accumbens ( Figure 3A; Figure 4 , Figure 7A–D ) . For each of the ‘canonical’ populations , between 12% and 20% of the input neurons were located in the ventral striatum whereas only 5% of the input neurons were located in the ventral striatum for TS-projecting dopamine neurons ( Figure 7D ) . 10 . 7554/eLife . 10032 . 016Figure 7 . Dense regions of inputs within the ventral striatum . ( A–C ) Coronal optical sections showing typical distributions of inputs in the ventral striatum . A comparison of the typical ‘patch’ structure of inputs from the ventral striatum ( to VS-projecting or DS-projecting dopamine neurons , for example ) with the atypical ‘patch-less’ structure of inputs from the striatum to TS-projecting dopamine neurons . ( D ) Percentage of input neurons within the ventral striatum ( core , medial shell , and lateral shell combined ) in each condition . Red dotted line indicates the percentage expected based on chance distribution throughout the brain , while green dotted line indicates the average percentage among all animals . ( E ) Horizontal optical section showing a typical distribution of inputs in the ventral striatum , with many input neurons in tight clusters . Bar represents 2 mm . ( F ) A zoomed view of E , taken from the indicated box . Clusters are dispersed along the A-P axis , so two planes are used to display them in coronal sections: the black arrow indicates the anterior plane and the white arrow indicated the posterior plane . ( G–H ) Coronal optical sections showing the two planes indicated above , as well as a graphical representation of the five patches in those planes . ( I ) Among labeled neurons in the ventral striatum , the percentage within the patches in each condition . Red dotted line indicates the percentage expected based on chance distribution throughout the nucleus accumbens , while green dotted line indicates the average percentage among all animals . ( J ) Among labeled neurons within the patches , the percentages within each patch in each condition . Green arrows point to patch 2 , cyan arrows point to patch 3 , and yellow arrows point to patch 5 . More detailed description of the patches shown in Figure 7—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 01610 . 7554/eLife . 10032 . 017Figure 7—figure supplement 1 . Density-based analysis of inputs to projection-specified dopamine neurons . All neurons from all animals were plotted in a reference space with 20 μm × 20 μm × 20 μm voxels , and a 3D Gaussian with kernel size of 60 μm × 60 μm × 60 μm was used to estimate density at each voxel . ( A ) A typical optical coronal section of raw fluorescence containing the ventral striatum colored such that black indicates fluorescence . ( B ) Average fluorescence among all animals displayed as a heat map . ( C ) Density estimate based on Gaussian smoothing of the image generated by plotting all extracted centroids from all animals displayed as a heat map . ( D ) A series of coronal sections demonstrating the 3D structure of the patches , obtained by finding the local maxima from C and expanding stepwise pixel-by-pixel until either 1/3 maximum intensity or another patch boundary was reached . Patch coloring matches Figure 7 . Bars represent 2 mm . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 017 Our labeling showed discrete peaks of high density within the nucleus accumbens that do not correspond to the boundaries of commonly delineated sub-areas of this region ( i . e . , ‘core’ or ‘shell’ ) ( Figure 7E–H ) . Our previous study reported the existence of these ‘ventral patches’ in consistent locations across animals ( Watabe-Uchida et al . , 2012 ) . In the present study , using precise 3D density plotting , we determined the exact number and locations of these patches: 5 total , with 3 in the anterior medial ventral shell , 1 in the medial core and 1 in the posterior lateral shell ( Figure 7G , H , Figure 7—figure supplement 1 ) . We observed these dense patches of inputs in all cases , with the exception of inputs to TS-projecting dopamine neurons ( Figure 7A–C , J ) . Of the input neurons in the nucleus accumbens , 16–26% were localized to these patches for all of the ‘canonical’ populations ( Figure 7I ) . By contrast , the few inputs that TS-projecting neurons did receive from the nucleus accumbens were not concentrated in patches ( Figure 7C , I , J ) . Thus , whereas neurons in the ventral patches may play a key role in regulation of dopamine neurons , TS-projecting dopamine neurons may be relatively disconnected from this regulatory system . We next compared the distribution of inputs from these patches between different populations of dopamine neurons . We found that each subpopulation received inputs differentially from different patches . For example , mPFC-projecting dopamine neurons received a larger ratio of inputs from the most medial patch , while Amy-projecting dopamine neurons received more of their inputs from the lateral patches ( Figure 7J ) . Each of these patches could potentially have a distinct functional role , and their contributions to dopamine neurons should be considered individually in future experiments . We observed that most populations of dopamine neurons receive a large number of monosynaptic inputs from the ventral striatum . Because dopamine neurons also innervate the ventral striatum , we wondered whether VS-projecting neurons receive a larger proportion of inputs from the VS than other populations and whether the other populations received a larger proportion of inputs from their respective projection sites . Because we labeled the AAV-FLEX-TVA virus injection sites with BFP ( to label the areas that the infected dopamine neurons project to , in each case ) , monosynaptic inputs residing within the BFP-labeled regions are more likely to have reciprocal connections with dopamine neurons . Because the CLARITY-based clearing process completely degraded the native fluorescence of BFP , we sliced the brains into 1 mm sections and stained them with an anti-BFP antibody to reveal the injection sites . We examined the distribution of inputs with respect to these injection sites . Surprisingly , we did not observe that monosynaptic inputs accumulated in BFP-labeled areas ( Figure 8A–F ) . 10 . 7554/eLife . 10032 . 018Figure 8 . Reciprocity of connection between dopamine neurons and neurons at their projection sites . CLARITY-based brain clearing did not preserve native BFP fluorescence , so brains were physically sectioned and stained with an anti-BFP antibody to label the injection site of AAV-FLEX-TVA in each animal . ( A–C ) The injection site ( red ) and input neurons ( green ) labeled in a physical coronal section for DS-projecting dopamine neurons . ( D–F ) The injection site ( red ) and input neurons ( green ) labeled in a physical coronal section for VS-projecting dopamine neurons . ( G ) A comparison of the percentage of inputs from the reciprocal region ( defined by brain region ) of each injection site ( in purple ) with the average percentage of inputs from that region among all other brains ( in black ) . For this analysis , the DS was split into ‘anterior DS’ for DS and ‘posterior DS’ for TS at Bregma −0 . 9 mm . There were no significant differences between pairs ( two-sample t-test ) . ( H ) A comparison of the percentage of inputs from the reciprocal site ( defined by region of BFP infection ) of each injection ( in red ) with the average percentage of inputs from that region among all other brains ( in black ) . There were no significant differences between pairs ( two-sample t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 018 To quantify this observation , we compared the number of potential reciprocal inputs to each subpopulation to the average number of inputs in this area among all other subpopulations . First , we compared how many inputs each population receives from the nucleus that it projects to . We found that there was some variability , but that no population received a statistically significantly higher fraction of inputs from the region that it projected to ( Figure 8G ) . Because some of our injections were limited to a portion of a region ( for example , our injection into the DS only covered the anterior medial DS ) , we did a further analysis using the injection sites instead of the targeted nuclei to quantify the data . With this analysis , we also found no statistically significant differences ( Figure 8H ) . In short , we observed that reciprocal monosynaptic connections are not a defining characteristic of inputs to dopamine neurons with differing projection targets ( Figure 9 ) . Rather , we found that most dopamine neurons ( besides TS-projecting dopamine neurons ) have similar inputs regardless of their projection target . 10 . 7554/eLife . 10032 . 019Figure 9 . Summary and model . ( A ) A working model suggesting that dopamine neurons receive inputs primarily from the region that they project to , based on the idea that input neurons provide an ‘expectation’ signal and dopamine neurons correct them ( i . e . , send them a prediction error signal based on the type of expectation that they encoded ) individually . ( B ) Our model , in which dopamine neurons in the ‘canonical’ pathway receive inputs primarily from a common set of regions ( and particularly heavy inputs from the ventral striatum ) and send a common prediction error signal to many parts of the forebrain . We propose that TS-projecting dopamine neurons could be part of a relatively separate pathway with a potentially unique function ( different from reward prediction error ( RPE ) calculation ) based on its unique distribution of inputs . Some common inputs such as the DS and ventral pallidum are omitted for clarity . DOI: http://dx . doi . org/10 . 7554/eLife . 10032 . 019
The brain performs complicated tasks using layers of hierarchical and/or parallel neural circuits . Alheid and Heimer proposed that a major organizational principle of the brain is the existence of parallel functional and anatomical macrosystems , each comprised of a cortical area , a cortical input nucleus ( such as an area in the striatum ) , an output nucleus ( such as a part of the pallidum ) , thalamus , and brainstem ( Alheid and Heimer , 1988; Zahm , 2006 ) . According to this view , each macrosystem ( such as the ventral striatum , the DS , or the extended amygdala ) would have its own dedicated dopamine system , resulting in a circuit organization resembling many parallel closed loops . This idea is plausible , considering the topographical organization of the dopaminergic nuclei ( Swanson , 1982; Yetnikoff et al . , 2014 ) , and it could explain the diverse function of dopamine neurons . If dopamine neurons encode RPE , each macrosystem could learn independently through its own feedback . To test this directly , we asked whether dopamine neurons projecting to a nucleus in one macrosystem receive a larger proportion of their inputs from that nucleus or from other nuclei in that same macrosystem . In stark contrast to this proposal , we found only loose reciprocity between dopamine neurons and their projection targets . If RPE is used to correct a neuron's behavior so that the neuron could represent predicted value more accurately in the future , a logical structure could be for prediction ( expectation ) to be sent to dopamine neurons and then for the same dopamine neurons to send back the error of the prediction ( RPE ) to the same neuron or the same system . Although many areas that receive dopamine projections also send projections to dopamine neurons , we found that reciprocal connections are not a basic principle explaining the distribution of monosynaptic inputs . In other words , reciprocal connectivity might not be the general mechanism of RPE computation and dopamine functions . Our data suggests that neurons in cortex , striatum , pallidum , habenula , or amygdala may not ‘entrain’ different populations of dopamine neurons . Rather , most dopamine neurons are more likely to gather information from a set of common sources and broadcast a common RPE to many brain areas ( Figure 9 ) , while specific populations with different sets of inputs , such as TS-projecting dopamine neurons , might have different functions . Previously , we found that many brain regions send monosynaptic inputs to dopamine neurons ( Watabe-Uchida et al . , 2012 ) . However , which of these brain areas are most important for RPE computation remains unclear . Because the ventral striatum was reported to receive a RPE signal from dopamine neurons ( Roitman et al . , 2008; Clark et al . , 2012; Hart et al . , 2014 ) , we reasoned that finding the monosynaptic inputs to VS-projecting dopamine neurons would narrow down the list of inputs that are potentially important to RPE computation . Although we were able to rule out some of the areas mentioned above which preferentially project to TS-projecting dopamine neurons , many brain areas still remain on the list of prominent inputs to VS-projecting dopamine neurons . Based on this logic , many brain areas have the potential to directly modulate RPE in dopamine neurons . Our results suggest that dopamine neurons projecting to the anterior vs posterior part of the striatum receive distinct inputs . Considering that dopamine plays essential roles in striatal functions , this result provides insight into the functioning of different parts of the striatum . Most commonly , the striatum is divided into two parts along the dorso–ventral axis: ventral striatum ( or nucleus accumbens in rodents , also called the ‘limbic’ striatum ) and DS . In addition , the striatum is divided into more sub-regions by splitting the ventral striatum into ‘core’ and ‘shell’ regions and splitting the DS into dorso-medial ( associative ) and dorso-lateral ( motor-related ) regions ( Francois et al . , 1994; Lynd-Balta and Haber , 1994; Haber et al . , 2000; Joel and Weiner , 2000; Redgrave et al . , 2010 ) . In primates , projections from different cortical areas define different striatal regions . Projections from dorsolateral prefrontal cortex ( DLPFC ) and pre-supplementary motor cortex preferentially target associative striatum , whereas projections from OFC are unique to the limbic striatum . In rodents , there is a substantial controversy over the existence and location of homologous structures to the DLPFC or OFC of the monkey ( Preuss , 1995; Uylings et al . , 2003; Vertes , 2004; Kita et al . , 2014 ) . Thus , it is difficult to define the subareas of the striatum in the mouse brain . Nonetheless , even if it is not perfectly homologous to the primate striatum , previous studies suggest that the rodent striatum could contain a functional gradient from ventromedial to dorsolateral in agreement with the organization of cortical projections ( Berendse et al . , 1992; Voorn et al . , 2004 ) . The three regions of the striatum that we targeted for investigation , VS , DS , and TS may reside roughly in this order in this gradient ( Figure 3—figure supplement 5 ) . Our target for DS ( or the dorsal head ) belongs to the ‘dorso-medial’ subarea , whereas TS ( or the posterior striatum ) consists of proportionally a more ‘dorso-lateral’ portion with some ‘dorso-medial’ and ‘ventral’ portions . Supporting the notion that these populations form a functional gradient , the input patterns to TS-projecting dopamine neurons are slightly more similar to that of DS-projecting dopamine than VS-projecting dopamine ( Figure 8 ) . However , we found that the similarity of inputs between DS- and VS- projecting dopamine neurons was much higher than similarity of inputs between DS- and TS- projecting dopamine neurons . This could potentially be explained by the fact that our injections into DS and VS were much closer together along the A-P axis ( Figure 3—figure supplement 5 ) . It may be fruitful to investigate a potential functional gradient along the A-P axis of the striatum in the future . Previous studies suggested that there is a functional difference between anterior and posterior striatum in monkey and human ( Miyachi et al . , 1997; Lehericy et al . , 2005 ) . For example , it has been shown that the anterior striatum is important for skill learning and the posterior part is important for skill execution in monkeys ( Miyachi et al . , 1997 ) . More recently , Kim and Hikosaka showed that the anterior caudate encodes flexible values , whereas the posterior caudate encodes stable values ( Kim and Hikosaka , 2013 ) . It would be interesting to determine whether this type of specialization for flexible or stable value coding applies also to the anterior and posterior parts of the putamen . It also remains unknown whether the striatum in rodents contains similar functional differences between the anterior and posterior parts . Increasing evidence suggests that dopamine neurons projecting to the ventral striatum signal RPE ( Roitman et al . , 2008; Clark et al . , 2012; Hart et al . , 2014 ) , however , it remains to be clarified whether dopamine neurons projecting to other brain areas send distinct signals . Our results raise the possibility that dopamine neurons projecting to the tail of the striatum convey a different type of information than RPE . One interesting possibility is that dopamine signals in the tail of the striatum lack mechanisms to update the value representation based on the consequences of actions , consistent with the idea that the dopamine recipient neurons residing in this part of the striatum store ‘stable values’ ( Kim and Hikosaka , 2013 ) . Dopamine is involved in different aspects of reward-based learning ( e . g . , goal-directed vs habit ) as well as a wider range of functions besides reward-based learning , including motor and cognitive functions . It is therefore possible that TS-projecting dopamine neurons are related to habitual behavior or one of these other functions , unlike other dopamine neurons that signal RPE . The most prominent feature of TS-projecting dopamine neurons is that there are many fewer inputs from the ventral striatum ( especially from the ventral patches ) , whereas the ventral striatum is one of the largest sources of inputs to the other dopamine neurons we investigated . Various models of RPE computations assume specific roles for inputs from particular areas . For instance , it has often been assumed that the striatum sends inhibitory ‘expectation’ signals to dopamine neurons that can ‘cancel out’ reward values when they are expected ( Houk and Adams , 1995; Kobayashi and Okada , 2007; Doya , 2008; Rao , 2010; Khamassi and Humphries , 2012 ) . Our finding that most populations of dopamine neurons received heavy innervation from the ventral striatum , which might encode ‘expectation’ or ‘state values’ in a RPE calculation , suggest that dopamine signals in these target areas are suppressed when reward is expected . On the other hand , our results indicate that TS-projecting dopamine neurons may not be subject to this type of regulation . Rather than the ventral striatum , TS-projecting dopamine neurons received more inputs from the STh , PSTh , and ZI . STh is known for its role in motor function and is a uniquely glutamatergic structure of the basal ganglia ( whereas all the other nuclei in basal ganglia are mainly inhibitory ) . STh receives a strong projection from the GP in addition to the cortex ( Canteras et al . , 1990; Francois et al . , 2004; Kita et al . , 2014 ) and sends projections to other basal ganglia nuclei , such as striatum , GP , EP and SNr ( Kita and Kitai , 1987; Degos et al . , 2008 ) . Interestingly , TS-projecting dopamine neurons receive inputs from STh as well as from areas that tend to be heavily interconnected with STh , such as GP , EP and SNr . These results suggest that , instead of inhibitory inputs from the ventral striatum ( ‘direct pathway’ ) , TS-projecting dopamine neurons receive more glutamatergic inputs from the STh ( ‘hyperdirect pathway’ ) , inhibitory inputs from pallidum ( ‘indirect pathway’ ) , and also inputs from two output nuclei: SNr and EP ( Figure 9 ) . Thus , cortical information may be transmitted to TS-projecting dopamine neurons through different pathways than other dopamine neurons . Finally , ZI , which is related to arousal , and PSTh , which is related to autonomic function , ( Goto and Swanson , 2004; Kita et al . , 2014 ) , are additional unique pathways from cortex to TS-projecting dopamine neurons . Previous studies have shown that there are functional and histological subdivisions of the ventral striatum ( Zahm and Brog , 1992; Pennartz et al . , 1994 ) . However , to our knowledge , there is no clear structure or function corresponding to the ventral patches that we described . In a previous study , we found that calbindin , a marker for patches in the DS , partially distinguished the ventral patches from surrounding areas ( Watabe-Uchida et al . , 2012 ) . These patches are , in essence , dense ‘hot spots’ for inhibitory inputs to dopamine neurons whereas neurons outside the patches are predominantly indirectly ( if at all ) connected to dopamine neurons . One interesting possibility is that these ventral patches are negatively or positively correlated with the observation of ‘hedonic hot spots’ within the striatum . These hot spots are so named because the ‘hedonic’ desire for food was enhanced by the local stimulation of μ-opioid receptors residing in the medial-dorsal part of the anterior ventral striatum but not by the stimulation of neighboring areas ( Castro and Berridge , 2014 ) . Overall , the type of information each part of the striatum relays to dopamine neurons is an open question . Our anatomical definition of the stereotypical locations of the ventral patches will help guide future studies of what information is channeled through these microdomains ( ventral patches ) of the ventral striatum . In Parkinson's disease , dopamine neurons in the SNc are progressively lost as the disease worsens . In particular , there is a prominent loss of dopaminergic axons in the posterior putamen and a corresponding loss of dopamine neurons in the ventrolateral SNc ( Kish et al . , 1988; Fearnley and Lees , 1991; Frost et al . , 1993; Morrish et al . , 1995; Damier et al . , 1999; Pavese and Brooks , 2009; Redgrave et al . , 2010 ) . Interestingly , dopamine neurons projecting to the tail of the striatum preferentially reside in the lateral part of SNc , suggesting that the tail of the striatum in mice may be homologous to the posterior putamen . One prominent feature of TS-projecting dopamine neurons is relatively strong input from STh . STh is uniquely excitatory compared to the other nuclei of the basal ganglia . Although the etiology of Parkinson's disease remains unclear , it has been proposed that uncontrolled , excessive excitation may eventually kill dopamine neurons ( Rodriguez et al . , 1998; Olanow and Tatton , 1999 ) . The strong monosynaptic excitation from STh could at least partially explain why dopamine neurons that project to the posterior putamen are specifically vulnerable . Because of the side effects of systemic L-DOPA administration , more and more patients have been receiving deep brain stimulation ( DBS ) as a treatment for Parkinson's disease ( Benabid et al . , 2009; Chaudhuri and Odin , 2010; Cyron et al . , 2010; Deniau et al . , 2010; Ponce and Lozano , 2010 ) . Currently , the main targets for DBS are STh , ZI , GPi ( also called EP in rodents ) and PPTg . Interestingly , our results showed that TS-projecting dopamine neurons receive monosynaptic inputs preferentially from all of these areas . In other words , effective DBS sites might reside in presynaptic sites that preferentially target TS-projecting dopamine neurons , rather than common inputs to all the dopamine neurons such as LH . As a result , DBS applied to these areas might influence only the circuits relevant to the TS-projecting dopamine system ( most of which is lost in patients ) without changing other aspects of dopamine functions , such as learning and reward-seeking behaviors . If this is true , the list of monosynaptic inputs to TS-projecting dopamine neurons will be a useful reference for seeking new DBS targets . Furthermore , expanding the framework for understanding the mechanism of DBS function could help improve its efficiency . In this study , we examined dopamine subpopulations that project to eight different projection targets . Two studies were published very recently that are highly related to our present work . One study found that monosynaptic inputs to dopamine neurons in VTA that project to medial and lateral NAc , mPFC and Amy were similar in all the areas they examined , with the exception of inputs from the dorsal raphe and within NAc ( Beier et al . , 2015 ) . Similarly , another study found that monosynaptic inputs to neurons in SNc that project to medial DS and lateral DS were similar with the exception of inputs within DS ( Lerner et al . , 2015 ) . Although neither of these studies examined TS-projecting dopamine neurons , the results are largely consistent with our finding that monosynaptic input patterns to most dopamine subpopulation are similar , and strengthen our finding that TS-projecting dopamine neurons are relatively unique . However , the patterns of inputs for VS-projecting neurons shown in the former study ( Beier et al . , 2015 ) and those for DS-projecting dopamine neurons in the latter study ( Lerner et al . , 2015 ) were quite different , in contrast to our finding . We previously observed different patterns of input between VTA and SNc dopamine neurons such as a strong preferential projection from DS to SNc ( Watabe-Uchida et al . , 2012 ) . In the present study , we found that most populations of projection-specific dopamine neurons are distributed both in VTA and SNc . We , therefore , injected rabies virus both in VTA and SNc for all eight conditions . The large difference in inputs between VS-projecting VTA neurons ( Beier et al . , 2015 ) and DS-projecting SNc populations ( Lerner et al . , 2015 ) may not be due to the difference in projection targets ( VS vs DS ) , but rather because of the location of their virus injection sites ( VTA vs SNc ) . It remains to be examined whether VTA dopamine neurons with a specific projection site receive a different set of monosynaptic inputs compared to SNc dopamine neurons that project to the same site ( i . e . , VTA → DS neurons vs SNc → DS neurons ) . Our study relies on the assumption that each dopamine subpopulation defined by projection target is a different population . In general , dopamine axons from both VTA and SNc are mainly unbranched and thus target one structure ( Fallon et al . , 1978; Swanson , 1982; Sobel and Corbett , 1984; Febvret et al . , 1991; Matsuda et al . , 2009 ) . However , there is some disagreement about the extent of collateralization of dopamine axons ( Yetnikoff et al . , 2014 ) ; some dopamine neurons may project to multiple brain areas ( Fallon and Loughlin , 1982; Loughlin and Fallon , 1984; Takada and Hattori , 1986 ) . Using CLARITY and whole brain imaging with membrane-bound GFP , Lerner et al . found that DS-projecting SNc neurons have negligible collateralization to NAc and no collateralization to other dopamine target areas such as prefrontal cortex or amygdala ( Lerner et al . , 2015 ) . On the other hand , using conventional light microscope , Beier et al . found that lateral NAc-projecting dopamine neurons have broad arborizations going to DS , whereas medial NAc-projecting dopamine neurons do not project to the DS ( Beier et al . , 2015 ) . They found that neither dopamine groups collateralize to mPFC or amygdala ( Beier et al . , 2015 ) . In spite of some discrepancies between studies , these studies show that dopamine neurons projecting to the target areas which we chose in the present study consist of largely non-overlapping subpopulations . However , because these studies could have overlooked sparse axon signals , we must interpret the results carefully , because any populations of dopamine neurons infected due to their projections to multiple areas would make the data appear more similar . Our viral tracing method may allow trans-synaptic spread between dopamine neurons , which would make the trans-synaptically labeled populations appear more similar . However , the intrinsic connections in VTA/SNc originate mainly from non-dopamine neurons without TH immunoreactivity ( Bayer and Pickel , 1990; Ferreira et al . , 2008; Jhou et al . , 2009 ) . With respect to potential trans-synaptic spread , aforementioned studies overcame this problem by using CAV-Cre in wild type mice , sacrificing cell type specificity ( Lerner et al . , 2015 ) or by establishing a technically advanced method using CAV virus combined with Cre and Flp recombinase ( Beier et al . , 2015 ) . Importantly , these studies also found that subpopulations of dopamine neurons had largely similar monosynaptic inputs . We found that dopamine neurons that were infected with AAV5-FLEX-TVA at axon terminals provided sufficiently high levels of TVA to allow for subsequent infection with a modified rabies virus ( i . e . , SADΔG ( envA ) ) . This is a good reminder of the extremely high efficiency of this rabies tracing system and simultaneously reinforces its limitations: although some studies have used this system to examine inputs to ubiquitous cell types such as GABA neurons ( Watabe-Uchida et al . , 2012; Weissbourd et al . , 2014; Beier et al . , 2015 ) , there is substantial risk for this adeno-associated virus ( AAV ) to infect not only Cre-expressing neurons in the injection site but also Cre-expressing neurons that project to that area , which may result in direct ( i . e . , non-transsynaptic ) infections by the rabies virus . Proper controls are always necessary for tracing using this system , especially because a very small TVA infection could allow for subsequent rabies infection . The acquisition and analysis of whole-brain tracing data has previously been accomplished by physically sectioning the brain at ∼100 μm intervals , imaging each ( or every third ) slice , manually defining regional boundaries in each section , and counting the number of cells in each region . This process is highly time-consuming and requires the dedicated attention of an expert in brain anatomy . To acquire and analyze this type of data with higher throughput , we decided to make use of novel methods in imaging and analysis . Several methods have recently been developed with the goal of acquiring an image of the whole mouse brain in an automated fashion . In broad terms , most strategies either employ serial sectioning 2-photon microscopy ( Ragan et al . , 2012; Oh et al . , 2014 ) or optical clearing ( Chung and Deisseroth , 2013 ) followed by light-sheet microscopy ( Tomer et al . , 2014 ) . Serial sectioning 2-photon microscopy has the advantage of using an opaque brain as starting material , meaning that it doesn't require any clearing steps . Light-sheet imaging of cleared tissues , by contrast , allows for much more rapid image acquisition since there is no time spent physically sectioning the tissue and also because scanning is done with a sheet rather than a line . This increased scan speed allows for the rapid imaging of whole mouse brains with sufficient resolution to image every neuron . Because we wanted to characterize the distributions of all labeled neurons across brains , we decided to use light sheet microscopy on cleared brains . Among methods to clear brains , a key distinction is between protocols that employ an organic solvent to reduce the internal differences in refractive index within a tissue and those that use an aqueous solution to wash away lipids and thereby reduce total scattering . Methods relying on organic solvents ( such as iDISCO ) denature native fluorescent proteins and require subsequent antibody staining ( Renier et al . , 2014 ) . Because we wanted to visualize GFP-labeled neurons , we decided to use CLARITY so that some native fluorescence ( i . e . , GFP and tdTomato ) would remain intact , although we found that CLARITY causes a loss of mCherry and BFP fluorescence . Furthermore , exposure to organic solvents dramatically alters the size and overall morphology of the brain , while CLARITY leaves it relatively intact . Therefore , although the autofluorescence is dimmer in CLARITY-cleared brains ( compared to iDISCO-cleared brains ) , brain structures are easier to identify . In summary , we found that the combination of CLARITY and light sheet microscopy allows for the rapid acquisition of whole brain rabies-tracing results . Paired with automated cell detection and alignment between brains , this method allows for a consistent , unbiased , and automated analysis of tracing experiments that would previously have required countless hours of attention from an expert in brain anatomy . Although the neural circuits controlling dopamine function are complex , a comprehensive understanding of the inputs/output structure of different populations of dopamine neurons will likely prove useful for understanding these circuits . In this study , we contributed to this effort in three different ways . First , we developed an automated pipeline for whole brain imaging , using a brain-clearing method ( CLARITY ) , light sheet microscopy , and semi-automated data analysis . This will help to lower the hurdle for future systematic anatomy studies , and increase their consistency and efficiency . Second , we defined the full sets of monosynaptic inputs to eight subpopulations of dopamine neurons specified by projection targets . Third , we found that dopamine neurons projecting to the tail of the striatum are unique in their monosynaptic inputs , suggesting that they might also be an outlier functionally compared both to dopamine neurons projecting to other parts of the striatum and to those projecting to other brain regions ( i . e . , cortex , GP , amygdala ) . This foundational work also opens the door to further investigations . For example , what are the functional differences between the anterior and posterior ( tail ) regions of the striatum , and what is the role of dopamine in each ? What information is calculated by dopamine neurons that project to different parts of the striatum , and how do the patterns of their inputs account for the differences in their signals ? How is each dopamine population related to the symptoms of various disorders such as Parkinson's disease ? We hope that our work will help guide these types of studies by providing them with a detailed circuit diagram as a starting point .
86 male adult mice were used . These mice were the result of a cross with C57BL/6J mice and DAT ( Slc6a3 ) -Cre mice such that they were heterozygous for Cre recombinase under the control of the dopamine transporter ( DAT ) ( Backman et al . , 2006 ) . Four male Vglut2 ( Slc17a6 ) -Cre ( Vong et al . , 2011 ) crossed with B6 . Cg-Gt ( ROSA ) 20Sortm9 ( CAG-tdTomato ) Hze/J ( Jackson Laboratory , Bar Harbor , ME , United States ) were used to help with the visualization of region boundaries . All procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and approved by the Harvard Animal Care and Use Committee . pAAV-EF1α-FLEX-TVA was made by inserting TVA950 ( Belanger et al . , 1995 ) truncated after transmembrane domain in pAAV-EF1α-FLEX ( Watabe-Uchida et al . , 2012 ) . pAAV-CA-BFP was made by cloning mTagBFP ( Evrogen , Moscow , Russia ) ( Subach et al . , 2008 ) in pAAV-CA , which was made by removing FLEX from pAAV-CA-FLEX ( Watabe-Uchida et al . , 2012 ) . AAVs were produced by the UNC Vector Core Facility ( Chapel Hill , NC , United States ) . In the first surgery , we injected 250–500 nl of AAV into the projection site ( AAV5-FLEX-TVA , 1 × 1012 particles/ml and AAV1-BFP , 9 × 1012 particles/ml ) and either AAV8-FLEX-RG ( 2 × 1012 particles/ml ) ( Watabe-Uchida et al . , 2012 ) ( to label inputs ) , or no virus ( to label only starter neurons ) into both the VTA and SNc . After 3 weeks , in the second surgery , 250–500 nl of SADΔG-EGFP ( EnvA ) ( 5 × 107 plaque-forming units [pfu]/ml ) ( Wickersham et al . , 2007 ) was separately injected into both the VTA and SNc . Brain samples were collected after 1 week . All surgeries were performed under aseptic conditions with animals anesthetized with isoflurane ( 1–2% at 0 . 5–1 . 0 l/min ) . Analgesia ( ketoprofen , 5 mg/kg , I . P . ; buprenorphine , 0 . 1 mg/kg , I . P . ) was administered for the 3 days following each surgery . The virus injection sites are as follows: ( VTA ) Bregma −3 . 1 , Lateral 0 . 6 , Depth 4 . 3 to 4 . 0 , ( SNc ) Bregma −3 . 1 , Lateral 1 . 2 , Depth 4 . 3 to 4 . 0 , ( VS ) Bregma 1 . 3 , Lateral 0 . 5 , Depth 3 . 8 , ( DS ) Bregma 1 . 0 , Lateral 1 . 5 , Depth 2 . 8 , ( TS ) Bregma −1 . 7 , Lateral 3 . 0 , Depth 2 . 35 , ( mPFC ) Bregma 2 . 1 , Lateral 0 . 75 , Depth 1 . 5 , Angle 20° , ( OFC ) Bregma 2 . 45 , Lateral 1 . 35 , Depth 1 . 75 , ( Amy ) Bregma −1 . 35 , Lateral 2 . 35 , Depth 4 . 0 , ( lHb ) Bregma −1 . 8 , Lateral 1 . 15 , Depth 2 . 95 , Angle 15° , ( GP ) Bregma −0 . 45 , Lateral 1 . 85 , Depth 3 . 5 . To prepare a hydrogel solution , water , paraformaldehyde ( PFA ) , phosphate buffered saline ( PBS ) , Bis , and Acrylamide were mixed on ice , and then VA-044 Initiator ( Fisher Scientific NC9952980 , Waltham , MA , United States ) was added . Then , 50 ml aliquots were made for storage at −20°C . Hydrogel was thawed at −4°C for 3 hr prior to each perfusion . During the perfusions , hydrogel solution was kept on ice . Immediately after the perfusions , brains were transferred to −4°C . The recipe used for hydrogel: 40 ml of 40% Acrylamide 10 ml of 2% Bis 1 gram VA-044 Initiator 40 ml of 10× PBS 100 ml of 16% PFA 210 ml of H20 After 2 days of storage in hydrogel at 4°C , oxygen was removed from tube using a food sealer and replaced with nitrogen . Immediately after oxygen/nitrogen exchange , tubes were placed in a water bath at 37°C . After 1 . 5 hr , the hydrogel solution solidified into a jelly-like texture . At this point , brains were gently removed from the hydrogel using gloved fingers . Brains were placed on a paper ( Kimwipe , Kimtech Science , Franklin , MA , United States ) and gently rolled over the paper so that excess hydrogel would come off of the brain . Great care was taken to ensure that all of the hydrogel was removed , but that none of the brain was damaged in the process . At this point , brains were put in PBST ( 0 . 1% Triton in PBS , pH = 8 ) overnight at 37°C on a rocker , covered by aluminum foil . After 1 day , brains were transferred to clearing solution . The recipe used for clearing solution: 123 . 66 g Boric Acid 400 g SDS 9 . 5 l H2O 60 ml NaOH ( 10 M ) Water , boric acid , and NaOH were mixed in a 5-gallon plastic container . Then , sodium dodecyl sulfate ( SDS ) was added 100 g at a time while mixing . Finally , solution was left to mix for several hours on a rocker at 37°C prior to use . Clearing was based on the original CLARITY protocol ( Chung and Deisseroth , 2013 ) . Simple and inexpensive parts were used in place of the originally proposed components . A Niagra 120 V ( Grey Beard Pumps #316 , Mt Holly Springs , PA , United States ) pump was used to circulate clearing solution . A Precision Adjustable 60 V/5 A ( Korad Technology #KA6005D , Shenzhen , China ) power supply was used to provide current at a constant voltage . A 5-gallon plastic container ( US Plastic #97 , 028 , Lima , Ohio , United States ) was used as a clearing solution reservoir and tubing was run though a second 5-gallon plastic container filled with water to cool the solution flowing through it . Chambers were constructed as previously described ( Chung and Deisseroth , 2013 ) using a Nalgene chamber ( Nalgene 2118-0002 , Rochester , NY , United States ) and platinum wire ( Sigma-Aldrich 267228 , St . Louis , MO , United States ) . Clearing was done in a room held at 37°C . Electrophoresis was performed with a constant current of 30 V , with an average temperature of 40°C , over the course of 60 hr . With the voltage held at 30 V , the current fluctuated between 0 . 5 A and 1 A , and generally stabilized at ∼0 . 75 A . The exact voltage/current relationship varied slightly depending on the chamber . The polarity of the electric field was switched every 12 hr . Clearing solution was replaced after every 120 hr of use ( so , after using it twice ) . Platinum wire was replaced after every 1200 hr of use ( so , after using it 20 times ) . After removing brains from the clearing chamber and placing them in PBST ( 0 . 1% Triton , pH = 8 ) for 24 hr , brains appeared snow white and expanded relative to their normal size . Brains were placed in an imaging solution called OptiView ( Isogai et al . , in press ) ( patent pending , application U . S . Serial No . 62/148 , 521 ) with refractive index 1 . 45 and pH 8 at room temperature on a rocker for 2 days before imaging . The imaging solution used was very similar to the recently described ‘RIMS’ imaging solution ( Yang et al . , 2014 ) . RIMS solution can be made as follows: 40 grams of Histodenz ( Sigma ) 30 ml PBS NaOH to pH 7 . 5 RI to 1 . 46 Images were acquired with the Zeiss Z . 1 Lightsheet microscope ( Carl Zeiss , Jena , Germany ) . Brains were glued to the tip of a 1 ml syringe ( without needle ) such that the posterior tip of the cerebellum was in contact with the syringe . Brains were then lowered into the imaging chamber . A 488 nm laser was used to excite GFP and a 561 nm laser was used to produce autofluorescence . Images were collected through a 5× objective with PCO-Edge scMOS 16 bit cameras ( PCO , Kelheim , Germany ) with 1920 × 1920 pixels . Each frame was 2000 × 2000 μm , so each pixel was roughly 1 . 04 μm . The step size between images was set to 5 . 25 μm , so the voxels were not quite isotropic . Brains were imaged horizontally from the dorsal side , and then rotated 180° for imaging from the ventral side . Each view was tiled with 7 × 6 tiles ( 14 , 000 × 12 , 000 μm ) and the two views were combined to create a continuous image . Autofluorescence images were subsequently downsized to 700 × 600 × 350 pixels for alignment to the reference space . In these downsized images , voxels have 20 μm spacing in all 3 dimensions . Images were segmented into ‘cell’ and ‘non-cell’ pixels with eight different segmentation algorithms . Each algorithm was trained with a selection of images from one of eight ‘parent regions’ including: ( 1 ) olfactory bulb , ( 2 ) cortex/hippocampus , ( 3 ) thalamus , ( 4 ) striatum/pallidum , ( 5 ) hypothalamus , ( 6 ) midbrain , ( 7 ) hindbrain , and ( 8 ) cerebellum . For each of these parent regions , a random set of 50 images from 10 different brains were used to train each of the segmentation algorithms manually using Ilastik , software that has been recently applied to segment EM images ( Sommer et al . , 2011; Maco et al . , 2014 ) . Because neuron appearance and background fluorescence appearance differs drastically through the brain , segmentation algorithms only reliably recognized cells within the region they were trained on . Ilastik was used to calculate six features for each pixel ( Gaussian , Laplacian of Gaussian , Gaussian Gradient Magnitude , Difference of Gaussians , Structure Tensor Eigenvalues , Hessian of Gaussian Eigenvalues ) with 3 radiuses ( 0 . 7 pixels , 1 . 6 pixels , and 5 pixels ) . In total , therefore , a vector with 18 values was produced for each pixel . During training , subsets of these vectors ( pixels ) would be marked as ‘cell’ or ‘non-cell’ by a human in each of the training images and then the rest of the vectors' identities were inferred using a random forest with 200 trees and four randomly chosen features per tree . Random sets of images from brains that were not part of the training set were used to manually verify the accuracy of segmentation . After training , all eight segmentation algorithms were applied to all of the images . This resulted in a binary version of the original data from the microscope with all pixels either classified as ‘1 ( cell ) ’ or ‘0 ( non-cell ) ’ . Because each cell was present in multiple Z positions , all sets of neighboring pixels that were labeled as ‘1 ( cell ) ’ pixels were collapsed in 3D using MATLAB to find the centroid of each detected cell . This afforded an extra opportunity to reduce noise by only counting ‘cells’ that were within a reasonable range of diameters ( 2 . 5 μm–50 μm ) , present in consecutive images , and with a reasonable shape ( i . e . , circularity of >0 . 1 ) . Finally , the centroids' positions were transformed into the reference space based on the result of autofluorescence alignment ( see below ) . At this point , the centroids were masked using the outline of the appropriate parent region and combined . For example , only the results of the cortex segmentation algorithm within the cortex were retained . The autofluorescence signals derived from 25 brains were averaged to produce a single ‘reference space’ to serve as a template for subsequent brains . Subsequent brains were routinely aligned to this reference space using Elastix ( Klein et al . , 2010 ) . We performed affine alignment followed by B-spline alignment based on mutual information , as previously proposed for human MRI image registration ( Metz et al . , 2011 ) . The resulting transformation ( from individual brain to reference space ) was applied to the centroids of detected cells in order to plot all cells in the same reference space . First , the 100 coronal slices of the Franklin and Paxinos atlas ( Franklin and Paxinos , 2008 ) were manually aligned to the nearest matching optical slices of the reference brain ( which has 700 coronal optical sections total ) . Then , boundaries of major regions were manually drawn onto each of these 100 slices . Finally , each region was manually smoothed in each dimension to produce continuous regions in 3D . Brains were sectioned using a vibrotome with 1 mm spacing . After CLARITY , brains were extremely difficult to section owing to their unusual structural properties ( soft , yet resistant to cutting ) . Slices were stained as previously described ( Chung and Deisseroth , 2013 ) . Briefly , primary antibody was applied for 2 days ( then washed off for 1 day ) and then secondary antibody ( Molecular Probes , Eugene , OR , United States ) was applied at 1:200 for 1 day ( then washed off for 1 day ) . Every step ( including washes ) was performed at 37°C on a rocker with PBST ( 0 . 1% Triton , pH = 8 ) containing sodium borate ( 1 M ) . After CLARITY , BFP signal was lost completely . A rabbit anti-tRFP antibody ( Evrogen #234 , 1:100 ) was used to detect BFP signal and a chicken anti-TH antibody ( Abcam #76442 , 1:100 , Camgridge , United Kingdom ) was used to detect TH . Brains were imaged to determine injection site and percentage of TH-positive starter cells using an inverted confocal microscope ( Zeiss LSM 700 ) . Stained 1 mm sections were placed in a glass-bottom well and then covered with a thin layer of imaging solution ( as described above ) to prevent them from drying during image acquisition . Images were acquired and stitched using Zeiss Blue . Further data analysis was performed using custom software written in MATLAB ( Mathworks , Natick , MA , United States ) . For statistical comparisons of the mean , we used 1-way analysis of variance ( ANOVA ) and the two-sample Student's t-test , unless otherwise noted . The significance level was corrected for multiple comparisons using a Holm-Sidak method . All error bars in the figures are the standard error of the mean ( s . e . m . ) . To quantify the percentage of inputs , we excluded areas that contained dopamine cell bodies that could potentially have been starter cells in and around VTA , SNc , SNr and RRF , to avoid potential contamination from dopamine neurons in the counts of inputs . We defined the primary infection sites by adding the sites of all the dopamine cell bodies in experiments without RG ( Figure 1A , Figure 3—figure supplement 1 ) . We did not exclude A10rv in the supramammillary area because DAT positive neurons in this area projected only to Amy and data did not change with or without this area . To quantify the similarity in input patterns , we calculated Pearson's correlation coefficients between the percent of input neurons across anatomical areas . To compare spatial distributions of input neurons in the forebrain , the mean of the coordinates of all input neurons on a given coronal section was obtained . Hierarchical clustering was done using the correlation coefficients as a distance metric and using the average linkage function . Using other linkage functions produced similar results . To define ‘ventral patches’ , we first pooled all neurons from all animals and plotted them in a reference space with 20 μm × 20 μm × 20 μm voxels . A 3D Gaussian with kernel size of 60 μm × 60 μm × 60 μm was then used to estimate the density of cells at each voxel . We defined ventral patches by finding the local maxima of the Gaussian-filtered 3D data and expanding stepwise pixel-by-pixel until either 1/3 of the maximum intensity or another patch boundary was reached . | Most neurons send their messages to recipient neurons by releasing a substance called a ‘neurotransmitter’ that binds to receptors on the target cell . The sites of this type of signal transmission are called synapses . Some small populations of neurons modulate the activity of hundreds or thousands of these synapses all across the brain by releasing ‘neuromodulators’ that affect how they work . These neuromodulators are essential because they broadcast information that is likely to be useful to many brain regions , like a ‘news channel’ for the brain . One important neuromodulator in the mammalian brain is dopamine , which contributes to motivation , learning , and the control of movement . Clusters of cells deep in the brain release dopamine , and people with Parkinson's disease gradually lose these cells . This makes it increasingly difficult for their brains to produce the correct amount of dopamine , and results in symptoms such as tremors and stiff muscles . Individual dopamine neurons typically send information to a single part of the brain . This suggests that dopamine neurons with different targets might have different roles . To explore this possibility , Menegas et al . classified dopamine neurons in the mouse brain into eight types based on the areas to which they project , and then mapped which neurons send input signals to each type . These inputs are likely to shape the activity of each type ( that is , their ‘message’ to the rest of the brain ) . The mapping revealed that most dopamine neurons do not receive substantial input from the area to which they project ( i . e . , they do not form ‘closed loops’ ) . Instead , most of their input comes from a common set of brain regions , including a particularly large number of inputs from the ventral striatum . However , Menegas et al . found one exception . Dopamine neurons that target part of the brain called the posterior striatum receive relatively little input from the ventral striatum . Their input comes instead from a set of other brain structures , and in particular from a region called the subthalamic nucleus . Electrical stimulation of the subthalamic nucleus can help to relieve the symptoms of Parkinson's disease . Therefore , the results presented by Menegas et al . suggest that this population of dopamine neurons might be particularly relevant to Parkinson's disease and that focusing future studies on them could ultimately be beneficial for patients . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Dopamine neurons projecting to the posterior striatum form an anatomically distinct subclass |
Toc75 plays a central role in chloroplast biogenesis in plants as the membrane channel of the protein import translocon at the outer envelope of chloroplasts ( TOC ) . Toc75 is a member of the Omp85 family of bacterial and organellar membrane insertases , characterized by N-terminal POTRA ( polypeptide-transport associated ) domains and C-terminal membrane-integrated β-barrels . We demonstrate that the Toc75 POTRA domains are essential for protein import and contribute to interactions with TOC receptors , thereby coupling preprotein recognition at the chloroplast surface with membrane translocation . The POTRA domains also interact with preproteins and mediate the recruitment of molecular chaperones in the intermembrane space to facilitate membrane transport . Our studies are consistent with the multi-functional roles of POTRA domains observed in other Omp85 family members and demonstrate that the domains of Toc75 have evolved unique properties specific to the acquisition of protein import during endosymbiotic evolution of the TOC system in plastids .
The biogenesis and function of plastids are reliant on the coordinate expression and selective import of >2500 different nucleus-encoded proteins ( Ling and Jarvis , 2015 , Shi and Theg , 2013 ) . The major pathway for protein import involves the synchronized action of two multi-protein membrane translocons , the TOC ( translocon at the outer envelope membrane of chloroplasts ) and the TIC ( translocon at the inner envelope membrane of chloroplasts ) complexes ( Paila et al . , 2015 , Lee et al . , 2014 ) , which reside within the chloroplast envelope . TOC and TIC complexes contain selective protein conducting channels , which align via physical association between the complexes to provide simultaneous membrane translocation of preproteins from the cytoplasm to the plastid interior ( Paila et al . , 2015 , Shi and Theg , 2013 , Ling and Jarvis , 2015 ) . The selectivity and directionality of protein import is controlled by sequential interactions between the targeting determinants within the cleavable , N-terminal transit peptides of preproteins and binding sites within the TOC and TIC complexes ( Li and Teng , 2013 ) , as well as import-associated molecular chaperones ( Flores-Pérez and Jarvis , 2013 , Paila et al . , 2015 , Schwenkert et al . , 2011 ) . The evolution of the TOC and TIC system during endosymbiosis required the adaptation or acquisition of protein conducting channels , receptors and molecular chaperones at both envelope membranes to provide a unidirectional conduit for the transport of nuclear encoded polypeptides into the organelle . Protein transport across the outer envelope membrane is mediated by a β-barrel membrane protein , Toc75 , which is proposed to constitute the major component of the TOC import channel ( Hinnah et al . , 2002 ) . Toc75 assembles with two integral membrane receptor GTPases , Toc159 and Toc34 , which recognize plastid preproteins in the cytoplasm and control access of preproteins to the Toc75 channel via their intrinsic GTPase activities ( Chang et al . , 2012 , Kessler and Schnell , 2009 , Kessler and Schnell , 2004 ) . In addition to its role as the TOC channel , Toc75 is proposed to participate in the targeting and insertion of membrane-anchored outer membrane proteins , including the TOC GTPases ( Wallas et al . , 2003 ) . Toc75 is encoded by a single gene in Arabidopsis ( TOC75-III , AT3G46740 ) and null mutants are embryo lethal ( Jackson-Constan and Keegstra , 2001 , Baldwin et al . , 2005 , Hust and Gutensohn , 2006 ) . Knockdown of TOC75-III expression by RNAi ( Huang et al . , 2011 ) and a hypomorphic toc75-III mutant ( Stanga et al . , 2009 ) both result in pale phenotypes and growth defects , consistent with the role of Toc75 in chloroplast biogenesis . Based on these studies , Toc75 appears to have evolved to perform multiple functions in protein import , in the assembly of TOC complexes , and in the biogenesis of the outer membrane ( Richardson et al . , 2014 ) . Toc75 belongs to the outer membrane protein of 85 kDa ( Omp85 ) superfamily of β-barrel integral membrane proteins ( Hsu and Inoue , 2009 ) . Omp85 proteins are exclusively localized to the outer membranes of Gram-negative bacteria , mitochondria and plastids ( Inoue and Potter , 2004 , Schleiff et al . , 2011 , Voulhoux and Tommassen , 2004 ) , and perform major roles in protein transport and the insertion and assembly of outer membrane proteins ( Voulhoux and Tommassen , 2004 , Schleiff et al . , 2011 ) . For example , BamA ( β-barrel Assembly Machinery protein A ) and Sam50 ( Sorting and Assembly Machinery 50 kDa ) , two members of the Omp85 family , function in outer membrane biogenesis in Gram-negative bacteria and mitochondria , respectively , by facilitating the folding and insertion of β-barrel proteins at the outer membrane from the periplasmic or intermembrane space ( Voulhoux et al . , 2003 , Noinaj et al . , 2013 , Schleiff and Soll , 2005 ) . The structural similarity of Toc75 to these conserved membrane transporters led to the hypothesis that Toc75 evolved from an ancestral Omp85 gene and was adapted during endosymbiosis of the cyanobacterial ancestor to function as the protein import channel in the plastid outer envelope ( Day et al . , 2014 , Topel et al . , 2012 , Bredemeier et al . , 2007 , Gentle et al . , 2005 ) . Toc75 exhibits two main structural features characteristic of Omp85 superfamily members . The N-terminal ~30 kDa region of Toc75 contains three repeats of POTRA ( polypeptide-transport associated ) domains , each characterized by a β1α1α2β2β3 secondary structural motif ( Clantin et al . , 2007 , Koenig et al . , 2010 , Paila et al . , 2015 ) . The ~45 kDa C-terminal region is predicted to contain 16 membrane-spanning β-strands and constitute the membrane-integrated β-barrel ( Paila et al . , 2015 , Day et al . , 2014 ) . While there is significant evidence that the C-terminal β-barrel domain of Toc75 acts as a component of the translocon channel for protein translocation at the outer membrane ( Hinnah et al . , 1997 ) , the functions of the N-terminal POTRA domains have not been clearly defined . In the case of BamA , specific POTRA repeats are indispensable for the insertion of β-barrel proteins at the E . coli outer membrane ( Browning et al . , 2013 ) . The POTRA domains of BamA extend into the periplasm , where they interact with other components of the BAM complex ( Bam B-E ) , a major periplasmic chaperone , SurA ( Bennion et al . , 2010 ) , and nascent outer membrane proteins ( Kim et al . , 2007 , Noinaj et al . , 2015 ) . Structural studies of BamA also suggest that specific POTRA repeats interact with the C-terminal domain to act as a possible gate to the membrane channel of the β-barrel ( Noinaj et al . , 2013 , Noinaj et al . , 2015 , Bakelar et al . , 2016 ) . In another example , deletion of a major portion of the POTRA domain of Sam50 in yeast mitochondria inhibits growth ( Habib et al . , 2007 ) , and subsequent studies provided evidence that the domain interacts with β-barrel precursors to promote their release from the SAM complex and insertion in the outer membrane from the intermembrane space ( Kutik et al . , 2008 , Stroud et al . , 2011 ) . The studies of BamA and Sam50 are consistent with models in which the POTRA repeats are assembled into multi-functional cassettes to mediate protein-protein interactions required for the assembly and specific targeting function of the membrane biogenesis machinery ( Koenig et al . , 2010 ) . The relationship between Toc75 and other Omp85 superfamily members raises interesting evolutionary and mechanistic questions of how the structural and functional features of the conserved POTRA and β-barrel domains of Toc75 have adapted during endosymbiosis to constitute the protein import channel . Structural and electrophysiological studies of Toc75 demonstrate that the protein forms a cation-selective channel with a pore size sufficient to transport unfolded polypeptide substrates ( Hinnah et al . , 1997; 2002 ) . In vitro pull-down experiments with recombinant POTRA domains of Toc75 suggest that they can interact with the TOC GTPases and chloroplast preproteins ( Ertel et al . , 2005 ) . This has led to the proposal that they function in TOC complex assembly , interactions with the TOC GTPases , preprotein recognition , or to provide a chaperone-like activity for the preprotein during membrane translocation ( Sánchez-Pulido et al . , 2003 , Ertel et al . , 2005 ) . In this study , we combine molecular genetic and biochemical analyses to examine the roles of the POTRA domains in protein import and membrane targeting in chloroplasts . We demonstrate that POTRA domains contribute essential functions to the Toc75 channel . Furthermore , expression of Toc75 lacking one or more of the three POTRA repeats leads to dominant negative phenotypes in Arabidopsis thaliana . Expression of specific deletion mutants alters the size and distribution of the TOC complexes and disrupts preprotein import into the organelle . Biochemical analyses indicate that the Toc75 POTRAs interact specifically with chloroplast preproteins and two proposed chaperones of the intermembrane space , Tic22-III and Tic22-IV . Our results demonstrate multiple roles for the POTRAs in protein import , including TOC complex assembly , preprotein translocation and the recruitment of chaperones to facilitate transport across the intermembrane space .
To investigate the role of POTRA domains in the function of Toc75 , we introduced a series of in-frame internal deletion mutations in a TOC75-III genomic construct to encode proteins lacking one ( Toc75ΔP1 ) , two ( Toc75ΔP1-2 ) or all three ( Toc75ΔP1-3 ) predicted POTRA domains ( Figure 1A ) , and tested their ability to complement the lethal phenotype of the toc75-III-1 null mutant . The TOC75-III genomic fragment retains both the native promoter and introns and was previously shown to complement toc75-III-1 ( Shipman-Roston et al . , 2010 ) . Heterozygous toc75-III-1 plants were transformed with a wild-type TOC75 gene construct ( TOC75 ) , TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 . Transformants were selected for hygromycin resistance linked to the toc75-III-1 T-DNA insertion and for the DsRed fluorescence marker linked to the TOC75 and POTRA-deletion constructs . The presence of TOC75 , TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 ( Figure 1B ) and the toc75-III-1 allele ( data not shown ) was confirmed by PCR of genomic DNA in T3 transformants homozygous for the DsRed marker . Plants homozygous for the TOC75 , TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 transgenes were selected in the heterozygous toc75-III-1 ( toc75III-1[+/-] ) background , and the progeny of these plants were segregated to test for complementation of the lethal phenotype of toc75-III-1 homozygous plants . At least five independent lines were selected for each construct . The segregation analysis of representative T3 lines is shown in Table 1 . As expected , toc75-III-1 ( +/- ) progeny carrying the homozygous TOC75 wild type transgene segregated at a near 3:1 ratio of hygromycin resistant:sensitive plants , consistent with complementation of the toc75-III-1 insertion ( Baldwin et al . , 2005 , Shipman-Roston et al . , 2010 ) ( Table 1A ) . Furthermore , plants homozygous for both toc75-III-1 and TOC75 alleles ( toc75-III-1[+/+] TOC75 [+/+] ) could be recovered ( Table 1A ) , indicating complementation of the lethal phenotype . By contrast , progeny of toc75-III-1 ( +/- ) lines homozygous for the TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 deletion constructs segregated at 2:1 or lower ratios of hygromycin resistant:sensitive plants , indicating that the POTRA deletions were unable to complement toc75-III-1 ( Table 1B ) . No homozygous toc75-III-1 lines expressing the POTRA-deletion constructs were recovered . On the basis of these data , we conclude that the POTRA domains are required for Toc75 function . 10 . 7554/eLife . 12631 . 003Figure 1 . Phenotypes of TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 in toc75-III-1 plants . ( A ) Schematic diagram of Toc75 and Toc75ΔP1 , Toc75ΔP1-2 or Toc75ΔP1-3 proteins used in this study . The dashed line represents the deleted POTRA domain region in various constructs . The numbers refer to the amino acid position , with 1 indicating the N-terminal residue of mature Toc75 . ( B ) PCR confirmation of the genotypes of heterozygous toc75-III-1 plants transformed with TOC75 and TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 constructs . Primers spanning the genomic regions encoding POTRA-1 , -2 or -3 were used to distinguish between TOC75 and TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 ( Table 2 ) . The approximate positions of the primers are indicated by the forward and reverse arrows . The positions of the PCR products for TOC75 and TOC75ΔP1 ( ΔP1 ) , TOC75ΔP1-2 ( ΔP1-2 ) and TOC75ΔP1-3 ( ΔP1-3 ) are indicated to the right of the figure . ( C ) Visual phenotypes of 7- and 22-day old TOC75 , TOC75ΔP1#2 , TOC75ΔP1-2#2 and TOC75ΔP1-3#2 ( 7-day only ) plants grown on soil . Bars = 1 cm ( D ) Chlorophyll content of 22-day old TOC75 , TOC75ΔP1#2 , TOC75ΔP1-2#2 plants . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 00310 . 7554/eLife . 12631 . 004Figure 1—figure supplement 1 . Phenotypes for individual transgenic toc75-III-1 lines expressing TOC75ΔP1 or TOC75ΔP1-2 . Visual phenotypes of 7- and 22-day old TOC75 , TOC75ΔP1 and TOC75ΔP1-2 plants grown on soil . Bars = 1 cmDOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 00410 . 7554/eLife . 12631 . 005Table 1 . Segregation analysis of toc75-III-1 plants transformed with full-length and POTRA deletion constructs of Toc75 expressed under the TOC75 promoter . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 005Parental plantsHygromycin resistant ( R ) Hygromycin sensitive ( S ) R:S ratioχ2- valueap0 . 05-valuebA . toc75-III-1 ( +/- ) 76411 . 850 . 180 . 67toc75-III-1 ( +/- ) TOC75 ( +/+ ) 82302 . 730 . 190 . 66toc75-III-1 ( +/+ ) TOC75 ( +/+ ) 119--B . toc75-III-1 ( +/- ) TOC75ΔP1 ( +/+ ) #179382 . 083 . 490 . 062toc75-III-1 ( +/- ) TOC75ΔP1-2 ( +/+ ) #176391 . 954 . 870 . 027toc75-III-1 ( +/- ) TOC75ΔP1-3 ( +/+ ) #172352 . 063 . 390 . 065C . toc75-III-1 ( +/- ) TOC75ΔP1 ( +/+ ) #267481 . 417 . 190 . 001toc75-III-1 ( +/- ) TOC75ΔP1-2 ( +/+ ) #264501 . 2821 . 62<0 . 001toc75-III-1 ( +/- ) TOC75ΔP1-3 ( +/- ) #261591 . 0337 . 37<0 . 001D . toc75-III-1 ( +/- ) TOC75myc ( +/+ ) 112382 . 950 . 0090 . 92toc75-III-1 ( +/+ ) TOC75myc ( +/+ ) 115--toc75-III-1 ( +/- ) TOC75ΔP1myc ( +/+ ) 63481 . 3119 . 7<0 . 001aGoodness-of-fit of the observed segregation ratios to the expected 2:1 ratio for toc75-III-1 ( +/- ) or expected 3:1 ratio for complementation of toc75-III-1 with the indicated TOC75 gene constructs was assessed by χ2 analysis . bp-Values were calculated using Graphpad Prism software version 4 . 00 . 10 . 7554/eLife . 12631 . 006Table 2 . List of primers used for genotyping plants . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 006Primers used for making POTRA deletions in TOC75 genomic construct for generating transgenic plantFor TOC75∆P1Primer 1CTTAGTGGTTTCAAGAAGTATTGGCAATCTGCTGATAGGPrimer 2CCTATCAGCAGATTGCCAATACTTCTTGAAACCACTAAGFor TOC75∆P1-2Primer 3CTTAGTGGTTTCAAGAAGTATATAACTCAGCTAGTTATTCAGPrimer 4CTGAATAACTAGCTGAGTTATATACTTCTTGAAACCACTAAGFor TOC75∆P1-3Primer 5CTTAGTGGTTTCAAGAAGTATCAGAAGTCAGCTGAAGCTPrimer 6AGCTTCAGCTGACTTCTGATACTTCTTGAAACCACTAAGPrimers used for myc insertions in TOC75 genomic construct for generating transgenic plantFor TOC75myc and TOC75∆P1myc using TOC75 and TOC75∆P1as templates , respectivelyPrimer 7GATGAAGAACAAAAACTTATTTCTGAAGAAGATCTGGAACAATCACCGGATPrimer 8ATCCGGTGATTGTTCCAGATCTTCTTCAGAAATAAGTTTTTGTTCTTCATCPrimers used for genotyping transgenic plantsFor TOC75 Primer 9TTCTTTGATCGACGGAGACPrimer 10CAGCAAACGAGATTGTAACACCFor TOC75∆P1Primer 11GGTTTCAAGAAGTATTGGCAATCTGCTGATPrimer 12GACATGTGTGTTCTTCACGGGTATTCTGATCTCTOC75∆P1-2Primer 13GGTTTCAAGAAGTATATAACTCAGCTAGTTPrimer 14GACATGTGTGTTCTTCACGGGTATTCTGATCTCFor TOC75∆P1-3 Primer 15GGTTTCAAGAAGTATCAGAAGTCAGCTGAAGTCPrimer 16GACATGTGTGTTCTTCACGGGTATTCTGATCTCPrimers used for generating transit peptide deletions using TOC75∆P1 and TOC75∆P1-2 cDNAPrimer 17GAAGGAGATATACATATGGATGAAGAACAATCACCGGPrimer 18CTACTTCTTGTTAGTGGCCCATATGTATATCTCCTTCTTAAAGPrimers used to construct POTRA domain constructs for solid phage binding assaysPrimer 19CTCGAGGGTGATGAAGAACAATCACCGGPrimer 20CTCGAGTTCTAGCTCCTTAAGCTTGATCTC10 . 7554/eLife . 12631 . 007Table 3 . List of primers used for RT-qPCR . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 007TOC75Primer 21TTCTTTGATCGACGGAGACPrimer 22CAGCAAACGAGATTGTAACACCTOC75∆P1Primer 23GGTTTCAAGAAGTATTGGCAATCTGCTGATPrimer 24ACATCTGCATAACCTCACCATACATOC75∆P1-2Primer 25GGTTTCAAGAAGTATATAACTCAGCTAGTTPrimer 26GCGTGGATTGACTTCAATGTTTOTAL TOC75Primer 27AAGCTTGGTAATGTGGTTGAAPrimer 28TCAACAATAATGCCCCCTTCTOC159Primer 29AGAACCAACCAACCCCTTCTPrimer 30ACCAAATTCGGCTTCTCCTTTOC33Primer 31GGTGCAAAACCTTGCTTGTTPrimer 32GGAAGAGCCTTTTCGTCCTTTIC22-IIIPrimer 33AAAACATGAGTTATCGCCCTGTPrimer 34TTGCTCAGTTGAAACCTCAAAATIC22-IVPrimer 35ATGCGTTAGAGCTCAAATCCTCPrimer 36CATCTCCATTTTCCTCAACACA Heterozygous toc75-III-1 plants are phenotypically indistinguishable from the wild-type plants ( Baldwin et al . , 2005 ) under normal growth conditions , implying that one copy of TOC75-III is sufficient for chloroplast function . Although most plants transformed with the Toc75 POTRA-deletion constructs ( e . g . TOC75ΔP1#1 , TOC75ΔP1-2#1 or TOC75ΔP1-3#1; Table 1B ) were phenotypically indistinguishable from toc75-III-1 plants ( data not shown ) , several transformed lines ( TOC75ΔP1#2 , TOC75ΔP1-2#2 or TOC75ΔP1-3#2 ) showed progressively increased pale phenotypes and reduced growth rates ( Figure 1C ) . The chlorophyll levels in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were reduced by 26% and 62% , respectively ( Figure 1D ) . TOC75ΔP1-3#2 plants were nearly albino and exhibited severe growth defects ( Figure 1C ) . Interestingly , the hygromycin resistance marker in progeny of the TOC75ΔP1#2 , TOC75ΔP1-2#2 and TOC75ΔP1-3#2 lines segregated at a ratio significantly lower than the expected 2:1 ratio ( Table 1C ) . These data suggest that expression of Toc75ΔP1 , Toc75ΔP1-2 or Toc75ΔP1-3 proteins in these lines competed with the normal function of native Toc75 , expressed from the single wild-type allele in the toc75-III-1 ( +/-background , resulting in disruption of the import apparatus and a consequent impact on plant growth and viability . The differences in the segregation ( Table 1 ) and phenotypes ( Figure 1C ) in the TOC75ΔP1#2 , TOC75ΔP1-2#2 and TOC75ΔP1-3#2 plants compared to TOC75ΔP1#1 , TOC75ΔP1-2#1 and TOC75ΔP1-3#1 plants , suggested that the levels of Toc75ΔP1 , Toc75ΔP1-2 or Toc75ΔP1-3 in these lines varied . To further examine the nature of the aberrant segregation , we focused on TOC75ΔP1#2 , TOC75ΔP1-2#2 lines . We were unable to pursue further characterization of TOC75ΔP1-3 because of the severe phenotype in lines expressing this construct ( Figure 1C ) . We compared the expression levels of Toc75ΔP1 or Toc75ΔP1-2 in lines showing normal 2:1 hygromycin resistance segregation ( TOC75ΔP1#1 and TOC75ΔP1-2#1 ) and those showing aberrant segregation ( TOC75ΔP1#2 and TOC75ΔP1-2#2 ) ( Table 1B and C ) . TOC75ΔP1#1 and TOC75ΔP1-2#1 plants were nearly indistinguishable from heterozygous toc75-III-1 plants , whereas TOC75ΔP1#2 and TOC75ΔP1-2#2 plants showed a lower germination rate and significant growth defects when grown on agar plates ( Figure 2A ) . Real-time reverse transcription quantitative PCR ( RT-qPCR ) ( Figure 2B ) and immunoblotting of extracts ( Figure 2C ) from seedlings of these lines indicated that the expression of the Toc75 POTRA deletion was higher in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants compared to TOC75ΔP1#1 and TOC75ΔP1-2#1 . The examination of additional transformants confirmed that the severity of defects in growth ( Figure 1 – supplement 1 ) and germination ( Figure 2 - supplement 1A and B ) correlated with the levels of Toc75ΔP1 or Toc75ΔP1-2 accumulation ( Figure 2 – supplement 1C ) . Although we did not examine the reasons for the differences in expression between the lines , the location of transgene insertion in the Arabidopsis genome could account for the variation ( Gelvin and Kim , 2007 ) . These data indicate that increased expression of the POTRA deletions induces significant phenotypic defects , consistent with a dominant-negative effect on native Toc75 function in the toc75-III-1 ( +/- ) background . Consistent with this observation , the levels of Toc75ΔP1 or Toc75ΔP1-2 mRNA in TOC75ΔP1#2 or TOC75ΔP1-2#2 plants were 3 . 5–4 . 5 times higher than those of native Toc75 mRNA , whereas the Toc75ΔP1 or Toc75ΔP1-2 protein levels were significantly lower than native Toc75 . This suggests that expression of TOC75ΔP1#2 and TOC75ΔP1-2#2 is tightly controlled post-transcriptionally or by proteolytic degradation of Toc75ΔP1 or Toc75ΔP1-2 , most likely in response to the negative impact of the POTRA deletions on chloroplast function . 10 . 7554/eLife . 12631 . 008Figure 2 . Dominant negative phenotypes exhibited by plants expressing Toc75ΔP1 or Toc75ΔP1-2 . ( A ) Phenotypes of wild type , toc75-III-1 ( +/- ) and TOC75 plants compared to plants expressing lower ( TOC75ΔP1#1 and TOC75ΔP1-2#1 ) or higher ( TOC75ΔP1#2 and TOC75ΔP1-2#2 ) levels of Toc75ΔP1 and Toc75ΔP1-2 , respectively . Plants were grown for 14 days on agar media containing 20 μg/ml hygromycin . ( B ) Toc75 , Toc75ΔP1 or Toc75ΔP1-2 mRNA levels in the transgenic plants indicated at the bottom of the graph as determined by quantitative real-time PCR . The primers used for RT-qPCR are listed in Table 3 . Data represent the mean of three replicates . Error bars represent standard deviation . ( C ) Immunoblots of protein extracts from TOC75 , TOC75ΔP1#1 , TOC75ΔP1-2#1 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants using anti-atToc75 sera . The positions of Toc75 , Toc75ΔP1 ( ΔP1 ) or Toc75ΔP1-2 ( ΔP1-2 ) are indicated to the right of the figure . Immunoblots of actin , a loading control , are shown in the bottom panel . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 00810 . 7554/eLife . 12631 . 009Figure 2—figure supplement 1 . Dominant negative phenotypes exhibited by plants expressing various levels of Toc75ΔP1 or Toc75ΔP1-2 . ( A ) and ( B ) Phenotypes of wild type , toc75-III-1 ( +/- ) and TOC75 plants compared to plants expressing lower ( TOC75ΔP1#1 , #3 and #4 ) or higher ( TOC75ΔP1#2 , #5 and #6 ) levels of Toc75ΔP1 or lower ( TOC75ΔP1-2#1 , #3 and #4 ) or higher ( TOC75ΔP1-2#2 , #5 and #6 ) levels of Toc75ΔP1-2 . A low and high exposure of the blots for the extracts from the TOC75ΔP1 plants is presented to allow visualization of the bands corresponding to Toc75ΔP1 ( ΔP1 ) . All plants were grown for 14 days on agar media containing 20 μg/ml hygromycin . ( C ) Immunoblots of protein extracts from lines expressing Toc75ΔP1 or Toc75ΔP1-2 shown in ( A ) and ( B ) . The positions of Toc75 , Toc75ΔP1 ( ΔP1 ) or Toc75ΔP1-2 ( ΔP1-2 ) are indicated to the right of the figure . Immunoblots of actin , a loading control , are shown in the bottom panel . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 009 The TOC complex mediates the initial recognition of preproteins followed by their translocation across the outer envelope membrane via the Toc75 channel ( Richardson et al . , 2014 ) . To test how deletion of the POTRA domains might impact preprotein translocation at the outer membrane , we measured time-dependent import kinetics in chloroplasts isolated from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . Isolated chloroplasts were incubated with [35S]-labeled chloroplast preprotein , preSSU , in the presence of 3 mM ATP to promote import for 2 , 5 or 10 min at 26° ( Figure 3A ) . Quantitative analysis indicates that preprotein import rates in chloroplasts from toc75-III-1 ( +/- ) and TOC75 plants were statistically indistinguishable ( Figure 3B ) , consistent with the normal phenotype of these plants . By contrast , the levels of [35S]preSSU import into chloroplasts expressing Toc75ΔP1 , and Toc75ΔP1-2 were reduced to ~35% of the levels observed in chloroplasts from toc75-III-1 ( +/- ) and TOC75 plants ( Figure 3B ) . A similar reduction in import was observed for an additional preprotein , the precursor of the E1α subunit of pyruvate dehydrogenase ( [35S]preE1α ) ( Figure 3C ) . PreE1α utilizes an import pathway involving Toc75 and a set of TOC receptor GTPase isoforms distinct from those involved in preSSU import ( Ivanova et al . , 2004 ) . These data demonstrate that Toc75ΔP1 and Toc75ΔP1-2 disrupt the function of protein import complexes involved in multiple import pathways , resulting in the defects in chloroplast biogenesis observed in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . 10 . 7554/eLife . 12631 . 010Figure 3 . Preprotein import is decreased in chloroplasts isolated from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants compared to toc75-III-1 ( +/- ) or TOC75 plants . ( A ) In vitro import with isolated chloroplasts from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . In vitro-translated [35S]preSSU was incubated with equivalent numbers of chloroplasts ( 107 chloroplasts ) from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants under import conditions in the presence of 3 mM ATP for the times indicated . Lane 1 contains 10% of the [35S]preSSU added to each reaction ( IVT ) . Dashed lines indicate that the panels in the figure were generated from different regions of the same SDS-PAGE gel using samples from the same experiment . The chloroplasts were analyzed directly by SDS-PAGE and phosphor imaging . ( B ) Quantitative analysis of the protein import assays in ( A ) . Data represent the mean of triplicate experiments , with bars indicating standard error . ( C ) Comparison of the import of in vitro-translated [35S]preSSUand [35S]preE1α into isolated chloroplasts from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . Import is presented as a percentage of the import of each preprotein observed in toc75-III-1 ( +/- ) chloroplasts . [35S]preSSU import data for the graph were derived from ( A ) . The inset shows a representative gel of the triplicate experiments used to generate the graph for [35S]preE1α . Lane 1 contains 40% of the [35S]pre E1α added to each reaction ( IVT ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 010 To understand whether the reduced import in chloroplasts expressing Toc75ΔP1 , and Toc75ΔP1-2 was due to defects in the recognition or translocation of preproteins at the outer membrane , we examined the interaction of [35S]preSSU with the import machinery under two distinct energy conditions . In the absence of ATP , preproteins bind to chloroplasts in a low-affinity , energy-independent interaction involving the Toc159 and Toc33 GTPase receptors ( Aronsson and Jarvis , 2011 , Inaba and Schnell , 2008 ) . In the presence of 0 . 1 mM ATP , preproteins are promoted to a higher affinity intermediate , which is inserted across the outer membrane and represents the initial translocation of preprotein across the envelope ( Fitzpatrick and Keegstra , 2001 , Inaba and Schnell , 2008 ) . The distinction between energy-independent binding and the formation of the higher affinity intermediate is apparent by an increase in the levels of preprotein that stably associates with chloroplasts ( Olsen and Keegstra , 1992 ) . As expected TOC75 and toc75-III-1 ( +/- ) chloroplasts showed 53% and 97% increases in chloroplast-associated preprotein , respectively , in the presence of 0 . 1 mM ATP compared to chloroplasts depleted of ATP ( Figure 4A and B ) . By contrast , the stimulation of chloroplast-associated [35S]preSSU in the presence of 0 . 1 mM ATP , in TOC75ΔP1#2 , and TOC75ΔP1-2#2 chloroplasts was ~25% and was statistically indistinguishable from energy-independent binding ( Figure 4A and B ) . Incubation of the chloroplasts from the 0 . 1 mM ATP-binding experiment with 3 mM ATP to promote import of the intermediate into the stroma resulted in translocation and processing of 70–85% of bound [35S]preSSU in the TOC75 and toc75-III-1 ( +/- ) control chloroplasts and TOC75ΔP1#2 , and TOC75ΔP1-2#2 chloroplasts ( Figure 4A and C ) . These data suggest that the POTRA deletions impact the transition from low-affinity , energy-independent binding to an energy-dependent , high-affinity insertion across the outer membrane , but not subsequent translocation across the inner envelope into the stroma . 10 . 7554/eLife . 12631 . 011Figure 4 . Deletion of POTRA-1 and -2 disrupt preprotein translocation across the outer membrane . ( A ) Energy-dependent and independent binding of [35S]preSSU to chloroplasts from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . In vitro-translated [35S]preSSU was incubated at 26°C for 5 min with chloroplasts ( 107 ) from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants in the absence of exogenous energy or in the presence of 0 . 1 mM ATP and GTP to promote translocation across the outer membrane and form an early import intermediate ( Inter ) . toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts ( 107 ) containing bound early import intermediate ( Inter ) were reisolated and resuspended in the presence of 3 mM ATP to promote full translocation of the bound preprotein into the stroma ( Chase ) . Lane 1 contains 10% of the in vitro translated [35S]preSSU added to each reaction ( IVT ) . Dashed lines indicate that the figure was generated from different regions of the same SDS-PAGE gel using samples from the same experiment . The chloroplasts were analyzed directly by SDS-PAGE and phosphor imaging . ( B ) Quantitation of the [35S]preSSU early import intermediate from ( A ) . ( C ) Quantitation of [35S]SSU imported from the chase experiment in ( A ) . Data represent the mean of triplicate experiments , with bars indicating standard error . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 011 The results from Figure 4 suggest that the POTRA deletions do not have a major impact on initial energy-independent binding of preSSU , but significantly impact preprotein translocation across the outer membrane . To test this hypothesis , we investigated the kinetic parameters of import in TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts to determine the affinity and maximum translocation capacity of TOC translocons for preSSU . We generated a preprotein binding curve with increasing concentrations of purified E . coli-expressed preSSU-FLAG-HIS in the presence of 0 . 1 mM ATP and visualized bound preprotein by immunoblotting with anti-FLAG ( Figure 5A ) . The apparent dissociation constants ( Kd ( app ) ) of preSSU-FLAG-HIS for TOC translocons in TOC75 and TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts were 158 . 6 ± 29 nM , 187 . 7 ± 46 nM and 126 . 5 ± 24 mM , respectively ( Figure 5B ) . This indicates that there were no measurable differences in the affinity of TOC translocons for preprotein in these plants . However , the maximum binding capacity was estimated as 91 . 4 ± 6 . 2 , 57 . 1 ± 5 . 6 and 48 . 2 ± 3 . 3 fmol preSSU/μg chloroplast protein in TOC75 and TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts , respectively ( Figure 5B ) . As shown in Figure 7 , the levels of TOC proteins were not reduced in the TOC75ΔP1#2 and TOC75ΔP1-2#2 plants , indicating that the reduced binding capacity was not due to a reduction in the levels of import components . These data are consistent with the hypothesis that deletion of the POTRA domains reduces the number of active TOC complexes at the outer envelope by interfering with their ability to mediate outer membrane translocation . 10 . 7554/eLife . 12631 . 012Figure 5 . Expression ofToc75ΔP1 and Toc75ΔP1-2reduces the number of import-competent TOC complexes . ( A ) Saturation binding of preprotein in chloroplasts from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . Isolated chloroplasts ( 107 ) from TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were incubated with increasing amounts of E . coli-expressed , urea-denatured preSSU-FLAG-HIS in the presence of 0 . 1 mM ATP and GTP at 26°C for 5 min , to promote formation of an early import intermediate . Chloroplasts were reisolated through Percoll silica gel , resolved by SDS-PAGE , and the early import intermediate form of preSSU-FLAG-HIS was detected by immunoblotting with anti-FLAG . Lane 1 contains 0 . 64 pmol of the preSSU-FLAG-HIS . ( B ) Quantitation of the data from ( A ) . Saturation binding analysis of the data in ( A ) is presented in the table inset . The maximum number of binding sites ( Maximum Binding ) and apparent Kd were calculated by nonlinear fitting of the data in ( A ) . Each data bar represents the mean ± SD ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 012 To investigate the nature of the import defect in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants , we examined whether the POTRA deletions disrupted Toc75 targeting and membrane integration . Toc75 is unique among known chloroplast outer envelope membrane proteins in being targeted to the membrane via a cleavable N-terminal bipartite targeting signal ( Tranel and Keegstra , 1996 ) . Toc75 is translated as an 89 . 2 kDa precursor ( pre-Toc75 ) and processed sequentially at the outer membrane and intermembrane space , to generate an 84 . 6 kDa intermediate form and the 75 kDa mature form , respectively ( Inoue et al . , 2001 , Inoue and Keegstra , 2003 , Shipman and Inoue , 2009 ) . Upon complete processing , the predicted sizes of mature Toc75ΔP1 and Toc75ΔP1-2 are 66 . 6 kDa and 52 . 7 kDa , respectively . To determine if the POTRA deletions were processed normally , we synthesized [35S]Toc75ΔP1 and [35S]Toc75ΔP1-2 in a cell-free in vitro translation system , mixed them with protein extracts from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants and compared the sizes of the radiolabeled and endogenous Toc75ΔP1 and Toc75ΔP1-2 by phosphorimaging and immunoblotting , respectively . The results in Figure 6A show that the SDS-PAGE migration of the radiolabeled and endogenous Toc75ΔP1 and Toc75ΔP1-2 are coincident , indicating that the POTRA-deletions were processed to their mature forms . Consistent with proper processing , fractionation of chloroplasts into soluble and membrane fractions demonstrated that Toc75ΔP1 and Toc75ΔP1-2 from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants , respectively , were membrane integrated ( Figure 6B ) . 10 . 7554/eLife . 12631 . 013Figure 6 . Toc75ΔP1 and Toc75ΔP1-2are properly targeted to the outer membrane with the POTRA domains oriented to the intermembrane space . ( A ) Toc75ΔP1 and Toc75ΔP1-2 accumulate as their mature forms in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . In vitro translated mature [35S]Toc75ΔP1 and [35S]Toc75ΔP1-2 were mixed with chloroplasts extracts from TOC75ΔP1#2 and TOC75ΔP1-2#2 plants and analyzed by SDS-PAGE . The mobility of [35S]Toc75ΔP1 and [35S]Toc75ΔP1-2 , detected by phosphorimaging ( Lanes 1 and 2 ) , was compared to the mobility of endogenous Toc75ΔP1 and Toc75ΔP1-2 , detected by immunoblotting with anti-atToc75 serum ( Lanes 3 and 4 ) . ( B ) Toc75ΔP1 and Toc75ΔP1-2 are integrated into the chloroplast envelope . Isolated chloroplasts ( T ) from heterozygous toc75-III-1 , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were hypotonically lysed and fractionated by centrifugation at 18 , 000 × g for 30 min at 4°C into membrane pellet ( M ) and soluble ( S ) fractions . Equivalent samples of each fraction from chloroplasts corresponding to 10 μg chlorophyll was resolved by SDS-PAGE and immunoblotted with anti-atToc75 serum . ( C ) Toc75myc and Toc75ΔP1myc constructs used in this study . The dashed line represents the deleted POTRA domain region in POTRA1-deleted constructs . The site of insertion of myc tag is shown in each panel . ( D ) Protease sensitivity of Toc75myc , and Toc75ΔP1myc proteins in isolated intact chloroplasts . Intact chloroplasts from TOC75 , TOC75myc and TOC75ΔP1myc seedlings were treated with trypsin or thermolysin in the absence ( - ) or presence ( + ) of 1% Triton X-100 . Reactions were incubated on ice for 30 min , and proteolysis was stopped with 2 . 5 mM PMSF , 0 . 05 mg/mL Nα-Tosyl-L-lysine chloromethyl ketone ( TLCK ) , 0 . 25 mg/mL soybean trypsin inhibitor , and 2 μg/mL aprotinin ( for trypsin ) or 20 mM EDTA ( for thermolysin ) . Chloroplasts were analyzed by immunoblotting with antibodies against various proteins as indicated . The asterisk indicates the position of a non-specific immunoreactive band . a and b denote bands corresponding to the ~55 kDa and ~46 kDa trypsin fragments of Toc75myc and Toc75ΔP1myc , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 01310 . 7554/eLife . 12631 . 014Figure 7 . Accumulation of outer envelope proteins in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . ( A ) Serial dilutions of protein extracts from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were immunoblotted using antisera corresponding to the proteins indicated at the left . ( B ) Relative quantitation of plastid proteins . The signal intensities falling within the linear range of chemiluminescence detection were normalized to the signal for actin in each sample and plotted as fold change relative to the levels of the corresponding protein in toc75-III-1 ( +/- ) plants . ( C ) Relative expression of various proteins analyzed by RT-qPCR from toc75-III-1 ( +/- ) , TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . The values are normalized to the internal levels of actin mRNA and plotted as the fold change relative to the levels of the corresponding mRNA in toc75-III-1 ( +/- ) plants . Each error bar represents the mean ± SD ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 014 To test if the POTRA deletions impacted the topology of Toc75 in the outer membrane , we examined the sensitivity of full-length Toc75 and Toc75ΔP1 to exogenous protease in isolated , intact chloroplasts . We were unable to examine Toc75ΔP1-2 due to relatively low levels of expression of this construct in transgenic plants and the consequent inability to reliably detect proteolytic fragments from the deletion construct . To facilitate our studies , we expressed Toc75 and Toc75ΔP1 containing a myc epitope inserted just upstream of the POTRA domains at the N-terminus of the mature proteins in toc75-III-1 plants ( Figure 6C ) . The TOC75myc transgene was able to fully complement the toc75-III-1 lethal phenotype ( Table 1D ) . Consistent with the TOC75ΔP1 transformants , the TOC75ΔP1myc lines showed both normal and pale phenotypes , and none were able to complement toc75-III-1 ( Table 1D ) . For subsequent studies , we selected a line , which showed a pale phenotype TOC75ΔP1myc ( data not shown ) similar to TOC75ΔP1#2 . Isolated chloroplasts from TOC75myc and TOC75ΔP1myc plants were treated with various concentrations of trypsin or thermolysin , resolved by SDS-PAGE and immunoblotted with anti-myc or antibodies specific to the Toc75 POTRA domains . Previous studies have shown that Toc75 is resistant to low-to-moderate levels of thermolysin , whereas trypsin generates several proteolytic fragments , including peptides corresponding to the N-terminal POTRA deletions ( Tranel and Keegstra , 1996 , Inoue and Potter , 2004 , Sveshnikova et al . , 2000 ) . As controls for protease activity , we examined the sensitivity of Toc33 and Tic110 to the protease treatments . Toc33 is anchored in the outer membrane , with a short C-terminal transmembrane helix and the bulk of the protein is exposed to the cytoplasm ( Kessler et al . , 1994 ) , and Tic110 is anchored to the inner membrane , with the bulk of the protein extending into the stroma ( Li and Schnell , 2006 ) . As predicted , Toc33 is digested by both trypsin and thermolysin , with complete degradation of the cytoplasmic domain observed at 200 μg/ml trypsin and 100 μg/ml thermolysin ( Figure 6D i ) . By contrast , Tic110 is completely resistant to the same concentrations of trypsin or thermolysin unless the membrane is disrupted with triton X-100 ( Figure 6D i ) . Toc75myc is degraded by treatments with increasing concentrations of trypsin ( Figure 6D ii ) . Reduction in full length Toc75myc ( 75 kDa ) is concomitant with the appearance of a ~55 kDa protease-resistant fragment , which reacts with anti-POTRA and anti-myc ( Figure 6D ii , band a ) . Disruption of the membrane with triton X-100 results in complete degradation of Toc75 , indicating that the protected fragment observed in intact chloroplasts is not intrinsically resistant to proteolysis ( Figure 6D ii ) . Treatment of chloroplasts from TOC75ΔP1myc plants with trypsin resulted in the degradation of both the 75 kDa Toc75 and the 68 kDa Toc75ΔP1myc proteins ( Figure 6D iii ) , and generated two anti-POTRA reactive proteolytic fragments of ~55 kDa and ~46 kDa ( Figure 6D iii , bands a and b ) . The difference in the size of the two anti-POTRA reactive fragments corresponds to the difference in size between native Toc75 and Toc75ΔP1 ( 8 . 6 KDa ) . The ~46 kDa fragment reacted with anti-myc ( Figure 6D iii , band b ) , confirming its identity as an N-terminal fragment containing the Toc75ΔP1myc POTRA domain fragment . These data demonstrate that the peptides are generated from the N-terminal region of Toc75 and Toc75ΔP1myc , respectively , and confirm that Toc75ΔP1 has the same topology as native Toc75 in the outer envelope membrane . Toc75 is implicated in the insertion of Toc33 and Toc159 into the outer membrane ( Richardson et al . , 2014 ) , and it participates in the initial stages of targeting of TIC components to the inner membrane ( Shi and Theg , 2013 ) . Consequently , we examined the expression and accumulation of TOC and TIC components in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants to determine if the import defects could result from a disruption in the levels of import components . Immunoblots indicated that overall levels of Toc75 , including Toc75 , Toc75ΔP1 and Toc75ΔP1-2 , were increased up to two fold in plants expressing POTRA-deleted Toc75 compared to wild type or TOC75 plants ( Figure 7A and B ) . Levels of Toc33 protein were significantly increased ( ~five fold ) in TOC75ΔP1 and TOC75ΔP1-2 plants , suggesting that accumulation of Toc33 is increased in response to expression of the POTRA deletion constructs ( Figure 7A and B ) . Toc159 levels were slightly increased ( <1 . 4-fold ) , but the increase was not statistically significant compared to toc75-III-1 ( +/- ) plants ( Figure 7A and B ) . The levels of OEP80 , another member of Omp85 superfamily in the outer membrane of chloroplasts ( Hsu et al . , 2012 , Day et al . , 2014 ) , were unchanged in plants expressing Toc75ΔP1 and Toc75ΔP1-2 ( Figure 7A and B ) , further indicating that the increased accumulation of Toc33 was not due to a general increase in outer membrane protein levels . A representative TIC component , Tic110 , did not show a change in abundance in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants ( Figure 7A and B ) . However , the levels of Tic22-III and Tic22-IV , two homologous protein import components of the intermembrane space ( Kouranov et al . , 1998 , Kasmati et al . , 2013 ) , were increased in TOC75ΔP1 and TOC75ΔP1-2 plants . Tic22-IV levels were moderately but significantly increased , whereas Tic22-III showed a ~three fold increase in both deletion lines ( Figure 7A and B ) . We performed RT-qPCR to determine if the increases in protein accumulation observed in the immunoblotting experiments correlated with increased transcription . Compared to toc75III-1 ( +/- ) and TOC75 plants , the levels of total transcripts encoding Toc75 and the POTRA deletions ( Toc75 , Toc75ΔP1 and Toc75 ΔP2 ) were increased by four- and six fold in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants , respectively ( Figure 7C ) . Toc33 transcript also increased , as did those of Tic22-III and Tic22-IV ( Figure 7C ) , whereas Toc159 transcript levels were unchanged ( Figure 7C ) . These data suggest that the changes observed in the levels of import components in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were at least partially due to changes in gene expression in response to expression of the Toc75 POTRA deletions . Toc75 forms stable complexes with Toc33 and Toc159 in stoichiometric ratios estimated at 4:4:1 or 3:3:1 ( Toc75:Toc34:Toc159 ) , based on the mobility of TOC complexes on blue-native gel electrophoresis ( Schleiff et al . , 2003 , Kikuchi et al . , 2006 , Chen and Li , 2007 ) . The over-accumulation of Toc75 and Toc33 in the POTRA deletion lines ( Figure 7 ) suggested that the mutations might disrupt the stoichiometry of TOC complexes , thereby resulting in the observed import defects ( Figure 3 ) . As a first step to determine if Toc75ΔP1 and Toc75ΔP1-2 were interacting with other TOC components , we immunoprecipitated detergent-solubilized chloroplast membranes from TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants with anti-Toc33-Sepharose and immunoblotted the samples with TOC antibodies . Toc159 , full-length Toc75 and Tic110 , a component of TIC complexes , co-immunoprecipitated with Toc33 from TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts . Toc75ΔP1 and Toc75ΔP1-2 also co-immunoprecipitate with Toc33 ( Figure 8A ) . OEP80 , an outer membrane protein that is not associated with TOC complexes , was not co-immunoprecipitated with anti-Toc33 ( Figure 8A ) . These data indicate that Toc75ΔP1 and Toc75ΔP1-2 can interact directly or indirectly with other import components of TOC-TIC complexes . 10 . 7554/eLife . 12631 . 015Figure 8 . Toc75ΔP1 and Toc75ΔP1-2 interfere with the stoichiometric assembly of TOC complexes . ( A ) Co-immunoprecipitation of Toc75ΔP1 and Toc75ΔP1-2 with Toc33 and Toc159 . Detergent-soluble chloroplast membranes from TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants were subjected to immunoaffinity chromatography on anti-atToc33 Sepharose . Twenty-five percent of the total extracts ( Start ) and unbound fractions ( Unbound ) or the eluate fractions ( Bound ) were resolved by SDS-PAGE and immunoblotted with antisera indicated to the left of each panel . ( B ) 2D Blue-native PAGE analysis of chloroplast membranes from TOC75 , TOC75ΔP1#2 and TOC75ΔP1-2#2 plants . Chloroplast membranes from TOC75 ( i ) , TOC75ΔP1#2 ( ii ) and TOC75ΔP1-2#2 ( iii ) plants were subjected to 4–12% BN-PAGE in the first dimension followed by 5–12% SDS-PAGE in the second dimension . Proteins were then transferred to a PVDF membrane and immunoblotted with the antisera indicated to the left of each panel . The positions of the major 1 . 3 MDa ( region a ) , 440 kDa ( region b ) and 70 kDa ( region c ) complexes are indicated at the top of each panel . Asterisks to the left of the middle panels indicate the positions of the bands corresponding Toc75 ( * ) , Toc75ΔP1 ( ** ) and Toc75ΔP1-2 ( *** ) . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 015 To determine if the interactions detected in Figure 8A correspond to the correct stoichiometric assembly of TOC components , we analyzed the complexes by two-dimensional ( 2D ) blue-native PAGE . As shown in Figure 8B i , Toc33 , Toc75 and Toc159 from TOC75 chloroplasts co-migrated in complexes , which peaked at ~1 . 3 MDa ( region a ) . A second complex , with a peak at ~440 kDa ( region b ) migrated between 400 and 880 kDa and contained Toc75 and Toc33 , but lacked detectable levels of Toc159 . The presence of the ~1 . 3 MDa and ~440 kDa complexes is consistent with the 800–1000 kDa and 300–450 kDa TOC complexes previously reported from pea chloroplasts ( Kikuchi et al . , 2006 ) on 2D blue-native gels . The minor differences in the mobilities of the TOC complexes reported here and in the previous study are likely due to differences in the sizes of TOC components in Arabidopsis versus pea chloroplasts . Both 1 . 3 MDa and 440 kDa complexes containing native Toc75 , Toc33 and Toc159 were also observed in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants ( Figure 8B ) . Toc75ΔP1 is detected in both 1 . 3 MDa and 440 kDa complexes in chloroplasts from TOC75ΔP1#2 plants ( Figure 8B ii ) , suggesting that Toc75ΔP1 assembles with the core TOC GTPase receptors . By contrast , Toc75ΔP1-2 was absent from the 1 . 3 MDa complexes and present only in complexes with a mobility similar or smaller than the 440 kDa complexes in TOC75ΔP1-2#2 plants ( Figure 8B iii ) . In conjunction with the observation that Toc75ΔP1-2 co-immunoprecipitates with Toc33 ( Figure 8A ) , the results from 2D blue-native gels suggest that Toc75ΔP1-2 is able to interact with Toc33 but is unable to assemble into the 1 . 3 MDa complexes containing all three core TOC components . Interestingly , the levels of Toc33 migrating at a low molecular weight ( ~70 kDa peak ) were increased significantly in TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts compared to TOC75 chloroplasts ( Figure 8B , region c ) . Neither Toc75 , the Toc75 POTRA deletions or Toc159 co-migrate with this form of Toc33 , suggesting that a considerable proportion of Toc33 is not assembled into native TOC complexes in the TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts . The dramatic increase in Toc33 expression relative to the other TOC components in TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts ( Figure 7 ) is likely to be responsible , at least in part , for the Toc33 that is not assembled into TOC complexes ( Figure 8B ) . Taken together , these data demonstrate that deletion of both the first and second POTRA domains of Toc75 disrupts assembly of native TOC complexes within the outer membrane . The increase in accumulation of the Tic22 isoforms in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants , raised questions regarding the relationship of these two import proteins . Previous work had demonstrated a potential interaction between the N-terminal POTRA domains of an Omp85 family member in the outer membrane of the cyanobacteria , Anabeana sp . PCC 7120 , and a chaperone protein in the periplasmic space , which is structurally related to Tic22 ( Tripp et al . , 2012 ) . These observations led us to examine a possible physical interaction between Toc75 and Tic22 in the chloroplast intermembrane space . A protein fragment encompassing all three Toc75 POTRA domains and a C-terminal hexahistidine tag ( POTRA-His ) was expressed and purified from E . coli and tested for binding to Tic22-III and Tic22-IV . The recombinant POTRA-His was immobilized on nickel-nitrilotriacetic acid ( Ni-NTA ) matrix and incubated with in vitro–translated [35S]Tic22-III and [35S]Tic22-IV . Binding was measured as the fraction of [35S]Tic22-III and [35S]Tic22-IV that co-sedimented with the immobilized POTRA domains . Transit peptide-dependent binding of preSSU to Toc75 POTRAs has been demonstrated previously ( Ertel et al . , 2005 ) , and therefore , we used [35S]preSSU and [35S]SSU as positive and negative controls for binding , respectively . As shown in Figure 9 , [35S]Tic22-III , [35S]Tic22-IV and [35S]preSSU bound to immobilized POTRA-His in a dose-dependent manner , with maximum binding observed at 60% , 35% and 77% of added radiolabeled proteins , respectively ( Figure 9A and B ) . None of the proteins exhibited significant binding to the Ni-NTA matrix alone ( Figure 9 A , lanes 2 and 7 ) , and <10% of added [35S]SSU , the negative control , associated with the resin at the highest concentrations tested ( Figure 9B ) . Soluble POTRA domains effectively competed for binding of [35S]Tic22-III , [35S]Tic22-IV and [35S]preSSU to immobilized POTRA-His in a dose-dependent manner ( Figure 9 C ) , whereas soluble SSU was unable to compete , further demonstrating the specificity of binding . These data demonstrate the ability of Tic22-III and Tic22-IV to directly bind to Toc75 POTRA domains and are consistent with a functional interaction between these proteins in the intermembrane space . 10 . 7554/eLife . 12631 . 016Figure 9 . Toc75 POTRA domains interact directly with chloroplast preproteins and Tic22 isoforms . ( A ) Binding of Tic22-III , Tic22-IV and preSSU to immobilized Toc75 POTRA domains . [35S]-labeled Tic22-III , Tic22-IV , preSSU and SSU were incubated with increasing amounts of Ni-NTA resin-immobilized Toc75 POTRA-His . Lane 2 in A shows the background binding of radiolabeled proteins to the Ni-NTA resin in the absence of POTRAs . In all cases , binding to Ni-NTA resin alone ( lane 2 ) was less than 10% of added radiolabelled protein . ( B ) Quantitation of the binding in ( A ) . Binding is presented as the percentage of added [35S]-labeled proteins recovered in each reaction after subtracting binding to Ni-NTA resin alone ( lane 2 ) . ( C ) Binding of POTRA domains to preSSU , Tic22-III or Tic22-IV is specific . Two hundred picomole immobilized POTRA-His was incubated with [35S]-labeled preSSU , Tic22-III or Tic22-IV in the absence or presence of increasing concentrations of soluble POTRA domains or SSU as a competitor . ( D ) Quantitation of the binding in ( C ) . Binding is presented as the percentage of maximal initial binding . Each data bar represents the mean ± SEM ( n = 3 ) . Lanes labeled IVT in panels A and C contain 10% of the in vitro translation product added to each reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 12631 . 016
The number and function of POTRA domains in Omp85 family members vary , but in all systems , they mediate interactions with transport substrates and/or factors that assist with Omp85 protein function ( Jacob-Dubuisson et al . , 2009 ) . POTRA domains are defined by repeats of conserved β1α1α2β2β3 secondary structural motifs ( Clantin et al . , 2007 , Koenig et al . , 2010 , Paila et al . , 2015 ) , and the variation in primary structure within the repeats contributes to the functional diversification and adaptation of family members for specific roles in the transport and insertion of nascent polypeptides at the outer membrane in bacteria and mitochondria ( Sanchez-Pulido et al . , 2003 , Koenig et al . , 2010 ) . Our analysis of Toc75 is consistent with a multifunctional cassette model for POTRA domains , and highlights the unique adaptation of the Toc75 POTRA domains to mediate interactions that are essential for assembly of the core TOC components and the transport of polypeptides across the outer membrane . We took advantage of the observation that expression of Toc75ΔP1 , Toc75ΔP1-2 and Toc75ΔP1-3 in the heterozygous toc75III-1 background resulted in plants with pale phenotypes and significant growth defects ( Figure 1C ) . The phenotypes were dependent on the expression levels of the deletion constructs , and severe effects were observed even at levels of Toc75ΔP1 , Toc75ΔP1-2 and Toc75ΔP1-3 that were far less than equimolar to endogenous native Toc75 ( Figure 1C ) . Given that TOC translocons are large multimeric complexes containing component ratios estimated at 4:4:1 or 3:3:1 ( Toc75:Toc34:Toc159 ) , the insertion of defective POTRA deletion subunits into these complexes likely interferes with the coordinate functions of the native subunits . Toc75ΔP1 and Toc75ΔP1-2 were targeted , integrated and fully processed at the outer membrane ( Figure 6A and B ) , with their POTRA domains extending into the intermembrane space ( Figure 6 ) , which is consistent with the topology of native Toc75 and all other known Omp85 family members ( Sanchez-Pulido et al . , 2003 , Stroud et al . , 2011 ) . This indicates that the POTRA domains do not play essential roles in targeting and membrane insertion of Toc75 itself . The orientation of the Toc75 POTRA domains in the intermembrane space also is consistent with the observation that they specifically interact with Tic22 isoforms , proteins localized to the intermembrane space ( Figure 9 ) . A separate study using N-terminal fusions of Toc75 to self-assembling split GFP ( saGFP ) concluded that the N-terminus of Toc75 was localized in the cytoplasm when transiently expressed in Arabidopsis protoplasts ( Sommer et al . , 2011 ) . However , the saGFP-Toc75 constructs used in the study were not shown to complement the lethal phenotype of toc75-III-1 mutants , and therefore , it is unclear if the fusions assumed a functional topology in the outer membrane . Our studies of protein import into isolated chloroplasts from the TOC75ΔP1#2 and TOC75ΔP1-2#2 plants demonstrate that the POTRA deletions resulted in a significant reduction in the rate of protein import into chloroplasts ( Figure 3 ) . The analysis of saturation-binding experiments under low [ATP] and [GTP] revealed that the decrease in import was due to the reduction in the total number of functional import sites rather than a significant reduction in the affinity of TOC complexes for the preprotein ( Figure 5 ) . Furthermore , the levels of TOC components were unchanged or increased in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants ( Figure 7 ) , suggesting that the decreased import capacity resulted from defective TOC function and not from decreases in the overall levels of the core TOC proteins . Our data reveal an interesting function for the POTRA domains in TOC complex assembly that accounts , at least in part , for the decreased levels of functional TOC complexes observed in TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts . Expression ( Figure 7 ) and 2D blue-native PAGE ( Figure 8 ) analyses revealed significant changes in the levels and assembly of the core TOC components , which likely results in a population of defective or non-functional translocons . Toc75ΔP1 appeared to retain the ability to assemble with both Toc33 and Toc159 GTPase receptors into 1 . 3 MDa and 440 kDa complexes ( Kikuchi et al . , 2006 ) , suggesting that POTRA1 is not required for stoichiometric assembly of TOC complexes . However , Toc75ΔP1-2 was only detected in novel complexes that migrated between 200–880 kDa . A major fraction of Toc33 , but not native Toc75 or Toc159 , exhibited a coincident broad range of distribution with Toc75ΔP1-2 ( Figure 8B ) . Together with the observation that Toc75ΔP1-2 co-immunoprecipitated with Toc33 , this suggests that Toc75ΔP1-2 is able to associate normally with Toc33 , but not Toc159 , evidenced by the lack of Toc159 in the 200–880 kDa complexes in TOC75ΔP1-2 chloroplasts . This implicates POTRA-2 alone , or in combination with POTRA-1 , in interactions between Toc75 and Toc159 , which is a requirement for proper TOC complex assembly and function . The role of Toc75 POTRAs in TOC assembly is reminiscent of the role of the five repeat POTRA domains of the BamA in the assembly of the BAM complex at the outer membrane . POTRA 5 of BamA interacts with the essential periplasmic lipoprotein , BamD , to form the core of the complex ( Malinverni et al . , 2006 , Sklar et al . , 2007 , Wu et al . , 2005 ) . The core complex recruits BamC and E ( Albrecht and Zeth , 2011 , Kim et al . , 2007 ) , and POTRA 2–4 of BamA interact with BamB ( Kim et al . , 2007 , Kim and Paetzel , 2011 , Noinaj et al . , 2011 , Dong et al . , 2012 ) to constitute the fully functional BAM complex . Despite the conservation of functional domains , Toc75 appears to be unique amongst Omp85 family members because it evolved to mediate protein import into plastids from the cytoplasm . The reversal of the transport activity relative to the bacterial and mitochondrial Omp85 family members occurred with the diversification of Toc75 from the cyanobacterial Omp85 ancestor and was driven by the association of Toc75 with the Toc33 and Toc159 receptors . Toc33 and Toc159 provide the GTP-dependent driving force for the unidirectional transport of preproteins from the cytoplasm into the organelle . The Toc33-Toc75-Toc159 core is a universal feature of the plant lineage , and the adaptation of POTRAs to facilitate these interactions was , therefore , a key event in the evolution of TOC complexes . Our data also indicate that disruption of Toc75 function in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants perturbs the known coordinate expression of the TOC components ( Sun et al . 2001 ) , resulting in significant over-expression of Toc33 relative to Toc75 and Toc159 . The major portion of the excess Toc33 migrated at ~70 kDa on blue-native PAGE of TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts and was not associated with the other TOC components ( Figure 8B ) . The excess Toc33 could contribute to the phenotypes observed in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants if it interacted with preproteins and interfered with import . However , our quantitative binding data indicate that energy-independent and energy-dependent bindings are reduced in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants ( Figures 4 and 5 ) , arguing against preprotein binding to excess Toc33 as a significant factor in TOC75ΔP1#2 and TOC75ΔP1-2#2 phenotypes . We also hypothesize that the Toc75 POTRA domains play a direct role in protein translocation across the outer membrane by selectively binding to chloroplast preproteins in the intermembrane space ( Figure 9 ) . This proposal is consistent with our observation that the isolated POTRA domains interact with a chloroplast preprotein , preSSU , in a transit-peptide and dose-dependent manner , and previous data demonstrating an interaction between the N-terminal region of Toc75 and preproteins targeted to chloroplasts ( Ertel et al . , 2005 ) . Several possibilities exist for the role of the POTRA-transit peptide interaction during protein import . Toc75 crosslinks to the transit peptides of preproteins during initial binding as well as during later stages of membrane translocation ( Ma et al . , 1996 ) , suggesting that the POTRA domains could contribute to preprotein recognition in conjunction with the TOC GTPases and/or stabilize preprotein binding to the translocon upon initial translocation through the Toc75 membrane pore . The latter possibility appears to be most likely , considering the localization of the POTRA domains in the intermembrane space , and the observation that Toc75ΔP1 or Toc75ΔP1-2 did not impact energy-independent binding of preprotein to chloroplasts ( Figures 4 and 5 ) . The substrate-binding activity of POTRA domains has precedence in other Omp85 family members . The POTRA domain of BamA binds nascent β-barrel proteins as they emerge into the periplasmic space from the Sec translocon in the bacterial inner membrane , suggesting that the BamA POTRAs form a platform for the polypeptide substrate to assume the necessary secondary structure required to initiate membrane insertion ( Gatzeva-Topalova et al . , 2008 , Kim et al . , 2007 , Knowles et al . , 2008 ) . POTRA-5 , located most adjacent to the β-barrel , is proposed to play an essential role in assisting the insertion of the substrate into the membrane directly or in coordination with the β-barrel and the oligomeric BAM complex ( Sinnige et al . , 2015 , Noinaj et al . , 2013 ) . The single POTRA of Sam50 β-barrel component of the mitochondrial SAM machinery also is proposed to assist in release of the β-barrel substrate into the mitochondrial outer membrane ( Kutik et al . , 2008 , Stroud et al . , 2011 ) . Recent crystal structures of BamA showed two different orientations of its POTRA domains , leading to the proposal that the domains are involved in a gating mechanism for regulating access of the substrate to the the β-barrel membrane insertase ( Noinaj et al . , 2013 , Bakelar et al . , 2016 ) . A crystal structure and molecular dynamics simulations of POTRA domains of OMP85 from Anabaena revealed a similar conserved feature in these proteins ( Koenig et al . , 2010 ) , and mutations in the unique loop region of the POTRA domains adjacent to the β-barrel in the cyanobacterial protein influenced the pore properties of the β-barrel ( Koenig et al . , 2010 , Ertel et al . , 2005 ) . It is tempting to speculate that a similar gating mechanism might function at Toc75 , thereby providing a mechanism by which transit peptide binding at TOC complexes controls the outer membrane channel by triggering reorientation of the POTRA domains . Finally , we also demonstrate specific interactions of the Toc75 POTRA domains with two homologous import components of the intermembrane space , Tic22-III and Tic22-IV . Tic22 from pea was previously shown to covalently cross-link to preprotein import intermediates ( Kouranov et al . , 1998 ) , and proteins with significant similarity to Tic22 from apicoplasts of Toxoplasma gondii and Plasmodium falciparum have been shown to exhibit chaperone activity ( Glaser et al . , 2012 ) . Interestingly , a Tic22-like protein from cyanobacteria was shown to covalently cross-link with an Omp85 family member of the cyanobacterial outer membrane ( Tripp et al . , 2012 ) , providing a precedent for the interaction of Omp85 family members with the Tic22 family of putative chaperones , and its potential conservation from the cyanobacterial endosymbiont . This interaction is also reminiscent of the interaction between the BamA POTRAs and a periplasmic chaperone , SurA , during β-barrel insertion at the outer membrane ( Bennion et al . , 2010 ) . The significant increase in the levels of Tic22 isoforms in TOC75ΔP1#2 and TOC75ΔP1-2#2 plants is consistent with a functional association between the Toc75 POTRA domains and the Tic22 chaperones and could represent a response to a disruption in chaperone-preprotein interactions during membrane translocation . To date , a chaperone activity for Tic22-III and Tic22-IV in plants has not been investigated , and it will be of great interest to understand how the Toc75 POTRA domains might coordinate interactions between preproteins and the putative chaperone to facilitate preprotein transit through the intermembrane space and engagement of the TIC complex . Our analysis of Toc75 is consistent with models in which the POTRA domains play versatile roles in TOC assembly and function , which account for the decreased levels of functional TOC complexes seen in TOC75ΔP1#2 and TOC75ΔP1-2#2 chloroplasts . As is the case in other Omp85 family members , each domain is likely to contribute individual activities , but also work in association with the other domains to mediate assembly of the translocon complex and coordinate preprotein translocation . In contrast to bacterial and mitochondrial outer membranes , chloroplasts contain two Omp85 family members: Toc75 and OEP80 ( Outer Envelope Protein 80 ) ( Hsu and Inoue , 2009 ) . Similar to Toc75 , OEP80 is essential in Arabidopsis ( Patel et al . , 2008 ) and is predicted to contain three N-terminal soluble POTRA domains and a C-terminal β-barrel ( Hsu and Inoue , 2009 , Hsu et al . , 2012 ) . OEP80 is not detected in isolated oligomeric TOC complexes , and it has been proposed to function in β-barrel protein biogenesis in a manner analogous to Sam50 and BamA ( Eckart et al . , 2002 ) . Although phylogenetic analyses indicated that Toc75 and OEP80 are more closely related to each other than either is to other bacterial Omp85 family members , distinct sequence signatures have been identified in both the POTRA and β-barrel domains of the chloroplast proteins to account for evolutionary divergence of these two β-barrel channels ( Day et al . , 2014 ) . The evolution of an import apparatus to mediate the transport of nucleus-encoded preproteins into plastids was an essential element of endosymbiosis . Our data are consistent with the divergent evolution of Toc75 and OEP80 from the ancestral cyanobacterial OMP85 gene to retain the pathway for β-barrel protein biogenesis , while providing a novel pathway for protein import into the organelle .
Seedlings of Arabidopsis thaliana Col-0 were grown on phytoagar plates containing 0 . 5 × Murashige and Skoog growth medium ( MS medium ) , 1% sucrose under long-day condition for 14-day at 22°C . Genetic complementation was carried out as described earlier ( Shipman-Roston et al . , 2010 ) with a few modifications . Briefly , the Arabidopsis TOC75 genomic fragment of 4 . 7 kb , including 1 kb upstream and 0 . 5 kb downstream of the coding sequence in pDONR221 was generated as described ( Shipman-Roston et al . , 2010 ) , producing the entry clone pDONRgTOC75wt . All deletion constructs were made using the Quickchange II XL site-directed mutagenesis kit ( Makarova et al . , 2000 ) using the primers listed in Table 2 . The expression clones were generated by an LR recombination reaction between entry clones and the Gateway destination vector , pBnRGW ( http://gateway . psb . ugent . be ) . The destination vector contains a fluorescent marker , DsRed , under a seed specific promoter and BASTA ( glufosinate ammonium ) resistance , allowing for rapid screening of seed and seedlings . Final constructs were confirmed by sequencing and introduced into heterozygous toc75-III-1 plants ( Baldwin et al . , 2005 ) by the Agrobacterium tumefaciens-mediated floral dip method ( Clough and Bent , 1998 ) . Transformed seed was initially screened for DsRed fluorescence , followed by growth on MS plates containing 20 µg/ml hygromycin to select for plants carrying the TOC75 variants and toc75-III-1 , respectively . Total RNA was isolated from 14-day-old Arabidopsis seedlings grown on plates using the RNeasy Plant Mini Kit ( Qiagen , Germany ) as per manufacturer’s instructions . RNA isolated was quantified by absorbance at 260 nm , and equivalent amounts were used to synthesize first-strand cDNA using SuperScript III and an oligo ( dT ) primer ( Invitrogen , Carlsbad , CA ) . cDNA was used for PCR reactions ( 35 cycles ) using gene-specific primers . Primers used for PCR amplification of TOC75 , TOC75ΔP1 , TOC75ΔP1-2 or TOC75ΔP1-3 are listed in Table 3 . Primers for RT-qPCR reactions were designed by Primer 3 v . 0 . 4 . 0 ( Untergasser et al . , 2012 ) with optimal melting temperatures of 58–60°C , primer lengths of 20–30 bp and amplicon lengths of 220–250 bp . RT-qPCR was performed in triplicate using DyNAmo Flash SYBR Green qPCR kit ( Thermo Scientific , Waltham , MA ) . Baseline and threshold cycles ( Ct ) were analyzed by Realplex 2 . 2 Software . Relative gene expressions were calculated with respect to internal reference of actin using the 2∆Ct method ( Schmittgen and Livak , 2008 ) . Intact chloroplasts were isolated from 14-day-old plants as described previously ( Schulz et al . , 2004 , Wang et al . , 2008 , Brock et al . , 1993 ) . Chloroplasts were counted with a hemocytometer using a phase-contrast microscope with 40x objective ( Sung and Chen , 1989 ) and equivalent numbers of chloroplasts were used in all assays . Chloroplasts were fractionated into soluble and membrane fractions by centrifugation at 18 , 000 x g at 4°C for 30 min . Proteins from the soluble fraction were precipitated with 15% trichloroacetic acid . Both fractions were resuspended in the sample buffer and analyzed by SDS-PAGE followed by immunoblotting with anti-atToc75 . Thermolysin and trypsin treatments of intact chloroplasts were performed as described earlier ( Pain and Blobel , 1987 , Chen and Schnell , 1997 , Inoue et al . , 2013 , Hsu et al . , 2012 ) . All protease treatments were performed in HS buffer ( 50 mM Hepes-KOH , 330 mM sorbitol , pH 7 . 5 ) for 30 min on ice with increasing protease concentrations up to 200 µg/ml trypsin and 100 µg/ml thermolysin in a reaction containing 5 x 108 chloroplasts/ml . Thermolysin was quenched with 20 mM EDTA and trypsin digestion was quenched by addition of protease inhibitor cocktail ( 2 . 5 mM PMSF , 0 . 05 mg/ml Nα-Tosyl-L-lysine chloromethyl ketone ( TLCK ) , 0 . 25 mg/ml soybean trypsin inhibitor and 2 μg/ml aprotinin ) . Intact chloroplasts were reisolated through 35% Percoll and washed with ice-cold HS buffer containing EDTA or trypsin inhibitor cocktail . 1% ( v/v ) Triton X-100 was included in reactions to disrupt membrane permeability . Total protein was extracted directly in SDS-PAGE sample buffer from Arabidopsis seedlings . Samples corresponding to equivalent amounts of total protein were resolved by SDS-PAGE , transferred to PVDF membranes , and subjected to immunoblotting with antisera to the indicated proteins . Immunoblotting was performed as described previously ( Ma et al . , 1996 ) using chemiluminescence detection . Antisera to atToc159 , atToc132 , atToc33 , atToc34 , atTic110 , LHCP , SSU and OEP80 were described previously ( Ivanova et al . , 2004 , Inaba et al . , 2005 , Wang et al . , 2008 , Hsu et al . , 2012 ) . atToc75 antiserum was a generous gift of Dr . Takehito Inaba at Miyazaki University . Quantitation of immunoblots was carried out using ImageJ ( v . 1 . 47 ) . Immunoaffinity chromatography of TOC core complex proteins under native conditions was performed as described previously ( Kouranov et al . , 1998 ) . The proteins in each fraction were precipitated with 10% trichloroacetic acid . Total membranes , unbound and eluate fractions were analyzed by SDS-PAGE and immunoblotting with antiserum of atToc159 , atTic110 , atToc75 and affinity purified atToc33 antibodies . [35S]Methionine-labeled Arabidopsis preSSU , preE1α , fully processed ( mature ) forms of Toc75ΔP1 and Toc75ΔP1-2 were generated in a coupled transcription-translation system containing reticulocyte lysate according to the manufacturer’s instructions ( Promega , Madison , WI ) . The in vitro translation product containing [35S]preSSU or [35S]preE1α was used directly for chloroplast import assays . Chloroplast early binding and import reactions were performed using [35S]methione-labeled preproteins and equal number of chloroplasts ( 107 chloroplasts ) , in a total volume of 50 μl of import buffer ( 330 mM sorbitol , 50 mM Hepes-KOH , pH 7 . 5 , 25 mM KOAc and 5 mM MgOAc ) for 5 min with 0 . 1mM ATP and GTP or 20 min at 26°C in the presence of 3 mM ATP for binding and import , respectively as described previously ( Chen et al . , 2002 , Wang et al . , 2008 ) . For preprotein binding reactions or early import intermediate formation , the in vitro translated [35S]preSSU was gel filtered on Sephadex G-25 to remove nucleotides ( Agne et al . , 2009 ) . For the chase experiments , chloroplasts ( 107 ) containing bound preprotein or the early import intermediate reactions were generated as above . Chloroplasts recovered after isolation over the 35% Percoll cushion were washed once and resuspended in import buffer . Preprotein translocation was initiated with the addition of 3 mM ATP , and samples were incubated at 26°C for 20 min . All samples were resolved by SDS-PAGE and analyzed by phosphorimaging ( Fuji Fla-5000 phosphorimager ) . Equivalent numbers of chloroplasts based on microscopic counting were loaded in all lanes . ImageQuant TL ( v 1 . 00 ) software was used for analysis . For saturation-binding experiments , preSSU-FLAG-HIS was expressed from pET21a:atpSC ( -1 ) -3xFLAG-HIS in E . coli BL21 ( DE3 ) and purified using Ni-NTA matrix ( Novagen ) under denaturing conditions as described previously ( Smith et al . , 2004 , Inoue et al . , 2013 ) . Import and binding were performed as previously described ( Wang et al . , 2008 ) . The preSSU-FLAG-HIS was denatured with 6 M urea and diluted into import buffer , before the addition of 107 chloroplasts to initiate the reaction . The maximum number of binding sites and apparent Kd for early binding were calculated by plotting specific binding vs . [S] , where specific binding was quantitated from immunoblots using ImageJ and [S] equaled the total concentration of preSSU-FLAG-HIS , using non-linear regression analysis of binding data using Graphpad Prism software version 4 . 00 ( San Diego , CA ) . 2D BN-PAGE was performed as described earlier ( Kikuchi et al . , 2006 , Hsu et al . , 2012 ) with a few modifications . Briefly , intact chloroplasts isolated from A . thaliana were resuspended in the hypotonic lysis buffer ( 10 mM HEPES-KOH , pH 7 . 5 , containing 1 mM MgCl2 and 10 μl/ml PIC [Protease Inhibitor Cocktail for plant extracts; P-9599 , Sigma-Aldrich , St . Louis , MO] ) for 10 min on ice , and centrifuged at 18 , 000 x g , 4°C for 30 min . The pellet was solubilized in the buffer ( 50 mM Bis ( 2-hydroxyethyl ) iminotris ( hydroxymethyl ) methane ( BisTris ) -HCl , pH 7 . 0 , containing 1% ( w/v ) decylmaltoside ( Calbiochem , San Diego , CA ) , 500 mM 6-amino-n-caproic acid , 10% ( v/v ) glycerol , and 10 μl/ml PIC ) at a concentration of 0 . 5 mg chlorophyll/ml for 20 min on ice . The insoluble materials were removed by centrifugation at 17 , 000 g for 20 min at 4°C . The supernatant was mixed with Coomassie Brilliant Blue G-250 solution ( 50 mM BisTris-HCl , pH 7 . 0 , containing 5% ( w/v ) Serva blue G and 500 mM 6-amino-n-caproic acid] in 100:3 . 25 ( v/v ) to give a ratio of the detergent ( decylmaltoside ) to the dye ( Serva blue G ) as 8:1 ( w/w ) . The sample was loaded onto a 4–12% ( w/v ) polyacrylamide gradient gel in 50 mM BisTris-HCl , pH 7 . 0 , containing 500 mM 6-amino-n-caproic acid ( 1 . 5 mm thickness , Mini-PROTEAN3 , Bio-Rad laboratories , Hercules , CA ) . Electrophoresis was carried out in the cathode buffer ( 50 mM tricine , 15 mM BisTris-KOH , pH 7 ) and the anode buffer ( 50 mM BisTris-HCl , pH 7 ) at a constant voltage of 30 V at 4°C for 14 hr . Spectra multicolor Broad Range Protein Ladder ( Thermo Scientific ) and ferritin ( Sigma , F-4503 ) were used as size standards . After BN-PAGE , the marker lanes were directly stained with Coomassie Brilliant Blue , and the sample lanes were excised and heated at 37°C for 30 min in 3 . 3% SDS , 4% 2-mercaptoethanol and 65 mM Tris-HCl , pH 6 . 8 . The gel strip was layered on top of a 1 . 5 mm-thick stacking gel , and the second dimension 5–12% gradient SDS-PAGE under reducing conditions was performed according to standard procedures . A glutathione S-transferase-POTRA fusion ( GST-POTRA-His ) was expressed in E . coli , purified by Glutathione Sepharose 4B chromatography and cleaved using TEV protease to generate soluble POTRA-His . Solid phase binding assays were performed as described previously ( Smith et al . , 2004 ) with some modifications . Soluble POTRA-His was bound to ~10 μl of packed Ni-NTA resin and washed thrice with 50 mM Hepes-KOH , pH 7 . 5 , 2 mM MgCl2 , and 40 mM KOAc ( HMK buffer ) with 10 mM imidazole and 0 . 1% Triton X-100 ( binding buffer ) . 1–3 μl of [35S]Tic22-III , [35S]Tic22-IV , [35S]preSSU and [35S]SSU were incubated in binding buffer with POTRA-bound resin in a final volume of 100 μl for 30 min at room temperature ( 23°C ) . After washing , resin-bound proteins were eluted with SDS-PAGE sample buffer containing 500 mM imidazole . All radiolabeled proteins from in vitro pull-down assays were resolved using SDS-PAGE , and radioactivity was analyzed by phosphorimaging ( Fuji Fla-5000 phosphorimager ) and quantitated using ImageQuant TL ( v 1 . 00 ) . Sequence data from this article can be found in the EMBL/GenBank data libraries under accession number Arabidopsis AGI locus identifier: AT3G46740 for Toc75-III . | Chloroplasts are a hallmark feature of plant cells and the sites of photosynthesis – the process in which plants harness the energy in sunlight for their own needs . The first chloroplasts arose when a photosynthetic bacterium was engulfed by another host cell , and most of the original bacterial genes have been transferred to the host cell’s nucleus during the evolution of land plants . As a result , modern chloroplasts need to import the thousands of proteins encoded by these genes from the rest of the cell . The chloroplast protein import system relies on a protein transporter in the chloroplast membrane that evolved from a family of bacterial transporters . However , the bacterial transporters were initially involved in protein export , and it was not known how the activity of these transporters adapted to move proteins in the opposite direction . Paila et al . set out to better understand the chloroplast protein import system and produced mutated forms of the transporter in the model plant Arabidopsis thaliana . These experiments revealed that a part of the transporter that is conserved in many other organisms , the “protein transport associated domains” , has been adapted for three key roles in protein import . First , this part of the transporter interacts with the other components of the import system that make the transporter more selective and control which direction the proteins are transported . Second , the domains interact with proteins during transport to help move them across the chloroplast membrane . Finally , the domains recruit other molecules called chaperones , which stop the protein from aggregating or misfolding during the transport process . These activities are similar to those for the bacterial export transporters , but clearly evolved to allow transport in the opposite direction – that is , to import proteins into chloroplasts . The next challenges are to explain how proteins destined for chloroplasts are recognized and transported through the chloroplast’s membrane . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"cell",
"biology"
] | 2016 | Multi-functional roles for the polypeptide transport associated domains of Toc75 in chloroplast protein import |
Shigella flexneri is the most common cause of bacterial dysentery in low-income countries . Despite this , S . flexneri remains largely unexplored from a genomic standpoint and is still described using a vocabulary based on serotyping reactions developed over half-a-century ago . Here we combine whole genome sequencing with geographical and temporal data to examine the natural history of the species . Our analysis subdivides S . flexneri into seven phylogenetic groups ( PGs ) ; each containing two-or-more serotypes and characterised by distinct virulence gene complement and geographic range . Within the S . flexneri PGs we identify geographically restricted sub-lineages that appear to have persistently colonised regions for many decades to over 100 years . Although we found abundant evidence of antimicrobial resistance ( AMR ) determinant acquisition , our dataset shows no evidence of subsequent intercontinental spread of antimicrobial resistant strains . The pattern of colonisation and AMR gene acquisition suggest that S . flexneri has a distinct life-cycle involving local persistence .
Once a major cause of mortality and morbidity in Europe and the US prior to the widespread provision of reliable sanitation systems and clean drinking water , bacterial dysentery caused by Shigella spp . , remains a significant infection in low-income countries ( Kotloff et al . , 2013 ) . While the Shigellae are , phylogenetically , Escherichia coli , they were originally classified as separate species based upon shared disease and biochemical phenotypes that marked them out as distinct from other E . coli strains—a distinction that is still reflected in their species nomenclature because of continued global medical importance . The ‘genus’ Shigella consists of four species ( Shigella flexneri , Shigella sonnei , Shigella boydii and Shigella dysenteriae ) causes approximately 165 million new infections globally per year ( Kotloff et al . , 1999; Ram et al . , 2008 ) , which have previously been estimated to result in up to 1 million deaths annually ( Kotloff et al . , 1999 , 2013 ) . The vast majority of cases and fatalities occur in low to middle-income countries in children under the age of 5 years ( Kotloff et al . , 1999; von Seidlein et al . , 2006 ) . The preponderance of these cases are attributable to endemic disease caused by the species S . flexneri ( Vinh et al . , 2009; Ud-Din et al . , 2013; Livio et al . , 2014 ) . Despite the importance of S . flexneri as an etiological agent of diarrheal disease globally , little is known about its detailed population structure . This knowledge gap is a substantial limitation as the ability to accurately track bacterial pathogens is a cornerstone of effective surveillance and downstream public health interventions . This poor understanding of S . flexneri is partly a result of a lack of high-resolution tools for subtyping this species . Traditionally S . flexneri strains are subdivided based on the antigenic variation of the O-antigen component of the bacterial lipopolysaccharide ( LPS ) using typing antisera and the slide agglutination method ( serotyping ) . Serotyping currently subdivides isolates into serotypes or subserotypes by the use of type-specific and group factors antisera ( Edwards and Ewing , 1986; Sun et al . , 2011 ) . Although serotyping still forms the central vocabulary for describing this species it is now widely agreed that many genes that determine serotype are encoded on horizontally transmissible genetic elements ( Allison and Verma , 2000 ) , thus facilitating serotype switching . Therefore the extent to which core characteristics of the species that are currently inferred based on serotype relate to the phylogenetic relationships of strains remains open to question . Through limited genetic and genomic studies we know that Shigella spp . represent distinct clades that fall within the E . coli species complex ( Pupo et al . , 2000; Yang et al . , 2005 , 2007; Ashton et al . , 2014; Sahl et al . , 2015 ) . These studies show the existence of two distinct S . flexneri lineages , with one lineage including only a single serotype ( S . flexneri 6 ) ( Choi et al . , 2007 ) that clusters within species S . boydii ( Yang et al . , 2007 ) and possesses a different LPS O-antigen . The second monophyletic S . flexneri group contains representatives of all other S . flexneri serotypes ( 1–5 , X , Y ) and is responsible for the majority of S . flexneri disease; it is this main lineage that is examined here . Whole genome sequencing has been used successfully to uncover key aspects of the provenance as well as global and regional epidemiology of other Shigella species . Work on the now rarely isolated ( von Seidlein et al . , 2006 ) , S . dysenteriae type 1 has shown it to be an epidemic pathogen characterised by sporadic , large scale outbreaks that spread rapidly and widely via a series of intercontinental transmissions , often associated with war or famine ( Rohmer et al . , 2014 ) . Studies examining the dominant endemic Shigella species of industrialised and newly industrialised countries , S . sonnei , have also revealed a species that is highly clonal ( Karaolis et al . , 1994; Holt et al . , 2012 ) . S . sonnei evolved in Europe ∼300 years ago and has recently moved out of this region via intercontinental spread as a single , rapidly evolving lineage , establishing new , local , populations in countries as they industrialise ( Holt et al . , 2012 , 2013 ) . The contrast between the distribution of the two endemic Shigella species , S . sonnei and S . flexneri , has posed the question why is it that two different species predominate under very different socio-economic conditions ( Kotloff et al . , 1999 ) ? S . flexneri remains the dominant species in low-income countries ( multiple studies report over 50% of all cases of shigellosis [Kotloff et al . , 1999 , 2013; Livio et al . , 2014] ) , whereas S . sonnei is the most commonly isolated Shigella species in industrialised countries ( 77% of cases as reported by Kotloff et al . , 1999 ) where there is better sanitation and access to clean food and drinking water . The reasons for these differences are poorly understood , as are specific transmission patterns that may account for the variation in the dominant Shigella species in these differing settings . To further our understanding of the evolution , population structure and phylogeography of S . flexneri we gathered a representative global collection of 351 isolates of S . flexneri , spanning serotypes 1–5 , X , Xv and Y , collected from the contemporary principal foci of endemic disease; Africa , Asia and South and Central America as well as historical isolates from reference collections dating back to 1914 ( Supplementary file 1 ) . We performed whole genome sequencing on the sample set to provide a basis for exploring the relationship of strain , serotype and geography . Our analysis reveals that S . flexneri is composed of phylogenetically distinct lineages , with each lineage holding similar levels of diversity to the entire S . sonnei species . We observe that the natural history of S . flexneri is characterised by long term ( in some cases over 100 years ) colonisation of individual countries demonstrating that it is far older , and far more diverse than S . sonnei . From this , our analysis uncovers further key differences between the population structure of these two species; providing new clues as to the reasons for the persistence , and most recently decline of S . flexneri .
To determine a detailed phylogeny of this species we mapped the sequence reads from the 351 S . flexneri strains to the concatenated reference genome of S . flexneri strain 301 , including its virulence plasmid ( VP ) , to detect single nucleotide polymorphisms ( SNPs ) . SNPs falling in transposases , IS elements , repeats or regions identified as being recombinant ( Figure 1—figure supplement 1 ) and unlikely to reflect the underlying phylogeny of the bacterium , were excluded from this analysis . We identified 67 , 981 SNPs among the 352 genomes ( including reference genome ) , with 63 , 186 of these on the chromosome and 4795 on the VP . Following the removal of recombinant regions ( almost entirely concentrated in phage , IS elements and S . flexneri Pathogenic Islands , Figure 1—figure supplement 1 ) , 55 , 662 SNPs remained in total , with 53 , 078 on the chromosome and 2584 on the VP . From these SNPs we inferred a maximum likelihood phylogeny showing that the population of S . flexneri is composed of several pylogenetically distinct lineages ( Figure 1 , Figure 1—figure supplement 2 ) . To unambiguously subdivide the species based on shared patterns of sequence variation , we used the software package Bayesian analysis of population structure ( BAPS ) ( Cheng et al . , 2011 , 2013 ) to identify robust phylogenetic groups ( PGs ) within S . flexneri . This subdivided the population into seven PGs , concordant with the phylogeny ( Figure 1 ) . Performing the BAPS analysis on alignments of the chromosome and VP collectively or individually resulted in the same pattern of clustering , reflecting that the VP phylogeny precisely mirrors that of the chromosome , indicating co-evolution ( Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 07335 . 003Figure 1 . Maximum likelihood phylogeny for Shigella flexneri isolates including serotypes 1–5 , X and Y produced from the results of mapping sequence reads against the genome of S . flexneri 2a strain 301 , with recombination removed . Phylogenetic groups ( PGs ) determined by Bayesian analysis of population structure clustering are boxed within dotted lines , with the geographic and serotype composition of isolates in each PG being inlaid as pie charts . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00310 . 7554/eLife . 07335 . 004Figure 1—figure supplement 1 . Location of segments detected as recombinant . Blue indicates a likely recombination within an individual isolate while red indicated recombination common to multiple isolates . Green text at the top indicates mobile elements determined by a manual examination of the reference S . flexneri strain 301 genome . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00410 . 7554/eLife . 07335 . 005Figure 1—figure supplement 2 . S . flexneri species tree , with the number of single nucleotide polymorphisms ( SNPs ) per branch . The SNP tree uses the same alignment as in Figure 1 , but is constructed from the SNPs that can be assigned to each branch . The ancestral states were reconstructed using ACTRAN . Insert—a table showing the number of SNPs between the most recent common ancestor ( MRCA ) of each of the PGs identified . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00510 . 7554/eLife . 07335 . 006Figure 1—figure supplement 3 . Co-evolutionary relationships of the S . flexneri genome and virulence plasmid ( VP ) . A maximum likelihood phylogeny of the S . flexneri chromosome ( left ) is shown adjacent to one of the VP ( right ) . Collared blocks and labels enclose independently identified BAPs clusters for sequence alignments of the chromosome and VP . Dotted lines indicate groups of isolates shared between clusters in phylogeny . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00610 . 7554/eLife . 07335 . 007Figure 1—figure supplement 4 . Maximum Clade Credibility trees generated using Bayesian evolutionary analysis by sampling trees ( BEAST ) for PG 1 . Dates of MRCA are shown overlying internal nodes followed by 95% HPD in parentheses . Tips display the country of origin for each isolate ( where available ) , coloured by region while the date given in red at the base of each group is the MRCA date obtained from the software Path-O-Gen , calculated based on the root-to-tip distance . The horizontal scale is in the unit of years in the past . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00710 . 7554/eLife . 07335 . 008Figure 1—figure supplement 5 . Maximum Clade Credibility trees generated using BEAST for PG 2 . Dates of MRCA are shown overlying internal nodes followed by 95% HPD in parentheses . Tips display the country of origin for each isolate ( where available ) , coloured by region while the date given in red at the base of each group is the MRCA date obtained from the software Path-O-Gen , calculated based on the root-to-tip distance . The horizontal scale is in the unit of years in the past . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00810 . 7554/eLife . 07335 . 009Figure 1—figure supplement 6 . Maximum Clade Credibility trees generated using BEAST for PG 3 . Dates of MRCA are shown overlying internal nodes followed by 95% HPD in parentheses . Tips display the country of origin for each isolate ( where available ) , coloured by region while the date given in red at the base of each group is the MRCA date obtained from the software Path-O-Gen , calculated based on the root-to-tip distance . The horizontal scale is in the unit of years in the past . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 00910 . 7554/eLife . 07335 . 010Figure 1—figure supplement 7 . Maximum Clade Credibility trees generated using BEAST for PG 5 . Dates of MRCA are shown overlying internal nodes followed by 95% HPD in parentheses . Tips display the country of origin for each isolate ( where available ) , coloured by region while the date given in red at the base of each group is the MRCA date obtained from the software Path-O-Gen , calculated based on the root-to-tip distance . The horizontal scale is in the unit of years in the past . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 010 Genetic clustering of the population and the phylogenetic tree revealed a species that is composed of seven discrete PGs , separated by considerable evolutionary distance ( ranging from 1138 to 8430 SNPs between PGs—Figure 1—figure supplement 2 ) , this analysis provided only limited information regarding the natural history of the organism . Most of the PGs contained organisms from several geographic regions , and significantly all PGs contained samples collected over a period of at least 60 years . To combine the geographic and temporal metadata with the genomic information , we used Bayesian evolutionary analysis by sampling trees ( BEAST ) ( Drummond and Rambaut , 2007 ) to reconstruct the temporal and geographical history of the major PGs . Inferring evolutionary rates and estimating the dates of the most recent common ancestor ( MRCA ) of individual PGs , we found that rates of mutation were similar among PGs ( between 6 . 46 × 10−7 substitutions per site per year [PG1] to 9 . 54 × 10−7 substitutions per site per year [PG2] ) . These mutation rates are consistent with previous estimates for S . flexneri ( 3 . 2 × 10−7 substitutions per site per year [Zhang et al . , 2014] ) , S . sonnei ( 6 . 0 × 10−7 substitutions per site per year [Holt et al . , 2012] ) and S . dysenteriae ( 6 . 52 × 10−7 substitutions per site per year [Rohmer et al . , 2014] ) . Using these data we predicted the MRCA of each PG , finding that the median age of the MRCA for all PGs was between 150 and 900 years ago ( Figure 1 , Figure 1—figure supplements 4–7 ) . Of the PGs , PG1 , 2 , 4 and 6 are the oldest lineages with median BEAST estimate of MRCAs of these groups dating to between 1341 and 1659 . Contrastingly , PG5 and PG3 , the latter containing the majority of serotype 2a isolates—one of the key suggested S . flexneri vaccine targets—are much younger , and these PGs have median estimated MRCAs dating to 1822 ( PG5 ) and 1848 ( PG3 ) ; contemporaneous with the MRCA of the emergence of S . sonnei from Europe . However , unlike S . sonnei , the emergence of new S . flexneri PGs did not appear to result in displacement or replacement of isolates from other PGs; rather the older PGs have persisted and continue to cause disease alongside the newer PGs , as evidenced by the fact that every PG contains at least one strain collected since 2008 ( Supplementary file 1 ) . One of the most striking features of the S . sonnei population was the recent intercontinental spread of the organism . Analysing our dataset using the discrete states phylogeographic analysis implemented in BEAST we observed a contrasting natural history of S . flexneri PGs 1–7 with respect to S . sonnei ( Holt et al . , 2012 ) . We observed considerable genomic diversity in S . flexneri isolates collected from the same region . Regions where S . flexneri is endemic simultaneously supported at least two populations of S . flexneri originating from distinct PGs; with the Indian subcontinent having geographically-monophyletic sublineages from four different PGs ( Figure 1 , Figure 1—figure supplements 4–6 , Figure 2 ) . None of these sublineages are recently introduced into these regions; most geographically monophyletic sublineages within the PGs had MRCAs in the range of 25–150 years . As well as observing isolates from different genomic backgrounds coexisting in the same geographic area , our data also show that sublineages from the same PG can also persist contemporaneously in the same geographical region , over an extended period of time ( Figure 1—figure supplements 4 , 6 ) . 10 . 7554/eLife . 07335 . 011Figure 2 . Correlation of isolate phylogeny with pathogenicity and antimicrobial resistance ( AMR ) determinants . The midpoint-rooted maximum likelihood phylogenetic tree shows PGs , with tips and terminal branches collared by continent of origin . Tracks adjacent to each isolate show the percentage BLAST identity of the best hit in the sample assembly against key virulence and AMR determinants . Isolates with mutations in the gyr/par genes have black bars in the relevant tracks . The virulence determinants shown are SHI-1 ( pic , sigA , set1AB ) , SHI-2 ( shiABCDE , iucABCD , iutA ) , sat , enterobactin ( entABECFD , fepABCDG ) , sitABCD , fecEDCBAR , stx1ab , fimZBCHGFDEAY , and the AMR genes are aac ( 3 ) -II , aadA1 , aadA2 , aadA5 , strA , strB , sat1 ( aminioglycosides ) , blaCTX-M-24 , blaOXA-1 , blaTEM-1 , ( β-lactams ) ermB , msrE , mphA , mphE , ( macrolides ) catA1 , catB1 , ( phenicols ) qacEΔ1 , qnrS1 , ( quinolones ) qepA , sul1 , sul2 ( sulphonamides ) tetA ( A ) , tetA ( D ) , tetA ( B ) ( tetracyclin ) , dfrA17 , dfrA3b , dfrA1 , dfrA5 , dfrA14 and dfrA8 ( trimethoprim ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 01110 . 7554/eLife . 07335 . 012Figure 2—figure supplement 1 . Results of molecular serotyping , displaying the distribution of MLST , molecular serotype , and the distribution of defining genes ( according to key , top left ) among isolates . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 01210 . 7554/eLife . 07335 . 013Figure 2—figure supplement 2 . Maximum likelihood phylogeny of an alignment of the concatenated nucleotide sequences the enterobactin locus of 13 genes ( 34 , 732 NT; containing entABCDEFS , fepABCDG and fes ) . The tree is drawn using PhyML , with a GTR model and contains 191 strains . Isolate labels are collared according to whole-genome based PG definition . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 01310 . 7554/eLife . 07335 . 014Figure 2—figure supplement 3 . Correlation of isolate phylogeny with AMR determinants , showing only the SRL-MDRE-associated loci aadA1 , blaOXA-1 , cat and tetA ( B ) . Grey circles indicate branches where acquisition events are predicted to have taken place . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 014 Despite this structure of long-term phylogeographic association on a population wide level , we also observe that some PGs were highly geographically restricted and appear to have spread to only a limited number of regions . For example , PG2 has a MRCA dated to 1550 ( 95% CI: 1340–1725 ) , with Asian isolates appearing in both sublineages of this cluster ( Figure 1—figure supplement 5 ) . The MRCA for the Asian isolates was estimated to be 1544; however the African isolates form a single cluster that has a more recent MRCA—dated to 1885 ( 95% CI: 1833–1923 ) ; a result that implies that PG2 S . flexneri was likely introduced into Africa from Asia at least 130 years ago . However , it is also important to note that while there is always a possibility of sample bias; within our dataset there is a clear signal that the African and Asian lineages are distinct , and have been for some time . Other samples may change this view , however , the sampling strategy we used was to collect samples globally—capturing a cross section of current disease with a historical view of the population of S . flexneri—without a specific focus on a single serotype grouping or geographic region . The fact that clear clusters within the population associated with geographic origin do appear would not be expected by chance . Perhaps surprisingly for a gastrointestinal pathogen of the E . coli complex , at least on a population level , we found very limited evidence of large-scale species-wide recombination within the genome ( Figure 1—figure supplement 1 ) , with S . flexneri displaying a recombination profile more in line with the lifestyle-restricted ST131 than with members of other gastrointestinal pathovars such as ETEC ( McNally et al . , 2013 ) . Within the virulence repertoire , where recombination could be identified , it was generally present among all isolates within a single PG ( s ) , indicating that the recombination was likely to be ancestral occurring prior to PG divergence . Furthermore , most evident recombination is restricted to previously identified mobile genetic elements ( Figure 1—figure supplement 1 ) . By examining known chromosomally encoded virulence determinants , it is clear that many known virulence functions are ubiquitous to all PGs ( Figure 2 ) . The distribution of virulence determinants indicated that many of the genes/islands whose loss is known to attenuate S . flexneri virulence ( such as the VP [Sansonetti et al . , 1982] and aerobactin [Lawlor et al . , 1987; Wyckoff et al . , 2009] ) were acquired early in its evolutionary history and have been subsequently retained . However , some variation in the composition of virulence determinants is also evident . For example , Sat , a Serine Protease Autotransporter , is found in PGs 2 , 4 , and 7 , but absent from PGs 1 , 3 , 5 or 6 ( Figure 2 ) . Although the latter PGs possess a genetic scar suggesting that Sat was present ancestrally and has been subsequently lost . We also find that the SHI-1 pathogenicity island—a distinguishing feature of serotype 2a ( the dominant serotype causing disease world-wide [Kotloff et al . , 1999] ) —was acquired once ( between 1848 and 1882 ) near to the root of PG3 . SHI-1 carries genes encoding three key virulence associated proteins—Pic , ( Henderson et al . , 1999 ) SigA ( Al-Hasani et al . , 2000 ) , and the S . flexneri enterotoxin ShET1 ( Fasano et al . , 1997 ) . Pic , which is also found in Enteroaggregative E . coli , encodes a secreted protease that has been shown to have mucinase activity and has been suggested to be involved in haemagglutination and serum resistance ( Henderson et al . , 1999 ) . SigA is thought to be involved in fluid accumulation ( Al-Hasani et al . , 2000 ) , while the S . flexneri enterotoxin ShET1 has also been shown to be involved in fluid accumulation in rabbit ileal loop models ( Fasano et al . , 1997 ) , and has been proposed as being the causative agent of the voluminous watery diarrhoea characteristic of a S . flexneri serotype 2a infection ( Noriega et al . , 1995; Fasano et al . , 1997 ) . Although SHI-1 is only present in PG3 ( 135 of 146 PG3 isolates included carried this island ) , it is present in the range of serotypes found within this phylogroup—including multiple isolates of serotypes 1a , 2a , 2b and 5a . This demonstrates that the toxin is not peculiar to strains carrying a particular serotype , but rather to strains within a PG that are predominantly serotype 2a; consistent with reports of a small number of non-2a serotypes reported to encode this toxin ( Noriega et al . , 1995 ) . Our genomic data further confirm that these cases were not simply mis-serotyped , but likely represent cases of sporadic serotype switching amongst isolates within PG3 . More generally , we observe evidence for recombination amongst the genes responsible for serotype ( see gtr Figure 1—figure supplement 1 ) , which remains the de-facto method for describing strains of Shigella . When cross comparing the classical and in silico molecular serotyping schemes ( Sun et al . , 2011 ) we saw no significant difference in the serotypes predicted ( assessed using a Kolmogorov–Smirnov test ) . However , our results do demonstrate that while certain serotypes make up the majority of certain clusters ( Figure 1 , Figure 2—figure supplement 1 ) , all of the PGs contained multiple serotypes . This indicates that although the core genome is remarkably stable , serotype switching does occur undermining the principle that serotype can be used either as a basis for describing the properties of a lineage or for epidemiological surveillance and tracking . These results also imply that vaccines targeting a particular backbone using a serotype ( such as a 2a vaccine to target SHI-1 encoding isolates from PG3 ) could be susceptible to the effects of serotype switching . In addition to our observations around SHI-1 and the VP we also observed some variation in virulence factor complement within PGs principally for Iron uptake functions; the sit ( Runyen-Janecky et al . , 2003 ) and fec loci ( Luck et al . , 2001 ) and enterobactin biosynthesis operon . Across our S . flexneri tree it is apparent that these have contrasting patterns of gain and loss; sit is ancestral , but was sporadically lost ( unexpectedly , as it has been reported to be in all clinical isolates previously [Runyen-Janecky et al . , 2003] ) , while fec has been gained sporadically by isolates within PG3 , as well as being gained ( and retained ) on separate occasions by isolates in PG1 , 2 and 6 . Of additional interest is the enterobactin operon; which encodes a high affinity iron siderophore ( O'Brien and Gibson , 1970 ) . It has previously been reported that in S . flexneri the enterobactin genes are ‘rarely utilized’ ( Schmitt and Payne , 1988; Wyckoff et al . , 2009; Reuter et al . , 2014 ) ; however , our data show that the enterobactin genes are present across PGs 3 , 5 , 7 and 4 , as well as a subset of related isolates within of PG1 . Examining the enterobactin gene specifically , we observed the same phylogenetic topology as that observed for the whole genome tree ( Figure 2 , Figure 2—figure supplement 2 ) ; suggesting this element is ancestral , and has been lost by a subset of PGs . The ubiquitous nature of these enterobactin genes suggests that these genes are more significant than ‘rarely utilized’; as they have been retained by multiple PGs , over a long timescale . In addition to these PG-wide patterns , within PG3 we observed six tightly clustered isolates that also carry a Shiga toxin 1a ( Stx1a ) -encoding phage ( Figure 2 ) which is identical to the recently reported φPOC-J13 ( accession number KJ603229 ) ( Gray et al . , 2014 ) . These isolates were collected between 2003 and 2008 from Latin America ( Haiti , French Guiana and the Dominican Republic ) . This is consistent with these findings charting the emergence of Stx1a-producing S . flexneri in the region ( Gray et al . , 2014 ) . Given that the reference phage clusters phylogenetically within our sequenced isolates it is likely that the phage was acquired once in this sublineage of PG3 . The dating analysis on the Stx-1-containing PG3 sublineage showed that the phage could have been acquired no later than 1998 ( Figure 1—figure supplement 6 ) suggesting that strains carrying this phage may have been present but undetected in the population for 5–10 years before the earliest reported isolation of Stx1a-producing S . flexneri . Antimicrobial resistance ( AMR ) has been shown to be a strong influence on the recent evolutionary history of many bacterial pathogens ( Mutreja et al . , 2011; Holt et al . , 2012; Okoro et al . , 2012; He et al . , 2013; Mather et al . , 2013 ) . Screening for the presence of known AMR genes ( Figure 2 ) we found that AMR gene distribution was highly variable both within and across PGs ( Figure 2 ) . Unlike S . sonnei , we observe little evidence that the acquisition of AMR-related loci is linked to the establishment or replacement of dominant lineages in any geographic location . While published evidence does exist of one major , on-going , clonal outbreak of S . flexneri in high income countries that has been shown to be the result of a combination of specific epidemiological , AMR and behavioural factors within the MSM community ( Baker et al . , 2015 ) , examining our dataset focused around historical isolates and those with an origin in low-income countries , we observe limited historical evidence of global expansion as a result of AMR . In contrast we observe the widespread acquisition of AMR determinants independently across the tree . The most widespread AMR loci are found on the multidrug resistance element ( MDRE ) , part of the Shigella resistance locus-pathogenicity island ( SRL-PAI ) with variants found in representatives from all PGs . Our analysis suggests that the SRL-PAI has been introduced independently on at least nine occasions ( Figure 2—figure supplement 3 ) . In six of these it has been maintained in subsequent lineages for protracted periods of time . There is , however , no evidence of isolates carrying SRL-PAI spreading outside of their geographic region , and , once acquired , in a number of cases it appears that the SRL-MDRE may have been subsequently lost ( Figure 2—figure supplement 3 ) . Beyond the SRL-PAI we observe 23 internal branches where non SRL-PAI resistance determinants have been acquired ( Figure 2 ) , depicting a distinctive wave-like pattern of resistance determinant acquisition over time ( Figure 3 ) . The earliest resistance genes observed in the population code for sulphonamide resistance ( 1950s ) , followed by streptomycin and tetracycline ( 1960s ) resistance , with β lactamases arising first in strains isolated in the 1970s ( Figure 3A ) . Notably , we observe that in all cases within S . flexneri the strains acquiring AMR determinants remain geographically restricted , although on a population level we also observe an upward trend in AMR determinant possession over time , in all PGs ( Figure 3B ) . We also observe that whilst AMR is widespread in the 21st century , we can isolate S . flexneri strains that do not harbour large numbers of AMR determinants; this is in marked contrast to observations within the S . sonnei population . 10 . 7554/eLife . 07335 . 015Figure 3 . Graphs showing the pattern of AMR presence within our dataset . ( A ) Graph showing the proportion of isolates from each decade that contain the AMR genes . ( B ) Graph showing the average number of resistance genes found in each isolate collected , by year . DOI: http://dx . doi . org/10 . 7554/eLife . 07335 . 015
Our results present the clearest and most complete overview of the temporal and geographic patterns that underpin the population structure of the species S . flexneri obtained to date . It provides a genetic framework and evolutionary context for studies looking within single countries or looking at individual outbreaks or lineages of interest . These data contrast with that of the other Shigella species for which we have global population data , S . sonnei and S . dysenteriae; with the pattern of long term colonisation , diversity and coexistence in S . flexneri lineages appearing to share more similarities with pathogenic E . coli variants such as ETEC ( von Mentzer et al . , 2014 ) than other Shigella species . It is evident from the presented data that , at a population level , S . flexneri is characterised by long-term colonisation of endemic regions , with limited evidence of intercontinental transmission within the last 30 years . Our data also suggest that endemic countries support a diverse population of S . flexneri , and have done for some time , implying long-term local-transmission/colonisation in those settings . This is in direct contrast to S . sonnei , where historically the population is characterised by recent colonisation of countries and dominant pandemic lineages , with frequent inter-country and inter-continental transmission and repeated strain replacement ( Holt et al . , 2012 ) . It is also evident that while AMR has had significant in the natural history of both species , the effects of AMR acquisition on the population are contrasting . For S . sonnei the acquisition of multiple AMR genes was strongly associated with lineage replacement by dominant resistant lineages , on a global scale . Whereas , for S . flexneri , we observe a population where there have been over 30 independent acquisitions of AMR determinants at a local level; but these acquisitions generally result in limited global spread ( Figure 2 ) . This implies that while AMR may be a significant factor in maintaining existing lineages within specific locales we do not see widespread evidence of AMR acquisition leading to the pandemic displacement of other pre-existing lineages . Moreover this observation also extends to the simultaneous acquisitions of multiple AMR genes en bloc , such as those associated with the SRL-PAI . The acquisition of this locus appeared to have occurred in lineages that , within this dataset , show no subsequent evidence of travel outside of their original geographic area . It may be that the transient nature of AMR gene carriage is the result of rapid gain and loss during periods of differential selection , consistent with S . flexneri persisting in the environment where it is conceivable that selection by AMR could be less significant . Despite this lack of evidence of rapid global spread of AMR , on a population level we did observe both an on-going trend of increasing resistance in S . flexneri , and evidence that this pathogen acquires resistance determinants rapidly , often keeping the AMR determinants for protracted periods . However , we also continue to observe recently collected strains with a low number of resistance determinants—suggesting that strains with a limited resistance repertoire are continuing to persist within the wider population . This observation suggests that although there is evidently some selective advantage to carrying AMR determinants , as evidenced by the increasing level of resistance across the pathogen population , it is not an absolute requirement for modern S . flexneri . The contrast between the observed population structure , natural history and role of AMR in the narrative of S . flexneri and S . sonnei may be related to the lifecycle of these two organisms . It has previously been shown that S . flexneri strains can be recovered from the environment in endemic countries ( Faruque et al . , 2002 ) unlinked to any outbreak , and that under experimental conditions S . flexneri can survive for several months in water at room temperature ( Uyanik et al . , 2008 ) , simulated environmental water ( Hendricks and Morrison , 1967; Islam et al . , 1996 ) and upon foodstuffs ( Islam et al . , 1993 ) . Furthermore , there are multiple examples of S . flexneri outbreaks where the outbreak strain was traced back to , and isolated from local water sources ( Swaddiwudhipong et al . , 1995; He et al . , 2009; Saha et al . , 2009 ) . In addition , at least one study has been undertaken that has demonstrated that within Bangladesh S . flexneri disease risk was persistent throughout the year and associated with environmental factors ( geographic location/provision of flood defences ) ( Emch et al . , 2008 ) . When these studies are combined with both our results and additional observations that the provision of clean water and good sanitation correlates with a reduction in the rates of S . flexneri disease ( Esrey et al . , 1985 ) , we believe that the implication is that in an environment where water is frequently persistently contaminated with faeces , S . flexneri , in effect , persists in the environment , and has been doing so over the long term , in countries around the world . A mechanism of environmental persistence would explain our observations around the longevity of lineages of S . flexneri in endemic countries and provide an obvious transmission mechanism . When considering the temporal and phylogeographic analyses presented here in this context , we think that the evidence supports the concept , suggested previously , that in endemic countries the lifecycle of S . flexneri disease is analogous to that of Vibrio cholerae ( Faruque et al . , 2002 ) ; with part of the pathogens infectious cycle being in contaminated water , from which it periodically emerges to cause disease . While our data suggests that an environmental stage may be important for S . flexneri—whose reduced incidence has previously been correlated with the provision of clean water and improved sanitation ( Esrey et al . , 1985 ) , the inverse is true for S . sonnei . It has been suggested that the O-antigen carried by Plesiomonas shigelloides ( another organism isolated from faecally contaminated water ) —that is immunologically indistinguishable from the S . sonnei O-antigen may help to explain the contrasting global distribution of S . sonnei and S . flexneri infections ( Sack et al . , 1994 ) . Exposure to P . shigelloides , through drinking contaminated ground water has been hypothesized to stimulate low level cross protective immunity against S . sonnei , explaining why it is rarely found in the resource poor settings but its incidence increases with water sanitation . If the P . shigelloides hypothesis is proved to be true , it may be that contaminated water holds the key to both the increase in S . sonnei and the decrease in S . flexneri observed as countries develop . Collectively , our analyses outline a historic pathogen with a stable core genome that comes equipped with a repertoire of virulence determinants that have enabled it to colonise , and persist , in multiple locations for hundreds of years . However , this life-cycle is at odds with current patterns of human development; with temporal analysis revealing limited evidence of intercontinental transmission and local colonisation within the last 30 years . It may be that this change is reflective of the fact that over the course of the 20th Century , many countries have industrialised; with increasing provision of clean water and sanitation possibly disrupting the traditional transmission route of S . flexneri . However , despite its decline in industrialised countries , where transmission of this pathogen by faecally contaminated food and water is well controlled , S . flexneri continues , as it has done for hundreds of years , to be a source of disease infecting millions in the low-income nations of the world .
Contemporaneous strains isolated from patients were collected from accredited diagnostic or Public Health laboratories located in Bangladesh ( ICDDRB , n = 114 ) , South Africa ( National Institute for Communicable Diseases , Johannesburg , South Africa , n = 29 ) , and , for Francophone Africa and Latin America , from the French National Reference Centre for E . coli , Shigella , and Salmonella ( Institut Pasteur , n = 54 ) and the Hospital for Tropical Diseases , Ho Chi Minh City , Vietnam ( n = 34 ) . A historical selection of isolates was obtained from the Institut Pasteur ( n = 41 ) and The International Vaccine Institute and Hanyang University , Korea ( n = 62 ) , to provide a stronger temporal signal for the BEAST analysis . Additionally , we also included the HPA type strains ( n = 16 ) ( Ashton et al . , 2014 ) and NCTC1 ( Baker et al . , 2014 ) . Serotyping was performed on isolates using the standard tests used locally by the submitting lab ( using either commercial or locally produced , polyclonal and monoclonal typing antisera ) , and the classical serotypes obtained were checked , and in most cases confirmed , using a molecular serotyping schema ( Sun et al . , 2011 ) . DNA was extracted by collecting laboratories and sequenced at the Wellcome Trust Sanger Institute to a minimum of 50-fold coverage using an Illumina HiSeq 2000 with multiplexing of 96 samples per flow cell using 100 bp paired end reads . Sequencing data was submitted to the ENA ( Genome accession numbers provided in Supplementary file 1 ) . To perform subsequent analyses around the AMR and virulence gene determinants , sequence data was assembled de novo using Velvet ( Zerbino and Birney , 2008 ) , with assemblies improved using iCORN ( Otto et al . , 2010 ) and the Velvet Columbus module . Phylogenetic and population genetic relationships were determined based on a similar approach to that employed by Holt et al . ( 2012 ) for S . sonnei . Sequence ( FASTQ ) files for each isolate in the study were mapped back against the reference S . flexneri serotype 2a strain 301 concatenated with its complete VP ( Accession numbers: AE005674 , AF386526 ) . A whole genome alignment was produced by SNP-calling isolates , and regions that were recombinant were identified using the software package Gubbins ( Croucher et al . , 2014 ) . The recombinant regions were subsequently manually checked and their content identified . Following removal of recombinant regions , mapping coverage was checked manually and variable sites were extracted . The resulting SNP alignment was used to infer a phylogeny using RAxML version 7 . 4 with a General Time Reversible model and gamma correction ( n = 4 ) for among site rate variation ( Stamatakis , 2006 ) . The alignment produced prior to the removal of recombinant regions ( 66 , 524 bp in length ) was passed to BAPS to identify phylogenetic clusters ( Corander et al . , 2004 , 2008 ) . The individual plasmid alignment was generated by extracting the part of the combined plasmid/chromosome alignment that contained plasmid sequence only . This alignment was then analysed on its own . Following SNP calling , this produced a SNP alignment of 4784 bp that was used to generate an unrooted phylogeny and perform BAPS clustering ( as above ) within the plasmid population . To examine the temporal and phylogeographic distribution of the samples we made use of BEAST , providing taxa subsets of the recombination-free alignment and providing geographic information and sample dates . The parameters were as follows: a Bayesian skyline model for population growth was used , with a log-normally distributed clock rate . For each PG BEAST was run across five chains of 100 , 000 , 000 generations each , sampled every 1000 generations . Convergence was determined by visual inspection of MCMC parameters across the chains . All parameter ESS values were ≥200 . The parameter and tree estimates were combined using the LogCombiner and TreeAnnotator components of the BEAST package , with the first 10% of states in each chain discarded as burn-in , and then produced a Maximum Clade Credibility tree from the combined file , which was visualized with FigTree . The presence of virulence and AMR determinants ( Figure 2 ) was determined by BLAST against reference loci as described previously ( Reuter et al . , 2014 ) . For the AMR genes ( Figure 2 ) the presence of mutations in the DNA gyrase and topoisomerase IV protein sequences were assessed by inspection of de novo assembled sequences . | Dysentery is a disease in which the intestine becomes inflamed due to infection by bacteria , viruses or other microbes . Of the bacteria that can cause dysentery , bacteria called Shigella are most often responsible . Humans can acquire Shigella through contaminated food or water . Over the last century , improvements to sanitation combined with access to clean drinking water and better food hygiene have decreased the number of cases of dysentery in many countries . However , the disease continues to be common in low-income countries , especially in young children . One species of Shigella bacteria , called S . flexneri , causes far more cases of dysentry than other species of Shigella . Across the world , there are many different strains of S . flexneri , but it is not clear how these strains are related to each other , or how variable the genes that they carry are—known as genetic diversity . Here , Connor , Barker , Baker et al . used a technique called whole genome sequencing to map the evolutionary relationships of over 300 S . flexneri samples collected from around the globe over the past 100 years . This revealed that the bacterial strains can be split into seven groups that each have distinct geographic ranges and combinations of genes that enable the bacteria to infect humans . Many of the strains of bacteria within these groups seem to have colonized , and remained in , quite small geographic areas over long periods of time . This is different to other Shigella species , which appear to have spread between continents far more frequently over much shorter timescales . Connor , Barker , Baker et al . 's findings reveal that S . flexneri is more diverse than other Shigella bacteria , and suggest that the ability of strains to persist in local areas may have contributed to the species' long-term success . These results point towards the importance of the provision of clean water in the fight against S . flexneri , and underline the need for a greater understanding of how disease-causing bacteria colonize and interact with the local environment . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health",
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | Species-wide whole genome sequencing reveals historical global spread and recent local persistence in Shigella flexneri |
Targeting the activation function-1 ( AF-1 ) domain located in the N-terminus of the androgen receptor ( AR ) is an attractive therapeutic alternative to the current approaches to inhibit AR action in prostate cancer ( PCa ) . Here we show that the AR AF-1 is bound by the cochaperone Bag-1L . Mutations in the AR interaction domain or loss of Bag-1L abrogate AR signaling and reduce PCa growth . Clinically , Bag-1L protein levels increase with progression to castration-resistant PCa ( CRPC ) and high levels of Bag-1L in primary PCa associate with a reduced clinical benefit from abiraterone when these tumors progress . Intriguingly , residues in Bag-1L important for its interaction with the AR AF-1 are within a potentially druggable pocket , implicating Bag-1L as a potential therapeutic target in PCa .
The androgen receptor ( AR ) plays a central role in the development of prostate cancer ( PCa ) . Current approaches aimed at reducing persistent AR signaling either inhibit the production of androgens , or compete with endogenous ligands for binding to the AR C-terminal ligand-binding domain ( LBD ) ( Helsen et al . , 2014 ) . While these therapies are initially effective in the advanced disease setting , PCa will eventually progress to a lethal , therapy-resistant state termed castration-resistant PCa ( CRPC ) . In most cases of CRPC , AR continues to play a dominant role . Newer androgen synthesis inhibitors such as abiraterone and AR antagonists such as enzalutamide have been developed as second-generation therapies for the treatment of CRPC ( de Bono et al . , 2011; Scher et al . , 2012; Beer et al . , 2014 ) . Nevertheless , while initially effective in the treatment of CRPC , most cases will develop resistance to these therapies ( Mostaghel et al . , 2014 ) . One of the major drawbacks of these drugs is that they target the AR LBD ( either directly or indirectly ) . Although the AR LBD is often regarded as the major site of AR regulation , most of the AR transactivation function is controlled through its N-terminal activation function 1 domain ( AF1 ) , which is subdivided into the tau-1 ( amino acids 100–359 ) and tau-5 regions ( amino acids 360–528 ) ( Claessens et al . , 2008 ) . Furthermore , multiple AR variants expressed in CRPC completely lack a LBD ( Guo et al . , 2009 ) , illustrating the need for alternative modes of AR inhibition . Targeting the AR AF-1 therapeutically has been challenging , due to its intrinsically disordered nature and lack of enzymatic activity or rigid binding clefts ( Lavery and McEwan , 2008 ) . However the lack of secondary or tertiary structure of intrinsically disordered regions ( IDRs ) of proteins , such as those found in the AR AF-1 domain , could be an advantage in providing a large surface area for protein-protein interactions ( Wright and Dyson , 2009 ) . IDRs can fold upon binding to their targets , allowing them to undergo conformational changes and participate in protein complex formations ( Dyson and Wright , 2005 ) . Although these interactions tend to be more transient and of lower affinity than complex formation between structured protein regions ( Latysheva et al . , 2015 ) , they have become exceedingly important for controlling the function of IDR-containing proteins . Proteins that bind the unstructured AR AF-1 domain may constitute regulatory targets for inhibiting AR action . Bag-1 ( Bcl-2-associated athanogene-1 ) is a multifunctional protein involved in a number of key cellular processes including proliferation , differentiation , cell cycle , transcription and apoptosis ( Townsend et al . , 2003a ) . Its main function is as a cochaperone and nucleotide exchange factor for Hsp70/Hsc70 ( Alberti et al . , 2003 ) . Four Bag-1 isoforms ( Bag-1L , −1M , −1 and −1S ) exist in humans and are generated from the same mRNA by a leaky scanning mechanism ( Takayama et al . , 1998; Yang et al . , 1998 ) . Although the Bag-1 family members differ in their N-terminal domains , their C-terminal ( BAG ) domains are conserved and essential for interaction with Hsp70/Hsc70 ( Brehmer et al . , 2001; Sondermann et al . , 2001 ) . Bag-1L , the largest family member , is the only one that possesses a nuclear localization sequence and is therefore localized in the nucleus ( Takayama et al . , 1998 ) where it enhances the transactivation function of several nuclear hormone receptors , including the AR ( Froesch et al . , 1998; Knee et al . , 2001; Shatkina et al . , 2003; Jehle et al . , 2014 ) . The function of Bag-1L on AR action is mediated through the direct interaction between two regions of each protein . We have recently shown that Bag-1L uses a N-terminal duplicated GARRPR motif to bind to a pocket near the AR LBD , termed binding function-3 ( BF-3 ) ( Jehle et al . , 2014 ) . Additionally , Bag-1L binds via its C-terminal BAG domain to the AR N-terminal domain ( NTD ) ( Shatkina et al . , 2003 ) . However , details of this interaction and its consequences are unknown . Here we show that the conserved BAG domain within the C-terminus of Bag-1L selectively interacts with the intrinsically disordered , but partially folded N-terminal AR tau-5 domain . Disrupting this interaction by knocking out or mutating the BAG domain of Bag-1L alters the structural properties of the AR NTD and receptor folding , and reduces the ability of AR to bind to chromatin and regulate transcription .
Bag-1L has been reported to upregulate the activity of AR ( Froesch et al . , 1998; Knee et al . , 2001; Shatkina et al . , 2003 ) . However , previous studies have not been able to clearly separate the function of Bag-1L from that of the other Bag-1 family members ( Krajewska et al . , 2006; Mäki et al . , 2007 ) . To specifically determine the function of Bag-1L , we employed a transcription activator-like effector nuclease ( TALEN ) approach that targets the first codon ( CTG ) of Bag-1L ( Figure 1A ) , resulting in the complete knock-out ( KO ) of this protein . We used this approach in hormone-dependent LNCaP cells where we observed , concomitant with the loss of Bag-1L , an upregulation of the other Bag-1 isoforms ( i . e . Bag-1S and Bag-1; Figure 1B ) ; this is consistent with the translation of the Bag-1 mRNA by a leaky scanning mechanism . No alterations of AR levels were observed in response to Bag-1L KO , even in the presence of dihydrotestosterone ( DHT ) . We also created rescue cell lines , re-expressing either an empty vector construct or ( wild-type ) Bag-1L ( Figure 1C ) , to exclude potential off-target effects . Cell growth , over a 5 day period , was determined for all cell lines ( Figure 1D ) . While loss of Bag-1L significantly reduced PCa growth , re-expression of Bag-1L ( but not that of the empty vector ) could reverse this phenotype , confirming that LNCaP cell growth is Bag-1L dependent . Since LNCaP cell growth is primarily driven by AR activity , we next tested whether Bag-1L has any effect on the AR cistrome or transcriptome . We performed ChIP-seq in hormone-depleted control , Bag-1L KO and the rescue LNCaP cell lines , treated for 4 hr with vehicle ( ETOH ) or 10 nM DHT . We observed only limited binding of AR in response to vehicle treatment , regardless of the presence or absence of Bag-1L ( Figure 1—figure supplement 1 ) . This is in concordance with previously published AR cistromes ( Wang et al . , 2007 ) . As previously reported ( Wang et al . , 2009 ) , AR binding increased in response to DHT in the control cells ( Figure 1E ) . In comparison , increase in AR binding was modest in the Bag-1L KO cells , and peaks were overall weaker than those observed in the control cells ( Figure 1F ) . This implies that Bag-1L is necessary for effective AR binding to chromatin . Concomitantly , Bag-1L re-expression , but not the re-expression of the empty vector construct , could restore the AR binding sites ( Figure 1F ) . The difference in AR binding in the presence and absence of Bag-1L was independent of the duration of DHT treatment , as confirmed by directed qPCR at three AR-bound enhancers ( Figure 1—figure supplement 2 ) . To test if the reduction in AR binding in response to Bag-1L loss has any repercussions on the AR-dependent transcriptome , we next performed RNA-seq in hormone-deprived control and Bag-1L KO cells , treated for 16 hr with ETOH or 10 nM DHT . RNA- and ChIP-seq data were correlated using GenomicRanges ( Bioconductor ) and direct AR target genes were defined as genes with DHT-induced differential expression ( p ( FDR ) ≤0 . 05 , fold change ≥1 . 5 ) that harbor a DHT-responsive AR binding site within 50 kb of their transcription start site ( TSS ) . Using this approach , we determined 599 direct AR target genes in the control and 306 in the Bag-1L KO cells ( Figure 1—source data 1 ) . Using the HALLMARK function from GSEA we established that most of those genes lost in response to Bag-1L KO are associated with ‘androgen response’ ( p=3 . 39×10−31; q = 1 . 7×10−29 ) ( Figure 1—figure supplement 3 ) , supporting our hypothesis that one of the functions of Bag-1L in PCa is to regulate AR transactivation . One of the consequences of the loss of a chaperone or cochaperone is an alteration of the folding properties of its client proteins ( Balchin et al . , 2016; Mayer and Bukau , 2005 ) . The impaired chromatin binding and transactivation function of the AR in cells lacking the cochaperone Bag-1L may therefore arise from an inability of the AR to adopt the appropriate conformation for its function . To test this , we employed fluorescence resonance energy transfer ( FRET ) using AR with its N- and C-termini tagged with CFP and YFP , respectively . This allowed us to measure the intra- and inter-molecular AR N/C-terminal interactions associated with the transactivation of the receptor ( Schaufele et al . , 2005 ) . Experiments were carried out in hormone-depleted control and Bag-1L KO cells treated with vehicle ( ETOH ) or 10 nM DHT for 2 hr , and FRET signals were quantified ( Figure 1G ) . We observed a much attenuated ( and overall lower ) FRET response to DHT in the Bag-1L KO compared to the control cells ( i . e . , mean difference of 0 . 24 ± 0 . 01 in the control versus mean difference of 0 . 09 ± 0 . 01 in the KO cells ) . Taken together , these results suggest that there are significant alterations in the inter-domain interaction and folding of the AR in the absence of Bag-1L that effect the ability of the receptor to efficiently respond to DHT , bind to chromatin and regulate gene expression . We have previously shown that Bag-1L enhances AR transactivation via direct interaction of the two proteins ( Jehle et al . , 2014 ) . We demonstrated that a novel GARRPR motif , found at the N-terminus of Bag-1L , interacts with the BF-3 pocket of the AR LBD . We further noted that the C-terminus of Bag-1L binds the AR NTD ( Shatkina et al . , 2003; Jehle et al . , 2014 ) . To demonstrate that these interactions contribute to AR action , we next performed a series of mammalian one-hybrid assays involving full-length Bag-1L and different domains of AR . Bag-1L enhances the activity of the previously identified AR LBD ( in a hormone-dependent manner ) as well as the AR AF-1 domain ( Figure 2A ) . Moreover , Bag-1L enhances AR activity via the tau-5 domain within the AR AF-1 ( Shatkina et al . , 2003 ) ( Figure 2B ) , a region known for its hormone-independent receptor activity ( Jenster et al . , 1995 ) . We have previously shown that the interaction of AR tau-5 is with the Bag-1L BAG domain , or its flanking regions ( Shatkina et al . , 2003 ) . To specifically delineate which residues , or combinations of residues are involved in this interaction , we employed a SPOT-synthesis technique that allows the screening of a large number of synthetic peptides ( Frank , 2002 ) . We synthesized short overlapping peptides ( 21 amino acids each ) spanning the entire C-terminal domain of Bag-1L onto a cellulose membrane and incubated it with bacterially-purified GST-tagged AR tau-5 . Once specific binding was established by immunoblotting , alanine substitutions were introduced into the synthesis of positively identified spots , until single amino acids were identified in the BAG domain that destroy the interaction with AR tau-5 ( Figure 2—figure supplement 1 ) . A triple mutation of K231/232/279A , referred to as CMut hereafter , was able to significantly decrease the interaction of AR and Bag-1L in a GST pull-down ( Figure 2C ) and co-IP experiment ( Figure 2D ) , and reduce the AR NTD activity in a mammalian one-hybrid assay ( Figure 2E ) . Additionally , Bag-1L CMut significantly altered the AR N/C-interaction compared to wild-type Bag-1L ( Figure 2F ) , similar to what we observed in the FRET experiments in response to DHT treatment in the Bag-1L KO cells ( Figure 1G ) . CMut Bag-1L- compared to wild-type Bag-1L-expressing cells additionally displayed a reduction in AR chromatin binding by ChIP-seq ( Figure 2—figure supplement 2 ) , similar to that described for the total loss of Bag-1L ( Figure 1F ) , and a corresponding reduction in hormone-dependent AR function ( Figure 2—figure supplement 3 ) . Moreover , we could show that several evolutionarily conserved residues within the BAG domain of Bag-1L ( primarily in helices 2 and 3 ) , when mutated , impair the ability of Bag-1L to enhance AR AF-1 transactivation to a similar extent as the CMut Bag-1L protein ( Figure 2—figure supplement 4 ) . This indicates that mutating the BAG domain disrupts the Bag-1L:AR response similar to what we observed for the complete loss of Bag-1L . To test if mutations in the BAG domain would disrupt the integrity of associated biochemical complexes , we next employed quantitative , stable isotope labeling with amino acids in cell culture ( SILAC ) combined with rapid immunoprecipitation mass spectrometry of endogenous proteins ( RIME ) ( Mohammed et al . , 2013 ) of LNCaP cells that stably express FLAG-HA-tagged wild-type or CMut Bag-1L . Association of AR , but not Hsp70 ( HSPA1 ) , was disrupted in Bag-1L biochemical complexes in the context of the triple mutation ( Figure 2G ) ; these data agree with the results of our GST pull-down ( Figure 2C ) and co-IP experiments ( Figure 2D ) . Several proteins which exhibited decreased association with CMut Bag-1L ( Figure 2—source data 1 ) are annotated with functional roles in protein synthesis , localization , or other aspects of proteostasis ( Powers and Balch , 2013; Taipale et al . , 2014; Labbadia and Morimoto , 2015 ) ( Figure 2—source data 2 ) . The dynamics we observed in the biochemical complex as a function of the Bag1L mutant is consistent with our hypothesis that Bag-1L is involved in the folding of AR ( Figures 1G and 2F ) , suggesting a broader role for the BAG domain in proteome homeostasis . The reduction of interactors for the mutant Bag-1L could , in principle , be the result of an altered BAG domain conformation brought about by the triple mutation . To test this , we recorded 13C correlation nuclear magnetic resonance ( NMR ) spectra to compare Cα and Cβ shifts ( Sattler et al . , 1999 ) , which are predominantly influenced by the secondary structure of a protein . Comparison of the Cα and Cβshifts revealed no significant changes in the wild-type and mutant BAG domain peptide ( Figure 2—figure supplement 5 ) , suggesting that the extent of α-helix formation is essentially unchanged for the two proteins . However , about one third of residues that make the three antiparallel , helix bundles of the wild-type BAG domain ( Briknarová et al . , 2001 ) shifted to new positions or demonstrated reduced signal intensities in 1H15N-HSQC NMR spectra in response to the K231/232/279A mutations ( Figure 2—figure supplement 6 ) . This is most likely due to a destabilization of the entire protein caused by the three mutations , a consequence of which is a significant change in the 3D-structure of Bag-1L and hence an altered interactome of CMut compared to wild-type Bag-1L ( Figure 2G ) . Differences in the structural consequences of the wild-type or mutant BAG domain interaction with the AR AF-1 was next tested using solution NMR spectroscopy . Addition of the wild-type BAG peptide resulted in the reduction of resonance intensities within the C-terminal part of AR AF-1 , indicating that these two molecules interact transiently ( Figure 3A ) . The residues of AF-1 most affected by this interaction corresponded to tau-5 and were previously identified as partially folded by NMR ( De Mol et al . , 2016 ) , suggesting that the wild-type BAG domain interacts preferentially with this sub-domain . Moderate decreases in intensity were also observed in tau-1 , in the region centered around residue 275 , which has the propensity to adopt extended conformations . This suggests that although Bag-1L through its BAG domain binds tau-5 , the interaction propagates to tau-1 . Equivalent experiments carried out with the BAG domain mutant showed a much-attenuated effect , indicating that the strength of the interaction was diminished by the mutations , which agrees with our previous results ( Figure 2 ) . Given the importance of the Bag-1L BAG domain for the interaction with AR , as well as its significance for AR function and activity , it is conceivable that inhibition of the Bag-1L:AR interaction through this domain might provide a powerful tool to suppress AR transactivation and PCa growth . We therefore employed the canSAR drug discovery platform ( Bulusu et al . , 2014; Tym et al . , 2016 ) to query whether the BAG domain of Bag-1L contains any druggable or ligandable cavities that may be utilized for drug discovery . We analyzed 44 3D structural snapshots of the BAG domain of human Bag-1/Bag-1L ( Figure 3—source data 1 ) . Although all 44 structures lack a classical ‘druggable’ site ( defined as sites that harbor geometric and physiochemical properties consistent with binding orally-bioavailable small molecules with strict drug-like properties ) ( Tym et al . , 2016 ) , we were able to identify a large ‘ligandable’ cavity , which is druggable using peptides or peptidomimetic drugs ( Figure 3B ) . Moreover , we could demonstrate that the three amino acids ( K231/232/279 ) important for AR binding lie within ( K231 and 232 ) or just outside the edge ( K279 ) of this cavity ( Figure 3C ) . This agrees with our independent findings from the pull-down and luciferase assay ( Figure 2C and E ) , where mutation of residues K231 and K232 jointly , but not K279 alone , reduced the binding to AR . However , in these experiments the biggest effect was achieved in response to a simultaneous mutation of all three residues , suggesting that the triple mutant behaves similarly to the complete loss of Bag-1L . The overlap between these experimental results and the computational prediction of the cavity suggests that pharmacological interference with the predicted cavity of Bag-1L is highly likely to impair AR action . To test if pharmacological interference of the BAG domain of Bag-1L would indeed impair AR action , we tested the efficacy Thio-2 , a tool compound BAG domain inhibitor . Thio-2 has previously been postulated to bind to the BAG domain and to inhibit the interaction between Hsp70 and Bag-1 ( Enthammer et al . , 2013; Papadakis et al . , 2016 ) . A consequence of this is the reduce growth of breast cancer cells , but so far this compound has not been tested in PCa . Thio-2 was indeed able to abrogate the endogenous Bag-1L:AR interaction in LNCaP cells ( Figure 4A ) , as well as inhibit the androgen-dependent LNCaP cell growth ( IC50 = 17 . 5 μM; Figure 4B ) . Moreover , Thio-2 , unlike the AR N-terminal inhibitors EPI-001 or ido-EPI-002 ( Andersen et al . , 2010 ) and the AR LBD inhibitor enzalutamide , was able to effectively suppress Bag-1L function in a mammalian-1H assay ( Figure 4C and D ) . This suggests that targeting Bag-1L through inhibition of its BAG domain , is a therapeutic possibility to abrogate AR function and reduce PCa growth . To test whether mutation in the BAG domain of Bag-1L ( i . e . Bag-1L CMut ) could indeed function as a substitute for Bag-1L inhibition or Bag-1L loss , we measured cell growth of Bag-1L KO cells stably expressing an empty vector control , or the wild-type or CMut Bag-1L ( Figure 5A ) . Equal protein expression and AR levels were confirmed by western blotting ( Figure 5—figure supplement 1 ) . While re-expression of wild-type Bag-1L was able to rescue the growth inhibition triggered by Bag-1L KO , re-expression of the vector or Bag-1L CMut failed to do so . Given this result , we postulated that over-expression of wild-type Bag-1L should therefore cause an increase in PCa cell growth . We were able to confirm that overexpression of wild-type , but not CMut Bag-1L , results in an increase in LNCaP growth ( Figure 5B ) , without any obvious alterations in AR levels ( Figure 5—figure supplement 2 ) . Similarly , wild-type Bag-1L overexpression also promotes growth in mouse xenograft experiments , in intact ( Figure 5C ) and castrated animals ( Figure 5D ) . However , overexpression of wild-type Bag-1L ( Figure 5—figure supplement 3 ) , did not cause an increase in CRPC growth in LNCaP-abl and LNCaP95 cells ( Figure 5E and F ) , most likely due to endogenously elevated Bag-1L levels in these cells compared to the parental LNCaP line ( Figure 5—figure supplement 4 ) . Overexpression of the CMut Bag-1L protein on the other hand , lead to a dominant negative effect on cell growth ( Figure 5E and F ) , without any obvious alteration in AR level ( Figure 5—figure supplement 5 ) . This suggests that intact and functional Bag-1L mediates growth of CRPC cell models and may be associated with castration resistance and PCa disease progression . Given our hypothesis that Bag-1L is a driver of castration resistance , we next investigated if Bag-1 ( and AR ) expression levels change with PCa progression . H-scores were determined by immunohistochemistry ( IHC ) of matched diagnostic ( archival ) hormone-sensitive PCa ( HSPC ) and CRPC biopsies ( Figure 6A ) of 43 patients . Bag-1 antibody specificity for IHC was confirmed using Bag-1 siRNA in HeLa cells ( Figure 6—figure supplement 1 ) , and antibody depletion in various cell lines ( Figure 6—figure supplement 2 ) and patient biopsies ( Figure 6—figure supplement 3 ) . Specificity of the AR antibody for IHC was previously described ( Welti et al . , 2016 ) . Since Bag-1L is the only Bag-1 family member localized to the nucleus , H-scores were determined separately for the nuclear and cytoplasmic compartments . The Mann-Whitney test was used to compare median H-scores by matched biopsy . While nuclear Bag-1 ( i . e . Bag-1L ) levels increased significantly ( p<0 . 0001 ) in the progression from HSPC to CRPC , there was little change in cytoplasmic Bag-1 expression ( p=0 . 14 ) ( Figure 6B ) . In comparison , both nuclear and cytoplasmic AR levels were significantly increased ( p<0 . 0001 ) as patients progressed to CRPC ( Figure 6C ) . Although both nuclear Bag-1 and AR expression increased substantially with PCa progression , expression levels of the two proteins were not correlated ( Spearman rank correlation coefficient; Figure 6—figure supplement 4 ) . Given our finding that Bag-1L is a key regulator of AR action ( Figure 1 ) , we next investigated the correlation between nuclear Bag-1 levels and clinical benefits from AR targeted therapy . Of the 43 patients with matched HSPC and CRPC biopsies , 38 had been treated with abiraterone in the CRPC setting . Of these , 9 ( 23 . 7% ) were negative and 29 ( 76 . 3% ) positive for nuclear Bag-1 staining in their primary tumors . There were no significant differences in baseline characteristics of these patient groups at diagnosis or at initiation of abiraterone therapy , except for patient performance status ( ECOG PS; p=0 . 05 ) ( Table 1 ) . Moreover , there was no difference in 50% PSA response rate ( 33 . 3% vs 29 . 6%; p=1 . 00; data not shown ) , but nuclear Bag-1 positive ( compared with Bag-1 negative ) patients had a shorter median time to PSA progression ( 2 . 8 vs 6 . 6 months; log rank test p=0 . 02 ) ( Figure 6D ) and radiological progression ( 4 . 9 vs 9 . 8 months; log rank test p=0 . 05 ) ( Figure 6E ) , and a reduced median time on abiraterone treatment ( 4 . 9 vs 10 . 4 months; log rank test p=0 . 03 ) ( Figure 6F ) . Furthermore , these patients had a decreased median overall survival on abiraterone treatment ( 15 . 6 vs 20 . 9 months ) , but this was not statistically significant ( log rank test p=0 . 10 ) ( Figure 6—figure supplement 5 ) . In contrast , nuclear AR levels , which were positive for all 38 patients at diagnosis , were not associated with time to PSA progression , radiological progression or overall survival ( Table 1—source data 1 ) . As most of the CRPC biopsies in this cohort were obtained after patients had developed abiraterone resistance ( 60 . 5% ) , neither Bag-1 nor AR expression was predictive of PSA or radiological progression on abiraterone or overall survival ( Table 1—source data 2 ) . Collectively , these results demonstrate that nuclear Bag-1 ( i . e . Bag-1L ) increases with PCa progression and is associated with reduced clinical benefit from abiraterone .
Bag-1L is a known regulator of AR action implicated in PCa progression , but its mechanism of action is poorly understood . We show here that Bag-1L is essential for AR transactivation and function , and PCa growth . This function is mediated through the direct interaction of Bag-1L and the AR . Mutations that disrupt this interaction , inhibition of the Bag-1L:AR interaction or loss of Bag-1L altogether , alter the structural properties of the receptor and result in changes in the AR cistrome and transcriptome ( Figure 7 ) . We show that this leads to reduced growth of hormone-dependent and -independent PCa cells in culture and tumors in xenograft mouse models . Human Bag-1L is one of four polypeptides translated from a single mRNA by a leaky scanning mechanism ( Takayama et al . , 1998; Yang et al . , 1998 ) and as a result it has been difficult to demonstrate its function in biological systems , independent of the other Bag-1 proteins . Previous studies on Bag-1L function have either employed siRNA approaches , which reduced the expression of all four isoforms , or Bag-1L overexpression systems ( Froesch et al . , 1998; Guzey et al . , 2000; Cutress et al . , 2003; Shatkina et al . , 2003; Jehle et al . , 2014 ) . In our present work , we have used genome editing techniques to specifically knock-out endogenous Bag-1L and additionally rescued the knockout by re-expressing wild-type Bag-1L . Using this approach , we could demonstrate that Bag-1L is required for the correct and efficient folding , chromatin binding and transcriptional activity of the AR . Moreover , utilizing mass spectrometry we could show that in addition to binding and regulating the activity of AR , Bag-1L binds stress response proteins , underscoring its function as a survival and antiapoptotic protein and regulator of the proteostasis network ( Townsend et al . , 2003b; Mosser and Morimoto , 2004 ) . We have previously reported that Bag-1L binds the AR through a novel GARRPR motif , found at its N-terminus , and a BF-3 pocket at the AR LBD ( Jehle et al . , 2014 ) . However , inhibition of this interaction by mutating the GARRPR motif had only a modest effect on AR chromatin binding , the AR-mediated transcriptome and PCa cell growth ( Jehle et al . , 2014 and unpublished data ) . Thus , we believe that the Bag-1L GARRPR:AR BF-3 interaction acts as a modulator , rather than a regulator of AR activity . This agrees with findings that the BF-3 pocket itself acts as an allosteric modulator for receptor activity ( Estébanez-Perpiñá et al . , 2007 ) . In addition to the GARRPR:AR BF-3 interaction , Bag-1L also binds the AR N-terminus via its BAG domain . A triple mutant ( K231/232/279; CMut ) within helices 1 and 2 in the Bag-1L BAG domain was sufficient to inhibit the Bag-1L/AR interaction , in an Hsp70-independent manner . Although other mutations , primarily within helix 3 of the BAG domain , also inhibited the binding of Bag-1L and AR , these sites additionally interacted with Hsp70/Hsc70 . The region in AR most affected by binding to Bag-1L maps to the partially folded region within tau-5 , which strongly overlaps the region affected by EPI-001 binding ( De Mol et al . , 2016 ) . EPI-001 is a recently developed experimental drug that targets the AR NTD ( Andersen et al . , 2010 ) . A derivative of this compound ( EPI-506 ) is currently being employed in clinical trials for CRPC patients resistant to abiraterone and/or enzalutamide ( ClinicalTrails . gov Identifier: NCT02606123 ) . However , EPI-001 and iodo-EPI-002 were not able to suppress Bag-1L function in a mammalian-1H assay . Given our findings that the BAG domain of Bag-1L binds and regulates AR ( tau-5 ) with high specificity , targeting the Bag-1L:AR interaction might be an alternative approach for the treatment of PCa . This notion is supported by our findings pointing to a druggable pocket within the BAG domain of Bag-1L , which overlaps our triple mutation ( CMut ) . An inhibitor ( Thio-2 ) that targets the BAG domain was shown to be efficacious in blocking the antiapoptotic action of Bag-1 in breast cancer and melanoma cells ( Enthammer et al . , 2013; Papadakis et al . , 2016 ) . In the present study we show that it is also effective in inhibiting the Bag-1L BAG:AR tau-5 interaction and suppressing PCa cell growth . Aberrant expression of Bag-1 has been described in a variety of human malignancies , such as breast , lung , cervical , colorectal and hepatocellular carcinoma ( Zapata et al . , 1998; Rorke et al . , 2001; Clemo et al . , 2008; Cutress et al . , 2003; Ni et al . , 2013 ) . Here we show that the nuclear Bag-1 ( i . e . Bag-1L ) , but not the other Bag-1 isoforms , is upregulated in PCa patients that progress from HSPC to CRPC . This is in agreement with previous reports that show that nuclear Bag-1 levels ( Krajewska et al . , 2006 ) , or Bag-1L specifically ( Mäki et al . , 2007 ) , correlates with PCa progression . However , our study is the first to analyze Bag-1 expression in matched HSPC and metastatic CRPC biopsies , rather than unmatched tissues from untreated or hormone-refractory tumors . We additionally show that nuclear Bag-1 levels at HSPC status associate with a reduced clinical benefit from abiraterone . This is in line with previous reports that overexpression of nuclear Bag-1 correlates with drug resistance ( Ni et al . , 2013 ) and that Bag-1 overexpression is commonly observed in drug-resistant cell lines ( Ding et al . , 2000; Chen et al . , 2002; Liu et al . , 2009 ) . Our study provides evidence of an association between Bag-1 levels and treatment response in PCa , highlighting the prognostic significance of Bag-1 in this disease . In conclusion , we demonstrate here the importance of Bag-1L for AR activity and function in PCa . Combined , our data support targeting the BAG-1L BAG domain:AR tau-5 interaction therapeutically in the treatment of PCa and CRPC .
TALEN Bag-1L KO and vector controls were created as described ( Cermak et al . , 2011 ) . In brief , the left Bag-1 TALEN ( targeting the sequence ‘GGGCGGTCAACAAGT’ ) was translated into the RVD code ‘NN NN NN HD NN NN NG HD NI NI HD NI NI NN NG’ and assembled in pZHY500 . The right Bag-1 TALEN arm ( targeting the sequence ‘CGGGGGGGCGCGGAGA’ ) , was translated into the RVD code ‘HD NG HD HD NN HD NN HD HD HD HD HD HD HD NN’ and assembled in pZHY501 . Subsequently , both arms were subcloned into pZHY013 ( a kind gift from Daniel Voytas ) to generate a heterodimeric Fok1 nuclease . The plasmids were then subcloned into pDest12 . 2 using the Gateway cloning technology ( Thermo Fisher Scientific , Waltham , MA ) and transiently transfected into LNCaP cells using Fugene 6 ( Promega , Madison , WI ) . Transfected cells were selected using G418 for 24 hr ( 800 μg/ml ) . Single clones where isolated by serial dilution and screened for Bag-1L deletion by western blotting . To verify the genomic deletion , part of exon 1 was PCR-amplified and cloned into pcDNA3 . 1 V5 His6 ( Thermo Fisher Scientific ) using the TOPO cloning kit ( Thermo Fisher Scientific ) , and sequenced using specific primers ( Bag1g14fw 5’-GCTGGGAAGTAGTCGGGC-3’; Bag1g252rev 5’-CTGGTGGGTCGGTCATGC-3’ ) . Stable TALEN rescue cell lines ( Bag-1L KO +Vector control , Bag-1L KO +Bag-1L WT and Bag-1L KO +Bag-1L CMut ) and stable wild-type ( Flag-HA-tagged ) Bag-1L overexpressing LNCaP cells were created as previously described ( Jehle et al . , 2014 ) , using expression plasmid poZN . HeLa cell lines were employed for transient transfection using PromoFectin ( PromoCell , Heidelberg , Germany ) according to the manufacturer’s instructions . LNCaP-abl and LNCaP95 cells were a kind gift from Zoran Culig ( Innsbruck Medical University , Austria ) ( Culig et al . , 1999 ) and Stephan Plymate ( University of Washington , Seattle , WA ) ( Hu et al . , 2012 ) , respectively . Unless otherwise stated , all cell lines were obtained from the American Type Culture Collection and their identities were confirmed by short tandem repeat profiling ( BioSynthesis , Lewisville , TX ) . They were all confirmed to be mycoplasma negative , using the MycoAlert mycoplasma detection kit ( Lonza , Portsmouth , NH ) . All cell lines and parental LNCaP cells were cultured in RPMI 1640 medium or DMEM ( for HeLa cells only ) supplemented with 10% FBS , penicillin ( 100 u/ml ) , streptomycin ( 100 u/ml ) and L-glutamine ( 2 mM ) . For experiments requiring hormone starvation , cells were grown for 72 hr in phenol red-free RPMI 1640 medium , supplemented with 10% charcoal-stripped FBS , penicillin ( 100 u/ml ) , streptomycin ( 100 u/ml ) and L-glutamine ( 2 mM ) . CRPC lines were continuously cultured under hormone starvation condition as described above . Cell proliferation experiments were carried out as previously described ( Groner et al . , 2016 ) . All animal experiments were performed per European and German statutory regulations . Animal protocols were approved by the ‘Regierungspräsidium’ Karlsruhe , Germany ( AZ 35–9185 . 81/G-43/14 ‘Bag-1L-Prostastakarzinomprojekt’ ) . LNCaP xenograft tumor studies were carried out as previously described for intact ( Maddalo et al . , 2012 ) or castrated mice ( Eder et al . , 2013 ) . Patients were identified from a population of men with metastatic CRPC treated at the Royal Marsden NHS Foundation Trust . All study participants had given written , informed consent and were enrolled in institutional protocols approved by a multicenter research ethics committee ( Ethics Committee Centre: London-Chelsea Research Ethics Committee , Reference no . 04/Q0801/60 ) . Forty-three patients with sufficient formalin-fixed , paraffin embedded ( FFPE ) , matched diagnostic ( archival ) HSPC and CRPC tissue were included in our study . HSPC tissue demonstrated adenocarcinoma and was obtained from either prostate needle biopsy ( 35 ) , transurethral resection of the prostate ( TURP; 3 ) , prostatectomy ( 4 ) or bone biopsy ( 1 ) . CRPC tissue was obtained from the same patients through biopsies of bone ( 25 ) , lymph node ( 10 ) , liver ( 5 ) , prostate ( TURP; 1 ) , bladder ( 1 ) or chest wall ( 1 ) . All tissue blocks were freshly sectioned and only considered for IHC analyses if adequate material was present ( ≥50 tumor cells ) . For Bag-1 IHC ( using antibody Y166 , Abcam ) , HSPC and CRPC FFPE biopsies were first deparafinised , followed by antigen retrieval ( microwaving in citrate buffer ( pH 6 . 0 ) for 18 min at 800 W ) . The Bag-1 antibody was diluted ( 1:250 ) in Dako REAL diluent ( Agilent Technologies , Santa Clara , CA ) and tissue was incubated for 1 hr . After washes , the bound antibody was visualized using the Dako REAL EnVision Detection System ( Agilent Technologies ) . Sections were counterstained with hematoxylin . AR protein expression was determined using the AR mouse monoclonal antibody ( AR441 , Agilent Technologies ) , as previously described ( Welti et al . , 2016 ) . Nuclear and cytoplasmic Bag-1 and AR expression was determined for each case by author D . N . R in a blinded fashion using the modified H-score ( HS ) method using formula: [ ( % of weak staining ) x 1] + [ ( % of moderate staining ) x 2] + [ ( % of strong staining x 3 ) to provide a range from 0 to 300 ( Detre et al . , 1995 ) . HS data was reported as median values with interquartile range ( IQR ) . Demographic and clinical data for each patient were collected retrospectively from the hospital electronic patient record system . These characteristics were compared by Bag-1 status at HSPC using Fisher’s exact test for categorical characteristics , the Chi-squared test for trend for ordinal characteristics and either an unpaired t-test for continuous data , if normally distributed ( e . g . hemoglobin and albumin ) , or a Mann-Whitney test . The characteristics were then compared by Bag-1 HS value at CRPC biopsy using linear regression models and either a t-test or Walt test . PSA progression was defined as an increase in the PSA level of 25% or more above the nadir ( and by ≥2 ng/ml ) ; patients who stopped abiraterone without PSA progression were censored . Radiological progression was defined as any radiological imaging reporting disease progression; patients who stopped abiraterone without radiological progression were censored . Overall survival was defined as time from initiation of abiraterone to date of death ( 35 patients ) or last follow up/contact ( 3 patients ) . Co-immunoprecipitation was carried out as described previously ( Jehle et al . , 2014 ) . Proteins were isolated using TIVE lysis buffer ( 50 mM Tris-HCl pH 7 . 8 , 2 mM EDTA , 150 mM NaCl , 1% NP-40 , protease inhibitors ) , or lysis buffer A ( 10 mM HEPES pH 7 . 5 , 10 mM KCl , 0 . 1 mM EGTA , 0 . 1 mM EDTA , 1 mM DTT , protease inhibitors ) plus 0 . 5% NP-40 and lysis buffer C ( 20 mM HEPES pH 7 . 5 , 420 mM NaCl , 1 . 5 mM MgCl2 , 0 . 2 mM EDTA , 25% Glycerol , 1 mM DTT , protease inhibitors ) for total , cytoplasmic and nuclear proteins respectively . Western blotting was carried out using standard protocols with following antibodies: Bag-1 ( FL-274 and F-7 , Santa Cruz Biotechnology , Dallas , TX or Y166 , Abcam , Cambridge , UK ) , Bag-1L ( was obtained from Andrew Cato; Crocoll et al . , 2000 ) , AR ( N-20 , Santa-Cruz Biotechnology ) , Hsp70 ( K-20 , Santa Cruz Biotechnology ) , Hsp70/Hsc70 ( W27 , Santa Cruz Biotechnology ) , Flag-tag ( M2 , Sigma Aldrich , St . Louis , MO ) , HA-tag ( F-7 , Santa Cruz Biotechnology or Abcam ) , β-actin ( C4 , Santa Cruz Biotechnology or Abcam ) , vinculin ( V9131 , Sigma Aldrich ) , Lamin B1 ( EPR8985 , Abcam ) . Bag-1L and AR inhibitors were produced by Jakob Troppmair ( Thio-2 ) , Xavier Salvatella ( EPI-001 and iodo-EPI-002 ) or purchased from Selleckchem ( enzalutamide ) . ChIP was carried out as described ( Jehle et al . , 2014 ) , using anti-AR antibody ( N-20; Santa-Cruz Biotechnology ) . Following primers were utilized for directed ChIP qPCR . KLK3 ( AREIII ) primers were previously published ( Jehle et al . , 2014 ) . TMPRSS2 Fwd: 5’-GCTCACACAGGATCAGAGCA-3’ TMPRSS2 Rev: 5’-TGCTCGTTAGTGGCACATTC-3’ NKX3 . 1 Fwd: 5’-TTTGGGCCACCCTGTAAATA-3’ NKX3 . 1 Rev: 5’-GGGTGGGAGGAGATGAAAAT-3’ ChIP-seq libraries were generated using the ThruPLEX DNA-seq kit ( Rubicon Genomics , Ann Arbor , MI ) and were sequenced on the Illumina NextSeq 500 platform at the Molecular Biology Core Facility ( Dana-Farber Cancer Institute ) . ChIP-seq data was processed using ChiLin2 ( Qin et al . , 2016 ) . All ChIP-seq data have been deposited at the GEO depository under accession number GSE89939 . Total RNA was extracted from cells using innuPREP RNA Mini ( Analytic Jena AG , Jena , Germany ) , following the manufacturer’s instruction . mRNA libraries were generated using the Illumina TruSeq stranded mRNA sample kit and 1 µg of total RNA per sample . Library preparation , sequencing on a HiSeq1500 Illumina platform , and data analysis were carried out at the NGS facility of the Institute of Toxicology and Genetics ( KIT ) . Fastq files were processed with CASAVA and mapped against the human reference genome GRCh37 using TopHat 2 . 0 . 11 ( Trapnell et al . , 2009 ) . Reads were quantified with HTSeq ( Anders et al . , 2015 ) , using the reference gene annotation from Ensembl . Differential expression analysis was performed using DESeq2 ( Love et al . , 2014 ) . All RNA-seq data have been deposited at the GEO depository under accession number GSE89939 . Cells were transiently transfected using Lipofectin ( Thermo Fisher Scientific ) with plasmid CFP-AR-YFP ( Schaufele et al . , 2005 ) , generously provided by Marc Diamond . Images were acquired on an Andor Revolution XD spinning disk laser scanning microscopy system ( BFi OPTiLAS ) using two color channels . CFP was excited at 405 nm ( 90 µW , 100 ms ) and its emission collected through a 447/60 nm bandpass filter ( center wavelength/width , AHF ) . YFP was excited at 488 nm ( 187 µW , 100 ms ) and observed through a 560/55 nm bandpass filter . A longpass dichroic mirror ( DCLP 530 , AHF ) was used to separate the emission light . In addition , a notch filter ( 532/10 nm , center wavelength/width , AHF ) was inserted in front of the dichroic mirror to reduce crosstalk between CFP and YFP . Additional control experiments were performed to minimize crosstalk and direct excitation of YFP . After background subtraction , FRET signals were calculated using Icorr . FRET= IFRET – 0 . 17 ICFP . Nuclear regions were identified manually based on YFP staining . Acquired images were analyzed using ImageJ ( Abramoff et al . , 2004 ) . Short overlapping peptides ( 21 amino acids each ) spanning the entire human Bag-1L BAG domain and its flanking regions were synthesized and spotted in duplicates onto amino-PEG cellulose membranes ( Intavis AG Bioanalytical Instruments , Cologne , Germany ) using an automated SPOT synthesizer ( MultiPep , Intavis AG Bioanalytical Instruments ) ( Frank , 2002 ) . The membranes were incubated with bacterially-purified GST-AR-τau-5 ( amino acids 360–528 ) . Specific binding was detected using an anti-GST antibody ( Santa Cruz Biotechnologies ) . Alanine substitutions were introduced into the synthesis of positively identified peptides using site-directed mutagenesis ( Agilent Technologies ) and the hybridization procedure was repeated until single amino acids were identified in the BAG domain that destroyed the interaction with AR τau-5 . GST pull-down was performed as previously described ( Jehle et al . , 2014 ) . Mammalian one-hybrid experiments were performed as described ( Shatkina et al . , 2003 ) , using constructs pG5ΔE4-38 luciferase ( Gal4 reporter gene ) , TK Renilla luciferase , Gal4 fusion genes ( pM-AR-AF-1 , pM-τ1AR , pM-τ5AR ) or MMTV luciferase , TK Renilla luciferase and ARΔLBD ( amino acids 1–682 ) with pcDNA3 Bag-1L ( wild-type and mutants ) . Mammalian two-hybrid assays were performed using constructs pG5ΔE4-38 luciferase and TK Renilla luciferase ( Jehle et al . , 2014 ) , and Gal4DBD-ARLBD and VP16-AR-AF-1 ( provided by Karin Knudsen ) . SILAC-labeled rapid IP-mass spectrometry of endogenous protein ( RIME ) was carried out essentially as previously described ( Mohammed et al . , 2013 ) . Bag-1L WT and CMut cells were grown in media supplemented with ‘light’ ( L-lysine-2 HCl , L-arginine-HCl ) or ‘heavy’ isotope labels ( 13C6L-lysine HCl , 13C615N4L-arginine-HCl ) and mixed at a 1:1 ratio ( 30 million cells per treatment arm ) prior to IP . Proteins were immunoprecipitated using 20 μg anti HA-tag ( Abcam ) or rabbit IgG ( Santa Cruz Biotechnology ) antibody-coupled Dynabeads ( Thermo Fisher Scientific ) . Immunoprecipitated proteins were digested directly on beads as described ( Mohammed et al . , 2013 ) . Recovered peptides were acidified with 10% TFA and subjected to batch mode RP-SCX to desalt peptides and remove traces of detergent ( Adelmant et al . , 2011 ) . After trapping on a self-packed pre-column ( 100 um I . D . packed with 4 cm POROS 10R2; Applied Biosystems , Foster City , CA ) , peptides were eluted with an HPLC gradient ( 0–35% B in 4 hr ) and resolved using a self-packed analytical column with integrated ESI emitter tip ( Ficarro et al . , 2009 ) ( 30 um I . D . packed with 50 cm Monitor C18 , Orochem , with ~1 um ESI tip , flow rate ~30 nL/min ) prior to electrospray ( voltage = 3 . 8 kV ) . Peptides were analyzed by nanoflow LC-MS/MS using a NanoAcquity UPLC system ( Waters Corporation , Milford , MA ) coupled to an Orbitrap Fusion mass spectrometer ( Thermo Fisher Scientific ) , as previously described ( Ficarro et al . , 2009 ) . The mass spectrometer was operated in data-dependent mode such that the top 15 precursor ions in each MS scan ( image current detection , 120K resolution , m/z 300–2000 , target 5e5 , max inject time = 500 ms ) were subjected to both CID ( quadrupole isolation , width 1 . 6 Da , first mass = 110 , CE = 30% , rapid scan , target = 5e3 , max fill time = 50 ms , electron multiplier detection ) and HCD ( quadrupole isolation , width 1 . 6 Da , first mass = 110 , CE = 30% , target = 5e4 , max fill time = 50 ms , image current detection with 15K resolution ) . Mascot was used to search peak lists ( . mgf ) generated by multiplierz ( Askenazi et al . , 2009; Parikh et al . , 2009 ) against a forward-reverse database of human proteins ( NCBI refseq ) . Search parameters specified SILAC quantitation ( K + 8 , R + 10 ) , variable methionine oxidation , and fixed carbamidomethylation of cysteine residues , as well as a precursor mass tolerance of 10 ppm and product ion mass tolerances of 0 . 6 Da and 25 mmu for CID and HCD spectra . After filtering results to 1% FDR , SILAC quantitation was performed using multiplierz scripts for genes or gene groups represented by at least 2 unique peptides . SILAC data were then normalized according to Bag-1L expression levels ( CMut vs . WT ) . For protein products detected in ‘forward’ ( WT light , CMut heavy ) and ‘reverse’ ( WT heavy , CMut light ) SILAC experiments , ratios were averaged to provide an aggregate CMut/WT ratio . Search results were each filtered to 1% FDR , combined and further filtered to yield proteins represented by 2 or more unique peptides . Proteins detected in the Bag-1L RIME experiment were considered background and excluded from further analysis unless total summed peptide signal intensities were ≥3 fold higher than those in the control experiment . Genes or gene groups were considered regulated by the mutation of Bag-1L if ratios deviated by more than 2 standard deviations from median-normalized IgG control ratios . Data were plotted using R version 3 . 2 . 4 . The 15N-labeled thrombin protease-cleavable GST-fused BAG domain ( wild-type or K231/232/279A mutant ) was cloned into pGEX-6P vector ( Addgene , Cambridge , MA ) and expressed in E . coli strain BL21 ( DE3 ) , grown in M9 minimal medium supplemented with 0 . 5 g/l 15NH4Cl . The 15N-BAG domains were cleaved off by rhinovirus 3C protease ( PreScission , GE Healthcare , Marlborough , MA ) . To remove residual GST , the protein solution was purified using a glutathione sepharose column and subjected to size exclusion chromatography ( Superdex 200 , HiLoad 16/60 , GE Healthcare ) . 15N-HSQC spectra for wild-type and mutant BAG domains ( at approximate 500 μM ) and standard triple resonance backbone experiments ( HNCA , HNCACB , CBCA ( CO ) NH ) for peak assignment were acquired at 23° C on a Bruker Avance I 600 spectrometer . The spectrometer was equipped with a broadband triple resonance probe head with 4 scans per increment and a total of 128 increments in the indirect dimension . For chemical shift calibration and to compare relative signal intensities , 0 . 2 mM DSS ( 2 , 2 Dimethyl-2-silpentane-5-sulfonic acid ) was added . Data were processed with NMRPipe ( Delaglio et al . , 1995 ) and analyzed using NMR VIEW ( Johnson , 2004 ) . Expression , purification and nuclear magnetic resonance assignment of isotopically labeled 15N-AF-1 has been previously described ( De Mol et al . , 2016 ) . [1H , 15N]-HSQC spectra for AR-AF-1 alone ( at a 25 μM concentration ) , at a 1:5 molar ratio with GST ( as a control ) or with unlabeled GST-BAG ( wild-type or mutant ) were acquired at 278K on a Bruker 800 MHz spectrometer . Data was processed using NMRPipe and NMRDraw , and analyzed using CcpNmr Analysis ( Vranken et al . , 2005 ) . Peak intensities were normalized and plotted as a function of residue number . The 3D structure of human Bag-1L was analyzed using the structure-based druggability algorithm ( Bulusu et al . , 2014 ) developed as part of the canSAR drug discovery resource ( Tym et al . , 2016 ) . In short , the algorithm identifies up to 10 cavities on a 3D-structure and measures ~30 geometric and physicochemical properties for each of these cavities . Such properties include the volume , enclosure , depth , and complexity of the cavity , as well as the number of hydrogen-bond donors and acceptors , polarity distribution and the projected ligand-binding energy of the cavity . | Prostate cancer is the second most common cancer in men around the world . The cancer relies on a protein called the androgen receptor in order to develop and grow . Currently , some of the most common treatments for prostate cancer , especially in its advanced stages , are drugs that block the activity of this receptor . However , such treatments are only successful for a limited period of time , and so alternative methods to inhibit this receptor are still needed . The androgen receptor must bind to a number of proteins to carry out its activity . These proteins include one called Bag-1L , which is also important for the development of prostate cancer . Stopping such a protein from binding with the androgen receptor might represent a new way to treat prostate cancer; but first it will be important to understand how this interaction actually regulates the activity of the receptor . Now , Cato et al . have analyzed samples of cancer cells that had been collected from 43 patients with prostate cancer and found that Bag-1L levels increase as the disease progresses . Looking at the patients’ medical records then revealed that therapies targeting the androgen receptor were less effective in people with high levels of Bag-1L . Conversely , altering , removing or inhibiting Bag-1L in prostate cancer cells grown in the laboratory made the receptor less active and made the cells grow slower . Further experiments went on to reveal that Bag-1L interacts with a regulatory region of the androgen receptor . Cato et al . note that this region remains largely unexplored therapeutically , because it has some unique structural properties that restrict how much it can interact with drug molecules . Targeting Bag-1L and stopping it from binding to this region of the androgen receptor would represent a different approach to inhibiting the androgen receptor and treating patients with prostate cancer . Together these new findings should provide pharmaceutical companies with much of the information they would require to immediately start screening for therapies that target Bag-1L . Ultimately , Cato et al . hope that any follow-up findings will benefit prostate cancer patients by improving the currently available treatments . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cancer",
"biology"
] | 2017 | Development of Bag-1L as a therapeutic target in androgen receptor-dependent prostate cancer |
Insecticide-treated nets ( ITNs ) for malaria control are widespread but coverage remains inadequate . We developed a Bayesian model using data from 102 national surveys , triangulated against delivery data and distribution reports , to generate year-by-year estimates of four ITN coverage indicators . We explored the impact of two potential 'inefficiencies': uneven net distribution among households and rapid rates of net loss from households . We estimated that , in 2013 , 21% ( 17%–26% ) of ITNs were over-allocated and this has worsened over time as overall net provision has increased . We estimated that rates of ITN loss from households are more rapid than previously thought , with 50% lost after 23 ( 20–28 ) months . We predict that the current estimate of 920 million additional ITNs required to achieve universal coverage would in reality yield a lower level of coverage ( 77% population access ) . By improving efficiency , however , the 920 million ITNs could yield population access as high as 95% .
To facilitate standardised and comparable monitoring of ITN coverage through time , WHO and the Roll Back Malaria Monitoring and Evaluation Reference Group ( RBM-MERG ) has over the past decade defined a series of indicators to capture two different aspects of ITN coverage: access and use . Gold standard measurements of these indicators are provided by nationally representative household surveys such as Demographic and Health Surveys ( DHS ) ( Measure , 2014 ) , Multiple Indicator Cluster Surveys ( MICS ) ( UNICEF , 2012 ) , and Malaria Indicator Surveys ( MIS ) ( RBM , 2014a ) . These surveys are carried out relatively infrequently , however , meaning they cannot be used directly for evaluating year-on-year coverage trends or for generating timely estimates of continent-wide coverage levels . In contrast , programmatic data such as the number of ITNs delivered and distributed within countries , while not describing coverage directly , are available for most countries and years ( WHO , 2013a ) . In a 2009 study , Flaxman and colleagues ( Flaxman et al . , 2010 ) used a compartmental modelling approach to link these programmatic and survey data , generating annual estimates of the two ITN indicators recommended at that time on access ( % households with at least one ITN ) and use ( % children < 5 years old who slept under an ITN the previous night ) . Since that study , there has been increasing recognition that a richer set of indictors is required to identify the complex nature of ITN coverage ( Kilian et al . , 2013 ) . An intra-household 'ownership gap' may exist whereby many households with some nets may not have enough for one net between two occupants ( the recommended minimum level of protection ( WHO , 2013b ) . Similarly , a 'usage gap' may exist whereby individuals with access to a net do not sleep under it . In response , the measurement of two additional indicators was recommended: % households with at least one ITN for every two people and % population with access to an ITN within their household ( assuming each net was used by two people ) ( RBM , UNICEF , WHO , 2013; RBM , 2011 ) . In addition , the indicator on usage was extended to include the entire population rather than only children under 5 years old . This updated set of four indicators , used individually and in combination , has the potential to provide a nuanced picture of ITN access and use patterns that can directly guide operational decision making ( Kilian et al . , 2013 ) . To achieve this , there is a need to develop modelling frameworks to allow all four to be tracked through time . Countries have an ongoing struggle to maintain high LLIN coverage in the face of continuous loss of nets from households due to damage , repurposing , or movement away from target areas . In response , systems need to be responsive to emerging coverage gaps by ensuring nets are distributed to households that need them and avoiding over-allocation ( i . e . distribution of nets to those that already have them ) . Together , the rate of net loss and the degree of over-allocation of new nets play a key role in determining how efficiently delivery to countries will translate into household coverage levels . These factors are not currently well understood but triangulation of survey and programmatic data allows new insights into both . The WHO define universal access to ITNs on the basis that two people can share one net . Using the working assumptions of a 3-year ITN lifespan and a 1 . 8 person-per-net ratio ( one-between-two but allowing for odd-numbered households ) , a simple calculation yields an indicative estimate of 150 million new nets required each year to provide universal coverage to an African population at risk of around 810 million ( WHO , 2013a ) . To support country planning and donor application processes ( RBM-HWG , 2014 ) , a more elaborate needs assessment approach has been developed by the RBM Harmonization Working Group ( RBM-HWG ) and implemented by 41 of the 47 endemic African countries ( RBM , 2014; Paintain et al . , 2013 ) . The tool takes into account the size and structure of national target populations , a 1 . 8 person-per-net ratio for mass campaigns , additional routine distribution mechanisms employed by countries , and volumes of previously distributed nets and their likely rates of loss through time . Countries have used these inputs to calculate requirements for new nets to achieve national coverage targets , leading to an estimated continent-wide need for 920 million ITNs over the 2014–2017 period ( approximately 230 million per year ) ( RBM , 2014 ) . This tool provides a transparent , intuitive and standardised mechanism for comparing forecasted needs against current financing levels and identifying likely shortfalls . However , calculated needs are sensitive to assumptions about how a given volume of new nets will translate into population coverage , and inefficiencies in the system such as such as over-allocation and rate of net loss are not accounted for explicitly in the current needs assessment exercise . The purpose of this study is to define a new dynamic modelling approach , triangulating all available data on ITN delivery , distribution and coverage in sub-Saharan Africa in order to ( i ) provide validated and data-driven time-series estimates for all four internationally recommended ITN indicators; ( ii ) explore and quantify different aspects of system efficiency and how these contribute to reduced coverage levels; and ( iii ) estimate future LLIN needs to achieve universal access by 2017 under different efficiency scenarios and how these compare to existing needs assessment estimates .
Figure 1A summarises the main inputs to and outputs from the stock-and-flow model for LLINs when aggregated at the continental level . Some 718 million LLINs have been delivered across the 40 endemic countries since their introduction in 2004 . As is well documented ( WHO , 2013a ) , annual LLIN deliveries increased year-on-year from 2004 to 2010 , reaching 145 million in that year , but then declined dramatically in 2011 and 2012 to less than half that amount before rising again to 143 million in 2013 ( green line ) . Taking into account rates of loss in households , these LLIN deliveries led to a continental net crop shown by the red line . We estimate that there were 252 million LLINs in sub-Saharan households by the end of 2013 , with that net crop growing approximately linearly from 2004 , with the exception of a slow-down resulting from the reduced supply of nets in 2011–2012 . Figure 1B shows equivalent distribution and resulting net crop estimates for cITNs , which constituted nearly all ITNs prior to 2005 but diminished rapidly in importance following the introduction of LLINs thereafter . 10 . 7554/eLife . 09672 . 003Figure 1 . Time series of ITN delivery , distribution , and estimated net crop in sub-Saharan households 2000–2013 for ( A ) LLINs and ( B ) cITNs . Manufacturer data on deliveries were available for LLINs only . cITNs , conventional insecticide-treated nets; HHs , households; ITNs , insecticide-treated nets; LLINs , long-lasting insecticidal nets; NMCP , National Malaria Control Programme . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 003 Figure 2 shows continent-level time-series estimates of the four internationally recommended ITN indicators , along with the 'access gap' indicator . All four indicators show a similar temporal trend: very low coverage levels and modest year-on-year increases for the first 5 years from 2000 , with a marked inflexion point in 2005 and much more rapid gains thereafter . Importantly , however , all four indicators show that the pace of increase has , overall , slowed since 2005 . By the end of 2013 , we estimate that around two-thirds ( 66% , 95% CI 62%–71% ) of households at risk owned at least one ITN . However , less than one-third ( 31% , 29%–34% ) owned enough for one ITN between two people . This much lower level of adequate ownership is reflected in the levels of access and use , with 48% ( 45%–51% ) of people at risk having access to an ITN within their household ( on a one-between-two basis ) and 43% ( 39%–46% ) sleeping under an ITN the previous night . Comparison of Figure 2A , B demonstrates that many households that own some ITNs do not own enough for one-between-two , and this is captured in the time-series for the 'ownership gap' ( Figure 2E ) . Encouragingly , this gap has been narrowed from 77% ( 76%–78% ) of net-owning households having insufficient nets in 2000 to 56% ( 54%–57% ) in 2013 . Analysis of the 'use gap' suggested a large majority ( 89% , 84%–93% ) of those with access to an ITN in the household slept under it the previous night , and we found no evidence of significant change in this proportion through time . 10 . 7554/eLife . 09672 . 004Figure 2 . Continental-level time series of estimated ITN coverage indicators for the years 2000–2013 . ( A ) % households with at least one ITN; ( B ) % households with at least one ITN for every two people; ( C ) % population with access to an ITN within their household; ( D ) % population who slept under an ITN the previous night; ( E ) 'ownership gap' , the % of ITN-owning households with insufficient ITNs for one-between-two . Black circles are the annual estimates; pink envelopes denote the 95% posterior credible interval . ITNs , insecticide-treated nets . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 004 The relatively smooth temporal trends seen at continental level obscure a great deal of complexity in the patterns of ITN scale-up occurring at national level ( Figure 3 ) . Nearly all countries began with very low coverage levels in 2000 and display a marked inflection point towards the middle of the decade , although there was considerable variation in the timing of onset of concerted scale-up activities . Importantly , the monotonic increases in coverage seen at the aggregated continental level are often replaced at national level with pronounced periods of rise and fall , and in many cases , 2013 does not represent the peak year . Variation in contemporary levels of coverage remains stark . The population with access to ITNs within the household , for example , was at or below 15% in seven countries in 2013 , while above 70% for the top four . 10 . 7554/eLife . 09672 . 005Figure 3 . Country-level time series of estimated ITN coverage indicators 2000–2013 . Each plot shows the four ITN coverage indicators: % households with at least one ITN ( black ) ; % households with at least one ITN for every two people ( red ) ; % population with access to an ITN within their household ( green ) ; % population who slept under an ITN the previous night ( blue ) . CAR = Central African Republic; DRC = Democratic Republic of Congo; ITNs , insecticide-treated nets; HH = household . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 005 Over the 14-year period since 2000 , on average 15% ( 12%–18% ) of all ITNs distributed to households were over-allocated ( owned by households already owning sufficient nets for one-between-two ) . Figure 4 illustrates how these over-allocation rates have changed through time . Around 7% ( 6%–9% ) of ITNs were over-allocated in 2000 , and this has risen steadily to 27% ( 22%–32% ) in 2013 . The year-on-year increase in over-allocation is to some extent an expected consequence of the overall growth in ITN provision: we found that over-allocation increased approximately 15 percentage points for each one-ITN-per-capita increase in net crop . Over-allocation also varied substantially between countries , for example ranging in 2013 from 50% ( 36%–65% ) in the Republic of the Congo to 11% ( 9%–15% ) in Côte D’Ivoire . 10 . 7554/eLife . 09672 . 006Figure 4 . Time series of over-allocation for the combined set of 40 sub-Saharan endemic countries , 2000–2013 . Over-allocation refers to insecticide-treated nets distributed to households already owning enough nets for one-between-two , measured as the percentage of over-allocated nets among all nets in households . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 006 Averaged over all years and all countries , we found the median retention time for LLINs in households was 23 ( 20–28 ) months . We found no statistically significant evidence of continent-wide temporal trends in retention times , but substantial between-country variation . Figure 5 plots the LLIN loss function representing the most recent three years ( 2011–2013 ) for each country individually ( blue lines ) , along with the aggregated continental-level curve ( red line ) . For reference , we also overlay on Figure 5 some alternative loss functions that have been proposed . Flaxman et al . ( orange line ) fitted very small annual loss rates ( 5% ) for years 1 , 2 and 3 - with all LLINs then assumed lost after 3 years ( Flaxman et al . , 2010 ) . The RBM-HWG proposed rate of loss ( green line ) is 8 , 20 and 50% of LLINs to remain after 1 , 2 and 3 years , respectively , with all nets being lost thereafter ( Networks , 2014 ) . As can be seen , we found rates of loss for the first 3 years to be greater than both these alternatives for all countries . Both alternatives impose a three-year maximum retention time and our decision not to do so meant that we modelled a small proportion of LLINs lasting some years beyond that point . 10 . 7554/eLife . 09672 . 007Figure 5 . Insecticide treated netretention . Estimated long-lasting insecticidal net retention curves for each country individually ( blue lines ) and combined ( red line ) , in both cases relating to the average of the most recent 3 years , 2011–2013 . Also shown for reference are the rate of loss recommended in the Roll Back Malaria Harmonization Working Group needs assessment exercise ( green line ) and the loss rate fitted by Flaxman et al . ( orange line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 007 Figure 6 shows the projected levels of coverage that we estimate would be achieved by the end of 2017 with LLIN deliveries for the 2014–2017 period varying from zero to 2 . 5 billion and under a range of different efficiency scenarios . The most important characteristic of our results is the pronounced shallowing of the delivery-coverage curves: proportionately smaller gains are made in coverage as more LLINs are delivered in an archetypal 'law of diminishing marginal returns' . This means that under a business-as-usual scenario , where current levels of over-allocation and LLIN loss persist , very large increases in LLIN delivery would be required to achieve high coverage . Under this scenario , we estimate that 1 billion LLINs ( i . e . an average of 250 million per year ) would be required to achieve 80% of the population with access to an LLIN in the household by the end of 2017 , although this would only translate into 70% population use . 10 . 7554/eLife . 09672 . 008Figure 6 . Projected 2017 coverage for sub-Saharan Africa in relation to number of LLINs delivered over 2014–2017 period . ( A ) % households owning at least one ITN; ( B ) % households owning enough ITN for one between two; ( C ) % population with access to ITN within the household; ( D ) % population sleeping under an ITN the previous night; ( E ) 'ownership gap' , the % of ITN-owning households with insufficient ITNs for one-between-two . For each indicator , we project likely coverage under four scenarios: current levels of over-allocation and net loss ( i . e . 'business as usual' ) ; with minimised over-allocation; with longer average net retention ( 3-yr median ) ; and with both minimised over-allocation and longer net retention . The vertical dashed lines indicate the number of LLINs calculated as required over the period under the country programmatic needs assessment supported by Roll Back Malaria Harmonization Working Group . LLINs , long-lasting insecticidal nets; ITNs , insecticide-treated nets . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 008 The extent to which coverage gains diminish as deliveries increase is mitigated substantially when over-allocation and ITN loss rate are reduced . In a scenario with minimised over-allocation ( where over allocation is set to zero ) , 80% population access in 2017 would be achievable with just 700 million nets ( 175 million per year ) . Reducing ITN loss rate to a 3-year median retention time would have a broadly similar impact , acting in isolation , to minimising over-allocation . If these two hypothetical efficiency gains were combined , however , 80% access could be reached in 2017 with around 560 million nets ( 140 million per year ) . We found that the relative importance of the over-allocation and LLIN loss rates changed as more LLINs were introduced . Increasing LLIN retention times was the most important factor at low levels of net delivery , but as more and more nets were provided , over-allocation became progressively more important . This is intuitive since it becomes increasingly difficult to avoid over-allocation as more households obtain adequate numbers of nets . For reference , we also plot on Figure 6 the 920 million additional LLINs calculated by countries as required for universal coverage of targeted populations by 2017 under the RBM-HWG needs assessment exercise . Under current levels of over-allocation and net loss , we estimate that by the end of 2017 this quantity of new LLINs would translate into 77% access ( among those populations targeted by countries for ITN coverage ) and , assuming current behaviour patterns continue , 68% sleeping under an ITN . Under the combined efficiency scenario with minimised over-allocation and 3-yr median ITN retention time , however , the 920 million nets would approach universal access ( slightly over 95% ) .
By linking manufacturer , programme and national survey data using a conceptually simple model framework , the intention has been to provide a transparent and intuitive mechanism for tracking net crops and resulting household coverage that reflects the input data while simultaneously providing a range of insights about the system itself . In doing so we have been able to ( i ) provide a new approach for estimating past trends and contemporary levels of ITN coverage; ( ii ) explore the effects of uneven net distribution between households and the rates of net loss once in households; and ( iii ) use these insights to estimate how many LLINs are likely to be required to achieve different coverage targets in sub-Saharan Africa . We have , for the first time , extended dynamic model-derived estimation of ITN coverage to all four internationally recognized indicators , along with the two 'gap' metrics . Our results reinforce a simple message: while gains in ITN coverage have been impressive , there remains an enormous challenge if the goal of universal access is to be achieved and sustained . The importance of the new expanded suite of indicators is also exemplified: while an encouraging two-thirds of households now own at least one ITN , less than half of these have enough to protect everyone who lives there . This ownership gap is narrowing but the disparity remains evident across nearly all countries . Conversely , there is little evidence that non-use of available nets contributes substantially to low coverage levels . We therefore reinforce earlier studies that suggest the overwhelming barrier to not sleeping under an ITN is lack of access rather than lack of use WHO , 2013a; Eisele et al . , 2009; 2011; Koenker and Kilian , 2014; Koenker et al . , 2014 ) . Of course , non-use may be important in certain local contexts , and finer-scale analysis can support identification of areas where behaviour change communication interventions may be appropriate to reduce it ( Kilian et al . , 2013 ) . We found substantial over-allocation of nets to households already owning a sufficient quantity , and that this became more pronounced as overall ownership levels increased through time . Mass distribution campaigns can , in principle , be designed to minimise over-allocation and maximise evenness of nets allocated to households strictly on the basis of households members and pre-existing nets . As other studies have highlighted , however , any possible commodity savings achieved by such strategies must be compared against the operational cost of these more complex distribution mechanisms ( Yukich et al . , 2013 ) . What is certain is that over-allocation becomes a major barrier to achieving universal coverage when levels of ITN provision are high because most new incoming nets are simply leading to surpluses in many households , while elsewhere there remains a shortfall . This may have a disproportionately high public health impact if those surplus nets are concentrated in households at lowest risk . Wealthier , better educated and more urban households may be better placed to obtain available nets but are often located in regions of lower transmission ( Steketee and Eisele , 2009; Webster et al . , 2005 ) . While beyond the scope of the present study , the approaches we have developed here could be extended to consider these issues of equity in coverage versus risk in more detail . One of the most important observations in our study is that LLINs may be lost from households at a substantially faster rate than is currently assumed . Importantly , we assess loss by comparing total inputs to countries ( from deliveries ) to total numbers in households ( net crop ) , and so we measure real losses rather than , for example , reallocation of nets between relatives ( Koenker et al . , 2014 ) . Longer retention times of the sort observed in some local studies are not supported by the body of evidence we have provided by triangulating large-scale net distributions and household survey data . This more rapid loss rate has potentially important implications for existing guidelines . Current RBM guidance is for mass ITN campaigns to be conducted every 3 years , complemented by continuous distribution of nets via routine channels in order to maintain coverage levels between those campaigns . However , whatever levels of coverage are achieved by a given campaign , we estimate that one-half of the campaign nets distributed , on average , will not be present in households just 2 years later . Our coverage time-series for many countries suggest that routine distribution channels are not yet compensating fully for this rate of loss , often displaying pronounced dips in coverage levels between mass campaigns . Maintaining higher continuous coverage therefore clearly requires some combination of more frequent campaigns , greater ongoing distribution between campaigns , or more durable nets and improved care behaviour by users that lead to longer overall retention times . We considered nets in households as simply present or absent , with no allowance for their condition . In reality , of course , nets may be retained by households ( and thus 'present' in our calculations ) even when they are badly torn , or have diminished insecticidal properties . As such , our estimates of 'coverage' would be revised downwards if additional measures of net efficacy were included . Our model is able to provide an estimate for every country and every year of the age-profile of ITNs in households . This raises the possibility of extending the predictions to incorporate modelled or observational data on average rates of net degradation in different contexts ( Briët et al . , 2012 ) to explore measures of entomologically effective coverage . Tools developed to assist countries to calculate LLIN requirements , have tended to define need using a simple ratio to populations at risk ( such as 1 . 8 people per net ) , and have made allowances for net loss from households using pre-defined rates of loss . We have been able to show that true LLIN requirements are likely to be considerably larger when the more rapid rates of loss are taken into account , along with the additional effect of likely over-allocation patterns . This more realistic framework not only provides the basis for more accurate needs assessments but identifies the relative importance of these different factors in determining the coverage that can be achieved for a given delivery level . Our analysis of future LLIN needs from the present time to 2017 demonstrates how these factors lead to a pronounced law of diminishing returns: as more nets are introduced to a population , proportional increases in coverage diminish , with over-allocation a particular problem at high net provision levels . Under business-as-usual , the number of nets required to approach full coverage is prohibitively large . Clearly , however , reducing current system inefficiencies and increasing net retention are not straightforward and already the subject of much attention by countries and international partners . Over-allocation is the complex result of different distribution strategies and varying levels of population access to services , and any solution comes with its own cost . Net retention can doubtless be increased by improved LLIN technology coupled with behaviour-change communication efforts , although it is also feasible that retention times may reduce when overall net provision increases ( with new nets displacing older ones ) . Additionally , we look only at the RBM definition of use and ignore the effectiveness of nets in repelling mosquitoes once they are being used . This is potentially an important confounder when considering retention times . While not aiming to provide solutions to these complex challenges , the results we present here provide an analytical framework in which the impact of theoretical efficiency gains can be assessed and this could be extended to include formal cost–effectiveness analysis . In conclusion , our results provide evidence that LLIN requirements to achieve universal coverage have been underestimated . If obtaining higher coverage remains an accepted goal of the international community , then larger LLIN volumes must be considered and planned for at national and international levels . We emphasise , however , that this would be best achieved in parallel with a renewed focus on maximising the efficiency of coverage achieved for each new net financed . Given that the pattern of diminishing coverage returns for each dollar spent is likely to be unavoidable , the cost–effectiveness of pursuing universal coverage rather than a lower operational target must ultimately be weighed against alternative malaria control investments .
Two important preceding studies have sought to model national-level ITN delivery , distribution , and coverage: the Flaxman et al . study ( Flaxman et al . , 2010 ) and the work of Albert Killian culminating in the NetCALC tool ( Networks , 2014 ) and a series of related publications ( Paintain et al . , 2013; Yukich et al . , 2013 ) . Although very different in implementation , both approached the problem in a similar two-stage process . First , a mechanism was defined for estimating net crop — the total number of ITNs in households in a country at a given point in time—taking into account inputs to the system ( e . g . deliveries of ITNs to a country ) and outputs ( e . g . the discard of worn ITNs from households ) . Second , empirical modelling was used to translate estimated net crops into resulting levels of coverage ( e . g . access within households ) . We have adopted a similar analytical outline , but the models we have developed for each stage differ structurally and conceptually from these earlier efforts . Our underlying principle has been to represent the ITN system in a simple and intuitive way and to parameterise that system using a data-driven approach that minimises the reliance on assumptions or small external datasets . In this Methods section , we describe: ( i ) the main data sources used; ( ii ) a new compartmental model for estimating net crop that also offers insights into rates of ITN loss from households; ( iii ) a new coverage model linking net crop to household net access and use that also assesses the efficiency of between-household distribution ( i . e . the extent of over-allocation ) ; and ( iv ) the use of our models to predict future ITN requirements to meet the goal of universal access . A schematic overview of our analytical framework is provided in Figure 7 , and additional methodological detail is provided in the Supplementary Information . 10 . 7554/eLife . 09672 . 009Figure 7 . Schematic showing overall analytical framework linking data , model components , and outputs . HH = household; ITN , insecticide-treated net; NMCP = National Malaria Control Programme . DOI: http://dx . doi . org/10 . 7554/eLife . 09672 . 009 We used three principal sources of data to fit our models . These are described briefly below and in more detail in Supplementary Information . Our main analysis covered 40 of the 47 ( WHO , 2013a ) malaria endemic countries of sub-Saharan Africa . We excluded six endemic countries on the basis that ITNs do not form an important part of their vector control programme , as reported by the respective NMCPs to the African Leaders Malaria Alliance , ALMA ( M . Renshaw , pers . comm . 3rd August 2014 ) . These were Botswana , Cape Verde , Namibia , São Tomé and Príncipe , South Africa and Swaziland . We also excluded the small island nation of Mayotte , for which no ITN delivery or distribution data were available . We limited all analyses to those populations categorized as being at risk by NMCPs ( WHO , 2013a ) . When interpreting NMCP distribution and household ownership data , we made the simplifying assumption that all reported ITNs were distributed among , and owned within , households situated in malaria endemic regions ( Burgert et al . , 2012 ) . Additionally , we used data from African Leaders Malaria Alliance ( ALMA ) on the proportion of populations at risk targeted for ITNs versus IRS , and downscaled targeted populations at risk accordingly . It should be noted that restricting the distribution of ITNs to populations at risk makes the assumption that no ITNs are distributed to populations not at risk . Like Flaxman et al . ( Flaxman et al . , 2010 ) , we represented national ITN systems using a discrete time stock-and-flow model . In this structure , a series of compartments were defined that contained a given number of nets at each time-step , with possible movement of nets from one compartment to another between time-steps ( see Supplementary Information ) . Nets delivered to a country by manufacturers were modelled as first entering a 'country stock' compartment ( stored in-country but not yet distributed to households ) . Nets were then available from this stock for distribution to households by the NMCP or other distribution channels . Years where NMCP distributions were smaller than available country stock represented potential ‘under-distribution’ , with nets left to stockpile rather than reaching households . However , because of the uncertainty associated with NMCP distribution data , these discrepancies could simply reflect under-reporting of distribution levels . To accommodate this uncertainty , we specified the number of nets distributed in a given year as a range , with all available country stock as one extreme ( the maximum nets that could be delivered ) and the NMCP-reported value ( the assumed minimum distribution level ) as the other . New nets reaching households joined older nets remaining from earlier time-steps to constitute the total household net crop , with the duration of net retention by households described by a loss function . In this representation , the net crop simply reflected the differences over time between inputs to and outputs from households . This meant that distribution , net crop , and the loss function together formed a closed system: the three must triangulate exactly and knowledge of any two components allowed the third to be calculated directly . Flaxman et al . ( Flaxman et al . , 2010 ) assembled data from six studies on ITN durability and rates of loss . Using a loss function fitted to these data , however , they found that the three components tended not to triangulate: net crops observed in surveys were too small , given the data on nets distributed to households and their modelled rate of loss . Their interpretation was that the number of ITNs distributed each year may be systematically over-reported by NMCPs , and a 'bias parameter' was included in the model , adjusting downward the volume of nets entering households in each country compared with reported levels . As described above , we took a different approach: with no a priori expectation that NMCP distribution reports exaggerate distribution levels . Rather than fitting the loss function to a small external dataset , we fitted this function directly to the distribution and net crop data within the stock-and-flow model itself . Conceptually , this reflected the view that the 560 country-years of distribution data triangulated against the 102 survey-derived national net crop values represented a more impartial and data-driven way of inferring rates of loss than using limited data from local ITN retention studies . Loss functions were fitted on a country-by-country basis , allowed to vary through time , and defined separately for cITNs and LLINs . We compared these fitted loss functions to existing assumptions about rates of net loss from households . The stock-and-flow model was fitted using Bayesian inference and Markov chain Monte Carlo ( MCMC ) , providing time-series estimates of national household net crop for cITNs and LLINs in each country along with evaluation of under-distribution , all with posterior credible intervals . A complete technical description is provided in the Supplementary Information . Levels of ITN access within households depend not only on the total number of ITNs in a country ( i . e . net crop ) , but on how those nets are distributed between households . In simple terms , a more even distribution yields a greater proportion of households owning nets than if those same nets are concentrated in fewer households . Many recent national surveys report the number of ITNs observed in each surveyed household . This allows , a histogram to be generated that summarises the net ownership pattern ( i . e . the proportion of households with zero nets , one net , two nets and so on ) . By analysing such data from multiple surveys , previous studies have demonstrated that histograms for different countries vary in a broadly predictable way according to national net crop ( Flaxman et al . , 2010; Yukich et al . , 2013 ) . By representing these histograms using a formal statistical distribution ( such as the negative binomial ) , and linking its parameters to net crop , predicted histograms can be generated for any country-year for which a net crop estimate is available ( Flaxman et al . , 2010; Yukich et al . , 2013 ) . These histograms , in turn , allow direct calculation of the first access coverage indicator ( % households owning one or more ITN ) . We took the view that this approach—linking net crop to a statistical distribution , and using the distribution to calculate access indicators—is preferable to the alternative of regressing the access indicators against net crop directly . The latter approach , used in the NetCalc tool ( Networks , 2014 ) , is simpler but provides less direct insight into the patterns of between-household ITN distribution that ultimately link net crops to access levels . One aspect that is known to strongly influence the relationship between net crop and household ownership distribution is the size of households found in different countries ( Networks , 2014; Yukich et al . , 2013 ) , which varies greatly across sub-Saharan Africa ( Swaziland , for example has an average household size of around three members , while in Senegal the average is nearly ten ) . Household size also , of course , determines whether a given number of owned nets will be sufficient to provide access to all residents . We extended earlier analyses ( Flaxman et al . , 2010; Yukich et al . , 2013 ) to explicitly account for household size: using a bivariate ( i . e . two- rather than one-dimensional ) histogram model to link net crop to ownership distributions for each household size stratum ( see Supplementary Information ) . We replaced the negative binomial distribution with a 2-d zero-truncated Poisson distribution and , for each household size stratum , fitted the distribution using two parameters: ( i ) the proportion of households with zero ITNs and ( ii ) the mean number of ITNs per ITN-owning household . Using the household-level data from 83 national surveys , we found that both parameters were strongly related to national net crop , allowing bivariate histograms to be generated for every country-year that were closely representative of the true ITN ownership distribution . Stratifying our analysis by household size had three important advantages over earlier approaches . First , the distribution of net ownership tended to vary substantially between households of different sizes within a given country and this variation would be missed if all households were considered together . Accounting for this enabled better fits to the data . This makes sense: all else being equal , larger households would be expected to own more nets than smaller ones and so distribution patterns would differ systematically . Second , the bivariate ownership histograms predicted for each country-year could be used to directly calculate all three indicators of household access . While a simple univariate histogram allows calculation of % households with at least one ITN , a bivariate histogram means the number of both ITNs and people in every household can be triangulated which , in turn , allows direct calculation of the two additional indicators: % households with at least one ITN for every two people and % population with access to an ITN within their household , along with the 'ownership gap' ( see Supplementary Information ) . Linking these bivariate histograms to our annual net crop estimates for each country meant we could predict time-series of the access indicators at the national level from 2000–2013 , with all parameters fitted in a Bayesian framework providing posterior credible intervals around each time-series . We also combined the country-level results to generate a set of continent-level indicator time-series , representing overall coverage levels among populations at risk in the 40 endemic countries . Third , the bivariate histograms allowed analysis of over-allocation: certain cells of the histogram represented households owning more ITNs than were required to achieve access on a one-between-two basis , and the proportion of the total net crop falling in this category was examined through time for every country . We took a different approach for the final indicator , % population who slept under an ITN the previous night . ITN use is less directly linked to national net crop and is primarily determined by the availability of nets within households ( Eisele et al . , 2009 ) . A total of 83 of the 102 national surveys contained data allowing the relationship to be explored between ITN use and each of the three access indicators with , perhaps unsurprisingly , % population with access to an ITN within their household displaying the largest correlation ( adjusted R2= 0 . 96 ) . We fitted this relationship across the 83 surveys using a simple Bayesian regression model ( see Supplementary Information ) and used it to predict time-series of the ITN use indicator for every country . The ratio of population use to access revealed the 'usage gap'—the fraction of the population with access to ITNs not using them—and between-country variation in this ratio was also explored . Our two-stage modelling framework represented the pathway from ITN delivery into countries through to resulting levels of net access and use in households . It also accounted for two potential factors that act to reduce access levels , and allowed these to be quantified through time for each country . Using this architecture , it was possible to simulate delivery of any hypothetical volume of ITNs to a given country over a given future time period , to predict the levels of access and use that would result , and to examine the impact of different amounts of over-allocation and net loss . The current needs assessment exercise that countries are undertaking ( RBM-HWG , 2014; RBM , 2014 ) is designed to identify the number of LLINs required to achieve coverage targets by 2017 . We used our model to estimate the levels of access likely to be achieved if these forecast LLIN commodity needs were met across the 2014–2017 period under a 'business as usual' scenario , that is , with current levels of over-allocation and net loss , and compared these predicted levels with the objective of universal access among target populations . We then generalized this experiment to predict the likely level of coverage ( for all four indicators ) achievable by 2017 under a broad spectrum of LLIN delivery levels , equivalent to a total for sub-Saharan Africa ( two of the 40 endemic countries in our study did not participate in the RBM-HWG needs assessment exercise [Djibouti and Equatorial Guinea] , and so our scenario analysis is based on the set of 38 remaining countries; to maintain comparability through time , we combined needs assessment data for mainland Tanzania and Zanzibar , and for Sudan and South Sudan ) of between zero and 2 . 5 billion nets across the 4-year period . Further , we ran these simulations under four scenarios: ( i ) 'business-as-usual' ( where current levels of over-allocation and net loss were maintained ) ; ( ii ) with no over-allocation ( new LLINs are distributed preferentially to those households with zero LLINs , then to those with less than one-between-two ) ; ( iii ) with reduced LLIN net loss by households ( using a modelled 3-year median retention time ) ; and ( iv ) with both no over-allocation and a 3-year median retention time . | Malaria is a major cause of death in many parts of the world , especially in sub-Saharan Africa . Recently , there has been a renewed emphasis on using preventive measures to reduce the deaths and illnesses caused by malaria . Insecticide-treated nets are the most prominent preventive measure used in areas where malaria is particularly common . However , despite huge international efforts to send enough nets to the regions that need them , the processes of delivering and distributing the nets are inefficient . This problem is compounded by the fact that little information is available on how many nets people actually own and use within each country . ` Bhatt et al . have now created a mathematical model that describes the use and distribution of nets across Africa since 2000 . This is based on data collected from national surveys and reports on the delivery and distribution of the nets . The model estimates that in 2013 , only 43% of people at risk of malaria slept under a net . Furthermore , 21% of new nets were allocated to households that already had enough nets , an inefficiency that has worsened over the years . Nets are also lost from households much more rapidly than previously thought . It’s currently estimated that 920 million additional nets are required to ensure that everyone at risk from malaria in Africa is adequately protected . However , Bhatt et al . ’s model suggests that given the current inefficiencies in net distribution , the extra nets would in reality protect a much smaller proportion of the population . Taking measures to more effectively target the nets to the households that need them could improve this coverage level to 95% of the population . The next challenge is to devise distribution strategies to send nets to where they are most needed . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"epidemiology",
"and",
"global",
"health"
] | 2015 | Coverage and system efficiencies of insecticide-treated nets in Africa from 2000 to 2017 |
Recent studies suggested an essential role for seryl-tRNA synthetase ( SerRS ) in vascular development . This role is specific to SerRS among all tRNA synthetases and is independent of its well-known aminoacylation function in protein synthesis . A unique nucleus-directing domain , added at the invertebrate-to-vertebrate transition , confers this novel non-translational activity of SerRS . Previous studies showed that SerRS , in some unknown way , controls VEGFA expression to prevent vascular over-expansion . Using in vitro , cell and animal experiments , we show here that SerRS intervenes by antagonizing c-Myc , the major transcription factor promoting VEGFA expression , through a tandem mechanism . First , by direct head-to-head competition , nuclear-localized SerRS blocks c-Myc from binding to the VEGFA promoter . Second , DNA-bound SerRS recruits the SIRT2 histone deacetylase to erase prior c-Myc-promoted histone acetylation . Thus , vertebrate SerRS and c-Myc is a pair of ‘Yin-Yang’ transcriptional regulator for proper development of a functional vasculature . Our results also discover an anti-angiogenic activity for SIRT2 .
In vertebrates from fish to humans , the vasculature is one of the most important and earliest networks to develop . Surprisingly , three independent forward genetics studies in zebrafish suggested an essential role for seryl-tRNA synthetase ( SerRS ) in vascular development ( Amsterdam et al . , 2004; Fukui et al . , 2009; Herzog et al . , 2009 ) . In fish embryos , disruption of sars ( gene encoding SerRS ) , through insertional mutagenesis ( Amsterdam et al . , 2004 ) or ENU mutagenesis-induced point/truncation mutations ( Fukui et al . , 2009; Herzog et al . , 2009 ) , caused excessive and abnormal blood vessel growth . As a member of the aminoacyl-tRNA synthetases family , SerRS is well-known for its essential function in aminoacylation of tRNASer for protein synthesis in the cytoplasm . However , the role of SerRS in vascular development is independent of its enzymatic activity ( Fukui et al . , 2009 ) , but dependent on its vertebrate-specific , non-catalytic , C-terminal domain UNE-S ( Guo et al . , 2010; Xu et al . , 2012; Guo and Schimmel , 2013 ) . The UNE-S domain contains a robust nuclear localization signal ( NLS ) sequence that , at least in human cells , directs a substantial amount of cellular SerRS into the nucleus ( Guo et al . , 2010; Xu et al . , 2012; Guo and Schimmel , 2013 ) . Remarkably , all non-null mutations of sars linked to vasculature abnormalities in the aforementioned genetics studies either have the NLS truncated or conformationally sequestered , and thus render deficient SerRS nuclear localization ( Xu et al . , 2012 ) . Conversely , zebrafish expressing engineered catalytically active but NLS-mutated SerRS exhibited the same abnormal blood vessel phenotype as observed in the sars mutant embryos ( Xu et al . , 2012 ) . Therefore , it has been clearly established that the essential role of SerRS in vascular development arises from its evolutionarily acquired nuclear presence . Interestingly , the vascular abnormalities associated with deficient SerRS nuclear localization were found to be accompanied with a high level of vegfa ( Vascular Endothelial Growth Factor A ) transcript in the mutant fish embryos ( Fukui et al . , 2009; Xu et al . , 2012 ) . This observation suggested that the nuclear function of SerRS in zebrafish is linked to attenuating the expression of Vegfa . However , the mechanism of the SerRS function has remained obscure . Because VEGFA is a key stimulator of vasculogenesis and angiogenesis for all vertebrates , and over-expression of VEGFA is not only associated with developmental vascular abnormalities , but also contributes to various diseases including cancer ( Drake and Little , 1995 ) , we were motivated to determine whether the VEGFA-regulating function of SerRS is conserved in higher vertebrates such as humans , and what is the mechanism by which nuclear SerRS controls VEGFA expression . It is well established that c-Myc is the major transcription factor promoting VEGFA gene expression in the nucleus , and thereby has a key role in vascular development . As a basic helix-loop-helix-leucine zipper ( bHLHZ ) protein , c-Myc functions through heterodimerization with the small bHLHZ partner MAX for binding to the Enhancer Box ( E-box ) DNA sequence ( 5′-CACGTG-3′ ) on its target genes ( Blackwood and Eisenman , 1991 ) . DNA-bound c-Myc recruits histone acetyltransferase to acetylate histone proteins to allow chromatin expansion and activate transcription ( Grandori et al . , 2000 ) . c-Myc knockout mice are embryonic lethal and exhibit , among others deformities , under-developed vasculature . Importantly , these deformities can be partially rescued by transgenic VEGFA expression ( Baudino et al . , 2002 ) . On the other hand , endothelial-specific c-Myc overexpression in mice also causes embryonic lethality arising from widespread edema , multiple hemorrhagic lesions and severe defects in the vascular network , accompanied by an elevated level of VEGFA ( Kokai et al . , 2009 ) . Together , these results suggest that the role of c-Myc in vascular development and in promoting VEGFA expression has to be tightly balanced . In the work described below , we have elucidated a novel mechanism by which this balance is achieved . We show a head-to-head competition between SerRS and c-Myc for the same VEGFA promoter binding site and , in addition , a direct recruitment of SIRT2 histone deacetylase by SerRS to erase the transcription-enhancing chromatin remodeling already instigated by c-Myc . These results and further experiments in a vertebrate model organism reveal that SerRS is a key balancing antagonist of c-Myc for regulation of VEGFA expression , as well as for proper development of a functional vasculature . In addition , our study provides the first report of an anti-angiogenic function for SIRT2 , which arises at least in part through its interaction with SerRS .
To study the mechanism of how nuclear SerRS represses VEGFA expression , we started by using human cells . A short hairpin RNA ( shRNA ) targeting the 3′ UTR of human SerRS mRNA was generated to knock down endogenous SerRS expression in human umbilical vein endothelial cells ( HUVECs ) and HEK 293 cells ( Figure 1—figure supplement 1A , B ) . For both cell types , the level of VEGFA transcript was more than doubled in cells expressing the shRNA against SerRS ( sh-SerRS ) vs a control shRNA ( sh-Con ) ( Figure 1A , Figure 1—figure supplement 2 ) . Considering that SerRS is an essential component of the translation machinery and that knockdown of SerRS would have a general effect on protein synthesis that may obscure the effect on VEGFA expression , we compensated the ‘knockout’ cells by expression of NLS-deleted SerRS ( ΔNLS ) that is fully active in aminoacylation but lost the ability to enter the nucleus ( Xu et al . , 2012 ) , or by expression of wild-type ( WT ) SerRS ( as a separate control ) . Remarkably , compared to WT SerRS-expressing cells , cells expressing ΔNLS SerRS resulted in three and fourfold higher expression of VEGFA in HUVECs ( Figure 1A ) and HEK293 cells ( Figure 1—figure supplement 2 ) , respectively . This result supports the idea that the role of nuclear SerRS in suppressing VEGFA expression is conserved from fish to humans . In addition , as measured in an in vitro endothelial tube formation assay , and consistent with the role of VEGFA in promoting angiogenesis , HUVECs expressing ΔNLS SerRS showed a much stronger propensity ( than WT SerRS-expressing cells ) to form a blood vessel-like tubular network ( Figure 1B–D ) . 10 . 7554/eLife . 02349 . 003Figure 1 . Nuclear SerRS suppresses VEGFA expression and angiogenesis . ( A ) VEGFA mRNA levels as detected by real-time RT-qPCR in HUVECs infected with lentiviral plasmids expressing nonspecific control shRNA ( sh-Con ) , SerRS-specific shRNA ( sh-SerRS ) , or sh-SerRS and wild type ( WT ) or NLS-deleted ( ΔNLS ) SerRS simultaneously . Values are means ± SEM ( n = 3 ) . ( B ) Endothelial tube formation assay to show that excluding SerRS from the nucleus promotes angiogenesis . Values are means ± SEM ( n = 3 ) . ( C and D ) Representative images of the tubular network formed by HUVECs expressing WT and ΔNLS SerRS , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 00310 . 7554/eLife . 02349 . 004Figure 1—figure supplement 1 . Manipulations of the expression of SerRS in HUVEC and HEK 293 cells . HUVEC ( A ) and HEK 293 ( B ) cells were infected with lentiviruses expressing nonspecific control shRNA ( sh-Con ) , shRNA targeting 3′-UTR of the endogenous SerRS gene ( sh-SerRS ) , or sh-SerRS with WT or NLS-deleted ( ΔNLS ) SerRS simultaneously . The expression levels were measure 48 hr post-transfection by Western blot analyses . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 00410 . 7554/eLife . 02349 . 005Figure 1—figure supplement 2 . Nuclear SerRS suppresses VEGFA expression in HEK 293 cells . VEGFA mRNA levels were detected by real-time RT-PCR in HEK 293 cells expressing indicated molecules . Values are means ± SEM ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 005 Using cellulose beads-linked calf thymus DNA , we found that purified SerRS , but not two other human tRNA synthetases ( GlyRS and LysRS ) , bound to DNA ( Figure 2—figure supplement 1A ) . Given this inherent capacity of SerRS to bind DNA , we performed a chromatin immunoprecipitation ( ChIP ) experiment with 10 primer pairs designed to scan the human VEGFA gene , from 4 kb upstream ( −4 kb ) to 4 kb downstream ( +4 kb ) of the transcription start site ( Figure 2A , B ) . We found that ectopically expressed SerRS bound to the promoter in the region from −1 . 5 kb to +1 of the start site . Importantly , this region encompasses the binding site of c-Myc on the VEGFA promoter ( Figure 2B; Kim et al . , 2007 ) . 10 . 7554/eLife . 02349 . 006Figure 2 . Identification of SerRS and c-Myc binding sites on the VEGFA promoter . ( A ) Flow chart of consecutive methods used for determining the SerRS binding site . ( B ) Chromatin immunoprecipitation ( ChIP ) scanning assay to probe the SerRS and c-Myc binding sites . The promoter region of the VEGFA gene scanned by 10 amplicons is shown on the top . The amounts of DNA immunoprecipitated by anti-SerRS or anti-c-Myc antibodies or by control IgG from HEK 293 cell lysates were measured by real-time quantitative PCR at each amplicon . The results are represented as percentages of the total input of the chromatin DNA and shown as means ± SEM ( n = 3 ) . ( C ) Luciferase assay to confirm the repressive activity of SerRS and narrow down the SerRS binding site on the VEGFA promoter . Three different lengths of the VEGFA promoter were used to drive luciferase expressions in HEK 293 cells transfected with plasmids expressing SerRS , GlyRS or empty vector . The normalized luciferase activities are shown as mean ± SEM ( n = 3 ) . ( D ) In vitro DNase I footprint assay to identify the SerRS binding site . A 308-bp DNA fragment ( −262 ∼ +46 on the VEGFA promoter ) radiolabeled at the 3′ end was incubated with purified recombinant c-Myc/MAX ( 1:1 molar ratio ) , SerRS or GlyRS each at 1 or 5 µM , and then subjected to DNase I digestion . The regions protected by c-Myc/MAX and by SerRS are indicated in red and blue boxes , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 00610 . 7554/eLife . 02349 . 007Figure 2—figure supplement 1 . Identification of the interaction between SerRS and DNA . ( A ) Human SerRS , but not LysRS or GlyRS , can bind to DNA . Purified , His6-tagged recombinant human LysRS , GlyRS , and SerRS proteins were incubated with calf thymus genomic DNA linked cellulose beads or empty beads . After wash , the beads were examined with Western blot analysis using anti-His antibody to detect protein binding . ( B ) Background luciferase activities driven by three different lengths of the VEGFA promoter . The promoter regions were inserted in front of the firefly luciferase reporter gene . The constructs ( together with the Renilla luciferase control reporter vector for transfection efficiency control ) were transfected into HEK 293 cells . The luciferase activities were measured 24-hr post-transfection , and the results are shown as relative firefly luciferase activities after normalized to Renilla luciferase activities and as means ± SEM ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 007 To investigate whether the binding is repressive in nature , we performed a luciferase assay in which the majority of this promoter region ( −1262 ∼ +46 ) was put in front of a luciferase reporter gene to test the transcriptional activity of SerRS . Strikingly , SerRS overexpression sharply reduced the elicitation of luciferase activity ( Figure 2C , Figure 2—figure supplement 1B ) . In contrast , overexpression of GlyRS , which lacks DNA-binding capacity , had no effect ( Figure 2C ) . In further work , we showed that the inhibitory effect of SerRS persisted as the promoter region was shortened to −262 ∼ +46 , suggesting that the SerRS-specific responsive element is within this 308-bp region ( Figure 2C ) . This DNA fragment was then radiolabeled and subjected to DNase I footprint analysis to further determine the exact SerRS binding sites . As shown in Figure 2D , purified SerRS protected a 25-bp region ( −62 ∼ −38 ) from DNase I digestion and did so in a concentration dependent manner . The direct interaction of SerRS with a slightly extended 27-bp DNA fragment ( −62 ∼ −36 ) was confirmed by an electrophoretic mobility shift assay ( EMSA ) ( Figure 3A ) . The binding affinity ( Kd ) of SerRS was 211 . 5 nM as measured by EMSA ( Figure 3A , B ) and 265 nM by surface plasmon resonance ( Figure 3—figure supplement 1 ) . Interestingly , truncations from both ends of the 27-bp DNA fragment weakened the interaction and , based on the EMSA analysis , the DNA minimal binding site of SerRS was determined to be 21 nt ( −59 ∼ −38 ) ( Figure 3C , D ) . 10 . 7554/eLife . 02349 . 008Figure 3 . Characterization of the interaction between SerRS and DNA . ( A and B ) In vitro EMSA assay to determine the binding affinity between SerRS and the 27-bp DNA . The 27-bp DNA fragment containing SerRS binding site on the VEGFA promoter ( −62 ∼ −36 ) were labeled by 32P at the 5′ end , and then incubated with purified SerRS or GlyRS at indicated concentrations . The SerRS–DNA complex was followed by electrophoresis on native acrylamide gels . ( C and D ) EMSA to determine the minimal SerRS binding site on the VEGFA promoter . Truncations of the DNA from either end weakened the SerRS–DNA interaction . Purified recombinant SerRS protein was used at the indicated concentrations . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 00810 . 7554/eLife . 02349 . 009Figure 3—figure supplement 1 . Determination of the binding affinity between SerRS and DNA by SPR . The binding affinity of SerRS with the 27-bp DNA from the VEGFA promoter as measured by surface plasmon resonance ( SPR ) analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 009 As mentioned above , c-Myc plays a pivotal role in vascular development by promoting VEGFA expression ( Baudino et al . , 2002; Kokai et al . , 2009 ) . In complex with its partner MAX , c-Myc directly binds to classic or nonclassic E-box sequences on DNA ( Blackwell et al . , 1993; Kim et al . , 2008 ) and recruits histone acetyltransferase to allow chromatin expansion and activate transcription . However , the exact binding site of c-Myc/MAX on the VEGFA promoter has not been reported . Using DNase I footprint analysis , we also identified the exact c-Myc/MAX binding region on the VEGFA promoter . The c-Myc/MAX binding site ( −53 ∼ −38 ) contains a nonclassical E-box sequence ( −49CATGCG−44 ) that completely overlaps with the SerRS binding site ( Figure 2D ) . To investigate the sequence-specificity of SerRS in DNA binding and the importance of the E-box sequence for SerRS binding , we designed 11 single or double mutations in the 27 bp DNA , including 5 in the E-box sequence ( Figure 4A ) . Two double mutants of the E-box , including one ( −49CTTACG−44 ) that would completely abolish c-Myc/Max binding ( Blackwell et al . , 1993 ) , did not affect the SerRS interaction ( Figure 4A ) ; on the other hand , five different single mutations outside the E-box ( on both the 5′ and the 3′ sides ) that would not affect c-Myc/Max binding greatly weakened SerRS binding ( Figure 4A ) , indicating that SerRS and c-Myc/Max have distinct DNA binding specificities . 10 . 7554/eLife . 02349 . 010Figure 4 . Further characterization of the interaction between SerRS and DNA . ( A ) EMSA assay to probe the DNA sequence specificity for interacting with SerRS . DNA mutations that do or do not impact SerRS binding are colored in red and green , respectively . ( B ) Domain mapping analysis and EMSA assay to reveal multiple DNA binding sites on SerRS . TBD: tRNA binding domain; CD: catalytic domain; UNE-S: C-terminal appended domain unique to vertebrates . ( C ) Deletion mutagenesis to further define DNA binding sites on SerRS . Deletion of either insertion I , insertion II , motif V2-G14 , or loop T413-V420 greatly weakens or abolishes the DNA interaction . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 010 We also investigated the DNA binding sites on SerRS through domain mapping and deletion mutagenesis . SerRS functions as a dimer in aminoacylation and the dimerization interface is mediated through the catalytic domain ( CD ) . The N-terminal tRNA binding domain ( TBD ) of SerRS is used for recognizing the long variable arm of tRNASer , while the C-terminal UNE-S domain directs SerRS into the nucleus . Deletion of TBD or UNE-S dramatically weakens or completely abolishes the DNA interaction ( Figure 4B ) . In fact , only the intact SerRS can bind to DNA ( Figure 4B ) , suggesting that multiple domains of SerRS contribute to the DNA interaction . To further define the DNA binding sites on SerRS , we made additional deletion mutants of SerRS . Deletion of each of the two higher eukaryote-specific insertions in TBD and CD , respectively , which does not negatively impact tRNA binding ( Xu et al . , 2013 ) , dramatically weakens the DNA interaction ( Figure 4C ) . Two additional deletions—ΔV2-G14 in TBD and ΔT413-V420 in CD—also abolish the DNA binding ( Figure 4C ) . Based on these results and our previously solved crystal structures of human SerRS ( Xu et al . , 2012 , 2013 ) , we modeled the SerRS–DNA interaction . As shown in Video 1 , motif V2-G14 and loop T413-V420 , located next to each other in 3D space , bind to one end of the DNA , while insertion I in TBD ( G75-N97 ) binds to the other end; insertion II in CD ( G254-N261 ) , as well as the UNE-S domain , which is disordered in the crystal structure of SerRS , would interact with the middle region of the DNA near the E-box . 10 . 7554/eLife . 02349 . 011Video 1 . Model for SerRS–DNA interaction . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 011 The overlapping DNA binding sites of SerRS and c-Myc on the VEGFA promoter and their opposing roles on VEGFA expression , suggest that SerRS may compete with c-Myc for DNA binding and thus inhibit c-Myc-driven VEGFA expression . It is important to note that the binding affinity of the SerRS–DNA interaction ( Figure 3A , B , Figure 3—figure supplement 1 ) is comparable to that of the c-Myc/MAX–DNA interaction ( 90 . 5 ∼ 229 nM ) ( Hu et al . , 2005 ) . Indeed , in HEK 293 cells and measuring by real time qRT-PCR , ectopic expression of WT ( but not NLS-deleted ) SerRS repressed the overexpression of VEGFA driven by c-Myc ( Figure 5A ) . Consistently , our EMSA and Western blot analyses clearly showed that SerRS , at the same concentrations , could compete in vitro with the c-Myc/MAX complex for binding to the 27-bp DNA fragment from the VEGFA promoter ( Figure 5B , Figure 5—figure supplement 1 ) . We note that SerRS cannot compete with the MAX/MAX homodimer for binding to the DNA at comparable concentrations ( Figure 5B , Figure 5—figure supplement 1 ) , presumably because of the tight DNA binding affinity of the MAX/MAX homodimer ( 19 . 2 ∼ 48 . 7 nM ) ( Hu et al . , 2005 ) . 10 . 7554/eLife . 02349 . 012Figure 5 . Competition between SerRS and c-Myc for DNA binding and their opposing effect in vascular development . ( A ) Competition between c-Myc and SerRS on VEGFA expression . HEK 293 cells were transfected with c-Myc alone or c-Myc with WT or ΔNLS SerRS . The mRNA levels of VEGFA were determined by RT-PCR . Values are shown as means ± SEM ( n = 3 ) . ( B ) Competition between c-Myc/MAX and SerRS for DNA binding in vitro as examined by EMSA . The 27-bp DNA was radio-labeled and incubated with purified recombinant c-Myc/MAX together with purified recombinant SerRS at indicated concentrations . The protein–DNA complexes were followed by electrophoresis on a native acrylamide gel . ( C ) Competition between ectopically expressed SerRS and c-Myc for DNA binding on the VEGFA promoter in HEK 293 cells as examined by ChIP . HEK 293 cells were co-transfected with plasmids expressing c-Myc and WT or ΔNLS SerRS or empty vector ( − ) 24 hr prior to ChIP analysis . The amounts of DNA immunoprecipitated by anti-SerRS or anti-c-Myc antibodies or by control IgG from HEK 293 cell lysates were measured by PCR using a primer set targeting the VEGFA promoter . The normalized results ( top panel ) are represented as fold change of immunoprecipitated DNA by anti-SerRS vs anti-c-Myc and are shown as means ± SEM ( n = 3 , *p<0 . 001 , **p<0 . 05 ) . The bottom panel shows representative gel images . ( D ) Competition between endogenously expressed SerRS and c-Myc for DNA binding on the VEGFA promoter in HUVECs . HUVECs were infected to express the indicated molecules 48 hr prior to ChIP analysis . The same ChIP experiment and data analysis were performed as described in ( C ) . *p<0 . 0001 , **p<0 . 005 . ( E ) Opposing effect of SerRS and c-Myc in zebrafish vascular development and their mutual phenotypic rescue . The percentage of Tg ( Fli1a:GFP ) zebrafish embryos showing different ISV phenotypes at 3 days post fertilization after the injection of morpholinos targeting SerRS ( SerRS-MO ) , Myca ( Myca-MO ) , or a control morpholino ( Control-MO ) are illustrated . Scale bars represent 0 . 125 mm . *p<0 . 0001 vs Control-MO , **p<0 . 0001 vs Myca-MO , ***p<0 . 0001 vs SerRS-MO . Control-MO was added to SerRS-MO or Myca-MO experiments in order to maintain a constant level of total morpholinos in each experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01210 . 7554/eLife . 02349 . 013Figure 5—figure supplement 1 . SerRS and c-Myc/MAX do not simultaneously bind to the DNA . Purified recombinant SerRS ( 500 nM ) and c-Myc/MAX ( 500 nM each ) proteins were mixed as indicated and incubated with cold or 32P-labeled 27-bp DNA fragment on the VEGFA promoter ( −62 ∼ −36 ) . The mixtures were subjected to electrophoresis in 6% native polyacrylamide gel and the protein–DNA complexes were detected by autoradiography and Western blot analysis using anti-SerRS and anti-c-Myc antibodies , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01310 . 7554/eLife . 02349 . 014Figure 5—figure supplement 2 . Effect of knocking down Myca or SerRS on Vegfa expression in zebrafish . The effects of injecting the indicated morpholinos on Vegfa expression in zebrafish embryos ( 3 dpf ) as measured by real-time RT-PCR . Data are shown as means ± SEM ( n = 10–15 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01410 . 7554/eLife . 02349 . 015Figure 5—figure supplement 3 . Design and efficiency of the antisense morpholino against Myca . ( A ) The design of Myca-MO and the expected mRNA/protein products . The Myca-MO targets the intron1-exon2 splicing site of Myca precursor mRNA to potentially generate two alternatively spliced mRNA products ( SB ) that would be translated into non-functional proteins . ( B ) RT-PCR analysis confirming the dramatic decrease of WT Myca mature mRNA and the expected alternative splicing variants ( SB ) as the result of injecting Myca-MO . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 015 We also demonstrated that SerRS competes with c-Myc for binding to the VEGFA promoter in whole cells . ChIP analysis showed that ectopically expressed WT , but not NLS-deleted , SerRS could compete c-Myc off of the VEGFA promoter in HEK 293 cells ( Figure 5C ) . Consistently , knocking down endogenous SerRS expression in HUVEC cells resulted in a dramatic increase of endogenous c-Myc binding to the VEGFA promoter ( Figure 5D ) . This increase was completely reversed when the cells were compensated with ectopically expressed WT , but not NLS-deleted , SerRS ( Figure 5D ) . These results suggest that SerRS is a potent endogenous inhibitor of c-Myc for binding to the VEGFA promoter , and vice versa . Given the competition between c-Myc and SerRS for binding to the VEGFA promoter and the opposing activity of c-Myc and SerRS in regulating VEGFA expression , we postulated that knocking down c-Myc , although toxic on its own , may have a rescue effect towards the vasculature abnormality caused by a SerRS deficiency . This possibility was investigated in zebrafish as a vertebrate model system . As expected ( Fukui et al . , 2009 ) , knocking down SerRS by injection of an antisense morpholino ( SerRS-MO ) resulted in a hyper-intersegmental vessel ( ISV ) branching phenotype in zebrafish ( Figure 5E ) , and the phenotype was accompanied with an elevated level of Vefga expression ( Figure 5—figure supplement 2 ) . Specifically , out of 130 fish embryos injected with SerRS-MO , 72 ( 55 . 4% ) exhibited hyper-ISV phenotype , as oppose to 2 . 9% ( n = 4 out of 140 ) of fish injected with a control morpholino ( control-MO ) . Although a small number of fish injected with SerRS-MO exhibited the opposite vascular defect ( hypo-ISV phenotype ) , the number is not significantly different from that in the control-MO group ( Figure 5E ) . Overall , without SerRS , there is an over-expansion of the vasculature . In contrast , knocking down c-Myc by injecting an antisense morpholino against Myca ( c-Myc homologue in zebrafish ) showed under-developed vasculature , which is accompanied with a reduced level of Vefga expression ( Figure 5—figure supplement 2 ) . In particular , 41 . 4% ( n = 24 out of 58 ) of Myca-MO-injected morphants exhibits hypo-ISV phenotype , as opposed to 10% ( n = 14 out of 140 ) of control-MO-injected morphants ( Figure 5E , Figure 5—figure supplement 3 ) . Remarkably , co-injection of Myca-MO with SerRS-MO efficiently rescued both hyper-ISV ( 10 . 7% , n = 18 out of 169 ) and hypo-ISV ( 10 . 1% , n = 17 out of 169 ) defects ( Figure 5E ) . Coincidently , co-injection of Myca-MO also partially reversed the high Vegfa expression level in SerRS-MO-injected zebrafish ( Figure 5—figure supplement 2 ) . Therefore , a counteracting effect between c-Myc and SerRS in vascular development was confirmed in a vertebrate system . These results highlight a ‘Yin-Yang’ regulation of SerRS and c-Myc on VEGFA expression and demonstrate that a delicate balance between them is essential for developing a functional vasculature . A large scale protein–protein interaction study indicated a potential interaction between SerRS and sirtuin 2 ( SIRT2 ) ( Ewing et al . , 2007 ) , a NAD+-dependent histone deacetylase of the sirtuin family that regulates a broad range of processes , including transcription , metabolism , neurodegeneration , and aging ( Finkel et al . , 2009 ) . Of the seven mammalian sirtuin isoforms , relatively little is known about SIRT2 . Considering that c-Myc activates gene expression by recruiting partners harboring histone acetyltransferase activity that modifies histones and leads to open chromatin structures ( McMahon et al . , 1998 , 2000; Amati et al . , 2001 ) , we postulated that an interaction between SerRS and SIRT2 might reverse this process to attenuate VEGFA expression . If this were true , then by implementing opposing deacetylase/acetyltransferase activities , the ‘Yin-Yang’ relationship between SerRS and c-Myc would also act at the level of chromatin modification . To test this hypothesis , we firstly performed coimmunoprecipitation to confirm the interaction . The ectopically expressed SIRT2 , but not SIRT1 , effectively pulled down SerRS , and vice versa ( Figure 6A ) . The interaction between endogenous SerRS and SIRT2 was also confirmed by coimmunoprecipitation ( Figure 6—figure supplement 1 ) . Moreover , a purified GST-SerRS fusion protein successfully pulled down purified SIRT2 , thus showing that the protein–protein interaction is direct ( Figure 6B ) . Further mapping analysis showed that the catalytic domain ( CD ) of SerRS is responsible for the interaction with SIRT2 ( Figure 6B ) . 10 . 7554/eLife . 02349 . 016Figure 6 . Demonstration and characterization of SerRS/SIRT2 interaction . ( A ) SerRS specifically interacts with SIRT2 but not SIRT1 . HEK 293 cells were co-transfected with plasmids expressing Flag-tagged SerRS and V5-tagged SIRT1 or SIRT2 . Cell lysate was immunoprecipitated with anti-V5 ( top panel ) , anti-Flag ( bottom panel ) antibodies or control IgG . The experiment was followed by Western blot analysis to detect the interaction between SerRS and SIRT1/SIRT2 using anti-Flag and anti-V5 antibodies . ( B ) GST-pull down assay to show that SerRS/SIRT2 interaction is direct and that the interaction is mediated by the catalytic domain of SerRS . Full-length SerRS or its domain fragments were fused with GST at N-termini to pull down purified His-tagged SIRT2 . SIRT2 was detected by Western blot analysis using anti-His6 antibody , and the GST fusion proteins attached on the Glutathione-Sepharose beads were analyzed using ponceau S staining . TBD: tRNA-binding domain; CD: catalytic domain; UNE-S: C-terminal appended domain . ( C ) Mapping study to identify the SerRS binding sites on SIRT2 . V5-tagged full-length SIRT2 or its truncated fragments was co-transfected with Flag-tagged SerRS into HEK 293 cells . SIRT2 proteins were immunoprecipitated with anti-V5 antibody and the SIRT2-bound SerRS proteins were detected by Western blot using anti-Flag antibody . ( D ) Illustration of the SerRS binding sites on the structure of SIRT2 . Two SerRS binding sites ( Gly52-Asp60 , Trp337-Ser356 ) are highlighted in red . The catalytic domain of SIRT2 is in green , while the partially disordered N- and C-terminal regions are in yellow and purple , respectively . The gray dash line represents a disordered internal region . ( E ) Effect of SerRS on SIRT2 deacetylation activity . Recombinant human SIRT2 ( 1 µM ) were incubated with purified SerRS ( concentration measured as monomer ) at the indicated ratios . The deacetylase activities of SIRT2 were measured by using a substrate peptide with one end coupled to a fluorophore and the other end to a quencher . An internal acetylated lysine residue serves as the substrate of SIRT2 , and the deacetylation allows the peptide to be cleaved by a lysylendopeptidase to release the fluorophore from the quencher to emit fluorescence . Therefore , the SIRT2 acitivity was measured by monitoring the fluororescence intensity ( excitation at 490 nm and emission at 530 nm ) . A reaction without NAD+ ( NAD+ [−] ) was performed as a negative control . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01610 . 7554/eLife . 02349 . 017Figure 6—figure supplement 1 . Endogenous interaction between SerRS and SIRT2 . Approximately 107 HEK 293 cells were lysed and the supernatant were immunoprecipicated with mouse monoclonal anti-SerRS antibody or the control mouse IgG . The co-immunoprecipitated proteins were analyzed by Western blot analysis using anti-SerRS and anti-SIRT2 antibodies , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01710 . 7554/eLife . 02349 . 018Figure 6—figure supplement 2 . SerRS recruits SIRT2 to the VEGFA promoter . ChIP assay to detect the impact of SerRS and GlyRS ( as a control ) on SIRT2 binding to the VEGFA promoter . HEK 293 cells were transfected with plasmids expressing the indicated shRNA for knocking down the expression of SerRS or GlyRS and subjected to ChIP analysis 48 hr post-transfection . The normalized results are represented as fold change of immunoprecipitated DNA by anti-SerRS vs anti-SIRT2 and are shown as means ± SEM ( n = 3 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 018 To test whether SerRS can recruit SIRT2 to the VEGFA promoter , we knocked down the expression of SerRS in HEK 293 cells and detected the binding of SIRT2 to the VEGFA promoter . Knockdown of SerRS , but not GlyRS , significantly reduced the amount to SIRT2 bound to the VEGFA promoter ( Figure 6—figure supplement 2 ) , demonstrating that SerRS specifically recruits SIRT2 to the VEGFA promoter . If SIRT2 were to be recruited by SerRS to reverse histone acetylation , their interaction should not negatively affect the enzymatic activity of the deacetylase . To address this question , we first mapped the SerRS binding site on SIRT2 by coimmunoprecipitation ( Figure 6C ) . SIRT2 is a 389-aa protein , whose crystal structure has been solved ( Finnin et al . , 2001 ) . The catalytic core of SIRT2 is flanked by ∼60 aa and ∼50 aa on the N- and C-terminal ends , respectively . Through a series of truncations from each end , we determined that both N- and C-terminal regions , outside the catalytic core of SIRT2 , interact with SerRS ( Figure 6C ) . Particularly , residues within the regions of G52-D60 and W337-S356 are critical , as the loss of either region abolished the interaction of SIRT2 with SerRS . Interestingly , both regions are located on the opposite side of the substrate-binding pocket ( Figure 6D ) , suggesting that bound SerRS should not interfere with the deacetylase activity of SIRT2 . To confirm the postulation , we directly assessed the effect of SerRS on the in vitro deacetylase activity of SIRT2 . Remarkably , at an equal molar ratio ( 1:1 ) , SerRS did not inhibit but rather promoted the deacetylase activity of SIRT2 ( Figure 6E ) . Because increasing the SerRS concentration to 1:2 ratio ( SIRT2: SerRS ) did not provide any additional enhancement ( but rather a small decline at early time points ) , the enhancement effect is likely to result from the specific interaction between SerRS and SIRT2 , at a location distal to the active site . Next , we tested whether SerRS can recruit SIRT2 to modify the histone modification on the VEGFA promoter . Because the evolutionally conserved deacetylase activity of SIRT2 has a strong preference for K16 of histone H4 ( Vaquero et al . , 2006 ) , we performed ChIP analyses using antibodies against acetylated H4K16 ( H4K16Ac ) . Remarkably , overexpression of SerRS , but not of GlyRS , reduced the level of H4K16Ac on the VEGFA promoter ( Figure 7A ) . Consistently , knocking down endogenous SerRS , but not endogenous GlyRS , had the opposite effect and significantly increased H4K16Ac on the VEGFA promoter ( Figure 7B , Figure 7—figure supplement 1A ) . The increase was reversed when the cells were compensated with WT , but not NLS-deleted , SerRS ( Figure 7B ) . These results demonstrated that nuclear SerRS acts to decrease the amount of acetylated H4 on the VEGFA promoter . Given the interaction between SerRS and SIRT2 , this effect on H4 acetylation is presumably through engagement of SIRT2 by SerRS . 10 . 7554/eLife . 02349 . 019Figure 7 . SerRS recruits SIRT2 to epigenetically silence VEGFA expression . ( A ) ChIP assay to show that overexpression of SerRS reduces histone H4 acetylation level on the VEGFA promoter . HEK 293 cells were transfected with plasmids expressing SerRS , GlyRS or empty vector . The cell lysates were subjected to local ChIP analysis using anti-H4K16Ac ( acetylated H4 at K16 ) , anti-H4 ( total ) , or anti-SerRS antibodies and a primer set targeting the VEGFA promoter . The amounts of DNA immunoprecipitated by anti-H4K16Ac antibody were normalized to those by anti-H4 antibody prior to fold change calculation . Inset: the normalized amounts of DNA immunoprecipitated by anti-SerRS . All data were shown as means ± SEM ( n = 3 ) . ( B ) ChIP assay to show that knock down of SerRS expression or exclusion of SerRS from the nucleus increases histone H4 acetylation level on the VEGFA promoter . HEK 293 cells were transfected with plasmids expressing the indicated molecules and subjected to local ChIP analysis as described above . As a control , GlyRS expression was knocked down but had no effect on H4 acetylation . ( C ) Effect of SIRT2 expression on the transcriptional repressor activity of SerRS as measured by VEGFA expression . HEK 293 cells were co-transfected with plasmids expressing shRNAs targeting SIRT1 , SIRT2 or control shRNA and plasmids expressing SerRS , GlyRS or empty vector for 36 hr . The VEGFA expression levels were determined by using real-time RT-qPCR and are shown as means ± SEM ( n = 3 ) . ( D ) Effect of SIRT2-specific inhibitor on the transcriptional repressor activity of SerRS as measured by VEGFA expression . HEK 293 cells were transfected with plasmids expressing SerRS , GlyRS or empty vector . SIRT2-specific inhibitor AGK2 ( 10 µM , final concentration ) or SIRT1-specific inhibitor EX-527 ( 1 µm , final concentration ) or solvent alone ( DMSO ) was added to the cell culture media 2 hr post-transfection . VEGFA expression levels were measured 24 hr post-transfection by using real-time RT-qPCR and are shown as means ± SEM ( n = 3 ) . ( E ) Functional correlation between SerRS and SIRT2 in zebrafish . The percentage of Tg ( Fli1a:EGFP ) zebrafish embryos showing different ISV phenotypes at 3 days post fertilization after the injection of morpholinos targeting SerRS ( SerRS-MO ) , Sirt1 ( Sirt1-MO ) , Sirt2 ( Sirt2-MO ) , or a control morpholino ( Control-MO ) are illustrated . Scale bars represent 0 . 25 mm . *p<0 . 0001 vs Control-MO , **p>0 . 1 vs Control-MO . ( F ) The effects of knocking down SerRS , Sirt2 , or Sirt1 in zebrafish on Vegfa expression were examined by real-time RT-qPCR at 1 day post fertilization after injection of morpholinos as indicated . Data are shown as means ± SEM ( n = 10–15 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 01910 . 7554/eLife . 02349 . 020Figure 7—figure supplement 1 . Knock-down efficiencies of shRNAs targeting GlyRS , SerRS , SIRT1 , and SIRT2 . ( A ) Western blot analysis to confirm the efficiency of shRNAs in knocking down endogenous GlyRS ( sh-GlyRS ) or SerRS ( sh-SerRS ) expression in HEK 293 cells . ( B ) Western blot analysis to confirm the efficiency of shRNAs in knocking down endogenous SIRT1 ( sh-SIRT1 ) or SIRT2 ( sh-SIRT2 ) expression in HEK 293 cells . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 02010 . 7554/eLife . 02349 . 021Figure 7—figure supplement 2 . Design and efficiency of the antisense morpholino against Sirt2 . ( A ) The design of Sirt2-MO and the expected mRNA/protein products . The Sirt2-MO targets the exon4-intron4 splicing site of Sirt2 precursor mRNA to potentially generate two alternatively spliced mRNA products ( SB ) that would be translated into non-functional proteins . ( B ) RT-PCR analysis confirming the dramatic decrease of WT Sirt2 mature mRNA and the expected alternative splicing variants ( SB ) as the result of injecting Sirt2-MO . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 021 To confirm that SIRT2 is a necessary cofactor for SerRS to repress VEGFA expression , we disrupted SIRT2 by RNAi . Indeed , knocking down the expression of SIRT2 , but not of SIRT1 , completely reversed the transcriptional repression activity of SerRS on VEGFA expression ( Figure 7C , Figure 7—figure supplement 1B ) . Consistently , inhibiting SIRT2 activity by AGK2 , a SIRT2-specific inhibitor ( Outeiro et al . , 2007 ) , also completely knocked out the transcriptional repression activity of SerRS , while EX-527 , a SIRT1 inhibitor ( Solomon et al . , 2006 ) , had little effect ( Figure 7D ) . Therefore , we have demonstrated that , by recruiting SIRT2 vs a histone acetyltransferase , the ‘Yin-Yang’ relationship between SerRS and c-Myc also acts at the level of chromatin modification . While SIRT1 has been shown to promote angiogenesis ( Potente et al . , 2007 ) , the role of SIRT2 in vascular development has not been clear . The critical role of SIRT2 in the mechanism of SerRS to inhibit VEGFA expression and vascular expansion suggests that SIRT2 could be associated with an anti-angiogenic function , and that knocking down SIRT2 may mimic the vasculature abnormality phenotype caused by a SerRS knockdown . We tested this hypothesis in the zebrafish system . Consistent with the previous report showing a VEGFA-independent pro-angiogenic role for SIRT1 in vascular development ( Potente et al . , 2007 ) , injection of an antisense morpholino against Sirt1 ( SIRT1 homologue in zebrafish ) generated the hypo-ISV phenotype ( Figure 7E ) and had no significant effect on Vegfa expression ( Figure 7F ) . In contrast , injection of a Sirt2-MO resulted in the same hyper-ISV phenotype ( 42 . 1% , n = 75 out 178 ) as with the injection of SerRS-MO ( 46 . 7% , n = 119 out of 255 ) ( Figure 7E , Figure 7—figure supplement 2A , B ) . Furthermore , the hyper-ISV phenotype , in both cases , was accompanied by a significantly elevated level of the vegfa transcript ( Figure 7F ) . Therefore , our studies have revealed an anti-angiogenic role of SIRT2 , which should be , at least in part , dependent on attenuating VEGFA expression through its interaction with SerRS .
Through in vitro , cell-based and animal experiments , we established that the essential role of SerRS in vascular development arises from its novel activities as a transcriptional repressor of VEGFA . There are two different aspects of this activity: first , SerRS directly binds to the VEGFA promoter; second , DNA-bound SerRS recruits the SIRT2 histone deacetylase to condense the chromatin at the VEGFA promoter , and thereby shut down the gene transcription . Importantly , in each aspect , these actions of SerRS directly compete with and thwart that of the VEGFA-promoting actions of c-Myc . While the opposing regulation of SerRS and c-Myc is applied on VEGFA expression , it is manifested at the organism level with respect to vascular development , making SerRS and c-Myc as a pair of ‘Yin-Yang’ regulators for proper development of a functional vasculature ( Figure 8 ) . 10 . 7554/eLife . 02349 . 022Figure 8 . The ‘Yin-Yang’ relationship of SerRS and c-Myc in vascular development . Nuclear SerRS binds to the VEGFA promoter at the identified SerRS binding site ( SBS ) and recruits the SIRT2 histone deacetylase to condense the local chromatin to shut down VEGFA expression . These tandem actions of SerRS symmetrically offset the VEGFA-promoting actions of c-Myc to maintain a delicate balance for the development of a functional vasculature . DOI: http://dx . doi . org/10 . 7554/eLife . 02349 . 022 With the same bHLHZ domain as in c-Myc and MAX , Mad family proteins ( comprised of Mad1 , Mxi1 , Mad3 , and Mad4 ) can compete with c-Myc for binding to MAX and also recruit histone deacetylases to reverse the action of c-Myc-bound acetyltransferase to shut down the expression of c-Myc target genes ( Grandori et al . , 2000 ) . Therefore , Mad proteins are generally considered as antagonists of c-Myc , especially with regard to the role of c-Myc in tumorigenesis ( Zhou and Hurlin , 2001 ) . However , the role of Mad proteins in vascular development appears to be non-essential . Disruption of members of the Mad family in mice does not exhibit any vascular phenotype ( Foley et al . , 1998; Schreiber-Agus et al . , 1998 ) . In addition , the temporal expression pattern varies between c-Myc and the Mad family proteins , with c-Myc being expressed during development , while expression of Mad proteins is mainly induced during terminal differentiation ( Queva et al . , 1998 ) . Thus , the Mad proteins are largely silent during the time that the vasculature is being established , which could explain their non-essential role in vascular development . In comparison , SerRS' ubiquitous expression as an essential tRNA synthetase could match better with that of c-Myc to provide the counterbalance . From an evolutionary perspective , a closed circulatory system with vasculature network is one of the hallmarks of vertebrates . Although c-Myc plays a key role in vascular development , the gene first appeared in Drosphila , an invertebrate with a primitive ‘open’ circulatory system . The appearance of the MAX and Mad family genes is even earlier , and the genes were first identified in roundworms such as C . elegans ( Atchley and Fitch , 1995; Prendergast , 1999 ) . In contrast , although SerRS is considered as one of the most ancient proteins , its UNE-S domain , which harbors the NLS signal to endow SerRS with its novel transcription repressor activity , only appeared in vertebrates . Thus , it seems that an ‘old’ tRNA synthetase evolved to function as a ‘new’ and essential antagonist against c-Myc for proper development of the advanced closed circulatory system of vertebrates . A critical component of the c-Myc-antagonizing role of SerRS is the recruitment of SIRT2 . In fact , a complete reversal of the inhibitory effect of SerRS on VEGFA expression was observed when SIRT2 was knocked out by RNAi or inhibited by EX527 ( Figure 7C , D ) . This observation indicates that the transcriptional repressor role of SerRS is ultimately through SIRT2 and that the direct blocking of c-Myc from the promoter by SerRS may have lesser significance for downregulating gene expression . Thus , an overlapping DNA binding site may not be a prerequisite for SerRS to antagonize c-Myc , as long as both SerRS and c-Myc bind to the same promoter . Considering the large number of genes that are regulated by c-Myc , one would not be surprised to find additional genes to be transcriptionally repressed by SerRS , presumably also through its collaboration with SIRT2 . It is worth noting that SIRT2 has been identified as a tumor suppressor ( Hiratsuka et al . , 2003; Lennerz et al . , 2005 ) , while c-Myc is a prominent oncogene that promotes tumor cell proliferation and tumor vascularization ( Baudino et al . , 2002 ) . We speculate that SerRS also functions as a tumor suppressor by collaborating with SIRT2 to antagonize c-Myc . Interestingly , human SARS is located on the short arm of chromosome 1 ( i . e . , 1p13 . 3 ) , which is frequently affected by rearrangements or allelic loss in a variety of human malignancies ( Morgan et al . , 1985; Mitchell and Santibanez-Koref , 1990; Mathew et al . , 1994; Munn et al . , 1995; Nagai et al . , 1995; Xu et al . , 2001; Caramazza et al . , 2009 ) . The frequent disruption of this chromosomal locus in human malignancies suggests the presence of tumor suppressor genes which , when perturbed , lead to increased cancer susceptibility . Because SerRS is also essential for survival through its function in protein synthesis , if it is a factor in any of these malignancies , then its function in translation would need to be preserved . This could be achieved by disruption of the UNE-S domain , which does not affect aminoacylation , but is essential for SerRS to antagonize c-Myc .
Custom-made rabbit anti-human SerRS antibody was raised against purified human recombinant SerRS and affinity-purified . Monoclonal anti-SerRS antibody for coimmunoprecipitation was purchased from Abnova ( Taipei , Taiwan ) . Anti-c-Myc , anti-SIRT2 , anti-SIRT1 , and anti-α-tubulin antibodies were purchased from Cell Signaling ( Danvers , MA , USA ) . Anti-V5 and anti-GlyRS antibodies were purchased from Invitrogen ( Grand Island , NY , USA ) and Abnova ( Walnut , CA , USA ) , respectively . Antibodies against histone H4 and acetylated H4 at Lys 16 ( H4K16Ac ) were purchased from Active Motif ( Carlsbad , CA , USA ) . Anti-SIRT2 antibody for chromatin immunoprecipitation was purchased from Thermo Fisher ( Rockford , IL , USA ) . For overexpressions in mammalian cells , human full-length , and NLS-deleted SerRS genes were cloned into the pFlag-CMV-2 vector ( Sigma-Aldrich , St . Louis , MD , USA ) , and human c-Myc , SIRT1 , and SIRT2 genes into the pCDNA6-V5/His-C vector ( Life Technologies , Grand Island , NY , USA ) . For recombinant protein purification , human SerRS , c-Myc , MAX , and SIRT2 genes were subcloned into pET-20b ( + ) vector ( Novagen , Darmstadt , Germany ) to express with a C-terminal his-tag in Escherichia coli . The SerRS proteins were purified in tandem by Ni-NTA affinity ( Qiagen , Valencia , CA , USA ) , HiTrap Heparin High Performance ( GE Healthcare , Pittsburgh , PA , USA ) , and HiLoad 16/600 Superdex 200 pg ( GE Healthcare ) columns . The GST-tagged SerRS constructs were subcloned into the pGEX-6P-1 vector ( GE Healthcare ) for expression in E . coli , and the proteins were affinity-purified using Glutathione-Sepharose 4B beads ( GE Healthcare ) . The purities of the recombinant proteins were assessed by Coomassie blue staining following 4–12% Mini Gel ( Life Technologies , Grand Island , NY , USA ) electrophoresis . Protein concentrations were determined using Bradford protein assay ( BioRad , Hercules , CA , USA ) . HEK 293 cells were cultured in DMEM supplemented with 10% fetal calf serum ( FCS ) and transfected with Lipofectamine 2000 ( Life Technologies , Grand Island , NY , USA ) . HUVEC cell were cultured in EGM complete medium ( Lonza , Allendale , NJ , USA ) supplemented with 8% FCS in gelatin-coated dishes and transfected using lentivirus . DNA expressing a short-hairpin RNA ( shRNA ) designed against human SerRS ( 5′-GGCATAGGGACCCATCATTGA-3′ ) , GlyRS ( 5′-GCATGGAGTATCTCACAAAGT-3′ ) , SIRT1 ( 5′-GAAGTTGACCTCCTCATTGTT-3′ ) ( Guarani et al . , 2011 ) , or SIRT2 ( 5′-GGACAACAGAGAGGGAGAAAC-3′ ) gene was inserted into the pLentiLox-hH1 plasmid , modified from the pLentiLox 3 . 7 plasmid to contain a H1 promoter ( between Xba I and Xho I sites ) to drive the shRNA expression . To compensate for the loss of endogenous SerRS expression , the coding region for GFP in the pLentiLox-hH1 plasmid was replaced with NLS-deleted or WT ( as control ) SerRS coding sequences . All designed shRNAs target sequences within the open reading frame except for the SerRS shRNA , which targets the 3′ untranslated region in ordered to selectively knockdown the endogenous gene but not the exogenous genes . The recombinant lentiviruses were produced in packaging 293 cells by cotransfecting the pLentiLox-hH1 plasmid with two helper packaging plasmids Δ8 . 9 and VSVG and subsequently concentrated by centrifugation at 50 , 000×g for 3 hr . Total RNA was isolated from cells by TRIzol Reagent ( Life Technologies , Grand Island , NY , USA ) . One milligram of the total RNA from each sample was reversely transcribed to cDNA by SuperScript II reverse transcriptase ( Life Technologies , Grand Island , NY , USA ) . All real-time PCR reactions were performed using the StepOnePlus Real-Time PCR system ( Applied Biosystems , Grand Island , NY , USA ) with SYBR Select Master Mix ( Applied Biosystems , Grand Island , NY , USA ) . The primer pairs for the PCR reactions were: 5′-GAGGGCAGAATCATCACGAAG-3′ and 5′-TGTGCTGTAGGAAGCTCATCTCTC-3′ for human VEGFA; 5′-CGTCACCAACTGGGACGA-3′ and 5′-ATGGGGGAGGGCATACC-3′ for human β-ACTIN; 5′-GGCTCTCCTCCATCTGTCTGC-3′ and 5′-CAGTGGTTTTCTTTCTTTGCTTTG-3′ for zebrafish vegfa ; 5′-TCACCACCACAGCCGAAAGAG-3′ and 5′-GTCAGCAATGCCAGGGTACAT-3′ for zebrafish β-actin . The PCR reaction program started at 95°C for 10 min , followed by 45 cycles of 95°C for 20 s and 60°C for 1 min . Each experiment was carried out in triplicate . The VEGFA gene expression was normalized to that of β-ACTIN . Statistical analyses were performed with the software SigmaPlot ( version 10 . 0 ) . Student's t test was used to analyze the changes between different groups . 48 hr before the tube formation assay , HUVEC cells were infected with lentiviruses that produce different shRNAs as indicated . Pre-thawed matrigel basement membrane matrix ( 0 . 15 ml ) ( BD Biosciences , San Jose , CA , USA ) was transferred to 48-well plates and incubated at 37°C for 30 min to form a thin layer of gel . The infected HUVEC cells ( 2 × 104 ) were seeded on the gel and then cultured in EBM Basal Medium ( without FCS ) at 37°C and 5% CO2 for 24 hr to form tubes . Images of the endothelia cell tubular network were taken with a Leica DC350F CCD camera attached to an inverted Leica DMIL microscope . The length of the tubes was measured by ImageJ software . Cells were fixed with formaldehyde ( 1% final concentration ) for 10 min at room temperature . The reaction was stopped by adding 125 mM of glycine . ChIP assays were performed according to the protocol of ChIP-IT Express Enzymatic kit ( Active Motif ) . After three washes , ChIPed DNA was analyzed on the StepOnePlus Real-Time PCR system using SYBR Select Master Mix ( Applied Biosystems ) . A primer set ( 5′-GGGCGGATGGGTAATTTTCA-3′ and 5′-CTGCGGACGCCCAGTGAA-3′ ) targeting the VEGFA gene near and upstream of the transcriptional start site was used . Nine additional primer sets for scanning the VEGFA promoter from −4 kb to +4 kb were described previously ( Kim et al . , 2007 ) . The luciferase activity was determined by using the Dual-Luciferase Reporter assay system ( Promega , Madison , WI , USA ) . The promoter regions ( −1262 ∼ +46; −762 ∼ +46; −262 ∼ +46 ) of VEGFA gene were PCR amplified and cloned into the pGL4 . 11[luc2P] vector ( between Kpn I and Xho I sites ) to create the pGL4-VEGFA firefly luciferase reporter plasmids . After 16 hr of incubation in 12-well plates , HEK 293 cells were transiently transfected with 100 ng of pGL4-VEGFA reporter plasmid and 500 ng of pFlag-SerRS , or pFlag-GlyRS or pFlag-CMV-2 empty vector as control . A Renilla luciferase control reporter plasmid pRL-SV40 ( 50 ng ) was co-transfected for normalizing the transfection efficiency among different experiments . The DNA of the VEGFA promoter from −262 to +46 bp was released from the pGL4-VEGFA plasmid by Kpn I and Xho I digestion . After purification by agarose gel electrophoresis , the 3′ end was radiolabeled using a standard Klenow fragment fill-in reaction with [α-32P]-dATP . The labeled DNA fragment was incubated with recombinant SerRS , c-Myc and MAX , or GlyRS in 20 µl binding buffer ( 20 mM HEPES pH 7 . 9 , 120 mM KCl , 8 mM MgCl2 , 0 . 2 mM EDTA , 0 . 5 mM DTT , 0 . 2 mg ml−1 bovine serum albumin [BSA] , 10 µg ml−1 poly [dG-dC] , and 5% glycerol ) for 1 hr at room temperature . DNase I ( New England Biolabs , Ipswich , MA , USA ) was then added to the mixture at a 2 . 5 U ml−1 final concentration and incubated for additional 40 min at room temperature . The reaction was stopped by adding 200 µl stop solution ( 20 mM Tris–HCl pH 7 . 5 , 0 . 1 M NaCl , 1% [wt/vol] SDS , 5 mM EDTA , and 50 µg ml−1 protease K ) to incubate for 30 min at 45°C . After extraction with phenol-chloroform and precipitation with ethanol , DNA fragments were resuspended in 80% formamide in 1x TE buffer and then denatured for 5 min at 95°C before separation by electrophoresis using 8% urea-polyacrylamide sequencing gels . Gels were dried and examined by autoradiography . The 27-bp DNA oligonucleotide corresponding to SerRS binding site on the VEGFA promoter and mutants were synthesized , annealed , and [32P]-labeled at the 5′ end by T4 DNA kinase ( New England Biolabs ) before purification using a sephadex G-25 spin column ( GE Healthcare ) . The labeled oligonucleotides ( 0 . 08 pmol ) were incubated with recombinant SerRS at indicated concentrations in binding buffer ( 20 mM Tris-HCl , pH 8 . 0 , 60 mM KCl , 5 mM MgCl2 , 0 . 1 mg ml−1 BSA , 10 ng µl−1 poly ( dG-dC ) , 1 mM DTT ) for 30 min at room temperature . The samples were loaded to 5% native polyacrylamide gel ( 17 . 5 cm in length ) and underwent electrophoresis at 250 V in running buffer ( 25 mM Tris , pH 8 . 3 , 190 mM glycine ) . Afterwards , the gel was dried and examined by autoradiography . HEK 293 cells were resuspended on ice with lysis buffer ( 20 mM Tris-HCl [pH 7 . 5] , 150 mM NaCl , 1 mM of EDTA , 1 mM EGTA , 1% Triton X-100 , 2 . 5 mM sodium pyrophosphate , 1 mM beta-glycerophosphate , 1 mM Na3VO4 , and protease inhibitor cocktail ) . Supernatants were incubated with indicated antibodies and protein-G-conjugated agarose beads ( Invitrogen ) for at least 2 hr . The beads were washed five times with wash buffer ( same as the lysis buffer , except that Triton X-100 was reduced from 1% to 0 . 1% ) and then subjected to SDS-PAGE and Western blotting analysis with indicated antibodies . GST pull-down assays were performed in the buffer containing 20 mM of HEPES ( pH 7 . 9 ) , 150 mM of NaCl , 0 . 5 mM of EDTA , 10% glycerol , 0 . 1% Triton X-100 and 1 mM of DTT . Equal amounts of GST or GST-SerRS fusion proteins were incubated with recombinant SIRT2 for two hours and pulled down by Glutathione-Sepharose 4B beads ( GE Healthcare ) . The deacetylase activity of SIRT2 was measured by using CycLex SIRT2 Deacetylase Fluorometric Assay Kit ( CycLex , Nagano , Japan ) . The reaction buffer contains 50 mM Tris-HCl ( pH 8 . 8 ) , 0 . 5 mM DTT , 0 . 25 mAU/ml Lysylendopeptidase , 1 µM Trichostatin A , 0 . 8 mM NAD+ , 20 µM Fluoro-Substrate peptide , 1 µM recombinant SIRT2 , and recombinant SerRS at different ratio indicated in Figure 6E . The reactions were performed at room temperature . The fluorescence intensities at 10-min intervals were read using FluoroMax-3 ( Jobin Yvon Inc , Edison , NJ , USA ) with excitation at 490 ± 10 nm and emission at 530 ± 10 nm . Biotin-labeled double-stranded DNA oligonucleotides corresponding to SerRS binding site on the VEGFA promoter ( 5'-GGCGGG GCGGAGCCATGCGCCCCCCCCTTTATA-biotin-3′ and 5'-AAAGGGGGGGGCGCATGGCTCCGCCCCGCC-3′ ) were synthesized , annealed , and purified by electrophoresis on 8% native acrylamide gel . Binding kinetics was analyzed using a Biacore 3000 instrument ( Biacore , Inc . , Piscataway , NJ , USA ) . The DNA was immobilized through biotin–streptavidin interaction on a SA sensor chip , and the interaction reached to 300 Response Unit . A flow cell without immobilized DNA was used as a blank reference control . The immobilized DNA was stable over the course of the experiment . Baseline drift was less than 5 RU/h after the chip was washed with HBS-EP buffer ( 0 . 01 M HEPES pH 7 . 4 , 0 . 15 M NaCl , and 0 . 005% surfactant P-20 [vol/vol] ) at 10 μl min−1 for 18 hr . SerRS proteins were injected using the KINJECT procedure for 300 s at 6 . 25 , 12 . 5 , 25 , 50 , 100 , 200 , 400 , 800 , and 1600 nM concentrations in HBS-EP buffer at 30 μl min−1 . Dissociation was monitored by flowing HBS-EP buffer for 480 s at 30 μl min−1 . The sensor chip was regenerated by a 60-s injection of 0 . 1% SDS , 10 mM NaOH to restore the original resonance signal of the surface . The injections were duplicated for each ligand concentration and were performed in random orders with buffer blanks injected periodically for double referencing . The variation between the replicates was less than 1% . Corrected response data were fitted with BiaevalTM 3 . 1 software and the apparent kinetic constants were calculated using data from the early parts of the association and dissociation phases . The fit was satisfactory for a simple 1:1 binding model . Transgenic Tg ( Fli1a: EGFP ) fish were maintained at 28 . 5°C under continuous water flow and filtration with automatic control for a 14:10 hr light/dark cycle . The night before injection , male and female fish were placed in a 1-L tank containing fish mating cage with an inner mesh and divider . Zebrafish embryos were obtained from natural spawning by removing the divider and stimulating with light . The embryos were kept at 28 . 5°C before and after microinjection . The antisense morpholinos ( MOs ) targeting SerRS or other genes were injected into the yolk of 1- to 2-cell stage embryos at the dosage of 4 ∼ 5 ng per embryo . The designs of SerRS-MO ( 5′-AGGAGAATGTGAACAAACCTGACAC-3′ ) ( Fukui et al . , 2009 ) and of Sirt1-MO ( 5′- TATTTTCGCCGTCCGCCATCTTCGC-3′ ) have been described previously ( Potente et al . , 2007 ) . The Myca-MO ( 5′-CATTTTGACACTTGAGGAAGGAGAT-3′ ) and Sirt2-MO ( 5′-CATCTGAGCAGAAACTCACATTTGC-3′ ) were designed de novo for this study . All MOs including a standard control MO ( 5'-CCTCTTACC TCAGTTACAATTTATA-3' ) were purchased from Gene Tools , LLC ( Philomath , OR , USA ) . After injection , embryos were incubated in E3 embryo medium supplemented with 0 . 003% 1-phenyl-2-thiourea ( PTU ) at 28 . 5°C to prevent pigment formation . Embryos were anesthetized with 0 . 168 mg ml−1 tricaine ( Sigma-Aldrich ) , mounted in 2% methylcellulose and photographed with a Nikon fluorescent microscope ( AZ100 ) equipped with a Nikon CCD camera ( Qimaging Retiga 2000R ) . All the experiments involving zebrafish had been conducted according to the guidelines established by the Institutional Animal Care and Use Committee ( IACUC ) at The Scripps Research Institute , IACUC approval number 09-0009 . Statistical analyses were performed with the software SPSS Statistics 19 . The effects of different morpholinos on ISV development were analyzed with χ2 test . | The network of blood vessels is one of the earliest structures to develop in a vertebrate embryo . A protein called Vascular Endothelial Growth Factor A ( or VEGFA for short ) is needed to promote the growth of these blood vessels , but too much VEGFA can cause blood vessels to grow too much and to grow abnormally . Like most of the DNA in the nucleus , the gene for VEGFA is tightly wrapped around proteins called histones and must be unwrapped before it can be expressed as a protein . For the VEGFA gene , this unwrapping process starts when a protein called c-Myc adds chemical tags to the histones . Recent research suggested that an enzyme called seryl-tRNA synthetase ( or SerRS for short ) also controls the expression of VEGFA . This came as a surprise because no other tRNA synthetase has a similar role during development . And although SerRS is known to enter the cell nucleus in vertebrates , researchers did not know what SerRS did in the nucleus to control the expression of VEGFA . Now , Shi et al . have discovered that SerRS controls blood vessel development in zebrafish embryos by counteracting the activity of c-Myc . It does this in two different ways: first , it directly blocks c-Myc from binding to and unpacking the DNA; and second , SerRS works with another enzyme to remove tags that are already on the histones . Shi et al . found that if the expression of this other enzyme ( called SIRT2 ) was reduced in zebrafish , the fish expressed more VEGFA and their blood vessels grew too much . Since blood vessel growth is important in the development of cancers , the findings of Shi et al . could also lead to a better understanding of how tumors develop , as well as how blood vessels develop normally . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2014 | tRNA synthetase counteracts c-Myc to develop functional vasculature |
Drosophila dorsal air sac development depends on Decapentaplegic ( Dpp ) and Fibroblast growth factor ( FGF ) proteins produced by the wing imaginal disc and transported by cytonemes to the air sac primordium ( ASP ) . Dpp and FGF signaling in the ASP was dependent on components of the planar cell polarity ( PCP ) system in the disc , and neither Dpp- nor FGF-receiving cytonemes extended over mutant disc cells that lacked them . ASP cytonemes normally navigate through extracellular matrix ( ECM ) composed of collagen , laminin , Dally and Dally-like ( Dlp ) proteins that are stratified in layers over the disc cells . However , ECM over PCP mutant cells had reduced levels of laminin , Dally and Dlp , and whereas Dpp-receiving ASP cytonemes navigated in the Dally layer and required Dally ( but not Dlp ) , FGF-receiving ASP cytonemes navigated in the Dlp layer , requiring Dlp ( but not Dally ) . These findings suggest that cytonemes interact directly and specifically with proteins in the stratified ECM .
The language of development has a small vocabulary of signaling proteins that consists in part of Fibroblast growth factor ( FGF ) and Bone morphogenic proteins such as Drosophila Decapentaplegic ( Dpp ) . This language may be used in most or all metazoan organs . Studies of Drosophila , chick , zebrafish , and cultured human cells show that the signaling proteins that regulate development are transported along actin-based filopodia ( cytonemes ) and exchange at synapses where the cells that produce them contact the cells that receive and respond to them ( Roy et al . , 2014 and reviewed in Kornberg and Roy , 2014; Pröls et al . , 2016 ) . The large distances between the source and receiving cells in some of these contexts ( as much as 100 µm in the wing disc and 150 µm in the chick limb bud ) highlights the question that this work investigates - how cytonemes extend to reach their targets . The cytonemes we characterized were made by the ASP , a tracheal tube that develops in the third instar larva under the influence of Dpp and FGF that are produced in the wing disc ( Roy et al . , 2011; Sato and Kornberg , 2002 ) . The cytonemes that mediate the exchange of these proteins contain the Dpp receptor Thickveins ( Tkv ) or the FGF receptor Breathless ( Btl ) , extending from the basal surface of the ASP cells and synapsing with Dpp- or FGF-producing disc cells , respectively ( Roy et al . , 2014 ) . The ASP lies underneath the basement membrane that envelops the wing disc ( Guha et al . , 2009 ) , and although the space they traverse has not been analyzed , it presumably has characteristics of prototypical extracellular matrix ( ECM , reviewed in Broadie et al . , 2011 ) . The number and distribution of ASP cytonemes depend on the production of Dpp and FGF in the disc and on their respective receptors in the ASP , but it is not known whether the cells between the producing and receiving cells ( henceforth called 'intermediate cells' ) also contribute to cytoneme-mediated signaling . Possible candidates that might have roles in these intermediate cells that we tested include components of the planar cell polarity ( PCP ) system , heparan sulfate proteoglycans ( HSPGs ) and integrins . PCP is an aspect of cell polarity that establishes a singular , shared bipolar orientation across an epithelial sheet ( reviewed in Goodrich and Strutt , 2011 ) . In the insect cuticle , it is responsible for the coordinated and consistent orientation of hairs and bristles . PCP is also manifested in the asymmetric subcellular localization of proteins such as Frizzled ( Fz , a seven-pass transmembrane protein ) , Dishevelled ( Dsh ) and Diego ( Dgo , cytosolic proteins ) to one side and Van Gogh ( Vang , a four-pass transmembrane protein ) and Prickle ( Pk , a cytosolic protein ) to the other . Fz , Dsh , Dgo , Vang and Pk are constituents of the core PCP pathway in Drosophila , and all are required for planar orientation and polarization . Absent any one and the hairs and bristles lack normal polarity . In Drosophila genetic mosaics , cells to one side of PCP mutant cells also have abnormal planar polarity ( Casal et al . , 2002; Taylor et al . , 1998; Vinson and Adler , 1987 ) , leading to the idea that PCP involves a system of cell-cell interactions that coordinate the polarity of neighboring cells and propagate orientation long-range . PCP components also have other roles . Studies of cells deficient for components of the PCP system in Drosophila ( Djiane et al . , 2005; Harumoto et al . , 2010 ) and mouse ( Tao et al . , 2009; Vandenberg and Sassoon , 2009 ) report loss of several features of normal apical-basal polarity . PCP components have been implicated in the deposition and remodeling of the ECM ( Dohn et al . , 2013; Goto et al . , 2005; Williams et al . , 2012a , 2012b ) , axon guidance ( Fenstermaker et al . , 2010; Mrkusich et al . , 2011; Shafer et al . , 2011; Gombos et al . , 2015 and reviewed in Yam and Charron , 2013 ) and cell migration ( Carreira-Barbosa et al . , 2003; Dohn et al . , 2013; Roszko et al . , 2015; Tatin et al . , 2013 ) . The glypicans Dally and Dally-like ( Dlp ) are components of the ECM that contribute essential functions to signaling ( Han et al . , 2004; Fujise et al . , 2003; Han et al . , 2005; Lin and Perrimon , 1999; Yan and Lin , 2007; Yan et al . , 2010 ) . They are glycophosphatidylinositol ( GPI ) -anchored and are modified with heparan sulfate glycosaminoglycan ( GAG ) chains . Although an intracellular function has been reported for Dlp in Hh-producing cells ( Callejo et al . , 2011 ) , it is assumed that both Dally and Dlp function externally . Models proposed for the roles of Dally and Dlp include binding as co-receptors of signaling proteins ( Fujise et al . , 2003; Kim et al . , 2011; Lin and Perrimon , 1999 ) and participating in either a process of surface diffusion that involves repeated cycles of signaling protein-HSPG binding , dissociation , sliding and re-association with adjacent HSPG binding sites ( Han et al . , 2004; Schlessinger et al . , 1995 ) or repeated cycles of ligand-HSPG binding , endocytic uptake , and transcytosis ( Yan and Lin , 2009 ) . Alternatively , the recent report that cytonemes do not pass over patches of cells that are deficient for GAG-modified HSPGs suggests that impairment of cytoneme-mediated transport may account for the mutant phenotypes ( Bischoff et al . , 2013 ) . Integrin function is required for axonal pathfinding , TGF-β signaling and for interactions of migrating cells and cell protrusions with the ECM ( Dominguez-Giménez et al . , 2007; Han et al . , 2012; Myers et al . , 2011; Robles and Gomez , 2006; Vuoriluoto et al . , 2008 and reviewed in Munger and Sheppard , 2011 ) . Although the basis for these roles has not been defined , we investigated whether integrins have an essential functional role in cytoneme-mediated signaling in the ASP system .
Clonal analysis of mutant wing disc cells defective for the PCP system revealed both cell autonomous and cell non-cell-autonomous effects ( Adler et al . , 2000; Casal et al . , 2002; Taylor et al . , 1998; Tree et al . , 2002 ) . To investigate whether PCP mutants also affect the cytonemes that interact with the wing disc , we examined ASP morphology in PCP mutant third instar larvae . The ASP grows laterally from the transverse connective ( TC ) across the disc toward the cells that express FGF , and at the wandering stage ( late L3 ) , the ASP extends across the band of Dpp-expressing cells ( Figure 1A–C ) . The late L3 ASP has a characteristic narrow proximal stalk , bulbous medial region and rounded distal tip . Many cytonemes emanate from its periphery , with long lateral ones that extend dorsally to the Dpp-expressing disc cells , and the longer ones at the tip that reach postero-laterally to the FGF-expressing disc cells . In pk and Vang ( Figure 1D–F ) , dachsous ( ds ) and fat ( ft ) mutants ( Figure 1—figure supplement 1A–C ) , and in larvae that express misshapen , dRhoA , drok , dRac1 , multiple wing hair , or Leukocyte-antigen-related-like RNAi ( Table 1 ) , ASP morphology was abnormal . Although the abnormalities of the pk and Vang mutant ASPs shown in Figure 1E and F are among the most extreme that were observed , all mutant ASPs with these genotypes were abnormal . We do not make inferences about the roles of the genes for which mutant phenotypes were not observed on the screen . We investigated the phenotypes of the pk and Vang mutants further . 10 . 7554/eLife . 18979 . 003Figure 1 . ASP cytonemes depend on Prickle and Van Gogh . ( A ) The drawing of a wing disc of a wandering stage third instar larva showing branches of disc-associated trachea ( white , outlined in blue ) with transverse connective ( TC ) and air sac primordium ( ASP ) indicated , and with the Dpp- ( green ) and FGF-expressing cells ( blue ) indicated . Sagittal section with ECM also indicated , below . ( B , C ) Unfixed preparations of wing disc expressing CD2:GFP driven by btl-LHG , and mCherry driven by bnl-Gal4 ( B ) and mCherry driven by dpp-Gal4 ( C ) . ( D–F ) ASPs marked by Cherry-CAAX ( driven by btl-LHG ) in control ( D ) and pk ( E ) and Vang ( F ) mutants typify normal ASPs with many cytonemes extending from the entire periphery ( D ) and morphologically abnormal ASPs with few cytonemes in the mutants . Bottom panels show extreme examples of duplicated ASPs in the mutants . Scale bar: 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 00310 . 7554/eLife . 18979 . 004Figure 1—source data 1 . Numbers of ASP cytonemes in control , pk mutants and Vang mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 00410 . 7554/eLife . 18979 . 005Figure 1—figure supplement 1 . Components of planar cell polarity modulate the activity of cytoneme-mediated signaling . ( A–C ) Compared to control ( A ) , ASPs of dachsous ( ds ) mutants were reduced ( B ) , and ASPs of fat ( ft ) mutants were enlarged ( C ) , and neither had normal cytonemes . Genotypes: ( A ) btl-LHG lexO-Cherry-CAAX/+; ( B ) dsUAO71/dsUAO71; btl-LHG lexO-Cherry-CAAX/+; ( C ) ft8/ft8; btl-LHG lexO-Cherry-CAAX/+ . Scale bar: 30 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 00510 . 7554/eLife . 18979 . 006Figure 1—figure supplement 1—source data 1 . Numbers of ASP cytonemes in ds and ft mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 00610 . 7554/eLife . 18979 . 007Table 1 . Expression of RNAi directed against PCP component genes . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 007PCP genesbtl-Gal4ap-Gal4frizzled––disheveled––Van Gogh–reducedprickle–smallflamingo––diego––fat––dachsous––four-jointed––Grunge–smallCasein Kinase 1ε ( discs overgrown ) ––G protein o α47A ( brokenheart ) ––Dishevelled Associated Activator of Morphogenesis– *–dRac1small–dRhoAsmallsmallmisshapenabnormal–Rho Kinase ( drok ) small–multiple wing hair ( mwh ) 2 ASP2 ASPnemo––mushroom body defect––kugelei––Leukocyte-antigen-related-like ( Lar ) no ASPabnormal“–” normal ASP . * lethal; repression by Gal80ts relieved for 18 hr at third instar . In pk and Vang mutants , growth was stunted , the morphology of every ASP was abnormal ( approximately 10% were duplicated ) , and the ASP cytonemes were less abundant and shorter than normal ( Figure 1E , F and Table 2 ) . These phenotypes implicate Pk and Vang in ASP development but do not distinguish between requirements in ASP cells , in signal-producing disc cells or in the intermediate disc cells over which the cytonemes extend . In order to discriminate between these alternatives , we reduced pk and Vang function by expressing pkRNAi and VangRNAi constructs specifically in tracheal or disc cells . Expression of the RNAi constructs in tracheal cells did not affect either ASP morphology or ASP cytonemes ( Figure 2A , B ) . In contrast , ASP morphology was abnormal and cytonemes were reduced in number and length when pkRNAi or VangRNAi were expressed in the dorsal region of the wing disc where the ASP is located ( Figure 2F , G ) . Similar results were obtained by ectopic over-expression of pk and Vang in the dorsal wing disc ( Figure 2—figure supplement 1 ) , consistent with previous studies showing that loss- and gain-of-function conditions for pk and other PCP genes have similar effects on planar polarity ( Adler et al . , 2000; Casal et al . , 2002; Taylor et al . , 1998; Tree et al . , 2002; Vinson and Adler , 1987 ) . Over expression of either fz or fmi had no apparent effect on ASP morphology or development ( Figure 2—figure supplement 1 ) . Expression of RNAi constructs directed against other genes of the PCP system ( Table 1 ) identified Grunge ( Gug ) to be another candidate function that is required specifically in disc cells ( Figure 2C , H ) . Gug encodes the Drosophila homolog of the Atrophin co-repressor ( Erkner et al . , 2002; Zhang et al . , 2002 ) and has been proposed to function in the Fat/Dachsous arm of the PCP system ( Fanto et al . , 2003 ) . RNAi directed against several other genes that are required for PCP function ( e . g . , dRhoA , mwh , Lar ) resulted in abnormal ASP phenotypes after expression in either the ASP or disc ( Table 1 ) . 10 . 7554/eLife . 18979 . 008Table 2 . Numbers of ASP cytonemes in PCP system , HSPG and integrin pathway , mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 008ASP cytonemes in prickle and Van Gogh mutantsGenotype# cytonemes per µm*t-test†btl-LHG , lexO-Cherry:CAAX/+0 . 66 ± 0 . 07p valuepkpk-sple-13/pkpk-sple-13; btl-LHG , lexO-Cherry:CAAX/+0 . 28 ± 0 . 072 . 09E-05Vang153/Vang153; btl-LHG , lexO-Cherry:CAAX/+0 . 35 ± 0 . 053 . 65E-05btl-Gal4 , UAS-CD8:mCherry/+;dally80/dally800 . 28 ± 0 . 059 . 24E-06btl-Gal4 , UAS-CD8:mCherry/+;dlpA187/dlpA1870 . 46 ± 0 . 069 . 85E-04ASP cytonemes in wingblister and multiple edematous wings mutantsGenotype# cytonemes per µm*t-test†btl-LHG , lexO-Cherry:CAAX/+0 . 68 ± 0 . 14P valuemewM6/+;; btlLHG , lexO-Cherry-CAAX/+0 . 57 ± 0 . 090 . 19‡wb3/+; btlLHG , lexO-Cherry-CAAX/+0 . 58+0 . 080 . 21‡mewM6/+; wb3/+; btlLHG , lexO-Cherry-CAAX/+0 . 30 ± 0 . 091 . 02E-03* cytonemes were counted around approximately one-half the perimeter of the ASP in images generated as projection stacks from approximately 20–25 optical sections . † significance for each genotype was calculated against the btl>Cherry:CAAX controls . ‡ not significant . 10 . 7554/eLife . 18979 . 009Figure 2 . The ASP depends on the components of the PCP system in the wing disc and not in the ASP . ASPs marked by membrane-tethered GFP and expressing RNAi constructs in either the ASP ( driven by btl-Gal4 ) or the dorsal compartment of the wing disc ( driven by ap-Gal4 ) and directed against the indicated genes . Genotypes: ( A ) btl-Gal4 UAS-CD8:GFP/+; UAS-pkRNAi/+; ( B ) btl-Gal4 UAS-CD8:GFP/UAS-VangRNAi; ( C ) btl-Gal4 UAS-CD8:GFP/+; UAS-gugRNAi/+; ( D ) btl-Gal4 UAS-CD8:GFP/UAS-dallyRNAi; UAS-dallyRNAi/+; ( E ) btl-Gal4 UAS-CD8:GFP/+; UAS-dlpRNAi/+; ( F ) ap-Gal4/+; btl-LHG lexO-CD2:GFP/UAS-pkRNAi; ( G ) ap-Gal4/UAS-VangRNAi; btl-LHG lexO-CD2:GFP/+; ( H ) ap-Gal4/+; UAS-gugRNAi/btl-LHG lexO-CD2:GFP; ( I ) ap-Gal4/UAS-dallyRNAi; btl-LHG lexO-CD2:GFP/UAS-dallyRNAi; ( J ) ap-Gal4/+; btl-LHG lexO-CD2:GFP/UAS-dlpRNAi . Scale bar: 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 00910 . 7554/eLife . 18979 . 010Figure 2—figure supplement 1 . Dependence of ASP cytonemes on Prickle and Van Gogh expression . Over-expression of pk or Vang in the ASP ( btl-Gal4 UAS-pk and btl-Gal4 UAS-Vang ) had no effect on the number of ASP cytonemes ( p value from student t test 0 . 51 and 0 . 42 , respectively , compared to control ) or on ASP morphology . Over-expression in the disc ( ap-Gal4 UAS-pk and ap-Gal4 UAS-Vang ) decreased the number of ASP cytonemes ( p value from student t test 0 . 0022 and 0 . 0020 , respectively , compared to control , and 0 . 0046 and 0 . 0012 compared to btl-Gal4 UAS-pk and btl-Gal4 UAS-Vang , respectively ) and affected ASP morphology . Over-expression of fz and fmi in either the ASP or disc was without apparent consequence ( p values from student t tests 0 . 20–0 . 9 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01010 . 7554/eLife . 18979 . 011Figure 2—figure supplement 1—source data 1 . Numbers of ASP cytonemes in flies over-expressing the PCP components . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 011 To examine the genetic requirements at cellular resolution , and specifically to determine whether cytonemes are influenced by the cells they encounter as they navigate to the cells they target , we generated mitotic recombination clones of pk and Vang mutant cells in wing discs . Ten pk and nine Vang clones were identified near the lateral side of the ASP or near the ASP tip; none of the clonal areas were traversed by cytonemes . Figure 3 shows eight examples . Cytonemes were present at the periphery of mutant clones , but no cytonemes were found inside mutant territory . These results suggest that Pk and Vang are essential in 'intermediate' disc cells in order for cytonemes to extend over their surface . 10 . 7554/eLife . 18979 . 012Figure 3 . Extension of ASP cytonemes depends on Prickle and Van Gogh in underlying wing disc cells . pk13 ( upper panels ) and Vang6 ( lower panels ) mutant clones in wing discs with ASPs marked with Cherry-CAAX . The mutant clones are outlined with dotted lines and do not express GFP . Scale bar: 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 012 Our previous studies showed that cytonemes that extend from the ASP to the wing disc are essential for Dpp and FGF signaling in the ASP ( Roy et al . , 2014 ) . We therefore sought to determine if signaling in the ASP was also affected by conditions that reduced expression of pk and Vang . To monitor Dpp and FGF signaling in the ASP , we used a sensor that generates green fluorescence in cells that actively transduce Dpp ( Dad-GFP; Ninov et al . , 2010 ) , and an antibody that detects dpERK , which accumulates in cells that actively transduce FGF ( Reich et al . , 1999 ) . Under normal conditions , cells in the medial region of the ASP activate Dpp signal transduction ( Figure 4A and Roy et al . , 2014 ) and cells at the tip activate FGF signal transduction ( Figure 4D and Roy et al . , 2014; Sato and Kornberg , 2002 ) . In pk and Vang mutants , Dpp and FGF signal transduction in the ASP were reduced ( Figure 4B , C , E , F ) , confirming the importance of these functions for signaling in the ASP . Cells in mutant ASPs appeared to be larger and reduced in number compared to controls , and their morphology appeared to be abnormal . 10 . 7554/eLife . 18979 . 013Figure 4 . Dpp and FGF signaling in the ASP depend on Prickle and Van Gogh . ( A–C ) Dad-GFP expression in control ( +/+; Dad-GFP/+ ) , pk mutant ( pk13/pk13; Dad-GFP/+ ) and Vang mutant ( Vang153/Vang153; Dad-GFP/+ ) ASPs marked with α-Dlg . ( D–F ) dpERK staining in control , pk13 mutant and Vang153mutant flies . The mutant ASPs ( B , C , E , F ) were selected because their morphology most approximated normal . Arrows point to the ASP cells expressing Dad-GFP ( A ) and dpERK ( D ) . Scale bar: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 013 The wing disc is enveloped by a collagen-containing ECM , and because the ASP is situated between the basal surface of the disc and ECM ( Guha et al . , 2009 ) , the ASP cytonemes do not normally penetrate from outside the ECM to reach the disc . Rather , they navigate along , or at some distance from the basal surface of the disc cells , and they presumably encounter the ECM in a manner that has not been defined . We previously observed that the structure of the ECM is sensitive to both increased and reduced Mmp2 expression , and that dis-regulation of Mmp2 reduces ASP growth and signaling and perturbs morphogenesis ( Guha et al . , 2009 ) . Expression of dallyRNAi or dlpRNAi in the disc perturbed ASP morphology and reduced the number of ASP cytonemes ( Figure 2I , J ) , an indication that the Dally and Dlp HSPGs are essential ECM components for ASP morphogenesis and signaling . Expression of these RNAi constructs had no effect when expressed in the ASP ( Figure 2D , E ) . To investigate if there is a link between Pk and Vang and the ECM , we characterized the Dlp , Dally and laminin in pk and Vang mutants . Using antibody directed against Dlp , we found that the level of Dlp was reduced approximately 60% in the areas over pk and Vang mutant clones ( Figure 5A–D ) . The correspondence between the clone borders and areas of reduced Dlp appeared to be less than a cell diameter . To monitor Dally , we analyzed a Dally:YFP protein trap allele ( because antibody against Dally was not available ) . As indicated by YFP fluorescence , Dally was present at similar levels over the wing disc , but fluorescence was reduced approximately 45% in the engrailed expression domain in discs in which either pkRNAi or VangRNAi was expressed in the posterior compartment under the control of engrailed-Gal4 ( Figure 5E–G ) . The reduction in Dally:YFP fluorescence appeared to coincide with the anterior-posterior compartment border . 10 . 7554/eLife . 18979 . 014Figure 5 . Dally and Dlp depend on Prickle and Van Gogh . Levels of Dlp were reduced in pk13 ( A , B ) and Vang6 ( C , D ) mutant clones ( outlined with dashed white lines and lacking green fluorescence ) . ( E–G ) Levels of fluorescence in the wing discs of the protein trap Dally:YFP line were relatively uniform in controls ( en-Gal4 UAS-RFP/+; dally:YFP/+ ) , but were reduced specifically in the posterior compartment in the presence of pkRNAi ( en-Gal4/+; dally:YFP/UAS-pkRNAi ) and VangRNAi ( en-Gal4/UAS-VangRNAi; dally:YFP/+ ) . ( H , I ) Levels of Laminin detected by α-Laminin antibody staining ( red ) were reduced in pk13 mutant clones ( outlined with dashed white line and lacking green fluorescence ) . Scale bar: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01410 . 7554/eLife . 18979 . 015Figure 5—source data 1 . Quantification of fluorescence intensity of α-Dlp staining , Dally:YFP and α-Laminin staining . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01510 . 7554/eLife . 18979 . 016Figure 5—figure supplement 1 . Quantification of Dally and Dlp transcripts in Prickle and Van Gogh mutant wing discs . qRT-PCR was carried out to monitor expression of Dally and Dlp in normal , pk and Vang wing discs . Bar graph plots qRT-PCR for the Dally and Dlp transcripts relative to actin . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01610 . 7554/eLife . 18979 . 017Figure 5—figure supplement 1—source data 1 . Threshold cycles for dally , dlp and actin transcripts in wild-type , pk mutants and Vang mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01710 . 7554/eLife . 18979 . 018Figure 5—figure supplement 2 . Abnormal apical/basal polarity in Van Gogh mutant cells . Image J-generated sagittal sections showing clones of Vang6 mutant cells ( areas lacking GFP fluorescence and bounded by white dashed lines ) had elevated levels of atypical Protein Kinase C ( red , α–aPKC antibody staining ) , which is normally localized to the apical membrane domain and is downregulated by the PCP system . Discs large ( α–Dlg antibody staining ) , which localizes basolaterally , was not apparently affected in the Vang6 mutant cells . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 018 pk loss-of-function clones also had reduced levels of laminin ( Figure 5H , I ) . The correspondence between the clone borders and areas of reduced laminin was not as tight as was noted for Dlp and Dally . Together , these observations indicate that pk and Vang are necessary to maintain normal levels of Dally , Dlp and laminin in the ECM . To investigate the link between the PCP components and levels of ECM proteins , we monitored Dally and Dlp transcripts by Q-PCR . We did not detect changes in amounts of either Dally or Dlp RNA when normal discs were compared to pk or Vang mutant discs ( Figure 5—figure supplement 1 ) . We also monitored cell components that localize to either the apical ( atypical PKC; aPKC ) or basolateral ( Discs large; Dlg ) compartments ( Figure 5—figure supplement 2 ) . Whereas cells in mutant Vang clones appeared to retain normal distributions of Dlg , which is consistent with the normal morphological appearance of the columnar disc cells , levels of aPKC were elevated and extended more basally than normal . The apparent downregulation of aPKC by the PCP system , which has been reported previously in studies of the Drosophila eye ( Djiane et al . , 2005 ) and wing ( Harumoto et al . , 2010 ) , indicates that some aspects of the apical/basal polarity are PCP-dependent . We do not know how the amounts of Dally and Dlp protein in the ECM are controlled or how the PCP components Pk and Vang , which concentrate apically , might influence the composition of the basal ECM . Possibly , mutant cells fail to export the HSPGs , fail to retain HSPGs that they or their neighbors make , or do not effectively sort or localize the HSPGs between their apical and basal compartments . To investigate whether Dally and Dlp are required for cytoneme-mediated Dpp and FGF transport and for Dpp and FGF signal transduction in the ASP , we characterized the morphology of the ASP in dally and dlp mutants . Whereas expression of membrane-tethered mCherry ( btl>CD8:Cherry ) in the ASP of controls revealed a large bulbous organ with cytonemes extending from both the tip ( arrow ) and lateral ( arrowhead ) regions ( Figure 6A ) , expression of mCherry in dally80 and dlpA187 mutants revealed ASPs that were reduced in size and abnormally shaped ( Figure 6B , C ) . In the dally80 ASP , cytonemes extended primarily postero-laterally from the tip; in contrast , long cytonemes extended only along the dorsal/ventral axis from the lateral surface in the dlpA187 ASP . 10 . 7554/eLife . 18979 . 019Figure 6 . ASP cytonemes and ASP signal transduction depend on HSPG expression in underlying disc cells . ( A–C ) CD8:Cherry expression in the ASP ( driven by btl-Gal4 ) marks the ASP and ASP cytonemes in control ( A ) , dally ( B ) and dlp ( C ) mutants . Arrows and arrowheads indicate tip and medial cytonemes , respectively . ( D–F ) FGF signaling monitored by levels of dpERK staining ( arrows ) were similar in control ( D; ap-Gal4/+; btl-LHG lexO-CD2:GFP/+ ) and dally knockdown ( E; ap-Gal4/UAS-dallyRNAi; btl-LHG lexO-CD2:GFP/UAS-dallyRNAi ) ASPs , but was reduced by expression of dlpRNAi ( F; ap-Gal4/+; btl-LHG lexO-CD2:GFP/UAS-dlpRNAi ) . ( G–I ) ASPs marked by α-Dlg staining and Dpp signaling monitored by Dad-GFP fluorescence ( arrows ) in control ( G; ap-Gal4/+; Dad-GFP/+ ) , dally knockdown ( H; ap-Gal4/UAS-dallyRNAi; Dad-GFP/UAS-dallyRNAi ) , and dlp knockdown ( I; ap-Gal4/+; Dad-GFP/UAS-dlpRNAi ) ASPs . ( J ) R plots depicting number and orientation of cytonemes in five control and five mutant discs . ( K–N ) ASP and ASP cytonemes ( arrows ) were marked by Btl:Cherry ( K , L ) or by Tkv:Cherry ( M , N ) and clones of dally ( K , M ) and dlp ( L , N ) are indicated by white dashed lines and absence of GFP . ( K ) hs-FLP/+; btl-Gal4 UAS-Btl:Cherry/+; FRT2A GFP/dally80 FRT2A; ( L ) hs-FLP/+; btl-Gal4 UAS-Btl:Cherry/+; FRT2A GFP/dlpA187 FRT2A; ( M ) hs-FLP/+; btl-Gal4 UAS-Tkv:Cherry/+; FRT2A GFP/dally80 FRT2A; ( N ) hs-FLP/+; btl-Gal4 UAS-Tkv:Cherry/+; FRT2A GFP/dlpA187 FRT2A . Scale bars: 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 01910 . 7554/eLife . 18979 . 020Figure 6—source data 1 . Figure 6 data - Cytoneme and fluorescence quantification . ( A–C ) Numbers of ASP cytonemes in dally and dlp mutants . ( D–I ) Quantification of Dad-GFP signal and dpERK staining in the ASPs of control , dallyRNAi and dlpRNAi flies . ( J ) Orientations of ASP cytonemes in control , dally and dlp mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 020 Signaling in the mutant ASPs was also affected , but the effects were specific to either the Dpp or FGF signaling pathway . Loss of Dally reduced Dpp signaling , but it did not alter FGF signaling ( Figure 6D , E , G , H ) . In contrast , loss of Dlp reduced FGF signaling but not Dpp signaling ( Figure 6F , I ) . Figure 6J presents Rose plots that show the number and orientation of ASP cytonemes in control , dally80and dlpA187 genotypes . This analysis confirms that whereas cytonemes extended postero-laterally from the ASP tip and along the dorsal/ventral axis from the medial ASP in approximately equal numbers under normal conditions , the dorsal-oriented cytonemes that mediate Dpp signaling were primarily reduced in dally80 mutants and the postero-laterally oriented cytonemes were primarily reduced in dlpA187 mutants . Thus , the absence of lateral cytonemes in the dally loss-of-function condition and the absence of tip cytonemes in the dlp loss-of-function condition correlate with the reduction of Dpp and FGF signaling , respectively . The presence of tip cytonemes in the dally loss-of-function condition correlates with FGF signaling in the ASP , and the presence of lateral cytonemes in dlp loss-of-function conditions correlates with Dpp signaling . To examine these differential effects of the dally and dlp mutants in finer detail , we generated dally and dlp mutant clones in the wing disc and monitored ASP cytonemes that were marked with either fluorescent FGF receptor ( btl>Btl:Cherry ) or fluorescent Dpp receptor ( btl>Tkv:Cherry ) . Whereas cytonemes marked with Btl:Cherry extended over dally mutant clones and were apparently unchanged relative to controls , they did not extend over dlp mutant cells ( Figure 6K , L ) . In contrast , cytonemes marked with Tkv:Cherry extended over dlp mutant cells without apparent perturbation , but they did not extend over dally mutant cells ( Figure 6M , N ) . These results confirm the distinct requirements that Dpp signaling and Tkv-containing cytonemes have for dally , and that FGF signaling and Btl-containing cytonemes have for dlp . To investigate how the Tkv- and Btl-containing cytonemes might engage the Dally and Dlp HSPGs , respectively , we examined the relative locations of these ECM proteins in wing disc preparations . Simultaneous expression of Tkv:GFP and Btl:Cherry in the ASP marked the Tkv- and Btl-containing cytonemes , and sagittal projections of serial confocal sections ( Figure 7A ) and frontal sections ( Figure 7B , C ) of a specimen isolated from an early to mid L3 stage larva showed that these two types of cytonemes are separated in the space over the wing disc: the Tkv-containing cytonemes traverse a region further from the basal surface of the disc than the Btl-containing ones . Both the Tkv-containing and Btl-containing cytonemes extend from the tip at this stage ( Roy et al . , 2014 ) . Similar preparations imaged to identify Dally and Dlp ( Figure 7D ) , and laminin , collagen and Dlp ( Figure 7E ) , showed that these proteins were stratified above the basal surface of the disc epithelium . Collagen was present in a discrete , distal-most layer . Laminin was also most prominent distally but it did not extend as far from the disc surface as collagen . Dally was also prominent distally , but Dally did not extend as far from the disc surface as collagen or laminin . Dlp was most prominent in the region more proximal to the basal surface of the disc epithelium . Figure 7F depicts the distributions of collagen , laminin , Dally and Dlp relative to the location of the Tkv- and Btl-containing cytonemes . The segregation of the Tkv-containing and Btl-containing cytoneme types to separate Dally and Dlp strata and the specific requirements for Dally and Dlp by Tkv-containing and Btl-containing cytonemes establishes the functional importance of ECM stratification . 10 . 7554/eLife . 18979 . 021Figure 7 . ASP cytonemes navigate in a stratified ECM . ( A–C ) Btl:Cherry and Tkv:GFP expressed simultaneously in the ASP ( btl-Gal4 UAS-Btl:Cherry/+; UAS-Tkv:GFP/+ ) of a early to mid stage L3 . ( A ) Sagittal representation from a composite projection and frontal optical sections ( B , C ) showing ASP cytonemes ( arrows ) marked by either Btl:Cherry or Tkv:GFP . Tkv:GFP-containing cytonemes lay in focal planes more distant ( distal ) from the wing disc than cytonemes with Btl:Cherry . ( D ) Sagittal representation from a composite projection showing and early to mid L3 stage wing disc expressing a dally:YFP protein trap and stained with α-Dlp antibody ( red ) and F-actin ( blue ) . The approximate position of ASP indicated by dashed white line; disc is below and not visible in this image . ( E ) Sagittal representation from a merged composite projection showing wing disc expressing a Viking:GFP protein trap that marks collagen ( green ) , and stained with antibodies against laminin ( red ) and Dlp ( blue ) . The approximate position of ASP is indicated by dashed white line . ( F ) Drawings showing our interpretations of the relative locations in an early to mid stage L3 disc of ASP ( gray ) , collagen ( green; restricted to the most distal layer ) , laminin ( textured , yellow; most abundant distally and extending proximally to the disc surface ) , Dally ( rose; broadly distributed ) and Dlp ( lavender; proximal to disc ) , and Tkv-containing ( green ) and Btl-containing ( red ) cytonemes . Scale bars: 20µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 021 The Drosophila genome encodes five α integrin subunits ( multiple edematous wings ( αPS1 , mew ) , inflated ( αPS2 , if ) , scab ( αPS3 , scb ) , αPS4 and αPS5 ) and two β integrin subunits ( myospheroid ( mys ) and βv ) . We tested for their role in the ASP by expressing RNAi constructs , and observed that whereas the ASP developed normally in the presence of scbRNAi , αPS4RNAi , αPS5RNAi and βvRNAi ( Figure 8—figure supplement 1A–D ) , ASP development was abnormal in the presence of mysRNAi , mewRNAi and ifRNAi ( Figure 8A–D ) . FGF and Dpp signaling in the ASP were reduced in the presence of mysRNAi , mewRNAi , or ifRNAi ( Figure 8E–L ) ; and relative numbers of cytonemes were also reduced ( Figure 8M ) . Expression of mysRNAi and mewRNAi had no apparent effect on the number of dividing or apoptotic cells , on cell shape or on the ability of ASP cells to activate Dpp signal transduction ( Figure 8—figure supplement 1E–N ) . We also examined control and integrin-depleted preparations to determine whether synaptic contacts between the ASP and wing disc are integrin-dependent . We applied the GRASP ( GFP Reconstitution Across Synaptic Partner ) technique that generates GFP fluorescence specifically at sites of close and stable cell-cell contacts ( Feinberg et al . , 2008 ) , and that we used previously to mark cytoneme synapses ( Huang and Kornberg , 2015; Roy et al . , 2014 ) . As shown in Figure 8 ( panels N-Q ) , GFP fluorescence was visible at the juxtaposition of the ASP and wing disc in controls , but fluorescence was markedly reduced in preparations from animals that expressed mysRNAi . This result is consistent with the reduced numbers of cytonemes in these preparations ( Figure 8B ) . 10 . 7554/eLife . 18979 . 022Figure 8 . Cytoneme-mediated transport requires integrin . ( A–D ) ASP cytonemes were shorter in ASPs that expressed RNAi against integrin subunits encoded by mys , mew and if . ( E–L ) ASP cells active in FGF signal transduction were detected with α-dpERK antibody ( E ) and cells active in Dpp signal transduction were detected by GFP fluorescence ( I ) in control but not in ASPs depleted of mys , mew or if ( F–H , J–L ) . Arrows point to the ASP cells that express dpERK . Genotypes: ( A , E ) btl-Gal4 UAS-CD8:GFP/+; ( B , F ) btl-Gal4 UAS-CD8:GFP/UAS-mysRNAi; ( C , G ) btl-Gal4 UAS-CD8:GFP/UAS-mewRNAi; ( D , H ) btl-Gal4 UAS-CD8:GFP/+; UAS-ifRNAi/+; ( I ) btl-Gal4 UAS-CD8:Cherry/+; dad-GFP/+; ( J ) btl-Gal4 UAS-CD8:Cherry/UAS-mysRNAi; dad-GFP/+; ( K ) btl-Gal4 UAS-CD8:Cherry/UAS-mewRNAi; dad-GFP/+; ( L ) btl-Gal4 UAS-CD8Cherry/+; dad-GFP/UAS-ifRNAi . ( M ) Bar graph plots the relative number of cytonemes in ASPs of control and RNAi-mediated integrin-depleted larvae . Error bars: standard deviation; **p<0 . 01; N . S . , not significant . ( N–Q ) Sagittal and frontal images showing green fluorescence of reconstituted GFP ( GRASP ) at contacts between ASP cytonemes and Dpp-expressing cells . ASPs outlined by dashed white lines . Genotypes: ( N , P ) btl-Gal4 dpp-LHG/+; UAS-CD4:GFP1-10 lexO-CD4:GFP11/+; ( O , Q ) btl-Gal4 dpp-LHG/UAS-mysRNAi; UAS-CD4:GFP1-10 lexO-CD4:GFP11/+ . ( R ) Localization of ILK:GFP ( arrows ) in ASP cytonemes ( btl-Gal4 UAS-CD8:Cherry/UAS-ILK:GFP ) . ( S ) Number and length of ASP cytonemes was reduced with expression of ilkRNAi ( btl-Gal4 UAS-CD8:GFP/UAS-ilkRNAi ) . Scale bars: 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 02210 . 7554/eLife . 18979 . 023Figure 8—source data 1 . Figure 8 data - Cytoneme and fluorescence quantification . ( A–D ) Numbers of ASP cytonemes in control and mysRNAi , mewRNAi and ifRNAi flies . Numerical data are represented as a graph in Figure 8M . ( E–L ) Quantification of dpERK staining and Dad-GFP fluorescence in the ASPs of control and mysRNAi , mewRNAi and ifRNAi flies . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 02310 . 7554/eLife . 18979 . 024Figure 8—figure supplement 1 . Integrin receptors in the ASP . ( A–D ) RNAi targeting integrin α subunits , scb and αPS4 , or β subunit , βv , did not reduce the number of ASP cytonemes . Genotypes: ( A ) btl-Gal4 UAS-CD8:GFP/+; UAS-scbRNAi/+; ( B ) btl-Gal4 UAS-CD8:GFP/UAS-αPS4RNAi; ( C ) btl-Gal4 UAS-CD8:GFP/+; UAS-αPS5RNAi; ( D ) btl-Gal4 UAS-CD8:GFP/+; UAS–βvRNAi/+ . ( E–J ) Number of mitotic cells ( E–G ) , and number of apoptotic cells and cell shape ( H–J ) was unchanged in ASPs with RNAi targeting mys or mew relative to control . ( E–G ) stained with a-pH3 antibody; ( H–J ) stained with α-Caspase-3 and α-Dlg antibodies . Genotypes: ( E , H ) btl-Gal4 UAS-CD8:GFP/+; ( F , I ) btl-Gal4 UAS-CD8:GFP/UAS-mysRNAi; ( G , J ) btl-Gal4 UAS-CD8:GFP/UAS-mewRNAi . ( K–N ) ASP cells with reduced expression of mys and mew integrins activated Dpp signaling ( monitored by α-pMad ( blue ) ) in the presence of ectopic Dpp:Cherry expression ( red ) . Genotypes: ( K ) btl-Gal4 UAS-CD8:GFP/UAS-mysRNAi; UAS-Dpp:Cherry/+; ( L ) btl-Gal4 UAS-CD8:GFP/UAS-mysRNAi; Gal80ts/+; ( M ) btl-Gal4 UAS-CD8:GFP/UAS-mewRNAi; UAS-Dpp:Cherry/+; ( N ) btl-Gal4 UAS-CD8:GFP/UAS-mewRNAi; Gal80ts/+ . Scale bar: 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 02410 . 7554/eLife . 18979 . 025Figure 8—figure supplement 1—source data 1 . Numbers of ASP cytonemes in scbRNAi , αPS4RNAi , βvRNAi and αPS5RNAi flies . Numerical data are represented as a graph in Figure 8M . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 025 To investigate whether integrin function is also necessary for cytoneme-mediated signaling , we characterized the distribution and role of Integrin-linked kinase ( Ilk ) , a protein that interacts with the cytoplasmic tail of β integrin and links the cytoskeleton and plasma membrane at sites of integrin-mediated adhesion ( Zervas et al . , 2001 ) . We expressed an Ilk:GFP fusion construct in the ASP and observed GFP fluorescence in puncta along and at the tips of ASP cytonemes ( Figure 8R ) . Expression of IlkRNAi in the ASP perturbed ASP morphogenesis and reduced the number of ASP cytonemes ( Figure 8S ) . We conclude that integrin function is required by ASP cytonemes and for signaling in the ASP . Laminin is an integrin ligand that has been implicated in many developmental and disease contexts , and it is present in the ECM of the wing disc ( Figure 7E , F ) where it is dependent on pk function ( Figure 5H , I ) . To determine if wing disc laminin is necessary for the ASP , we depleted laminin in the disc by expressing an RNAi construct that targets the α subunit of laminin that is encoded by the wing blister ( wb ) gene . wb has been shown to interact genetically with mys ( Martin et al . , 1999; Schöck and Perrimon , 2003 ) , which encodes a β integrin subunit; the Wb laminin , which has an RGD domain , has been shown to interact with both αPS1 and αPS2 integrins ( Gotwals et al . , 1994; Graner et al . , 1998 ) . The wb knockdown genotype had a shortened and malformed ASP with reduced numbers of cytonemes ( Figures 8M , 9A ) , and low levels of Dpp ( Figure 9E ) and FGF signaling ( Figure 9I ) . Mutants that lack functional wb or mew are inviable , but wb/+ and mew/+ heterozygotes survived and developed normal ASPs that had normal levels of Dpp and FGF signaling ( Figure 9B , C , F , G , J , K ) . In contrast , mew/+; wb/+ double heterozygotes produced abnormal ASPs that had reduced numbers of cytonemes and reduced Dpp and FGF signaling ( Figure 9D , H , L and Table 2 ) . These results suggest that the Wb laminin is a ligand for integrin-dependent cytoneme-ECM interactions . 10 . 7554/eLife . 18979 . 026Figure 9 . Genetic interaction between laminin mutants and integrin mutants . ( A , E , I ) RNAi targeting expression of the laminin gene wb in the wing disc -altered ASP morphology and reduced ASP cytonemes ( A ) and reduced Dpp ( dad-GFP fluorescence; E ) and FGF signaling ( α-dpERK staining; I ) . Genotypes: ( A , I ) ap-Gal4/+; btl-LHG lexO-CD2:GFP/UAS-wbRNAi; ( E ) ap-Gal4/+;dad-GFP/UAS-wbRNAi . ( B–D , F–H , J–L ) ASP morphology and number of ASP cytonemes ( B , C ) , Dpp signaling ( F , G ) and FGF signaling ( J , K ) was normal in mew and wb heterozygotes but not in mew wb double heterozygotes ( D , H , L ) . Arrows indicate signaling cells and dashed white lines indicate ASPs . Genotypes: ( B ) mewM6/+; btl-LHG lexO-Cherry-CAAX/+; ( C ) wb3/+; btl-LHG lexO-Cherry-CAAX/+; ( D ) mewM6/+; wb3/+; btl-LHG , lexO-Cherry-CAAX/+; ( F ) mewM6/+; dad-GFP/+; ( G ) wb3/+; dad-GFP/+; ( H ) mewM6/+; wb3/+; dad-GFP/+; ( J ) mewM6/+; ( K ) wb3/+; ( L ) mewM6/+; wb3/+ . ( J–L ) Staining was with α-Dlg and α-dpERK antibodies . Scale bar: 25 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 02610 . 7554/eLife . 18979 . 027Figure 9—source data 1 . Figure 9 data - Cytoneme and fluorescence quantification . ( A ) The number of ASP cytonemes of ap>wbRNAi flies . ( B–D ) The numbers of ASP cytonemes in mew/+ , wb/+ and mew/+; wb/+ flies . ( E , I ) Quantification of Dad-GFP fluorescence and dpERK staining in ap>wbRNAi flies . ( F–H ) Quantification of Dad-GFP fluorescence in mew/+ , wb/+ and mew/+; wb/+ flies . ( J–L ) Quantification of dpERK staining in mew/+ , wb/+ and mew/+; wb/+ flies . DOI: http://dx . doi . org/10 . 7554/eLife . 18979 . 027
The ASP is a powerful system for studies of cell-cell signaling . Its large cells , easy accessibility for visual analysis , dependence on and sensitivity to paracrine signaling , and robust and abundant cytonemes offer many ways to exploit Drosophila’s genetic toolkit . We took advantage of these attributes to address the general question of targeting , asking how ASP cytonemes navigate across the wing disc in order to synapse specifically with cells that express either Dpp or FGF . We focused on the 'intermediate' disc cells that are situated between ASP cells that extend cytonemes and the disc cells they target , and we identified several functions that the intermediate cells must express in order for ASP cytonemes to mediate signaling . The results show that the extracellular environment in which cytonemes navigate is organized and structured and provides essential support that is specific to different cytoneme types . Cytonemes are dynamic , and their orientations change during development as the tissues that make them grow and the spatial relationship between the cells that produce signaling proteins and the responding cells changes ( Bischoff et al . , 2013; Roy et al . , 2011 ) . They also change in response to ectopic over-expression of signaling proteins ( Guha et al . , 2009; Roy et al . , 2011; Sato and Kornberg , 2002 ) . These properties suggest that cytonemes are dependent on an active targeting mechanism that conceivably might involve: ( 1 ) a random search; ( 2 ) chemotaxis informed by an attractant; or ( 3 ) structural information encoded in their environment . Although the chemotaxis model seems unlikely because cytonemes that deliver signaling proteins orient to receiving cells ( Bischoff et al . , 2013; Callejo et al . , 2011 ) , and although the results we report do not rule out a role for random search , our results indicate that the extracellular environment created by intermediate cells is essential and raise the question how cytonemes interact with it . We found that the PCP components pk and Vang are essential for ASP signaling and that ASP cytonemes did not extend over intermediate cells that lack pk or Vang function ( Figure 3 ) . The strong mutant phenotypes in the ASP had not been noted in previous studies of PCP mutants , the most likely reason being that the dorsal air sacs are not visible in intact flies and flies that lack dorsal air sacs are viable ( Guha et al . , 2009 ) . Because components of the PCP system of the disc has been principally associated with epithelial planar polarity and with polarized subcellular distributions of its core components , the question arises what roles its components might have that affect cytonemes that extend in the extracellular environment . In other contexts , some PCP components have been implicated in processes that may not be directly involved in epithelial planar polarity - in integrin signaling ( Lewellyn et al . , 2013 ) , microtubule polarity ( Ehaideb et al . , 2014 ) , axon guidance ( Matsubara et al . , 2011; Shimizu et al . , 2011; reviewed in Yam and Charron 2013 ) , polarized extensions of epithelial cells ( Peng et al . , 2012 ) , neural tube closure ( Kibar et al . , 2001 ) , directed cell movements ( Tatin et al . , 2013 ) and non-canonical Wnt signaling ( reviewed in Karner et al . , 2006 ) . Various mechanisms have been proposed , including integration of cell-matrix and cell-cell interactions ( Dohn et al . , 2013 ) , polarized assembly of fibronectin ( Goto et al . , 2005 ) , and Mmp-dependent remodeling of the ECM ( Williams et al . , 2012 ) . Polarized rows of laminin have been observed in the basement membrane at the basal surface of follicle cells in the Drosophila ovary ( Gutzeit et al . , 1991 ) that may involve receptor-mediated alignment of the cytoskeleton with the ECM ( Frydman and Spradling , 2001 ) . Although the functional link between the PCP system and these processes has not been established and although the possibility remains that the roles of the PCP components in these systems are unrelated to their roles in planar polarity , our working assumption is that axon pathfinding , polarized extensions , directed cell migrations and cytoneme navigation may share a requirement for features in the substrates they encounter that are dependent on cell polarity . We discovered that pk and Vang mutant cells had reduced amounts of Dally , Dlp and laminin ( Figure 5 ) . This finding calls into question the extent to which the mutant phenotypes might be due to ECM defects and an indirect consequence of Pk or Vang mis-localization . The related findings that ECM collagen , laminin , and HSPGs are stratified in layers , and that the cytoneme subtypes that contain either Tkv or Btl navigate specifically in the Dally and Dlp ECM layers , respectively ( Figures 6 , 7 ) , indicates that the ECM is organized , partitioned and regulated to an unanticipated degree and that the ECM might be directly involved in the process of cytoneme extension and pathfinding . Specific requirements of Dally for Dpp signaling ( Dejima et al . , 2011 ) and of Dlp for FGF signaling ( Yan and Lin 2007 ) have been reported previously and have been interpreted as evidence that Dally and Dlp function as co-receptors; modifications of the Dally and Dlp heparan sulfate backbones have been proposed as structures that provide specificity for the respective signaling systems ( Kamimura et al . , 2006 and reviewed in Nakato and Kimata , 2002 ) . Our observation that Tkv- and Btl-containing cytoneme subtypes were not present in areas that lacked Dally or Dlp , respectively ( Figure 6 ) , suggests that each HSPG provides a substrate that supports cytoneme extension and/or stability . This implies that cytonemes interact directly and specifically with the HSPGs and that the roles of these ECM components are not limited to co-receptor functions . A generally held model posits that axons , migrating cells and cells in developmental fields interpret secreted guidance cues and signaling proteins that encode positional information in the form of chemogradient distributions that are stored in the ECM . HSPGs were proposed to function both to distribute the guidance cues and signaling proteins , and as co-receptors for signaling proteins ( Fujise et al . , 2003; Yan and Lin , 2007; Yan et al . , 2010 ) . Our discovery that cytonemes traffic signaling proteins between cells ( Roy et al . , 2014 ) revealed that signaling proteins that have been observed between signal producing and receiving cells are cytoneme-associated and are neither extracellular nor ECM-bound ( Roy and Kornberg , 2015 ) . Our findings are consistent with the idea that secreted signaling proteins and signaling protein receptors are not distributed in the extracellular environment and are not bound to the ECM , but rather that the cytonemes that mediate Dpp and FGF signaling contact the ECM directly in ways that involve both integrins and specific HSPG interactions .
Flies were raised on standard cornmeal and agar medium at 25°C , unless otherwise specified . bnl-Gal4/TM6B , btl-LHG , UAS-CD8:GFP , UAS-CD8:Cherry , lexO-CD2:GFP , UAS-Dpp:Cherry , UAS-tkv:Cherry , UAS-btl:Cherry and UAS-btl:GFP were previously described ( Roy et al . , 2014 ) . dpp-LHG and lexO-mCherry-CAAX , from K . Basler ( Yagi et al . , 2010 ) ; UAS-ILK:GFP , from N . Brown ( Zervas et al . , 2001 ) ; UAS-mysRNAi and UAS-mewRNAi ( Han et al . , 2012 ) ; FRT42D Vang6 , from M . Mlodzik ( Wu and Mlodzik , 2008 ) ; hs-FLP; FRT2A GFP , from M . Buszczak . hs-FLP; FRT42D GFP , from M . Fuller ( Morillo Prado et al . , 2012 ) ; lexO-CD4-GFP11 and UAS-CD4-GFP1–10 , from K . Scott; dpp-Gal4/CyO and btl-Gal4 ( Sato and Kornberg , 2002 ) ; dad-GFP ( Ninov et al . , 2010 ) ; UAS-pkRNAi , from S . Eaton; dally80 , dlpA187 and UAS-dallyRNAi ( II and III ) , from H . Nakato ( Akiyama et al . , 2008 ) ; dally-GFP ( Kyoto Stock Center ) ; UAS-VangRNAi from the Vienna Drosophila RNAi Center; pk13 ( pkpk-sple13 ) , Vang153 ( Vangstbm-153 ) , ft8 , dsUAO71 , ap-Gal4 , UAS-ifRNAi and UAS-dlpRNAi from the Bloomington Stock Center . Wing imaginal discs and their associated trachea were dissected in cold phosphate-buffered saline ( PBS ) , placed on a coverslip with the columnar layer facing the coverslip , and the coverslip was mounted upside-down on a depression slide as previously described ( Huang and Kornberg , 2015 ) . Images were acquired with an upright Leica TCS SPE confocal microscope using LAS-AF software . To obtain PCP loss-of-function clones in wing discs: for notum clones , hs-FLP; FRT42D GFP females were crossed with FRT42D pk13/CyO; btl-LHG , lexO-mCherry-CAAX/TM6B males or FRT42D Vang6/CyO; btl-LHG , lexO-mCherry-CAAX/TM6B males; for wing blade clones , hs-FLP; FRT42D-GFP females were crossed with pk13 FRT42D/CyO or Vang6 FRT42D/CyO males . For MARCM clones , hs-FLP; tubP-Gal80 FRT40A; tub-Gal4 UAS-GFP/TM6 , Tb females were crossed with ft8 FRT40A/CyO , act:GFP males . Progeny were heat-shocked at 38°C for 1 hr between 48 and 72 hr AEL . To generate dally or dlp mutant clones , hs-FLP; FRT2A GFP females were crossed with btl-Gal4 , UAS-btl:Cherry; dally80 FRT2A/TM6B , btl-Gal4 , UAS-tkv:Cherry; dally80 FRT2A/TM6B , btl-Gal4 , UAS-btl:Cherry; dlpA187 FRT2A/TM6B males , or btl-Gal4 , UAS-tkv:Cherry; dlpA187 FRT2A/TM6B males . Progeny were heat-shocked at 38°C for 1 hr at the L2 stage ( ~48 hr AEL ) , allowed to develop at 25°C , and dissected at the late wandering L3 stage . L3 larvae were dissected in cold PBS and wing discs together with Tr2 trachea were fixed in 4% formaldehyde . After extensive washing , the samples were permeablized with TritonX-100 and then blocked in 10% donkey serum , and incubated with primary antibodies that had been diluted in blocking buffer . The following primary antibodies were used: α-dpERK ( Sigma ) ; α-laminin ( from J . Fessler; Fessler et al . , 1987 ) ; α-pMad ( from E . Laufer and P . ten Dijke ) ( Persson et al . , 1998 ) ; α-Senseless ( from H . Bellen ) , α-Discs large and α-Dally-like ( Developmental Studies Hybridoma Bank ) ; α-cleaved Caspase-3 ( Asp175 ) and α-phosphohistone H3 ( Ser10 ) ( Cell Signaling Technology , Danvers , MA ) . Secondary antibodies were conjugated to Alexa Fluor 405 , 488 , 555 , or 647 . Samples were mounted in Vectashield . ASP cytonemes were counted and measured in z-section stacks of images from five ASPs for each genotype . Lengths represent distance from each tip along the connecting shaft to the widening base at the plasma membrane . The ratios represent the mean value for number of cytonemes per unit length along the circumference of the ASP together with standard deviation . R plots were made for five preparations of each genotype and rose diagrams were generated by R software . For intensity measurements using Image J , a rectangle in the wing disc or ASP that included approximately ten cells was chosen . The average values ( with background fluorescence subtracted ) are presented . Total RNA was extracted from wing discs of 30 larvae using with the Zymo Research RNA MicroPrep ( Cat . #R1060 ) . Reverse transcription was carried out using the Applied Biosystem High Capacity RNA-to-cDNA ( Cat . #4387406 ) . qPCR reactions were performed with a BioRad C1000 Touch Thermal Cycler and SYBR Green ( Bioline ) . qPCR results were analyzed according to the comparative threshold cycle ( Ct ) method , where the amount of target , normalized to an endogenous actin reference and relative to an experimental control , is given by 2–△△Ct . Ct represents the PCR cycle number at which the amount of target reaches a fixed threshold . The △Ct value is determined by subtracting the reference Ct value ( i . e . actin ) from the target Ct value . △△Ct was calculated by subtracting the △Ct experimental control value . | The embryos of animals develop in a controlled manner that ensures that their tissues and organs form properly and at the right time . These processes depend on molecules called morphogens that are distributed throughout the embryo in specific ways and that are dispersed via extensions that protrude from the surfaces of cells . These extensions , called cytonemes , transport the morphogens across the distances that separate cells and transfer these molecules to target cells via direct contact . However , it was not known how cytonemes navigate to their targets . The fruit fly Drosophila is commonly used to investigate how animals develop organs and tissues . Previous studies have shown that the development of one of the fly’s organs – the air sac primordium –relies on morphogens transported by cytonemes . Now , Huang and Kornberg reveal that these cytonemes navigate to their targets by using the composition of the mesh-like framework – referred to as the extracellular matrix – that surrounds animal tissues as a guide . Further experiments showed that the extracellular matrix between the cells that produce the morphogens and the cells of the air sac primordium is roughly arranged into layers . These layers contain different molecules and the cytonemes navigate within specific layers . These findings reinforce the idea that the extracellular space is organized and regulated , and show that the extracellular matrix is essential for developmental signaling . Future challenges include understanding how the layers of the extracellular matrix form and how information is encoded in these layers for the cytonemes to decipher as they navigate to their targets . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2016 | Cells must express components of the planar cell polarity system and extracellular matrix to support cytonemes |
Integrins require an activation step prior to ligand binding and signaling . How talin and kindlin contribute to these events in non-hematopoietic cells is poorly understood . Here we report that fibroblasts lacking either talin or kindlin failed to activate β1 integrins , adhere to fibronectin ( FN ) or maintain their integrins in a high affinity conformation induced by Mn2+ . Despite compromised integrin activation and adhesion , Mn2+ enabled talin- but not kindlin-deficient cells to initiate spreading on FN . This isotropic spreading was induced by the ability of kindlin to directly bind paxillin , which in turn bound focal adhesion kinase ( FAK ) resulting in FAK activation and the formation of lamellipodia . Our findings show that talin and kindlin cooperatively activate integrins leading to FN binding and adhesion , and that kindlin subsequently assembles an essential signaling node at newly formed adhesion sites in a talin-independent manner .
Integrins are heterodimeric transmembrane receptors that mediate cell adhesion to the extracellular matrix ( ECM ) and to other cells ( Hynes , 2002 ) . The consequence of integrin-mediated adhesion is the assembly of a large molecular network that induces various signaling pathways , resulting in cell migration , proliferation , survival and differentiation ( Winograd-Katz et al . , 2014 ) . The quality and strength of integrin signaling is controlled by the interaction between integrins and substrate-attached ligands , which is , in turn , regulated by the on- and off-rates of the integrin–ligand binding process . The on-rate of the integrin–ligand binding reaction ( also called integrin activation or inside-out signaling ) is characterized by switching the unbound form of integrins from an inactive ( low affinity ) to an active ( high affinity ) conformation . The affinity switch proceeds from a bent and clasped low affinity conformation to an extended and unclasped high affinity conformation with an open ligand-binding pocket ( Askari et al . , 2010; Springer and Dustin , 2012 ) . This change in affinity is believed to be induced through the binding of talin and kindlin to the β integrin cytoplasmic domain ( Moser et al . , 2009; Shattil et al . , 2010 ) and divalent cations to distinct sites close to the ligand-binding pocket ( Gailit and Ruoslahti , 1988; Mould et al . , 1995; Xia and Springer , 2014; Mould et al . , 2003 ) . The stabilisation of integrin–ligand complexes is mediated by integrin clustering and catch bond formation between integrin and bound ligand . The stabilizing effect of clustered integrins is achieved by the ability of dissociated integrin–ligand complexes to rebind before they leave the adhesion site ( van Kooyk and Figdor , 2000; Roca-Cusachs et al . , 2009 ) , while catch bonds are receptor–ligand bonds whose lifetime increases with mechanical force ( Kong et al . , 2009; Chen et al . , 2010; Kong et al . , 2013 ) . Both mechanisms extend the duration and increase the strength of integrin-mediated adhesion and signaling ( also called outside-in signaling ) ( Koo et al . , 2002; van Kooyk and Figdor , 2000; Maheshwari et al . , 2000; Roca-Cusachs et al . , 2009; Coussen et al . , 2002 ) , and depend on the association of integrins with the actin cytoskeleton via talin ( Roca-Cusachs et al . , 2009; Friedland et al . , 2009 ) , and probably kindlin ( Ye et al . , 2013 ) . The talin family consists of two ( talin-1 and -2 ) and the kindlin family of three isoforms ( kindlin-1-3 ) , which show tissue-specific expression patterns ( Calderwood et al . , 2013; Moser et al . , 2009; Shattil et al . , 2010 ) . The majority of studies that defined integrin affinity regulation by talin and kindlin were performed on αIIbβ3 and β2-class integrins expressed by platelets and leukocytes , respectively . These cells circulate in the blood and hold their integrins in an inactive state until they encounter soluble or membrane-bound agonists ( Evans et al . , 2009 , Bennett , 2005 ) . The prevailing view is that agonist-induced signaling pathways activate talin-1 and the hematopoietic cell-specific kindlin-3 , which cooperate to induce integrin activation ( Moser et al . , 2008; Han et al . , 2006 ) and clustering ( Cluzel et al . , 2005; Ye et al . , 2013 ) . Integrin affinity regulation in non-hematopoietic cells such as fibroblasts and epithelial cells is poorly understood . It is not known how integrin activation is induced on these cells ( no integrin-activating agonists have been identified ) and it is also controversial whether talin and kindlin are required to shift their integrins into the high affinity state . While there are reports showing that talin and kindlin are required for integrin activation in epithelial cells ( Montanez et al . , 2008; Margadant et al . , 2012 ) , it was also shown that in myoblasts and mammary epithelial cells activation of β1 integrins , adhesion and spreading on multiple ECM substrates can proceed in the absence of talin ( Conti et al . , 2009; Wang et al . , 2011 ) . Likewise , it was reported that focal adhesion kinase ( FAK ) -deficient fibroblasts develop small , nascent adhesions ( NAs ) at the edge of membrane protrusions without visible talin and that integrins carrying a mutation in the talin-binding site can still nucleate and stabilize NAs ( Lawson et al . , 2012 ) . Also fibroblasts lacking talin-1 and -2 were shown to adhere to fibronectin ( FN ) and initiate isotropic spreading ( Zhang et al . , 2008 ) . Another intriguing study demonstrated that overexpression of kindlin-2 in Chinese hamster ovary ( CHO ) cells inhibits rather than promotes talin head-induced α5β1 integrin activation ( Harburger et al . , 2009 ) . Given the fundamental importance of talin and kindlin for integrin activation in hematopoietic cells , the findings of these studies are unexpected and imply that either integrin affinity regulation is substantially different in fibroblasts and epithelial cells or the experimental approaches used to manipulate protein expression and localization were imperfect . To directly evaluate the functions of talin and kindlins for FN-binding integrins on fibroblasts , we used a genetic approach and derived fibroblasts from mice lacking either the Tln1 and -2 or the Fermt1 and -2 genes . We show that integrin affinity regulation depends on both talin and kindlin , and that kindlin has the additional function of triggering cell spreading by binding directly to paxillin in a talin-independent manner .
To obtain cells lacking the expression of talin-1 and kindlin-2 , we intercrossed mice carrying loxP flanked ( floxed; fl ) Tln1 or Fermt2 alleles ( Figure 1A ) , isolated kidney fibroblasts and immortalized them with the SV40 large T antigen ( parental fibroblasts ) . The floxed alleles were deleted by adenoviral Cre recombinase transduction resulting in T1Ko and K2Ko fibroblasts . Loss of talin-1 or kindlin-2 expression in fibroblasts was compensated by talin-2 or the de novo expression of kindlin-1 , respectively , allowing adhesion and spreading , although to a lesser extent compared with control cells ( Figure 1—figure supplement 1A , B ) . To prevent this compensation , we generated mice with floxed Tln1 and nullizygous Tln2 alleles or with floxed Fermt1 and -2 alleles ( TlnCtr; KindCtr ) from which we isolated , immortalized and cloned kidney fibroblasts with comparable integrin surface levels ( Figure 1A and Figure 1—figure supplement 2 ) . The floxed alleles were deleted by transducing Cre resulting in talin-1 , -2 ( TlnKo ) and kindlin-1 , -2 ( KindKo ) deficient cells , respectively ( Figure 1A–C ) . Since the TlnCtr and KindCtr control cells showed similar morphologies and behaviour in our experiments , we display one control cell line in several result panels . Cre-mediated deletion of the floxed Tln1 or floxed Fermt1/2 genes was efficient ( Figure 1B ) and resulted in cell rounding , weak adhesion of a few cells , and reduced cell proliferation despite the immortalisation with the oncogenic large T antigen ( Figure 1C and Figure 1—figure supplement 3 ) . To minimize cell passage-induced abnormalities , we used cells only up to 12 passages after Cre-mediated gene deletions . 10 . 7554/eLife . 10130 . 003Figure 1 . Kindlin and talin are required for integrin-mediated cell adhesion . ( A ) Scheme showing gene loci before and after ablation of the Tln1 , -2 and Fermt1 , -2 genes . Orange diamonds indicate loxP sites and rectangles exons; untranslated regions are marked grey . ( B ) Western blot of TlnKo and KindKo cells . Keratinocyte lysates ( Kerat . ) served to control kindlin-1 expression . ( C ) Bright field images of TlnCtr , KindCtr , TlnKo and KindKo cells . ( D ) Quantification of cell adhesion on indicated substrates 30 min after seeding by counting DAPI stained cells; n=3 independent experiments , error bars indicate standard error of the mean; t-test significances are calculated between untreated TlnKo or KindKo cells and the corresponding TlnCtr and KindCtr or Mn2+-treated TlnKo or KindKo cell lines on same substrates; only significant differences are shown . ( E ) Quantification of Mn2+-stimulated cell adhesion for indicated times on FN; cells were quantified by absorbance measurement of crystal violet staining; n=3 independent experiments; lines represent sigmoidal curve fit; error bars indicate standard deviation; significances for indicated pairs after 2 . 5 min were calculated by two-tailed t-test and significances for indicated pairs of the overall kinetics were calculated by two-way RM ANOVA . Bar , 10 µm . COL , collagen; DAPI , 4' , 6-diamidino-2-phenylindole; FN , fibronectin; GAPDH , glyceraldehyde-3-phosphate dehydrogenase; LN , laminin-111; RM ANOVA , repeated measures analysis of variance; VN , vitronectin . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00310 . 7554/eLife . 10130 . 004Figure 1—figure supplement 1 . Talin-1- and kindlin-2-deficient fibroblasts . ( A ) Western blots showing talin-2 expression in floxed talin-1 ( T1F ) and T1Ko fibroblasts and de novo expression of the Fermt1 gene in kindlin-2-null ( K2Ko ) fibroblasts . Keratinocytes ( Kerat . ) expressing high levels of kindlin-1 served as control for the anti-kindlin-1 antibody . GAPDH served as loading control . ( B ) Talin-1- and kindlin-2-deficient fibroblasts partially spread ( bright field imaging , left panels ) and form paxillin-positive adhesion sites ( immunostaining , right panels ) . Bars , 10 µm . GAPDH , glyceraldehyde-3-phosphate dehydrogenase . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00410 . 7554/eLife . 10130 . 005Figure 1—figure supplement 2 . Integrin expression profiles of TlnCtr and KindCtr cells . Cell surface expression of different integrin subunits on TlnCtr and KindCtr cells was measured by flow cytometry and presented as histograms . Fluorescence-activated cell sorting histograms of cells lacking expression of all integrins ( pKO ) served as negative control and are shown in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00510 . 7554/eLife . 10130 . 006Figure 1—figure supplement 3 . Cell proliferation of TlnKo and KindKo cells . TlnKo and KindKo cells show a significantly reduced increase in cell numbers , which were determined by cell counting at indicated time points ( error bars indicate standard deviation; significances are given for indicated pairs and were calculated by two-way ANOVA ) . ANOVA , analysis of variance . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00610 . 7554/eLife . 10130 . 007Figure 1—figure supplement 4 . Cell adhesion of TlnKo and KindKo cells on different FN concentrations . Cell adhesion was measured 20 min after seeding the indicated cell lines on plastic surfaces coated with the indicated FN concentrations . Cells were PFA fixed and quantified by absorbance measurement of crystal violet staining ( n=3 independent experiments; lines represent hyperbolic curve fits; error bars indicate standard deviation; significances for indicated pairs of the overall kinetics were calculated by two-way RM ANOVA ) . FN , fibronectin , PFA , paraformaldehyde; RM ANOVA , repeated measures analysis of variance . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 007 To define the adhesion defect , we performed plate and wash assays for 30 min on defined substrates and found that neither TlnKo nor KindKo cells adhered to FN , laminin-111 ( LN ) , type I collagen ( COL ) and vitronectin ( VN ) ( Figure 1D ) . To test whether the inability of TlnKo and KindKo cells to adhere to ECM proteins is due to an integrin activation defect , we bypassed inside-out activation by treating cells with Mn2+ , which binds to the integrin ectodomain and induces unbending and unclasping of integrin heterodimers ( Mould et al . , 1995 ) . Treatment with Mn2+ induced partial adhesion of TlnKo and KindKo cells to FN , while partial adhesion to LN and VN was only induced in TlnKo cells ( Figure 1D ) . Time course experiments revealed that Mn2+-induced adhesion of TlnKo and KindKo cells to FN was already significantly lower 2 . 5 min after plating and remained significantly lower compared with control cells ( Figure 1E ) , suggesting that talin and kindlin cooperate to initiate and maintain normal Mn2+-induced adhesion to FN . In line with these findings , dose-response profiles showed that TlnKo and KindKo cells have severe adhesion defects at low ( 1 . 25 µg ml–1 ) as well as high ( 20 µg ml–1 ) substrate concentrations ( Figure 1—figure supplement 4 ) . These findings indicate that talin and kindlin promote integrin-mediated adhesion to FN and proliferation , and that the integrin-activating compound Mn2+ can only partially substitute for the adhesion promoting roles that talin and kindlin accomplish together . The inability of Mn2+ to fully rescue the adhesion defect of TlnKo and KindKo cells raised the question whether integrin surface levels change after deletion of the Tln1/2 and Fermt1/2 genes . We quantified integrin surface levels by flow cytometry and found that the levels of β1 and β3 were significantly reduced in KindKo and unaffected in TlnKo cells ( Figure 2A and Figure 2—figure supplement 1 ) . The levels of α2 and α3 integrin were reduced in both cell lines , α6 was elevated in TlnKo and decreased in KindKo cells , and the α3 levels were significantly more decreased in KindKo than in TlnKo cells ( Figure 2A ) explaining the absent adhesion of both cell lines to COL and their differential adhesion behaviour on LN ( Figure 1D ) . The β5 levels were similarly up-regulated in KindKo and TlnKo cells , and the α5 and αv integrin levels were slightly reduced but not significantly different between TlnKo and KindKo cells ( Figure 2A ) . The differential adhesion of Mn2+-treated TlnKo and KindKo cells to VN ( Figure 1D ) , despite similar surface levels of αv integrins , points to particularly important role ( s ) for kindlin-2 in αv integrins-VN adhesion and signaling ( Liao et al . , 2015 ) . Serendipitously , the reduced expression of β1-associating α2 , α3 and α6 subunits in KindKo cells , which impairs adhesion to LN and COL enables α5 to associate with the remaining β1 subunits and leads to comparable α5β1 levels on TlnKo and KindKo cells ( Figure 2—figure supplement 2 ) explaining their similar adhesion to FN ( Figure 1D , E and Figure 1—figure supplement 4 ) . Therefore , we performed all further experiments with FN . 10 . 7554/eLife . 10130 . 008Figure 2 . FN binding by TlnKo and KindKo cells . ( A ) Quantification of integrin surface expression levels relative to the TlnCtr and KindCtr cell lines; independent experiments: n=10 for β1; n=4 for β3 , α5 , αv; n=3 for remaining integrin subunits; error bars indicate standard error of the mean; significances are calculated between TlnKo and KindKo cells indicated by brackets , or between TlnKo or KindKo cells and corresponding control cells indicated by the significances above or below bars . ( B , C ) Box plot representation of adhesion forces generated by cells interacting with surface immobilized FN fragments . Cells were immobilized on ConA-coated AFM cantilevers and pressed onto surfaces coated with the FN-RGD or integrin-binding deficient FN-ΔRGD fragments for varying contact times , either in the absence ( B ) or presence of Mn2+ ( C ) . Coloured and grey boxplots represent adhesion forces from at least 10–15 independent experiments with single cells; + signs represent mean; the significance between adhesion on FN-RGD versus FN-ΔRGD is given on top of each boxplot and was calculated with a Mann–Whitney U test; brackets indicate two-way RM ANOVA comparisons of the whole adhesion kinetics . ( D ) FN staining after plating cells on a FN-coated dish for 24 hr . ( E ) Quantification of cell adhesion on FN 30 min after seeding; values are normalized to TlnCtr and KindCtr; n=3 independent experiments; error bars indicate standard error of the mean . Bar , 10 µm . AFM , atomic force microscopy; ConA , Concanavalin A; FN , fibronectin; K2GFP , green fluorescent protein-tagged kindlin-2; RGD , Arg-Gly-Asp; RM ANOVA , repeated measures analysis of variance; THD , talin-1 head domain; Tln1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00810 . 7554/eLife . 10130 . 009Figure 2—figure supplement 1 . Integrin expression profiles of TlnCtr , TlnKo , KindCtr and KindKo cells . Representative FACS histograms of different integrin subunits expressed on TlnCtr , TlnKo , KindCtr and KindKo cells are shown . FACS histograms of cells lacking expression of all integrins ( pKO ) served as negative control and are shown in grey . FACS , fluorescence-activated cell sorting . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 00910 . 7554/eLife . 10130 . 010Figure 2—figure supplement 2 . TlnKo and KindKo cells display comparable α5β1 integrin cell surface levels . Live TlnCtr , KindCtr , TlnKo and KindKo cells were incubated with antibodies against α5 integrin ( α5 ) or with an unrelated isotype control ( iso ) on ice to immunoprecipitate α5 integrin from their cell surface . Following immunoprecipitation , the proteins were analyzed by western blotting to determine the levels of β1 and α5 integrin . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01010 . 7554/eLife . 10130 . 011Figure 2—figure supplement 3 . β1 integrin activation in TlnCtr , TlnKo , KindCtr , KindKo cells . ( A ) FACS quantification of 9EG7 antibody binding to the indicated cell lines in the presence of 0 . 3 µM FN ( FN ) or 5 mM MnCl2 and 0 . 3 µM FN ( FN+Mn2+ ) ( n=3 independent experiments; 9EG7 binding was normalized to total-β1 surface levels and 100% represents the average of TlnCtr and KindCtr under control buffer condition; error bars indicate standard error of the mean; significances are calculated between Ctr and indicated condition ) . ( B ) FACS quantification of total β1-antibody and 9EG7-antibody binding to TlnCtr and KindCtr cells and cells reconstituted with Tln1V , K2GFP or THD . 9EG7 binding was normalized to total β1 surface levels and control cell lines were set to 100% ( n>3 independent experiments; significances are given for indicated pairs; error bars indicate standard error of the mean ) . FACS , fluorescence-activated cell sorting; FN , fibronectin; K2GFP , green fluorescent protein-tagged kindlin-2; THD , talin-1 head domain; Tln1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01110 . 7554/eLife . 10130 . 012Figure 2—figure supplement 4 . Re-expression of talin-1 or kindlin-2 in TlnKo and KindKo cells . Western blot analysis of cell lysates from TlnKo and KindKo cells reconstituted with Tln1V , THD or K2GFP expression plasmids . GAPDH , glyceraldehyde-3-phosphate dehydrogenase; K2GFP , green fluorescent protein-tagged kindlin-2; THD , talin-1 head domain; Tln1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 012 Since we excluded different surface levels of FN-binding integrins as a cause for the severely compromised adhesion of TlnKo and KindKo cells to FN , we tested whether talin and kindlin are required to activate FN-binding α5β1 integrins . To directly assess integrin activation , we made use of an antibody against the 9EG7 epitope , which specifically recognizes Mn2+ and/or ligand activated β1 integrins ( Bazzoni et al . , 1995 ) . The amount of 9EG7 epitope exposure relative to total β1 integrin exposure corresponds to the integrin activation index , which can be measured by flow cytometry using 9EG7 and anti-total β1 integrin antibodies . These measurements revealed that TlnCtr and KindCtr cells bound 9EG7 antibodies , while TlnKo and KindKo cells lacked 9EG7 binding in the absence of Mn2+ ( Figure 2—figure supplement 3A ) . Mn2+ treatment significantly increased 9EG7 binding by TlnCtr and KindCtr cells , which was further elevated in the presence of FN-Arg-Gly-Asp ( RGD ) ligand known to stabilize the high affinity state of integrins ( Figure 2—figure supplement 3A ) . Mn2+-treated TlnKo and KindKo cells bound significantly less 9EG7 antibodies than control cells , which marginally increased with FN-RGD ( Figure 2—figure supplement 3A ) . Moreover , the normalization of the 9EG7 binding to the total β1 integrin surface levels also indicated a significantly lower influence of Mn2+ and FN-RGD on the integrin activation index of KindKo as compared to TlnKo cells ( Figure 2—figure supplement 3A ) . These findings confirm that both , talin and kindlin are required for β1 integrin activation and to stabilize Mn2+-induced unbending/unclasping of α5β1 integrins . Our findings so far suggest that talin and kindlin are required to activate FN-binding integrins and maintain Mn2+-induced activation of FN-binding integrins . To further analyze whether ligand-induced stabilisation of high-affinity integrin conformations ( also termed ‘ligand-induced integrin activation; Du et al . , 1991 ) can form in the absence of talin or kindlin , we used atomic force microscopy ( AFM ) -based single cell force spectroscopy ( SCFS ) . We attached control , TlnKo or KindKo cells to Concanavalin A ( ConA ) -coated cantilevers , allowed the cells to contact surfaces coated with either wild type FN-III7-10 ( FN-RGD ) or an integrin-binding-deficient FN-III7-10 fragment lacking the RGD binding motif ( FN-ΔRGD ) for increasing contact times , either in the absence or presence of Mn2+ and then detached them from the substrate by lifting the cantilever ( Figure 2B , C ) . In the absence of Mn2+ TlnCtr and KindCtr cells developed significant adhesion to FN-RGD within 5 s contact time . After a contact time of 20 s around 2 nN force was required to disrupt adhesion to FN-RGD , and after 50 and 120 s , respectively , 3 and 6 nN were required ( Figure 2B ) . TlnKo and KindKo cells failed to develop measurable adhesions to FN-RGD after contact times of 5 , 20 , 50 and 120 s ( Figure 2B ) . Treatment with Mn2+ induced a slight and similar increase of force required to disrupt adhesion of control , TlnKo and KindKo cells to FN-RGD after 5 s contact time ( Figure 2C ) . However , with increasing contact times , the AFM profiles of TlnKo and KindKo cells differ in the presence of Mn2+ . While the adhesion force increased concomitantly with longer contact times in TlnCtr , KindCtr and TlnKo cells , adhesion forces of KindKo cells plateaued after 50 s and showed no further increase towards 120 s contact time . The latter finding suggests that kindlin stabilizes integrin–ligand complexes with time , by inducing integrin clustering and/or by modulating the off-rate of integrin ligand bonds , for example , through associating with the integrin-linked kinase ( ILK ) -Pinch-Parvin ( IPP ) complex that links kindlin to the F-actin cytoskeleton ( Cluzel et al . , 2005; Ye et al . , 2013; Montanez et al . , 2008; Fukuda et al . , 2014 ) . We next tested whether their impaired integrin function affects the assembly of FN into fibrils , which requires association of active α5β1 integrin with the actin cytoskeleton ( Pankov et al . , 2000 ) , and whether re-expression of talin and kindlin reverts the defects of TlnKo and KindKo cells . While neither TlnKo nor KindKo cells were able to assemble FN fibrils , re-expression of full-length Venus-tagged talin-1 ( Tln1V ) in TlnKo or GFP-tagged kindlin-2 ( K2GFP ) in KindKo cells ( Figure 2—figure supplement 4 ) rescued FN fibril assembly and adhesion to FN ( Figure 2D , E ) . Furthermore , neither overexpression of the talin-1 head ( THD ) nor K2GFP in TlnKo cells , nor Tln1V or THD in KindKo cells rescued adhesion to FN or 9EG7 binding ( Figure 2E and Figure 2—figure supplement 3B ) . Altogether , our results demonstrate that both talin and kindlin are required ( 1 ) for ligand-induced stabilisation of integrin-ligand complexes , ( 2 ) to stabilize Mn2+-activated α5β1 integrins , and ( 3 ) to induce integrin-mediated FN fibril formation . It has been reported that a significant number of talin-2 small interfering RNA ( siRNA ) -expressing talin-1–/– fibroblasts adhere to FN and initiate isotropic cells spreading ( Zhang et al . , 2008 ) . To test whether spreading can also be induced in adherent TlnKo and KindKo cells , we bypassed their adhesion defect with Mn2+ , seeded them for 30 min on FN and stained with an antibody against total β1 integrin and the β1 integrin activation epitope-reporting 9EG7 antibody . As expected , TlnCtr or KindCtr cells clustered 9EG7-positive β1 integrins in NAs and focal adhesions ( FAs ) , whose frequency and size increased upon Mn2+ treatment ( Figure 3A , B ) . In contrast , the sporadic and very weakly adherent TlnKo and KindKo cells were small , round and formed small and finely dispersed β1 integrin aggregates over the entire cell ( Figure 3A ) and lacked 9EG7-positive signals ( Figure 3B ) in the absence of Mn2+ treatment . Upon Mn2+ treatment 37 ± 1% ( n=684 , mean ± standard deviation of three independent experiments ) of the TlnKo cells showed isotropic membrane protrusions ( circumferential lamellipodia ) with small , dot-like aggregates of β1 integrin , kindlin-2 , paxillin and ILK at the membrane periphery ( Figure 3A and Figure 3—figure supplement 1 ) , which eventually detached from the substrate leading to the collapse of the protruded membrane ( Video 1 ) . Furthermore , 9EG7-positive β1 integrins accumulated along the lamellipodial edge and beneath the nucleus of TlnKo cells ( Figure 3B ) . The remaining cells were spheroid , with half of them showing short , finger-like protrusions , which were motile due to their poor anchorage to the substrate . In the case of KindKo cells , we analysed 652 cells in three independent experiments and found that only 7 ± 1% ( mean ± standard deviation ) of the cells established lamellipodia , which formed around the entire circumference in 2 ± 0 . 4% ( mean ± standard deviation ) of the cells . Around 93 ± 1% of the KindKo cells were spheroid ( mean ± standard deviation ) and frequently had finger-like , motile protrusions with small dot-like signals containing β1 integrin and talin but rarely paxillin or ILK ( Figure 3A and Figure 3—figure supplement 1 ) . Importantly , re-expression of Tln1V in TlnKo cells or K2GFP in KindKo cells normalized FA formation and spreading on FN ( Figure 3—figure supplement 2 ) . These findings indicate that kindlin-2 expressing TlnKo cells can initiate the formation of large lamellipodia and assemble β1 integrins in lamellipodial edges . 10 . 7554/eLife . 10130 . 013Video 1 . Spreading KindCtr , TlnKo and KindKo cells on FN . Assembled time lapse movies of KindCtr , TlnKo and KindKo cells . Cell spreading was recorded 5 min after seeding on FN . KindCtr cells were already well spread and only a minor size increase was observed over the following minutes . The TlnKo cells formed a circumferential lamellipodium that rapidly collapsed and subsequently the cells formed finger-like protrusions of varying size and failed to reestablish a fully formed circular lamellipodium . The KindKo cells failed to form a lamellipodium and formed finger-like protrusions that were not always adherent . Bar , 10 µm . FN , fibronectin . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01310 . 7554/eLife . 10130 . 014Figure 3 . Integrin distribution in TlnKo and KindKo cells . ( A ) Confocal images of the ventral side of adherent cells stained for β1 integrin and F-actin in the absence or presence of Mn2+ stimulation . Notice the increase in the spreading area ( w/o Mn2+: 1696 ± 360 µm2 , Mn2+: 2676 ± 466 µm2 ) and in the average size ( w/o Mn2+: 0 . 64 ± 0 . 1 µm2 , Mn2+: 0 . 89 ± 0 . 1 µm2 ) and number ( w/o Mn2+: 105 ± 38 , Mn2+: 246 ± 8 ) of focal adhesions in KindCtr cells after Mn2+ stimulation and the increase of spreading area in the TlnKo ( w/o Mn2+: 77 ± 1 µm2 , Mn2+: 572 ± 37 µm2 ) and KindKo cells ( w/o Mn2+: 76 ± 27 µm2 , Mn2+: 152 ± 8 µm2 ) ( n=3 , mean ± standard deviation ) . ( B ) Confocal images from the ventral side of adherent cells stained for the 9EG7 epitope in the absence or presence of Mn2+ stimulation . ( C ) TIRF-dSTORM images of β1 integrin ( grey scale image ) obtained from immunostaining of non-permeabilized cells overlaid with anti-paxillin staining following permeabilization ( red , normal resolution ) . Boxed areas are displayed in a five-fold magnification . ( D ) Images show heat map representations of dSTORM localizations per µm2 and sec , indicative for integrin clustering defined by local integrin densities . The colour range indicates localizations s–1 µm–2 with low values shown in dark red colours and high densities from yellow to white colours . Bars , 10 µm ( A , B ) ; 5 µm ( C , D ) ; 500 nm ( for the magnification in C , D ) . TIRF , total internal reflection fluourescence; dSTORM , direct stochastic optical reconstruction microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01410 . 7554/eLife . 10130 . 015Figure 3—figure supplement 1 . Localization of FAs proteins in Mn2+-treated KindCtr , TlnKo and KindKo cells . Confocal images of the ventral plasma membrane of adherent , Mn2+-treated KindCtr , TlnKo and KindKo cells stained for ILK , paxillin , talin , and kindlin-2 ( green ) , always together with phalloidin to visualize F-actin ( red ) . For TlnKo and KindKo , three-fold magnifications of indicated areas are shown . For kindlin-2 staining , acetone-methanol fixation was used . Bar , 10 µm . FAs , focal adhesions; ILK , integrin-linked kinase . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01510 . 7554/eLife . 10130 . 016Figure 3—figure supplement 2 . Rescue of FA formation and spreading after expression of Tln1V in TlnKo cells or K2GFP in KindKo cells . Confocal images of KindKo and TlnKo cells reconstituted with K2GFP or Tln1V expression plasmids , respectively . Bar , 10 µm . FA , focal adhesion; K2GFP , green fluorescent protein-tagged kindlin-2; Tln1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01610 . 7554/eLife . 10130 . 017Figure 3—figure supplement 3 . Distribution of β1 integrins in spheroid-shaped TlnKo cells . dSTORM image and density map of β1 integrins in non-spread , spheroid-shaped TlnKo cells shows aggregation of integrin in the cell body and finger-like protrusions . Bars , 5 µm and 500 nm ( for high magnification ) . dSTORM , direct stochastic optical reconstruction microscopy . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 017 To further characterize the distribution of β1 integrins in the lamellipodial edges of TlnKo cells , we visualized them by combining direct stochastic optical reconstruction microscopy ( dSTORM ) and total internal reflection fluorescence microscopy ( TIRFM ) . Mn2+-treated and non-permeablized cells were seeded on FN , stained with anti-total β1 integrin antibodies , and then permeabilized , immunostained for paxillin and imaged with normal resolution TIRFM and dSTORM ( Figure 3C ) . Each localization detected by dSTORM was plotted as a Gaussian distribution around its centre with an average spatial accuracy of ~20 nm ( resolution limit of dSTORM imaging ) . Since two or more localizations from single or multiple dyes in close proximity cannot be distinguished , the number of localizations does not directly reflect integrin numbers . However , all antibody molecules display the same average behaviour with respect to the number of localizations per second in all areas of the cell . This allowed to average the number of localizations per second and µm2 and to plot them in a heat map representation ( Figure 3D ) , which directly reflects the density of stained β1 integrin molecules and thus the degree of integrin clustering . The β1 integrin staining of TlnCtr and KindCtr cells revealed small round structures of ~50 nm diameter indicating clusters of integrins larger than the resolution limit ( Figure 3C; high magnification; see arrowheads ) . Furthermore , high numbers of localizations were enriched in paxillin-positive FAs and in NAs at the lamellipodial edge ( Figure 3C; see arrowheads ) . In these areas , we observed a high average density of 60–120 localizations s–1 µm–2 , while outside of the adhesion sites ~0–20 localizations s–1 µm–2 were detected , indicating a high degree of β1 integrin clustering within and a low degree of clustering outside of adhesion sites ( Figure 3D ) . TlnKo cells with circumferential lamellipodia showed a high density of blinking with up to 100 localizations s–1 µm–2 at lamellipodial edges ( Figure 3C , D; see arrowheads ) , which appeared less compact than in control cells . KindKo cells showed >120 localizations s–1 µm–2 in the periphery and finger-like membrane protrusions ( Figure 3C , D; see arrowheads ) , which were also observed in TlnKo cells that adopted a spheroid rather than an isotropic spread shape ( Figure 3—figure supplement 3 ) . The exclusive presence of these large and entangled β1 integrin aggregates on TlnKo and KindKo cells with small , spheroid shapes and protrusions suggests that they were induced by spatial constraints rather than specific signaling . These findings demonstrate that , in contrast to KindKo cells , Mn2+-treated kindlin-2-expressing TlnKo cells induce circumferential membrane protrusions with β1 integrins at the protrusive edges . Our data so far indicate that the expression of kindlin-2 enables initial , isotropic spreading and the accumulation of integrins in lamellipodia of Mn2+-treated TlnKo cells . To identify binding partner ( s ) of kindlin-2 that transduce this function to downstream effectors , we performed yeast-two-hybrid assays with kindlin-2 as bait using a human complementary DNA ( cDNA ) library containing all possible open reading frames and a human keratinocyte-derived cDNA library . Among the 124 cDNAs identified from both screenings , 17 coded for leupaxin and 11 for Hic-5 . Immunoprecipitation of overexpressed green fluorescent protein ( GFP ) -tagged paxillin family members , paxillin , Hic-5 and leupaxin in HEK-293 cells with an anti-GFP antibody efficiently co-precipitated FLAG-tagged kindlin-2 ( K2flag ) ( Figure 4A ) . Conversely , overexpressed GFP-tagged kindlin family members ( kindlin-1 , kindlin-2 , kindlin-3 ) co-precipitated Cherry-paxillin ( Figure 4—figure supplement 1 ) . Since fibroblasts express high levels of paxillin ( Figure 4—figure supplement 2 ) , we performed all further interaction analysis with paxillin . Immunoprecipitations of GFP-tagged paxillin or kindlin-2 truncation mutants ( Figure 4—figure supplement 3A ) revealed that the interaction between kindlin-2 and paxillin was dramatically reduced in the absence of the Lin-11 , Isl-1 and Mec-3 ( LIM ) 1-4 , LIM2-4 or LIM3-4 domains of paxillin ( Figure 4B ) , or the pleckstrin homology ( PH ) domain ( K2ΔPHGFP; lacking amino acids 380-477 ) or the N-terminus of kindlin-2 including the F0 , F1 , and the N-terminal part of the F2 domains ( K2NTGFP; terminating at the end of F1; spanning amino acids 1-229 ) ( Figure 4C ) . As expected , the interaction between kindlin-2 and ILK ( Montanez et al . , 2008 ) , which is mediated via a recently identified sequence in the linker domain between the end of the N-terminal F2 and the beginning of the PH domain ( amino acids 353–357 ) ( Fukuda et al . , 2014; Huet-Calderwood et al . , 2014 ) , was abolished by the K2NTGFP truncation but unaffected by the deletion of the PH domain ( K2ΔPHGFP ) or the deletion of the N-terminal F0 and F1 domains ( K2CTGFP , spanning amino acids 244-680 ) ( Figure 4C ) . Importantly , immunoprecipitation of KindCtr lysates with antibodies against paxillin co-precipitated kindlin-2 ( Figure 4D ) , confirming interactions between the endogenous proteins . Pull down experiments with recombinant full-length paxillin or paxillin-LIM3 domain and recombinant kindlin-2 demonstrated that binding to LIM3 and full-length paxillin is direct , Zn2+-dependent and abrogated with ethylenediaminetetraacetic acid ( EDTA ) ( Figure 4E and Figure 4—figure supplement 3B ) . KindKo cells were transduced with K2GFP or K2ΔPHGFP expression constructs , seeded on FN for different times and stained for β1 integrin , paxillin and F-actin . The experiments revealed that the expression of K2GFP in KindKo cells rescued spreading and induced robust paxillin recruitment to β1 integrin-positive NAs ( Figure 4F , G ) . In contrast , expression of K2ΔPHGFP failed to recruit paxillin to β1 integrin-positive adhesion sites at the rim of membrane protrusions ( Figure 4F , G ) and induce normal cell spreading ( Figure 4—figure supplement 4A ) despite proper , although weaker , localisation to β1 integrin-positive adhesion sites ( Figure 4—figure supplement 4B , C ) . Interestingly , mature FAs in K2ΔPHGFP-expressing cells were prominent after 30 min and contained significant amounts of paxillin , indicating that paxillin is recruited to mature FAs in a kindlin-2-independent manner ( Figure 4F ) . 10 . 7554/eLife . 10130 . 018Figure 4 . Kindlin binds and recruits paxillin to NAs . ( A ) GFP-IP of lysates from HEK 293T cells overexpressing GFP-tagged paxillin , Hic5 and leupaxin constructs ( Pxn , paxillin; Hic5; Lpx , leupaxin ) and K2flag reveal interaction of kindlin-2 with all three paxillin family members . ( B ) GFP-IP of lysates from HEK 293T cells overexpressing GFP-tagged paxillin truncation mutants and K2flag identifies the paxillin LIM3 domain as kindlin-2-binding domain . ( C ) GFP-IP of lysates from HEK 293T cells overexpressing GFP-tagged kindlin-2 truncation/deletion mutants and Cherry-tagged paxillin ( PxnCH ) identifies the kindlin-2 PH domain as paxillin binding domain . ( D ) Co-IP of endogenous paxillin and kindlin-2 from KindCtr cells . ( E ) Purified His-tagged paxillin-LIM3 domain pulls down recombinant kindlin-2 in a Zn2+-dependent manner . ( F ) K2GFP and K2ΔPHGFP expressing KindKo cells seeded on FN for the indicated times and stained for paxillin and β1 integrin . ( G ) Fluorescence intensity line scans from K2GFP- ( n=11 cells ) and K2ΔPHGFP- ( n=17 cells ) expressing KindKo cells cultured on FN for 10 min and stained for paxillin and β1 integrin; error bars indicate standard error of the mean . Bar , 10 µm . EDTA , ethylenediaminetetraacetic acid; FN , fibronection; GAPDH , glyceraldehyde-3-phosphate dehydrogenase; GFP , green fluorescent protein; ILK , integrin-linked kinase; IP , immunoprecipitation; K2GFP , green fluorescent protein-tagged kindlin-2; LIM , Lin-11 , Isl-1 and Mec-3; NAs , nascent adhesions; PH , pleckstrin homology . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01810 . 7554/eLife . 10130 . 019Figure 4—figure supplement 1 . Kindlin-1 , -2 and -3 interact with paxillin . GFP-IP of lysates from HEK293T cells expressing GFP , K1GFP , K2GFP and K3GFP followed by western blotting for Cherry-tagged paxillin ( PxnCH ) and GFP . GFP , green fluorescent protein; IP , immunoprecipitation; K1GFP , green fluorescent protein-tagged kindlin-1; K2GFP , green fluorescent protein-tagged kindlin-2; K3GFP , green fluorescent protein-tagged kindlin-3 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 01910 . 7554/eLife . 10130 . 020Figure 4—figure supplement 2 . Expression of paxillin family members in different cell lines . qPCR of paxillin ( Pxn ) , Hic5 , and leupaxin ( Lpxn ) from cDNAs generated from wild type fibroblasts ( Fibrobl . ) , keratinocytes ( Kerat . ) , RAW 264 . 7 macrophages ( RAW ) and T cells ( TC ) . Results are normalized to the isoform with highest expression in the respective cell types ( n=3 independent repeats; error bars show standard error of the mean ) . cDNA , complementary DNA; GAPDH , glyceraldehyde-3-phosphate dehydrogenase; qPCR , quantitative polychromase chain reaction . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02010 . 7554/eLife . 10130 . 021Figure 4—figure supplement 3 . Direct interaction between paxillin and kindlin-2 . ( A ) Domain organization of kindlin-2 ( F0 , 1 , 3: FERM domains 0 , 1 , 3; F2a , F2b: N-terminal ( F2a ) and C-terminal ( F2b ) halves of FERM domain 2 connected by a PH domain and small linkers on each side of the PH domain ) , and paxillin ( L1-4: LIM domains 1-4; grey stripes represent LD-rich motifs ) . The ILK interaction site in kindlin-2 ( N-terminal linker region located before the PH domain ) and the FA-targeting region of paxillin ( LIM3 domain ) are indicated in red; the black lines show the length of the truncation mutants . ( B ) Full-length paxillin pulls down recombinant kindlin-2 in a Zn2+-dependent manner . FERM , Four-point-one , ezrin , radixin , moesin; ILK , integrin-linked kinase; LD , leucine-aspartate repeat; LIM , Lin-11 , Isl-1 and Mec-3; PH , pleckstrin homology . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02110 . 7554/eLife . 10130 . 022Figure 4—figure supplement 4 . K2ΔPHGFP fails to recruit paxillin to β1 integrin-positive adhesions in KindKo cells . ( A ) Boxplots show the distribution of spreading areas for K2GFP and K2ΔPHGFP seeded on FN for the indicated times ( n> 65 cells per time point ) . Significances for indicated pairs were calculated by a Mann–Whitney U test . ( B ) Confocal images of K2GFP- and K2ΔPHGFP-expressing KindKo cells seeded on FN for 10 min and stained for total β1 integrin and paxillin . ( C ) Confocal images of K2GFP- and K2ΔPHGFP-expressing KindKo cells seeded on FN for 10 min and stained for 9EG7 . Bars , 10 µm . GFP , green fluorescent protein; FN , fibronectin; K2GFP , green fluorescent protein-tagged kindlin-2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 022 These findings indicate that the PH domain of kindlin-2 directly binds the LIM3 domain of paxillin and recruits paxillin into NAs but not into mature FAs . Our findings revealed that kindlin-2 is required to recruit paxillin to NAs . Paxillin in turn , was shown to bind , cluster and activate FAK in NAs , which leads to the recruitment of p130Cas , Crk and Dock followed by the activation of Rac1 and the induction of cell spreading , and , in concert with growth factor signals , to the activation of Akt-1 followed by the induction of cell proliferation and survival ( Schlaepfer et al . , 2004; Bouchard et al . , 2007; Zhang et al . , 2014; Brami-Cherrier et al . , 2014 ) . We therefore hypothesized that the recruitment of paxillin and FAK by kindlin-2 triggers the isotropic spreading and expansion of TlnKo cells . To test this hypothesis , we seeded our cell lines on FN or poly-L-lysine ( PLL ) in the presence or absence of epidermal growth factor ( EGF ) and Mn2+ . We found that EGF induced similar phosphorylation of tyrosine-992 ( Y992 ) of the epidermal growth factor receptor ( pY992-EGFR ) in control , TlnKo and KindKo cells . The phosphorylation of tyrosine-397 of FAK ( pY379-FAK ) in KindCtr cells was strongly induced after the adhesion of control cells on FN and was not further elevated after the addition of EGF and Mn2+ ( Figure 5A and Figure 5—figure supplement 1 ) . TlnKo cells also increased pY397-FAK as well as pY31-Pxn and pY118-Pxn levels upon adhesion to FN , however , significantly less compared to control cells ( Figure 5A and Figure 5—figure supplement 1A-C ) . Furthermore , EGF and Mn2+ treatments further increased pY397-FAK levels in TlnKo cells and localized pY397-FAK to peripheral NA-like adhesions ( Figure 5A , B and Figure 5—figure supplement 1A-C ) . In sharp contrast , KindKo cells seeded on FN or treated with EGF and Mn2+ failed to induce pY397-FAK , pY31-Pxn , pY118-Pxn ( Figure 5A and Figure 5—figure supplement 1A-C ) and localize pY397-FAK to peripheral membrane regions ( Figure 5B ) . Importantly , re-expression of Talin1-Venus in TlnKo and Kindlin2-GFP and KindKo cells fully rescued these signaling defects ( Figure 5—figure supplement 1B , C ) . Furthermore , stable expression of K2GFP in KindKo cells rescued pY397-FAK and pS473-Akt levels ( Figure 5C ) and co-precipitated paxillin and FAK with K2GFP ( Figure 5—figure supplement 2 ) . In contrast , stable expression of K2ΔPHGFP in KindKo cells failed to co-precipitate paxillin and FAK ( Figure 5—figure supplement 2 ) and induce pY397-FAK and pS473-Akt ( Figure 5C ) . 10 . 7554/eLife . 10130 . 023Figure 5 . The kindlin/paxillin complex induces FAK signaling and cell spreading . ( A ) FAK and EGFR activation after seeding serum-starved KindCtr , TlnKo and KindKo cells on PLL or FN and treating them with or without EGF and Mn2+ . ( B ) Immunofluorescence staining of activated ( Tyr-397 phosphorylated ) FAK and F-actin in cells seeded on FN and treated with Mn2+ for 30 min ( FAKGFP indicates exogenous expression of FAKGFP fusion protein ) . ( C ) FAK and Akt activation in KindKo cells stably transduced with K2GFP or K2ΔPHGFP either seeded on FN or kept in suspension . GFP indicates similar expression of transduced GFP-tagged constructs . GAPDH levels served to control loading . ( D ) Levels of phosphorylated signaling mediators downstream of FAK in Mn2+-treated , serum-starved or EGF-treated KindCtr , TlnKo and KindKo cells . GAPDH levels served to control loading . ( E ) Quantification of lamellipodia formation of FN-seeded TlnKo and KindKo cells treated with Mn2+ and either DMSO or the FAK inhibitor PF-228 ( n=3 independent repeats; >100 cells/condition; error bars indicate standard error of the mean; significances are given in comparison to the DMSO control ) . ( F ) FAK activity in TlnKo and KindKo cells stably transduced with FAKGFP ( n=3 independent experiments ) . ( G ) Quantification of lamellipodia formation in TlnKo and KindKo cells stably transduced with FAKGFP ( n=3 independent experiments; significances are given in comparison to untreated control; error bars indicate standard error of the mean ) . Bar , 10 µm . DMSO , dimethyl sulfoxide; EGF , epidermal growth factor; EGFR , epidermal growth factor receptor; FAK , focal adhesion kinase; FAKGFP , green fluorescent protein-tagged FAK; FN , fibronectin; GAPDH , glyceraldehyde-3-phosphate dehydrogenase; GFP , green fluorescent protein; PLL , poly-L-lysine . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02310 . 7554/eLife . 10130 . 024Figure 5—figure supplement 1 . FAK phosphorylation in TlnCtr , TlnKo , TlnKo+T1V , KindCtr , KindKo and KindKo+K2GFP cells . ( A ) Densitometric quantification of western blot signals of lysates from untreated , EGF- and Mn2+-treated KindCtr , TlnKo and KindKo cells seeded either on FN or PLL and probed with anti-Tyr-397 phosphorylated FAK ( pY397-FAK ) antibodies ( n=3 independent repeats; significances are calculated with respect to PLL adherent cells; error bars indicate standard error of the mean ) . ( B , C ) Western blotting of indicated signaling proteins in untreated , EGF- and Mn2+-treated TlnCtr , TlnKo and TlnKo cells re-expressing Venus-tagged talin-1 ( TlnKo+T1V ) ( B ) , and KindCtr , KindKo and KindKo cells re-expressing GFP-tagged kindlin-2 ( KindKo+K2GFP ) ( C ) seeded either on FN or PLL . EGF , epidermal growth factor; FAK , focal adhesion kinase; K2GFP , green fluorescent protein-tagged kindlin-2; FN , fibronectin; GFP , green fluorescent protein; PLL , poly-L-lysine; T1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02410 . 7554/eLife . 10130 . 025Figure 5—figure supplement 2 . Kindlin-2 forms a ternary complex with paxillin and FAK . GFP-IP in lysates of K2GFP- , K2ΔPHGFP- or GFP-reconstituted KindKo cells overexpressing Myc-tagged FAK ( FAK-Myc ) and Cherry-tagged paxillin ( PxnCH ) . K2GFP but not K2ΔPHGFP forms a ternary complex with paxillin and FAK . FAK , focal adhesion kinase; GFP , green fluorescent protein; IP , immunoprecipitation; K2GFP , green fluorescent protein-tagged kindlin-2 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02510 . 7554/eLife . 10130 . 026Figure 5—figure supplement 3 . Activity of signaling mediators downstream of FAK in TlnCtr , TlnKo , TlnKo+T1V , KindCtr , KindKo and KindKo+K2GFP cells . ( A , B ) Western blotting of indicated signaling proteins in untreated ( first three lanes ) , EGF- and Mn2+-treated TlnCtr , TlnKo and TlnKo cells re-expressing Venus-tagged talin-1 ( TlnKo+T1V ) ( A ) , and KindCtr , KindKo and KindKo cells re-expressing GFP-tagged kindlin-2 ( KindKo+K2GFP ) ( B ) seeded on FN . EGF , epidermal growth factor; FN , fibronectin; GAPDH , glyceraldehyde-3-phosphate dehydrogenase; GFP , green fluorescent protein; K2GFP , green fluorescent protein-tagged kindlin-2; T1V , Venus-tagged full length talin-1 . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 02610 . 7554/eLife . 10130 . 027Figure 5—figure supplement 4 . Cell spreading of FAK overexpressing TlnKo and KindKo cells . ( A , B ) Cell spreading area of TlnKo ( A ) and KindKo cells ( B ) seeded on PLL , FN or on FN after FAKGFP overexpression measured by image quantification ( n=3; independent repeats are pooled; >100 cells/condition and repeat; resulting areas are shown as binning histograms; significances are calculated between non-transfected cells and FAKGFP expressing cells plated on FN ) . FAK , focal adhesion kinase; FAKGFP , green fluorescent protein-tagged FAK; FN , fibronectin; GFP , green fluorescent protein; PLL , poly-L-lysine . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 027 In line with previous reports showing that the paxillin/FAK complex can trigger the activation of p130Cas ( Zhang et al . , 2014 ) and , in cooperation with EGFR signaling , the activation of Akt ( Sulzmaier et al . , 2014 , Deakin et al . , 2012 ) , we observed Y410-p130Cas , pT308-Akt , S473-Akt and pT202/pY204 Erk1/2 phosphorylation after Mn2+ and/or EGF treatment of FN-seeded control and rescued cells , and to a slightly lesser extent TlnKo cells ( Figure 5D , Figure 5— figure supplement 3A , B ) . In contrast , FN-seeded KindKo cells failed to activate p130Cas and showed reduced Akt and Erk1/2 phosphorylation in response to EGF ( Figure 5D , Figure 5—figure supplement 3A , B ) . Finally , we tested whether the impaired activity of FAK contributed to the spreading defect of KindKo cells by chemically inhibiting FAK activity in TlnKo cells or by overexpressing FAK in KindKo cells ( Figure 5E–G ) . The experiments revealed that inhibiting FAK reduced lamellipodia formation of TlnKo cells to an extent that was similar to untreated KindKo cells ( Figure 5E ) . Conversely , overexpression of FAKGFP in KindKo cells resulted in high active FAK , increased lamellipodial formation and increased cell spreading in TlnKo and KindKo cells ( Figure 5F , G and Figure 5—figure supplement 4A , B ) . Altogether , these findings show that the kindlin-2/paxillin complex in NAs recruits and activates FAK to induce cell spreading and increase the strength of Akt signaling .
While the functions of talin and kindlin for integrin activation , adhesion and integrin-dependent signaling in hematopoietic cells are firmly established , their roles for these processes in non-hematopoietic cells are less clear . To clarify this issue , we established mouse fibroblast cell lines that lacked either talin-1/2 ( TlnKo ) or kindlin-1/2 ( KindKo ) and tested whether they were able to activate integrins and mediate substrate adhesion and signaling . In line with previous reports ( Bottcher et al . , 2012; Margadant et al . , 2012 ) , the deletion of Tln1/2 or Fermt1/2 genes changed the surface levels of laminin- and collagen-binding integrins . Since surface levels of α5 and αv integrins remained unchanged between TlnKo and KindKo cells , we were able to establish the specific roles of talin and kindlin for the function of FN-binding integrins under identical conditions . A major finding of our study demonstrates that integrin affinity regulation ( activation ) is essential for fibroblast adhesion and depends on both talin and kindlin-2 ( Figure 6A , D ) . The unambiguity of this finding was unexpected in light of several reports showing that integrin activation and integrin-mediated adhesion still occurs in talin-depleted cells , or is inhibited when kindlin-2 is overexpressed ( Harburger et al . , 2009; Wang et al . , 2011; Lawson et al . , 2012 ) . The previous studies that addressed the functional properties of talin used siRNA-mediated protein depletion , a combination of gene ablation and siRNA technology , or approaches to interfere with talin recruitment to NAs either by ablating the talin upstream protein FAK or by expressing an integrin that harbors a mutation in the talin binding site . Since the majority of approaches deplete rather than eliminate proteins from cells and adhesion sites , the respective cells were most likely recruiting sufficient residual protein to adhesion sites to allow integrin activation , cell adhesion and spreading , and the assembly of adhesion- and signaling-competent NAs . It is possible that not all integrin molecules have to be occupied by talin and therefore low levels of talin suffice , particularly in NAs that were shown by fluorescence correlation spectroscopy to contain only half the number of talin relative to α5β1 integrin and kindlin-2 molecules ( Bachir et al . , 2014 ) . However , when the entire pool of talin is lost or decreased below certain thresholds ( Margadant et al . , 2012 ) integrins remain inactivate and consequently adhesion sites do not form . With respect to kindlin , it was reported that overexpressed kindlin-2 in CHO cells inhibits rather than promotes talin head domain-induced α5β1 integrin activation ( Harburger et al . , 2009 ) . An integrin inhibiting effect of kindlin-2 is inconsistent with our study , which identified a crucial role for kindlin in integrin activation , as well as with other studies also demonstrating that kindlin-2 promotes integrin functions ( Montanez et al . , 2008 ) . It could well be that the reported inhibition of α5β1 by kindlin-2 represents an artifact that arose from protein overexpression . 10 . 7554/eLife . 10130 . 028Figure 6 . Model for the roles of talin and kindlin during inside-out and outside-in signaling of α5β1 integrin . Integrin subunits are modelled according to Zhu et al . ( 2008 ) , with the α5 subunit in green and the β1 subunit in blue showing the bent and clasped low affinity and the extended and unclasped high affinity conformations; the 9EG7 epitope is marked as red dot at the β1 leg and the FN ligand as beige dimers . ( A ) α5β1 integrin fails to shift from a bent to an extended/unclasped , high affinity state in the absence of talin-1/2 or kindlin-1/2; the bent/clasped conformation brings the EGF-2 domain of the β subunit in close contact with the calf domain of the α5 subunit and prevents exposure of the 9EG7 epitope . ( B ) In the absence of talin ( TlnKo ) and presence of Mn2+ , kindlin-2 allows adhesion by stabilizing the high affinity conformation of a low number of integrins and the direct binding of paxillin , leading to nucleation of integrins , recruitment of FAK , FAK-dependent signaling and lamellipodia formation . ( C ) In the absence of kindlins ( KindKo ) , talin stabilizes the high affinity conformation of a low number of integrins but does not enable paxillin recruitment and lamellipodia formation . ( D ) In normal fibroblasts , binding of kindlin and talin to the β1 tail is associated with the stabilisation of the unclasped α5β1 heterodimer and 9EG7 epitope exposure . ( E ) Kindlin recruits paxillin and FAK through the kindlin-PH domain and ILK/Pinch/Parvin ( IPP; not shown ) in a talin-independent manner and induces cell spreading , proliferation and survival . ( F ) The high affinity conformation of α5β1 integrin is stabilized by linkage of the β1 tail to the actin cytoskeleton through talin ( and potentially the IPP complex; not shown ) . The arrow length indicates integrin conformations existing at equilibrium . EGF , epidermal growth factor; FAK , focal adhesion kinase; FN , fibronectin; ILK , integrin-linked kinase; IPP , integrin-linked kinase-Pinch-Parvin; SFK , src family kinases . DOI: http://dx . doi . org/10 . 7554/eLife . 10130 . 028 Integrin activation can be induced with Mn2+ , whose binding to the ectodomain of β subunits directly shifts integrins into the high affinity state without the requirement for inside-out signals ( Mould et al . , 1995 ) . We observed that Mn2+-treated TlnKo and KindKo cells expressed the activation-dependent epitope 9EG7 and adhered to FN , albeit at significantly lower levels and efficiencies than the normal parental or rescued cells ( Figure 6B , C ) . This observation strongly indicates that talin and kindlin also cooperate to maintain the extended and unclasped conformation of active integrins . Although it is not known how talin and kindlin keep integrins in an active state , it is possible that they stabilize this conformation by linking the unclasped β integrin cytoplasmic domain to the plasma membrane and/or to cortical actin , which may firmly hold separated integrin α/β subunits apart from each other . The expression of mutant talins and kindlins in our cells should allow us to examine these possibilities in future . Finally , our study also revealed that Mn2+-treated TlnKo cells began to form large , circumferential lamellipodia that eventually detached from FN , leading to the collapse of the protruded membrane . This initial isotropic spreading was significantly less frequent in KindKo cells , and has also been observed in talin-2-depleted talin-1–/– cells on FN , although these cells did not require Mn2+ for inducing spreading , which is likely due to the presence of residual talin-2 that escaped siRNA-mediated depletion ( Zhang et al . , 2008; Zhang et al . , 2014 ) . These findings strongly suggest that integrin binding to FN enables kindlin-2 in TlnKo cells to cluster β1 integrins ( as shown for αIIbβ3 by kindlin-3 in Ye et al . , 2013 ) and to trigger a signaling process that initiates spreading . To find a mechanistic explanation for the kindlin-2-mediated cell spreading , we used the yeast-two-hybrid technology to identify paxillin as a novel and direct binding partner of kindlin-2 . The interaction of the two proteins occurs through the LIM3 domain of paxillin , which was previously identified as integrin adhesion-targeting site ( Brown et al . , 1996 ) , and the PH domain of kindlin-2 . It is not unusual that PH domains fulfill dual roles by binding phospholipids and proteins , either simultaneously or consecutively ( Scheffzek and Welti , 2012 ) . The expression of a PH domain-deficient kindlin-2 in KindKo cells rescues adhesion to FN and FA maturation , however , significantly impairs spreading and plasma membrane protrusions . This finding together with the observations that paxillin-null fibroblasts and embryonic stem cells have defects in spreading , adhesion site remodeling and formation of lamellipodia ( Hagel et al . , 2002 , Wade et al . , 2002 ) indicates that the kindlin-2/paxillin complex induces the elusive signaling process , leading to initial spreading of TlnKo and talin-depleted cells ( Zhang et al . , 2008 ) . Indeed , the kindlin-2/paxillin complex in NAs recruits FAK ( Deramaudt et al . , 2014 , Thwaites et al . , 2014; Choi et al . , 2011 ) , which cooperates with growth factor receptors ( such as EGFR ) to induce signaling pathways that activate Erk and Akt to promote proliferation and survival , as well as Arp2/3 and Rac1 to induce actin polymerization and membrane protrusions ( Figure 6B , E ) . Kindlin-2 also recruits ILK , which binds in the vicinity of the kindlin-2 PH domain and links integrins to actin and additional signaling pathways ( Figure 6E ) . The short-lived nature of the initial spreading of TlnKo and talin-depleted ( Zhang et al . , 2008 ) cells shows that talin concludes the integrin-mediated adhesion process in NAs ( Figure 6F ) and induces the maturation of FAs . The formation of paxillin-positive FAs in cells expressing the PH domain-deficient kindlin-2 suggests that the recruitment of paxillin to FAs occurs either in a kindlin-independent manner or through a modification of kindlin in a second binding motif .
The floxed kindlin-1 ( Fermt1flox/flox ) , floxed talin-1 ( Tln1flox/flox ) and the constitutive talin-2-null ( Tln2–/– ) mouse strains have been described ( Rognoni et al . , 2014; Nieswandt et al . , 2007; Conti et al . , 2009 ) . The floxed kindlin-2 ( Fermt2flox/flox ) mouse strain generated via recombinant recombination in embryonic stem cells ( Fassler and Meyer , 1995 ) carries loxP sites flanking exon 15 , which contains the stop codon and the polyadenylation signal of the murine Fermt2 gene . Homologous recombination and germ line transmission were verified by Southern blots , and the frt-flanked neo cassette was removed with a transgenic mouse strain carrying a deleter-flipase gene . Floxed talin-1 and talin-2-null mice , and floxed kindlin-1 and kindlin-2 mice were intercrossed to generate Tln1flox/flox Tln2–/– and Fermt1flox/flox Fermt2flox/flox mice . The cell lines used in this study are mouse fibroblasts derived from the kidneys of 21 d old mice , immortalized by retrovirally transducing the SV40 large T antigen , cloned ( TlnCtr and KindCtr ) and finally infected with an adenovirus to transduce the Cre recombinase resulting in talin-null ( TlnKo ) and kindlin-null ( KindKo ) cells . The parental cell lines were authenticated based on morphological criteria and the surface experession of specific integrins . All cells were cultured under standard cell culture conditions using Dulbecco's modified Eagle's medium ( DMEM ) supplemented with 8% fetal calf serum ( FCS ) and Penicillin/Streptomycin but not subjected to mycoplasma contamination testing . Flow cytometry was carried out with a FACSCantoTMII cytometer ( BD Biosciences , Franklin Lakes , NJ , USA ) equipped with FACS DiVa software ( BD Biosciences ) using standard procedures . Data analysis was carried out with the FlowJo program ( version 9 . 4 . 10 ) . Fibroblasts were incubated with primary antibodies diluted in FACS-Tris buffered saline ( FACS-TBS; 30 mM Tris , pH 7 . 4 , 180 mM NaCl , 3 . 5 mM KCl , supplemented with 1 mM CaCl2 , 1 mM MgCl2 , 3% BSA , 0 , 02% NaN3 ) for 1 hr on ice , washed twice with cold FACS-TBS and finally incubated with the secondary antibody for 45 min on ice . Total RNA was extracted with the RNeasy Mini extraction kit ( Qiagen , Germany ) from cultured cells , cDNAs were prepared with an iScript cDNA Synthesis Kit ( BioRad , Germany ) and real-time polymerase chain reaction ( PCR ) was performed with an iCycler ( BioRad ) . Each sample was measured in triplicate and values were normalized to Gapdh . Primer sequences for Lpxn and Pxn were from PrimerBank ( Spandidos et al . , 2010 ) ( Lpxn: 26080416a1; aPxn: 114326500c2; bPxn: 22902122a1 ) , GAPDH primers were described before ( Rognoni et al . , 2014 ) and Hic5 primers were newly designed ( Hic5-fwd: 5’-ttcctttgcagcggttgttcc-3’; Hic5-rev: 5’-ggttacagaagccacatcgtggg-3’ ) . The following antibodies or molecular probes were used at indicated concentrations for western blot ( WB ) , immunofluorescence ( IF ) or flow cytometry ( FACS ) : kindlin-1 ( home made ) , ( Ussar et al . , 2008 ) WB: 1:5000 , IF: 1:1000; kindlin-2 ( MAB2617 from Millipore , Germany ) WB: 1:1000 , IF: 1:500; talin ( 8D4 from Sigma , Germany ) WB: 1:1000; talin ( sc-7534 from Santa Cruz , Germany ) IF: 1:500; talin-1 ( ab57758 from Abcam , UK ) WB: 1:2000; talin-2 ( ab105458 from Abcam ) WB: 1:2000; GAPDH ( 6C5 from Calbiochem , Billerica , MA , USA ) WB: 1:10 , 000; Paxillin ( 610051 from BD Transduction Laboratories , Franklin Lakes , NJ , USA ) WB: 1:1000 , IF: 1:400; integrin β1-488 ( 102211 from Biolegend , San Diego , CA , USA ) IF: 1:400 , FACS: 1:200; integrin β1 ( MAB1997 from Chemicon , Billerica , MA , USA ) FACS: 1:400; integrin β1-647 ( 102213 from Biolegend ) IF: 1:200; integrin β1 ( home-made ) , ( Azimifar et al . , 2012 ) IF: 1:400; integrin β3-biotin ( 553345 from PharMingen , Franklin Lakes , NJ , USA ) FACS: 1:200; integrin β3 ( M031-0 from Emfret , Germany ) IF: 1:200; integrin α2-FITC ( 554999 from BD Biosciences ) FACS: 1:100; integrin α3 ( AF2787 from R&D , Germany ) FACS: 1:200; integrin α5-biotin ( 557446 from Pharmingen ) FACS: 1:200 , IP 1µg; integrin α5 ( 4705 from Cell Signaling , Germany ) WB: 1:1000; integrin α6-FITC ( 555735 from Pharmingen ) FACS 1:100; integrin αv-biotin ( 551380 from Pharmingen ) FACS: 1:200; β1-integrin 9EG7 ( 550531 from BD Biosciences , San Diego , CA , USA ) IF: 1:200; FACS: 1:200; fibronectin ( AB2033 from Millipore ) IF: 1:500; IgG2a rat isotype control ( 13-4321 from eBioscience , Germany ) FACS: 1:200; IP 1µg; Tritc-Phalloidin ( P1951 from Sigma ) IF: 1:400; Flag-tag-HRP ( 8592 from Sigma ) WB: 1:10 , 000; GFP ( A11122 from Invitrogen , Germany ) WB: 1:2000; Cherry ( PM005 from MBL , Woburn , MA , USA ) WB:1:1000; Myc ( 05-724 from Millipore ) WB 1:2000; FAK ( 06-543 from Upstate , Billerica , MA , USA ) WB: 1:1000; FAK ( 3285 from Cell Signaling ) WB ( 1:1000 ) ; phospho-Y397 FAK ( 3283 from Cell Signaling ) WB: 1:1000; phospho-Y397 FAK ( 44624G from Biosource , Waltham , MA , USA ) WB: 1:1000 , IF: 1:400; ILK ( 611803 from Transduction Labs ) WB: 1:5000; IF: 1:500; phospho-Y992 EGFR ( 2235 from Cell Signaling ) WB: 1:2000; phospho-Y31 Paxillin ( 44720G from Invitrogen ) WB: 1:1000; phospho-Y118 Paxillin ( 44722G from Invitrogen ) WB: 1:1000; p130Cas ( P27820 Transduction Labs ) WB: 1:1000; phospho-Y410 p130 Cas ( 4011S from Cell Signaling ) WB: 1:1000; Akt ( 9272 from Cell Signaling ) WB: 1:1000; phospho-S473 Akt ( 4060 from Cell Signaling ) WB: 1:1000; phosho-T308 Akt ( 9275 from Cell Signaling ) WB: 1:1000; Erk1/2 ( 9102 from Cell Signaling ) WB: 1:1000; Erk1/2 phosphorylated T202 Y204 ( 4376 Cell Signaling ) WB: 1:1000 . The following secondary antibodies were used: goat anti-rabbit Alexa 488 ( A11008 ) , goat anti-mouse Alexa 488 ( A11029 ) , goat anti-rat Alexa 488 ( A11006 ) , goat anti-mouse Alexa 546 ( A11003 ) , donkey anti-mouse Alexa 647 ( A31571 ) , goat anti-rabbit Alexa 647 ( A21244 ) , ( all from Invitrogen ) FACS: 1:500 , IF: 1:500; streptavidin-Cy5 ( 016170084 ) FACS: 1:400; goat anti-rat horseradish peroxidase ( HRP ) ( 712035150 ) ( both from Dianova , Germany ) WB: 1:10 , 000 , donkey anti-rabbit Cy3 ( 711-165-152 ) ( from Jackson ImmunoResearch , West Grove , PA , USA ) IF: 1:500 , goat anti-mouse HRP ( 172-1011 ) and goat anti-rabbit HRP ( 172-1019 ) ( both from BioRad ) WB: 1:10 , 000 . The FAK inhibitor PF-228 ( PZ0117 from Sigma ) was dissolved in dimethyl sulfoxide at 10 mM and used at 1:2000 . The recombinant expression of kindlin-2 , full-length paxillin ( paxillin-FL ) and paxillin-LIM3 in Escherichia coli Rosetta cells ( Merck Millipore ) was induced with 1 mM or 0 . 2 mM IPTG , respectively , at 18°C for 22 hr . After cell lysis and clarification of the supernatant , kindlin-2 was purified by Ni-NTA affinity chromatography ( Qiagen ) . Eluate fractions containing kindlin-2 were pooled , cleaved with SenP2 protease and purified by size-exclusion chromatography ( Superdex 200 26/600 , GE Healthcare , UK ) yielding unmodified murine kindlin-2 . The paxillin constructs were purified by Ni-NTA affinity chromatography ( Qiagen ) , and subsequent size-exclusion chromatography ( SEC650 , BioRad ) to obtain N-terminally tagged His10-SUMO3-paxillin-FL and His10-SUMO3-paxillin-LIM3 domain , respectively . For immunostaining , cells were cultured on plastic ibidi-µ-slides ( 80826 from Ibidi , Germany ) coated with 20 µg ml–1 FN ( Calbiochem ) . Cells were routinely fixed with 4% paraformaldehyde ( PFA ) ( w/v ) in phosphate buffered saline ( PBS; 180 mM NaCl , 3 . 5 mM KCl , 10 mM Na2HPO4 , 1 . 8 mM K2H2PO4 ) for 10 min at room temperature ( RT ) or with –20°C cold acetone–methanol when indicated . If necessary , cells were solubilized with staining buffer ( PBS supplemented with 0 . 1% Triton X-100 ( v/v ) and 3% BSA ( w/v ) ) or with –20°C cold methanol for kindlin-2 staining . Background signals were blocked by incubating cells for 1 hr at RT in staining buffer . Subsequently , they were incubated in the dark with primary and secondary antibodies diluted in staining buffer . Fluorescent images were aquired with a LSM 780 confocal microscope ( Zeiss , Germany ) equiped with a 100×/NA 1 . 46 oil objective and with a DMIRE2-SP5 confocal microscope ( Leica , Germany ) equiped with a 40×/NA 1 . 25 or 63×/NA 1 . 4 oil objective using Leica Confocal software ( version 2 . 5 build 1227 ) . Brightfield images were aquired with an Axioskop ( Carl Zeiss ) 40×/NA 0 . 75 objective and DC500 camera with IM50 software ( Leica ) . Z-stack projection and contrast adjustments ImageJ ( v1 . 47 ) were used for further image analysis . Super-resolution imaging was carried out by direct stochastic optical reconstruction microscopy ( dSTORM ) ( van de Linde et al . , 2011 ) , which is based on precise emitter localization . To induce reversible switching of the Alexa 647 label and reduce photobleaching , imaging was performed in imaging solution ( 50% Vectashield ( v/v ) ( H-1000; Vector Laboratories , Burlingame , CA , USA ) , 50% TBS ( v/v ) , pH=8 . 0 ) supplemented with 50 mM β-mercaptoethylamine ( Sigma-Aldrich; M9768 ) . dSTORM was implemented on a custom built total internal reflection fluourescence ( TIRF ) system ( Visitron Systems , Germany ) based on a Zeiss Axiovert 200M with fiber-coupled lasers . Sample were excited with a 640 nm laser in a TIRF mode using a Zeiss α Plan-Fluar 100×/NA 1 . 45 oil objective . The emitted light was detected in the spectral range 660–710 nm through a Semrock FF02-685/40-25 bandpass filter ( Semrock Inc . , Rochester , NY , USA ) . Images were recorded with a Photometrics Evolve Delta emCCD camera ( Photometrics , Huntington Beach , CA , USA ) , with its EM gain set to 250 . Additional magnification by a factor of 1 . 6 resulted in a pixel size of 100 nm . For each final image , a total of 20 , 000 frames with an exposure time of 14 ms were recorded . A standard TIRF imaging of the same sample in the green channel ( anti-paxillin ) was achieved by illumination with a 488 nm laser and detection in the spectral range 500–550 nm through a Chroma Et 525/50 bandpass filter ( Chroma Technology Corporation , Bellows Falls , VT , USA ) . Simultaneous dual-colour imaging of both the green and the red channel was realized with a Hamamatsu W-View Gemini image splitter ( Hamamatsu Photonics , Bridgewater , NJ , USA ) mounted between the microscope and the camera . Image analysis was carried out with the ImageJ plugin ThunderSTORM ( Ovesny et al . , 2014 ) and standard tools of ImageJ . Heat maps of density of blink events were created using the 2D-Frequency Count/Binning module of OriginPro 9 . 1 ( OriginLab Corporation , Northampton , MA , USA ) . Tipless , 200 µm long V-shaped cantilevers ( spring constants of 0 . 06 N m–1; NP-O , Bruker , Billerica , MA , USA ) were prepared for cell attachment as described ( Friedrichs et al . , 2010 ) . Briefly , plasma cleaned cantilevers were incubated in 2 mg ml–1 ConA ( Sigma ) in PBS at 4°C overnight . Polydimethylsiloxan ( PDMS ) masks were overlaid on glass bottoms of Petri dishes ( 35 mm FluoroDish , World Precision Instruments , Sarasota , FL , USA ) to allow different coatings of the glass surface ( Te Riet et al . , 2014 ) . PDMS-framed glass surfaces were incubated overnight with 50 µg ml–1 FN-RGD and 50 µg ml–1 FN-∆RGD in PBS at 4°C . Overnight serum-starved fibroblasts ( TlnCtr , KindCtr , TlnKo , KindKo ) grown on FN-coated ( Calbiochem ) 24 well plates ( Thermo Scientific , Denmark ) to confluency of ~ 80% were washed with PBS and detached with 0 . 25% ( w/v ) trypsin/EDTA ( Sigma ) . Detached cells were suspended in single-cell force spectroscopy ( SCFS ) medium ( DMEM supplemented with 20 mM HEPES ) containing 1% ( v/v ) FCS , pelleted and further resuspended in serum-free SCFS medium . Detached cells were left suspended in SCFS media to recover from detachment for ~1 hr ( Schubert et al . , 2014 ) . For the activation or chelation assay , the detached cells were incubated in SCFS media supplemented with 0 . 5 mM Mn2+ or 5 mM EDTA , respectively , for ~1 hr and SCFS was performed in the presence of the indicated supplement . SCFS was performed using an AFM ( NanoWizard II , JPK Instruments , Germany ) equipped with a CellHesion module ( JPK Instruments ) mounted on an inverted optical microscope ( Zeiss Axiovert 200M ) . Measurements were performed at 37°C , controlled by a PetriDish Heater ( JPK Instruments ) . Cantilevers were calibrated using the equipartition theorem ( Hutter and Bechhoefer , 1993 ) . To attach a single cell to the cantilever , cell suspensions were pipetted to the region containing the FN-∆RGD coating . The ConA functionalized cantilever was lowered onto a single cell with a velocity of 10 µm s−1 until reaching a contact force of 5 nN . After 5 s contact , the cantilever was retracted from the Petri dish by 50 µm and the cantilever-bound cell was left for incubation for >10 min . For adhesion experiments , the cantilever-bound cell was brought into contact with the FN-∆RGD coated support at a contact force of ~2 nN for 5 , 20 , 50 and 120 s and then retracted while measuring the cantilever deflection and the distance travelled . Subsequently , the cell adhesion to the FN-RGD coated support was characterized as described . In case cantilever attached cells showed morphological changes ( e . g . spreading ) they were discarded . The approach and retract velocity of the cantilever was 5 µm s–1 . The deflection of the cantilever was recorded as force-distance curves . Adhesion forces were extracted from retraction force-distance curves using the AFM data processing software ( JPK Instruments ) . GFP-IPs were performed using µ-MACS anti-GFP magnetic beads ( 130-091-288 from Miltenyi , Germany ) . To pulldown recombinant kindlin-2 35 µg of purified His10-LIM3 or 10 µg of purified His10-paxillin-FL were incubated with 100 µl of 50% Ni-NTA-Agarose slurry ( Qiagen ) in pulldown buffer ( 20 mM Tris , pH 7 . 5 , 200 mM NaCl , 1 mM TCEP , 0 . 05% Tween20 ) for 1 hr at 4°C . After a first wash with 20 column volumes ( CV ) of pulldown buffer supplemented with 1 mM ZnCl2 and a second wash with 20 CV of pulldown buffer , 14 µg of purified kindlin-2 were added to 100 µl of Ni-NTA-agarose slurry and incubated for 30 min at 4°C . Subsequently , the Ni-NTA beads were washed three times with 20 CV of pulldown buffer supplemented with 25 mM imidazole and either 1 mM ZnCl2 or 1 mM EDTA . The beads were eluted with 50 µl pulldown buffer supplemented with 500 mM imidazole and analysed on a 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis ( SDS-PAGE ) . For immunoprecipitation of kindlin-2 or paxillin , control fibroblasts were lysed in lysis buffer ( 50 mM Tris , pH 8 . 0 , 150 mM NaCl , 1% Triton X-100 , 0 . 05% sodium deoxycholate , 10 mM EDTA ) . Lysates were incubated with kindlin-2 or paxillin antibodies for 2 hr at 4°C while rotating . Isotype-matched IgG was used as a negative control . After this , lysates were incubated with 50 µl protein A/G Plus Agarose ( Santa Cruz ) for 2 hr at 4°C . Following repeated washes with lysis buffer , proteins were eluted from the beads using Laemmli buffer and analyzed by western blotting . For the immunoprecipitation of α5 integrin from the cell surface of live cells , α5 integrins were labeled with a biotinylated anti-α5 integrin antibody ( PharMingen #557446 ) or an isotype control ( eBioscience # 13-4321 ) for 1 hr on ice . After two washes in ice-cold PBS to remove unbound antibody , cells were lysed in IP buffer ( 50 mM Tris , pH 7 . 5 , 150 mM NaCl , 1% Triton X-100 , 0 . 1% sodium deoxycholate , 1mM EDTA , and protease inhibitors ) and cleared by centrifugation . α5 integrin immuno-complexes were pulled-down by incubation with streptavidin-sepharose ( GE Healthcare ) overnight at 4°C with gentle agitation . After several washes with lysis buffer , proteins were subjected to SDS-PAGE and western blot analysis with antibodies against α5 and β1 integrin . Cells were grown to 70% confluency and then detached using trypsin/EDTA . Suspended cells were serum starved for 1 hr in adhesion assay medium ( 10 mM HEPES , pH 7 . 4; 137 mM NaCl; 1 mM MgCl2; 1 mM CaCl2; 2 . 7 mM KCl; 4 . 5 g L–1 glucose; 3% BSA ( w/v ) ) before 40 , 000 cells per well were plated out in the same medium supplemented with 8% FCS , and 5 mM Mn2+ if indicated . Plastic ibidi-µ-slides ( Ibidi; 80826 ) were coated with 10 µg ml–1 FN ( Calbiochem ) for adhesion or 20 µg ml–1 FN for spreading assays , 10 µg ml–1 LN ( 11243217001 from Roche , Germany ) , 10 µg ml–1 COL ( 5005B from Advanced Bio Matrix , Carlsbad , CA , USA ) , 10 µg ml–1 VN ( 07180 from StemCell , Canada ) or 0 . 1% PLL ( w/v ) ( Sigma; P4707 ) diluted in PBS . Seeded cells were centrifuged at 600 rpm in a Beckman centrifuge for 30 min at 37°C before they were fixed with 4% PFA ( w/v ) in PBS and stained with Phalloidin-TRITC and DAPI . For cell adhesion assays , nuclear staining of the whole well was imaged using a 2 . 5x objective and cell numbers were counted using ITCN plugin for imageJ ( Byun et al . , 2006 ) . For cell spreading assays , 12 confocal images of different regions of Phalloidin and DAPI stained cells were aquired using a Leica confocal microscope , cell spreading was quantified using imageJ . For time dependent and ligand concentration dependent adhesion on FN , 40 , 000 cells were plated on 96-well plates , vigorously washed after the indicated timepoints with PBS and fixed with 4% PFA . Cell attachment was meassured by crystal violet staining ( 0 . 1% in 20% methanol ) of cells in a absorbance plate reader at the wavelength of 570 nm . A hole of 15 mm diameter was drilled into the bottom of a 35 mm falcon tissue culture dish ( 353001 , Becton Dickinson ) and a coverslip ( Ø 25 mm , Menzel-Gläser , Germany ) , rinsed with ethanol , was glued to the dish with silicon glue ( Elastosil E43 , Wacker , Germany ) . After coating coverslips with 20 µg ml–1 FN ( Calbiochem ) overnight at 4°C , cells were plated and imaged in an inverted transmission light microscope ( Zeiss Axiovert 200 M , Carl Zeiss ) equipped with a climate chamber . Phase contrast images were taken with a ProEM 1024 EMCCD camera ( Princeton Instruments , Acton , MA , USA ) through a Zeiss Plan Neofluar 100x objective ( NA 1 . 3 , Ph3 ) . Frames were acquired at 30 sec or 1 min intervals and converted to time lapse movies using ImageJ . K2ΔPHGFP was cloned by PCR using the K2GFP cDNA ( Ussar et al . , 2006 ) as template and the Kind2fwd ( 5’-ctcgaggaggtatggctctggacgggataag -3’ , Kind2PHrev 5’-tggtcttgcctttaatatagtcagcaagtt -3’ ) , Kind2PHfwd ( 5’-ctatattaaaggcaagaccatggcagacag -3’ ) and Kind2rev ( 5’- tctagatcacacccaaccactggtga-3’ ) primers . The two fragments containing homologous regions ( indicated with bold letters in the primer sequences ) were fused by another round of amplification using the most 5’ and 3’ primers ( Kind2fwd and Kind2rev ) . The resulting PCR product was cloned into the K2GFP vector . The N- and C-terminal truncation constructs of kindlin-2 were cloned by PCR using K2GFP as template . The primer sequences were: Kind2-NT-fwd 5’-ctgtacaagtccggactc-3’ , Kind2-NT-rev 5’-gcggccgcctattttgctttatcaagaagagc-3’ , Kind2-CT-fwd 5’-ctcgagctatggataaagcaaaaaccaaccaag-3’ , Kind2-CT-rev 5’-gttatctagagcggccgc-3’ . Stable expression of K2ΔPHGFP and FAKGFP- or Myc-FAK ( a gift from Dr . Ambra Pozzi; Vanderbilt University , Nashville , USA ) cDNAs was achieved with the sleeping beauty transposase system ( Bottcher et al . , 2012 ) . Kindlin-1-GFP and Kindlin-3-GFP constructs have been described ( Ussar et al . , 2008; Moser et al . , 2008 ) . For stable expression of murine talin-1 and THD ( amino acids1-443 ) , the corresponding cDNAs were N-terminally tagged with Venus and cloned into the retroviral pLPCX vector . The constructs for GFP-tagged paxillin-LIM truncation mutants were generated by PCR from GFP- and Cherry-tagged α-paxillin ( Moik et al . , 2013 ) and cloned into the retroviral pLPCX vector . The primer sequences were: stop codon in bold: ΔLIM1-4fwd 5’- caccgttgccaaatgagggtctgtggagcc -'3 , ΔLIM1-4rev 5’- ggctccacagaccctcatttggcaacggtg -'3 , ΔLIM2-4fwd 5’- cagcctcttctccccatgacgctgctactactg -'3 , ΔLIM2-4rev 5’- cagtagtagcagcgtcatggggagaagaggctg -'3 , ΔLIM3-4fwd 5’- aagattacttcgacatgtttgcttgacccaagtgcggc -'3 , ΔLIM3-4rev 5’- gccgcacttgggtcaagcaaacatgtcgaagtaatctt -'3 , ΔLIM4fwd 5’- ggcgcggctcgtgactgtgctccgg -'3 , ΔLIM4rev 5’- ccggagcacagtcacgagccgcgcc -'3 ) . The cDNA of murine Hic5 was amplified from a cDNA derived from murine vascular smooth muscle cells , cloned into pCR2 . 1-TOPO ( Invitrogen ) and subcloned into pEGFP-C1 vector . Murine leupaxin cDNA ( cloneID: 5065405 from Thermo Scientific , Germany ) was PCR-amplified ( Lpxn-fwd: 5’- ctcgagcaatggaagagctggatgccttattg -3’; Lpxn-rev 5’- gaattcctactgtgaaaagagcttagtgaagc -3’ ) and subcloned into the pEGFP-C1 vector . To express recombinant murine kindlin-2 and paxillin-LIM3 ( A473-S533 ) cDNAs were fused with N-terminal tandem tags consisting of 10x-Histidine followed by a SUMO3-tag and cloned into pCoofy17 . The primer sequences for amplifying the paxillin-LIM3 domain were: LIM3fwd 5’-aaccggtggagctcccaagtgc-3’ and LIM3rev 5’-ttctcgagttacgagccgcgcc-3’ . The plasmid carrying FNIII7-10 cDNA has been described previously ( Takahashi et al . , 2007 ) . For Y2H analysis , the kindlin-2 cDNA was PCR amplified using the primers K2-Bamfw: 5’-gggatcccactgggcctaatggctctggacgggataagg-3’ and K2-Salrev: 5’-gtgtcgacgtcacacccaaccactggtgagtttg-3’ and cloned into the pGBKT7 plasmid to obtain a kindlin-2 version that was N-terminally fused with the Gal4-DNA binding domain . Screening of this construct against a human full ORF library was conducted by the Y2H protein interaction screening service of the German Cancer Research Center in Heidelberg , Germany . Experiments were routinely repeated at least three times and the repeat number was increased according to the effect size or sample variation . Unless stated differently , all statistical significances ( *P<0 . 05; **P<0 . 01; ***P<0 . 001; n . s . , not significant ) were determined by two-tailed unpaired t-test . In the boxplots , the middle line represents the median , the box ends represent the 25th and 75th percentiles and the whisker ends show the 5th and 95th percentiles . Statistical analysis were performed with Prism ( GraphPad , La Jolla , CA , USA ) . | A meshwork of proteins called the extracellular matrix surrounds the cells that make up our tissues . Integrins are adhesion proteins that sit on the membrane surrounding each cell and bind to the matrix proteins . These adhesive interactions control many aspects of cell behavior such as their ability to divide , move and survive . Before integrins can bind to the extracellular matrix they must be activated . Previous research has shown that in certain types of blood cells , proteins called talins and kindlins perform this activation . These proteins bind to the part of the integrin that extends into the cell , causing shape changes to the integrin that allow binding to the extracellular matrix . However , it is not clear whether talin and kindlin also activate integrins in other cell types . Fibroblasts are cells that help to make extracellular matrix proteins , and are an important part of connective tissue . Theodosiou et al . engineered mouse fibroblast cells to lack either talin or kindlin , and found that both of these mutant cell types were unable to activate their integrins and as a result failed to bind to an extracellular matrix protein called fibronectin . Even when cells were artificially induced to activate integrins by treating them with manganese ions , cells lacking talin or kindlin failed to fully activate integrins and hence did not adhere well to fibronectin . This suggests that talin and kindlin work together to activate integrins and to maintain them in this activated state . When treated with manganese ions , cells that lacked talin were able to flatten and spread out , whereas cells that lacked kindlin were unable to undergo this shape change . Theodosiou et al . found that this cell shape is dependent on kindlin and its ability to bind to and recruit a protein called paxillin to “adhesion sites” , where integrins connect the cell surface with the extracellular matrix . Kindlin and paxillin then work together to activate other signaling molecules to induce the cell spreading . The next challenge is to understand how talin and kindlin are activated in non-blood cells and how they maintain integrins in an active state . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | Kindlin-2 cooperates with talin to activate integrins and induces cell spreading by directly binding paxillin |
How huntingtin ( HTT ) triggers neurotoxicity in Huntington’s disease ( HD ) remains unclear . We report that HTT forms a transcription-coupled DNA repair ( TCR ) complex with RNA polymerase II subunit A ( POLR2A ) , ataxin-3 , the DNA repair enzyme polynucleotide-kinase-3'-phosphatase ( PNKP ) , and cyclic AMP-response element-binding ( CREB ) protein ( CBP ) . This complex senses and facilitates DNA damage repair during transcriptional elongation , but its functional integrity is impaired by mutant HTT . Abrogated PNKP activity results in persistent DNA break accumulation , preferentially in actively transcribed genes , and aberrant activation of DNA damage-response ataxia telangiectasia-mutated ( ATM ) signaling in HD transgenic mouse and cell models . A concomitant decrease in Ataxin-3 activity facilitates CBP ubiquitination and degradation , adversely impacting transcription and DNA repair . Increasing PNKP activity in mutant cells improves genome integrity and cell survival . These findings suggest a potential molecular mechanism of how mutant HTT activates DNA damage-response pro-degenerative pathways and impairs transcription , triggering neurotoxicity and functional decline in HD .
Huntington’s disease ( HD ) is an autosomal dominant neurodegenerative disorder caused by a CAG triplet repeat expansion in exon 1 of the HTT gene that is translated into polyglutamine ( polyQ ) sequences in the huntingtin ( HTT ) protein which leads to progressive deterioration of cognitive and motor functions ( The Huntington’s Disease [MACDONALD , 1993; Ross and Tabrizi , 2011; Vonsattel and DiFiglia , 1998] ) . The polyQ expansion in the mHTT protein leads to progressive degeneration most overly affecting γ-aminobutyric acid ( GABA ) -releasing striatal neurons and glutamatergic cortical neurons , although neuronal dysfunction and tissue atrophy in other brain regions is also present ( Vonsattel and DiFiglia , 1998; Ross and Tabrizi , 2011 ) . Altered conformation of the mutant protein is reported to reduce normal function of the protein as well as facilitate aberrant protein-protein interactions or subcellular localization , leading to neurotoxicity . Among the numerous molecular interactions and signaling pathways implicated in HD pathomechanism , transcriptional dysregulation ( Jimenez-Sanchez et al . , 2017; Ross and Tabrizi , 2011; Valor , 2015 ) , mitochondrial ( mt ) dysfunction ( Shirendeb et al . , 2011; Siddiqui et al . , 2012 ) , DNA strand break accumulation , and atypical ataxia telangiectasia-mutated ( ATM ) pathway activation , involved in the DNA damage response ( Bertoni et al . , 2011; Giuliano et al . , 2003; Illuzzi et al . , 2009; Xh et al . , 2014 ) , have emerged as key players in HD-related neuronal dysfunction . Genetic or pharmacological ablation of ATM activity to ameliorate the consequence of aberrant ATM activation decreased neurotoxicity in HD animal models and HD induced pluripotent stem cells , respectively ( Xh et al . , 2014 ) , supporting the emerging view that inappropriate and chronic DNA damage-response ( DDR ) pathway activation is a critical contributor to HD pathogenesis . Although , recent genome-wide association ( GWA ) studies and genetic data from other sources suggest that DNA damage and repair pathways are central to the pathogenesis of HD and other diseases associated with CAG repeat expansion ( Bettencourt et al . , 2016; Lee et al . , 2015 ) , the perplexing questions that remain to be elucidated include how polyQ expansion induces DNA strand breaks , activates the DDR pathway , and disrupts transcription . It is also unclear whether transcriptional dysregulation and atypical ATM activation are mechanistically interconnected . We recently reported that the wild-type ( wt ) form of the deubiquitinating enzyme ataxin-3 ( wtATXN3 ) enhances the activity of polynucleotide kinase-3'-phosphatase ( PNKP ) , a bifunctional DNA repair enzyme with both 3'-phosphatase and 5'-kinase activities that processes unligatable DNA ends to maintain genome integrity and promote neuronal survival . In contrast , mutant ATXN3 ( mATXN3 ) abrogates PNKP activity to induce DNA strand breaks and activate the DDR-ATM→p53 pathway , as observed in spinocerebellar ataxia 3 ( SCA3; Chatterjee et al . , 2015; Gao et al . , 2015 ) . Furthermore , we recently reported that PNKP plays a key role in transcription-coupled base excision repair ( TC-BER ) and transcription-coupled double strand break repair ( TC-DSBR ) ( Chakraborty et al . , 2015; Chakraborty et al . , 2016 ) . Here our data demonstrate that wtHTT is a part of a transcription-coupled DNA repair ( TCR ) complex formed by RNA polymerase II subunit A ( POLR2A ) , basic transcription factors , PNKP , ATXN3 , DNA ligase 3 ( LIG 3 ) , cyclic AMP response element-binding ( CREB ) protein ( CBP , histone acetyltransferase ) , and this complex identifies lesions in the template DNA strand and mediates their repair during transcriptional elongation . The polyQ expansion in mHTT impairs PNKP and ATXN3 activities , disrupting the functional integrity of the TCR complex to adversely impact both transcription and DNA repair . Low PNKP activity leads to persistent accumulation of DNA lesions , predominantly in actively transcribing genes , resulting in unusual activation of the ATM-dependent p53 signaling pathway . Increased PNKP activity in mutant cells improved cell survival by substantially reducing DNA strand breaks and restricting ATM→p53 pathway activation . Likewise , low ATXN3 activity increases CBP ubiquitination and degradation thereby negatively influencing CREB-dependent transcription . These findings provide important mechanistic insights that could explain how mHTT may trigger neurotoxicity in HD .
Both wtHTT and mHTT interact with transcription factors and co-activators including CBP ( McCampbell et al . , 2000; Nucifora et al . , 2001; Steffan et al . , 2000 ) , TATA-binding protein ( TBP; Huang et al . , 1998 ) , p53 ( Bae et al . , 2005; Steffan et al . , 2000 ) , the general transcription factors TFIID and TFIIF ( Zhai et al . , 2005 ) , and specificity protein 1 ( Sp1; Dunah et al . , 2002 ) . POLR2A also interacts with HTT and is detected in nuclear inclusions in the HD brain ( Huang et al . , 1998; Suhr et al . , 2001 ) . It is hypothesized that wtHTT , which shuttles into the nucleus , assists in the assembly of transcription factor and co-activator complexes to regulate target gene expression , and that polyQ expansion perturbs the functional integrity of these complexes ( Kumar et al . , 2014; Luthi-Carter and Cha , 2003; Ross and Tabrizi , 2011 ) . How mHTT disrupts the activities of specific promoters and whether mHTT-mediated transcriptional dysregulation is linked to DNA damage accumulation and aberrant DDR pathway activation remains unknown . Given that HTT interacts with huntingtin-associated protein 1 ( HAP-1; Li et al . , 1995 ) , while ATXN3 interacts with HAP-1 ( Takeshita et al . , 2011 ) and PNKP ( Chatterjee et al . , 2015; Gao et al . , 2015 ) , we asked whether ATXN3 and PNKP might interact with HTT to form a TCR complex and if this is affected by polyQ expansion . We isolated nuclear protein extract ( NE ) and cytosolic protein extract ( CE ) from SH-SY5Y cells and the fractions were analyzed by western blot ( WB ) to determine purity of nuclear protein fractions ( Figure 1A ) . We immunoprecipitated ( IP’d ) endogenous wtHTT from the NE of SH-SY5Y cells , and WBs of the immunocomplexes ( ICs ) showed the presence of HAP-1 , ATXN3 , CBP , TAFII 130 ( TAF4 ) , POLR2A , PNKP , and LIG 3 ( Figure 1B and Figure 1—figure supplement 1 ) . Similarly , IP of endogenous ATXN3 from NEs revealed these proteins in the ATXN3-IC ( Figure 1C and Figure 1—figure supplement 2 ) . Finally , IP of PNKP from NEs confirmed that they were also present in the PNKP-IC ( Figure 1D and Figure 1—figure supplement 3 ) . To verify the specificity of these interactions in vivo , we analyzed the ICs for the presence of apurinic-apyrimidinic endonuclease 1 ( APE1 ) , another critical DNA base excision repair ( BER ) enzyme that works independently of PNKP-mediated BER pathways ( Wiederhold et al . , 2004 ) . APE1 was not detected ( Figure 1B to D ) , suggesting interaction specificity and selectivity . Finally , IP of POLR2A again revealed these proteins in the IC ( Figure 1E and Figure 1—figure supplement 4 ) . For further confirmation , we IP’d Myc-tagged HTT from the NEs of PC12 cells expressing Myc-tagged FL-wtHTT-Q23 or FL-mHTT-Q148 . WB confirmed the presence of ATXN3 , PNKP , POLR2A , CBP , and LIG three but not APE1 in the Myc-IC ( Figure 1F and G , Figure 1—figure supplements 5 and 6 ) , suggesting that HTT , POLR2A , CBP , ATXN3 , LIG 3 , and PNKP form a multiprotein TCR complex . Proximity ligation assays ( PLAs ) were then performed to verify interaction specificity ( Gao et al . , 2015 ) . The reconstitution of fluorescence in neuronal cells ( Figure 1H to M ) and postmortem human brain sections ( Figure 1—figure supplement 7 ) suggested substantial interaction among these proteins . Importantly , the majority of the PLA signals was from the nuclei but substantial amount of signals were from the periphery or cytoplasm . Immunostaining the cells with mitochondrial markers suggested that HTT forms similar complexes in the mitochondria ( data not shown ) . Importantly , about 60–70% of the PLA signal was nuclear in control brain , while the complexes were predominantly in the perinuclei or cytoplasm of HD brain sections ( Figure 1—figure supplement 7 ) . Since PNKP and HTT are present in the mitochondria ( Mandal et al . , 2012; Orr et al . , 2008 ) , the extranuclear signals detected in the control subjects are presumably from mitochondrial HTT-ATXN3-PNKP complexes . WB analysis of subcellular protein fractions from neuronal cells show the presence of HTT , ATXN3 and PNKP in mitochondria ( data not shown ) . Moreover , co-staining the cells or brain sections with mitochondrial markers suggested presence of HTT , ATXN3 and PNKP in mitochondria ( data not shown ) . These findings indicate that HTT may form a similar complex in mitochondria regulating mtDNA repair and transcription . The possible in vivo association of these proteins was further assessed by immunostaining HTT , PNKP , and ATXN3 in postmortem brain tissue from patients with HD and control subjects . Confocal microscopy revealed colocalization of HTT with PNKP and ATXN3 in HD and control brain ( Figure 2A & B; arrows ) . Colocalization of ATXN3 with PNKP was observed in both groups ( Figure 2C; arrows ) . Marked HTT/PNKP colocalization was also observed in brain sections from HD knock-in ( zQ175; Menalled et al . , 2012 ) and WT control mouse brain tissue ( data not shown ) . PNKP contains an N-terminal fork head-associated ( FHA ) domain , C-terminal fused 3'-phosphatase ( PHOS ) domain , and 5'-kinase ( KIN ) domain . The PHOS domain hydrolyzes 3'-phosphate groups , while the KIN domain promotes addition of a phosphate group to the 5'-OH at damaged sites for error-free repair ( Karimi-Busheri et al . , 1999 ) . To identify the specific PNKP domain ( s ) that interact with HTT , full-length PNKP ( FL-PNKP ) ; the FHA , PHOS , and KIN domains; the FHA + PHOS domains; or the PHOS + KIN domains were expressed as a FLAG-tagged peptide , as illustrated in Figure 3A . We individually expressed these domains in SH-SY5Y cells ( Figure 3B; upper panel ) and isolated the NEs . IPs of these domains with a FLAG antibody and subsequent WB analysis of the IC showed the presence of HTT in the FLAG- ( FL-PNKP ) -IC and FLAG- ( PHOS + KIN ) -IC ( Figure 3B; Lower panel , lanes 1 and 6 , arrow ) . HTT was not detected in FLAG-ICs when the individual FHA , PHOS , or KIN domains were IP’d ( Figure 3B; lanes 2–5 , arrow ) . This suggests that the C-terminal catalytic domain of PNKP interacts with HTT , but the individual FHA , PHOS , and KIN domains are not sufficient . We separately expressed the PNKP domains in cells , isolated the NEs , and IP’d endogenous HTT . WBs revealed the presence of full-length and PHOS + KIN domains in the HTT-IC ( Figure 3C; Lower panel , lanes 1 and 6 ) . When we expressed the PNKP domains in PC12 cells expressing Myc-wtHTT-Q23 or Myc-mHTT-Q148 ( Figure 3D & E; upper panels ) , IP of the Myc-HTT and WB revealed the full-length protein or PHOS + KIN domain ( Figure 3D & E , Lower panels , lanes 1 and 6 ) . These data suggest that both wtHTT and mHTT interact with the C-terminal catalytic domain of PNKP . The N-terminal-truncated fragment of mHTT ( NT-mHTT ) containing the polyQ expansion is encoded by exon 1 of the HTT gene . Transgenic mice expressing exon one or a truncated fragment extending beyond the first exon ( N171 ) with NT-mHTT recapitulate HD-like neurological and behavioral abnormalities ( Mangiarini et al . , 1996; Schilling et al . , 1999 ) . To test whether this fragment interacts with PNKP , we expressed NT-wtHTT-Q23 or NT-mHTT-Q97 ( 1–586 base pairs ) as a GFP-tagged peptide in SH-SY5Y cells ( Figure 4A; upper panel ) , isolated the NEs , and IP’d the GFP-NT-HTT fusion peptide with a GFP antibody . WBs showed the presence of PNKP , ATXN3 , and HTT in the GFP-IC ( Figure 4A; lower panel , lanes 4 and 6 ) . We next IP’d this fragment from PC12 cells expressing Myc-NT-wtHTT-Q23 or NT-mHTT-Q148 and found ATXN3 , PNKP , POLR2A , CBP , and LIG three in the Myc-IC . Importantly , APE1 was not detected in the IC , again suggesting interaction specificity ( Figure 4B ) . To identify which PNKP domain interacts with NT-HTT , we expressed various domains as FLAG-tagged peptides in SH-SY5Y cells expressing either Myc-NT-HTT-Q23 or Myc-NT-HTT-Q97 ( Figure 4C & D; upper panels ) and IP’d Myc-tagged fragments from the NEs . WBs revealed FL-PNKP and PNKP- ( PHOS + KIN ) domains in the Myc immunocomplex ( Figure 4C & D; lanes 1 and 6 ) , suggesting that the N-terminal fragment of HTT interacts with the C-terminal catalytic domain of PNKP . However , from the WB analyses we could not establish whether the interaction of the mutant HTT fragment ( NT-mHTT ) with PNKP- ( PHOS + KIN ) domain is stronger than the interaction with the N-terminal fragment of WT HTT ( NT-wtHTT; Figure 4C & D; lanes 6 ) . The PNKP- ( FHA + PHOS ) domain also showed a relatively weaker interaction with the N-terminal fragment of HTT ( Figure 4C & D; lanes 5 ) indicating that the FHA-PHOS domain of PNKP alone interacts with the N-terminal fragment of HTT . To further assess these possible interactions , we performed bi-molecular fluorescence complementation ( BiFC ) assays as we previously reported ( Gao et al . , 2015 ) . We cloned either the full-length or C-terminal catalytic domain of PNKP at the N-terminus of cyan fluorescent protein ( CFP ) into plasmid pBiFC-VN173 to construct plasmids pVN-PNKP and pVN- ( PHOS +KIN ) , respectively . We also cloned the N-terminal fragment of wtHTT and mHTT cDNA ( encoding 23 and 97 glutamines , respectively ) at the C-terminus of CFP in plasmid pBiFC-VC155 to construct pVC-NT-HTT-Q23 and pVC-NT-HTT-Q97 , respectively ( detailed descriptions of these plasmids are provided in the STAR Methods ) . Cotransfection of plasmid pVN-PNKP with the parent plasmid pBIFC-VC155 did not reconstitute fluorescence ( Figure 4E; Panel 1 ) , whereas cotransfection of pVN-PNKP with either pVC-NT-HTT-Q23 or pVC-NT-HTT-Q97 did ( Figure 4E; Panels 2 and 3 ) . Similarly , cotransfection of pVN- ( PHOS +KIN ) with pBIFC-VC155 did not produce fluorescence ( Figure 4E; Panel 4 ) , whereas cotransfection of pVN- ( PHOS +KIN ) with either pVC-NT-HTT-Q23 or pVC-NT-HTT-Q97 robustly reconstituted fluorescence ( Figure 4E , Panels 5 and 6 ) . Although these data suggest that the N-terminal of mHTT interacts with PNKP , these experiments do not inform the relative strengths of interaction between these peptides . Nonetheless , the IP and BIFC studies suggest that the truncated-N-terminal fragments of both WT and mHTT interact with the C-terminal catalytic domain of PNKP . The interaction of these peptides with the PHOS-KIN domain of PNKP is relatively stronger than with the PHOS domain alone . However , more rigorous structural and biophysical measurements using purified proteins/peptides will be required to understand the true nature of these protein-protein interactions , the relative binding efficacies and to identify the direct interacting partners in this complex . Moreover , since the HTT-TCR complex is not fully characterized , the presence of additional unidentified components of the complex could significantly alter these interactions in vivo . Given that PNKP interacts with mHTT , we measured the 3'-phosphatase activity of PNKP in induced pluripotent stem cells ( iPSCs ) differentiated to neurons enriched for medium striatal neuronal populations from HD and unaffected control subjects using a modification of Telezhkin et al . ( 2016 ) . HD iPSC-derived neurons ( mHTT-109Qs ) were compared to control neurons ( wtHTT-33Q; HD iPSC HD iPSC Consortium , 2017 ) and activity was found to be 70–80% lower in the NE of HD neurons , while PNKP protein levels did not change ( representative experiment , Figure 5A to C ) . Similar differences were found for neurons with adult onset alleles ( Q50 and Q53 ) compared to controls ( Q18 and Q28 ) . In these comparisons there was substantially reduced ( 70% to 80% ) PNKP activity in HD neurons ( Q50 and Q53 ) compared with control neurons ( Q18 or Q28 ) ( Figure 5—figure supplement 1 ) , supporting an impairment in human neurons in the presence of mHTT . We next measured PNKP activity in PC12 cells expressing exogenous full-length wtHTT ( FL-wtHTT-Q23 ) and full-length mHTT ( FL-mHTT-Q148 ) ( Igarashi et al . , 2003; Tanaka et al . , 2006 ) . We found that it was about 30–40% higher in the NE of PC12 cells expressing wtHTT , and about 70% lower in the NE of PC12 cells expressing FL-mHTT-Q148 compared to control cells , while PNKP protein levels did not change ( Figure 5—figure supplement 2A to C ) . These data suggest that wtHTT and mHTT stimulate and abrogate PNKP activity , respectively . Since wtHTT interacts with and stimulates PNKP activity , we examined the extent to which HTT depletion alters PNKP activity . We found that in HTT-depleted cells , PNKP activity was reduced by >70% ( Figure 5—figure supplement 2D to G ) , suggesting that wtHTT plays key roles in stimulating PNKP activity , maintaining the functional integrity of the TCR complex , and repairing DNA damage . PNKP activity was 80–90% decreased in the striatum ( STR ) and cortex ( CTX ) , and marginally ( 5% ) decreased in the cerebellum ( CRBL ) of male heterozygous asymptomatic zQ175 mice at 7 weeks , whereas PNKP protein levels were not different from WT ( Figure 5D to F ) . An identical trend was observed in female littermates ( data not shown ) . Because the N-terminal of mHTT interacts with PNKP , we investigated whether N-terminal truncated fragment of mHTT interferes with PNKP activity in PC12 cells or N171-82Q mice ( Schilling et al . , 1999; Tanaka et al . , 2006 ) . PNKP activity was ~30% higher in PC12 cells expressing NT-wtHTT-Q23 and >80% lower in cells expressing NT-mHTT-Q148 ( Figure 5—figure supplement 3A & B ) . Similar to full-length HTT , PNKP activity was decreased in SH-SY5Y cells expressing NT-mHTT with variable glutamine expansions ( Figure 5—figure supplement 3C & D ) . Moreover , PNKP activity was >80% decreased in the STR and CTX of N171-82Q brain compared to control ( Figure 5—figure supplement 3E & F ) . To test if mHTT specifically blocks PNKP activity rather than interfering with DNA repair per se , we examined how it modulated the repair of two nicked DNA duplexes: one without a 3'-phosphate end that requires DNA polymerase and ligase activities but not PNKP activity for repair , and another duplex with a 3'-phosphate end that requires PNKP and DNA polymerase and ligase activities for complete repair . We observed that NEs from cells expressing mHTT or from zQ175 mouse brain did not hamper repair of the duplex that required DNA polymerase and ligase activities but did not require PNKP activity . In contrast , NEs from these cells and mice did abrogate repair of the duplex that requires PNKP ( Figure 5G to J ) , suggesting that mHTT specifically blocks PNKP activity but does not interfere with the activities of other repair enzymes in the TCR complex . In response to DNA strand break accumulations , ATM is activated by phosphorylation which phosphorylates p53 , which in turn activates pro-apoptotic gene transcription ( Chipuk et al . , 2004; Nakano and Vousden , 2001; Oda et al . , 2000 ) . Consistently , we found chronic activation of the DDR-ATM-p53 pathway in HD neurons Figure 5—figure supplement 4A & B ) and in zQ175 CTX ( Figure 5—figure supplement 4C & D ) compared with respective controls . mHTT expression has been shown to activate p53 in HD , whereas deleting p53 in the HD transgenic brain rescues behavioral abnormalities ( Bae et al . , 2005 ) . Consistently , markedly increased mRNA expression of p53 target genes ( e . g . , Bcl2L11 , Pmaip1 , Bid , Pidd1 and Apaf1 ) were observed in the STR but not in CRBL of zQ175 mice compared to controls ( Figure 5—figure supplement 4E & F ) . We next expressed the N-terminal truncated fragment of mHTT encoding Q97 ( NT-mHTT-Q97 ) in SH-SY5Y cells overexpressing PNKP and carried out a comet assay ( Olive and Banáth , 2006 ) . Analysis of mutant cells showed more strand breaks , which were substantially rescued after PNKP overexpression ( Figure 5—figure supplement 5A to C ) , suggesting that mHTT-mediated ablation of PNKP activity contributes to increased DNA strand breaks . Consistently , we noted activation of ATM-p53 signaling in cells expressing NT-mHTT-Q97 ( Figure 5—figure supplement 5D & E ) , and PNKP overexpression reduced mHTT-mediated DDR-ATM pathway activation ( Figure 5—figure supplement 5F ) . PC12 cells expressing the full-length mHTT encoding 148Qs ( FL-mHTT-Q148 ) showed increased caspase-3 activity and PNKP overexpression reduced caspase-3 activation Figure 5—figure supplement 5G ) . Consistently , PC12 cells expressing FL-mHTT-Q148 also showed higher cell toxicity and PNKP overexpression significantly rescued cell toxicity ( Figure 5—figure supplement 5H & I ) . Collectively , these results suggest that mHTT-mediated activation of the ATM-p53 pathway and associated cell toxicity is at least partially due to PNKP inactivation by mHTT . Emerging evidence suggests that the TCR complex plays a pivotal role in editing strand breaks in actively transcribing template DNA to maintain genome integrity and cell survival , and its inactivation leads to preferential accumulation of DNA breaks in the transcriptionally active genome ( Chakraborty et al . , 2015; Chakraborty et al . , 2016; Hanawalt and Spivak , 2008 ) . Since mHTT abrogates the activity of PNKP , a key component of the TCR complex ( Chakraborty et al . , 2016 ) , we compared the associations of HTT and TCR proteins with transcriptionally active versus inactive genomes and asked whether the former accumulates more strand breaks in the HD brain . Chromatin immunoprecipitation ( ChIP ) revealed significantly higher HTT occupancy on actively transcribing genes in the brain ( e . g . , neuronal differentiation factor 1 and 2 [Neurod1 and Neurod2] , neurogenic basic-helix-loop-helix protein neurogenin 1 [Neurog1] , tubulin beta three class III [Tubb3] , neuron-specific enolase 2 [Eno2γ] , and DNA polymerase beta [Pol b] ) over genes that are not transcribed in the brain but actively transcribed in skeletal or cardiac muscle ( e . g . , myogenic differentiation factor 1 [Myod1]; myogenic factor 4; myogenin [Myog]; and myosin heavy chain 2 , 4 , 6 , or 7 [Myh2 , Myh4 , Myh6 , or Myh7]; ( Figure 6A & B ) . Increased association between HTT with the transcriptionally active genome and mHTT-mediated abrogation of PNKP activity indicate that the wtHTT-TCR complex repairs lesions during transcriptional elongation , but polyQ expansion might impair the TCR and facilitate DNA damage accumulation . To test this theory , we performed Long-amplicon quantitative polymerase chain reaction ( LA-qPCR ) analysis , a versatile technique to measure nuclear and mitochondrial DNA damage ( Gao et al . , 2015; Santos et al . , 2006 ) to assess DNA strand breakage in actively transcribing and non-transcribing genes in the transgenic mouse cortex ( CTX ) . The results revealed 60–70% lower PCR-amplification of actively transcribing genes in asymptomatic ( seven wks ) zQ175 mouse CTX compared to age-matched WT controls ( Figure 6C & D ) . In contrast , the amplification efficacy for non-transcribing genes in the zQ175 CTX was only marginally ( 10–15% ) reduced ( Figure 6E & F ) , indicating less DNA damage accumulation . Consistent with the levels of PNKP activities observed in the striatum ( STR ) and cerebellum ( CRBL ) , the LA-qPCR analysis revealed substantial DNA damage accumulation in STR but negligible DNA damage accumulations in the CRBL ( Figure 6—figure supplement 1 ) . Moreover , immunostaining of the HD patients’ brain and HD transgenic mouse brain sections with anti-phospho-53BP1 antibody , a DNA damage marker , showed increased presence of DNA damage as compared to control ( Figure 6—figure supplement 2 ) . Consistently , preferential accumulation of DNA strand breaks was observed in iPSC-derived HD primary neurons ( Q50 and Q53 ) than controls ( Q18 and Q28 ) ( Figure 6—figure supplement 3 ) . Increased DNA break accumulation was also observed in actively transcribing genes vs . non-transcribing genes in the N171-82Q transgenic CTX than the age-matched controls Figure 6—figure supplement 4 ) . These data support our hypothesis that the HTT-TCR complex repairs strand breaks during transcription , and that this function is impaired by polyQ expansion , resulting in persistent strand break accumulation predominantly affecting actively transcribing genes in HD . Given that the deubiquitinase ATXN3 is present in the TCR complex , interacts with mHTT , and is sequestered in polyQ aggregates in HD brain , we postulated that compromised ATXN3 activity might increase ubiquitination and decrease levels of TCR components , adversely impacting complex functionality and transcription . To explore this possibility , we examined whether mHTT stimulates ubiquitination and degradation of specific TCR complex proteins . WB analyses of NEs from HD and control iPSC-derived primary neurons revealed a significant decrease in soluble CBP protein levels in HD neurons , whereas ATXN3 , PNKP , POLR2A , and CREB levels were not affected ( Figure 7A & B ) . Quantitative reverse transcription PCR analyses did not show a significant change in CBP mRNA levels upon mHTT expression ( data not shown ) , suggesting that mHTT does not interfere with the expression of CBP in HD . This finding indicates that CBP might be degraded more in HD . However , an alternative possibility is that CBP becomes insoluble when post-translationally modified . Substantially reduced levels of CBP was also observed in the soluble fraction of proteins from cells expressing exogenous mHTT ( data not shown ) . To determine whether abrogating ATXN3 activity causes reduced CBP levels , we measured TCR protein levels in ATXN3-depleted cells . Similar to HD iPSC-derived primary neurons , markedly lower CBP levels were observed in the soluble protein extract from the ATXN3-depleted cells ( Figure 7C & D ) . Consistent with a previous report ( Cong et al . , 2005; Jiang et al . , 2003 ) , CBP levels were dramatically ( ~80% ) reduced in the zQ175 CTX but only marginally ( ~20% ) decreased in the CRBL ( Figure 7E & F ) . To test whether ATXN3 interacts with CBP , we co-expressed Myc-ATXN3 and FLAG-CBP and IP’d Myc-ATXN3 from the NEs . WBs showed CBP in the Myc-IC ( Figure 7G ) . Conversely , IP of the FLAG-CBP and subsequent WB revealed ATXN3 ( Figure 7H ) . The PLA results also suggested intracellular interaction between ATXN3 and CBP ( Figure 7I ) . Confocal microscopy showed distinct colocalization of ATXN3 and CBP in HD and control brain sections ( Figure 7J , arrows ) . A recent study also showed significantly increased ubiquitination and reduced level of CBP in HdhQ7/Q111 HD transgenic mouse brain ( Giralt et al . , 2012 ) . To test whether mHTT expression increases CBP ubiquitination in zQ175 mouse brain , we IP’d CBP from the NE of zQ175 and control mouse brain . Consistent with a previous report ( Jiang et al . , 2003; Giralt et al . , 2012 ) , WBs of the CBP IC showed increased CBP ubiquitination in the transgenic brain ( Figure 7K; lower panel ) . These data suggest that decreased ATXN3 activity due to its interaction with mHTT in the TCR complex may increase ubiquitination , and increased ubiquitination of CBP may cause aberrant localization of CBP , negatively impacting its solubility in HD . It is also possible that increased ubiquitination may facilitate degradation of CBP in HD .
Our findings reveal a critical proximal event by which polyQ expansions in mHTT induce DNA damage to activate the DDR ATM→p53 pro-apoptotic signaling cascade and disrupt tissue-specific transcriptional activity – key pathogenic features consistently described in HD ( Giuliano et al . , 2003; Bae et al . , 2005; Illuzzi et al . , 2009; Bertoni et al . , 2011; Xh et al . , 2014 ) . A significant association of wtHTT with PNKP , ATXN3 , POLR2A and associated transcription factors suggest that wtHTT may act as a scaffold factor to assemble various core components of the TCR complex . Our data suggest that formation of this functional TCR complex with wtHTT is essential for sensing and editing DNA lesions in the template strand during transcriptional elongation in post-mitotic differentiated neurons and may contribute in maintaining genome integrity and neuronal survival . Our results further indicate that interaction of PNKP with wtHTT stimulates its DNA end-processing activity to facilitate neuronal DNA repair . The role of wtHTT in maintaining TCR complex functionality and genome integrity is further validated by the fact that depletion of endogenous wtHTT protein dramatically depletes PNKP activity with a concurrent increase in DNA damage accumulation . In contrast , mHTT with polyQ expansions interacts with several key components of the complex but abrogates the activities of PNKP and ATXN3 , thereby disrupting DNA repair and transcription leading to a possible early trigger for neurotoxicity and functional decline in HD ( Figure 8 ) . Our data demonstrate that mHTT interaction with PNKP and the resultant decline in PNKP’s enzymatic activity was evident in CTX and STR of HD transgenic mouse models but insignificant in the CRBL . Consistent with these findings , CTX and STR , the most affected brain regions in HD displayed extensive DNA strand breaks with impaired DNA repair capacity . In contrast , PNKP activity and genome integrity was marginally affected in CRBL , the brain region that is reported to be relatively unaffected in HD . However , no alterations in the steady-state levels of PNKP , ATXN3 and other key TCR components were observed between CRBL vs . CTX or STR . Our present data do not explain why mHTT expression specifically impacts PNKP activity and genome integrity in the CTX and STR but spares CRBL . Further investigation is required to understand region-specific decreases in PNKP activity and DNA break accumulation in the HD brain . Complete characterization of the TCR complex in different brain regions may provide valuable insight into the selective neuronal vulnerability to mHTT-mediated toxicity . It will be interesting to understand the mechanism that imparts protection to CRBL against mHTT and could provide another node in the signaling pathway that could potentially be developed as a therapeutic target . Furthermore , more rigorous biophysical and structural studies are necessary with purified peptides/proteins to characterize the true nature of the protein-protein interactions and interacting partners to understand how this putative HTT-RNA polymerase complex maintains neuronal genome integrity and survival . It is notable that the HTT-TCR complex preferentially associates only with the transcriptionally active genome both in vitro and in vivo . This suggests that the complex could be actively involved in repairing lesions in the template DNA strand during transcription and thus maintaining sequence integrity of the transcriptionally active genomes over the non-transcribing genome . This HTT-TCR complex may provide an additional layer of protective mechanism to maintain the sequence integrity of highly transcriptionally active genes in post-mitotic neurons . Depletion of PNKP activity by mHTT and subsequent accumulation of DNA strand breaks in the transcriptionally active genome extends our previous report ( Chakraborty et al . , 2016 ) . Collectively , these data support our hypothesis that polyQ expansions in HTT result in the preferential accumulation of strand breaks in the transcriptionally active genome . Persistent DNA strand breaks in the actively transcribing genes may stall or impair transcription elongation , preventing adequate expression of a wide variety of neuronal genes and may contribute to the complexity and variability of HD pathology . Moreover , unrepaired lesions may further induce chronic activation of ATM→p53 signaling , as evidenced by increased phosphorylation of ATM , H2AX , and p53 in the HD brain ( Figure 5—figure supplement 1 ) . Chronic ATM-p53 pathway activation resulted in an increased expression of several of the p53 target genes may facilitate neuronal apoptosis in HD . These data support the hypothesis that decreased PNKP activity could be an important proximal event that triggers early neurotoxicity in HD; however , it remains to be tested whether restoration of PNKP activity and DNA repair efficiency can rescue genome integrity and structural and behavioral defects in HD models . Our results provide evidence that ATXN3 is another key regulatory component of the TCR complex . Our data suggest that the mHTT-mediated decrease in ATXN3 activity either enhances degradation of specific TCR complex components or prevents appropriate formation of the TCR complex in HD . We propose that abrogating ATXN3 activity is a potential mechanism by which mHTT decreases CBP activity and thus adversely impacts the transcription of CREB-dependent genes in HD . ATXN3 binds and deubiquitinates both mono- and polyubiquitin chains in target proteins ( Burnett et al . , 2003; Chai et al . , 2004 ) . ATXN3 inactivation in mice increases protein ubiquitination ( Schmitt et al . , 2007 ) , so diminished ATXN3 activity could influence CREB-regulated gene expression as described in HD ( Wyttenbach et al . , 2001 ) . Since disruption of CREB activity in the brain triggers neurodegeneration ( Mantamadiotis et al . , 2002 ) , mHTT-mediated decreases in ATXN3 , CBP , and CREB activities might compromise neuronal function and trigger neurotoxicity , further amplifying pro-degenerative output in HD . The identification of HTT , CBP , PNKP , and ATXN3 as key regulatory components of the TCR complex and our description of how polyQ expansions disrupt the complex’s functional integrity provide important insight into how mHTT could coordinately disrupt CREB-mediated transcription , increases DNA strand breaks , and activates ATM→p53 signaling . Collectively , these events compromise neuronal survival in HD . We hypothesize that POLR2A-mediated transcription might temporarily pause at DNA lesions , leading to mono-ubiquitination of specific TCR complex components , which signals complex assembly to stimulate and/or coordinate lesion repair in normal cells . ATXN3 deubiquitinates the components after repair , and normal transcription resumes . In contrast , the TCR complex stalls at strand breaks in mHTT-carrying cells , and due to reduced ATXN3 activity , specific complex component ( s ) are polyubiquitinated and accumulate aberrantly in polyQ inclusions ( Figure 8 ) . We propose that mHTT-mediated ATXN3 inactivation might impair CBP/CREB-dependent transcription , while reduced PNKP activity might result in DNA break accumulation and DDR pathway activation . The combination of chronic DDR signaling and dysregulation of CREB-dependent genes could trigger selective neuronal degeneration , a hallmark of HD . Defective DNA repair in post-mitotic neurons is an emerging causative factor of cognitive decline in neurodegenerative diseases ( Madabhushi et al . , 2014; Madabhushi et al . , 2015; Rass et al . , 2007 ) . Consistent with our findings , point mutations in PNKP result in microcephaly and seizures ( Shen et al . , 2010 ) , whereas a frame-shift mutation in the PNKP gene was identified in a neurodegenerative disorder characterized by epilepsy ( Poulton et al . , 2013 ) . Therefore , mHTT-mediated ablation of PNKP activity could lead to impaired DNA repair , persistent accumulation of DNA strand breaks that may in part contribute to neurotoxicity and neuronal dysfunction in HD . This study provides multiple lines of evidence suggesting that mHTT-mediated loss of DNA repair and deubiquitinating activity could possibly be critical proximal events that impair the TCR . This could provide a mechanistic link between transcriptional dysregulation leading to aberrant activation of ATM-dependent pro-degenerative pathways and early neurotoxicity in HD . Although the final biological output triggered by impaired TCR and unrepaired DNA strand breaks in HD remains to be fully described , the present data indicate a potential mechanism by which polyQ expansions in mHTT could disrupt the functional integrity of TCR complex and compromises transcriptional regulation and genomic integrity in post-mitotic neurons . Molecular strategies that interfere with the interaction of mHTT with the TCR complex could reduce neurotoxicity and slow functional decline in HD . Alternatively , molecular approaches to stimulate PNKP activity could be a reasonable way to combat transcriptional dysregulation and inappropriate activation of pro-apoptotic signaling in HD . Our findings could help elucidate the cell type-specific pattern of pathology in HD . We propose the possibility that the compromised TCR efficiency in the basal ganglia or cortex could render these neuronal populations more vulnerable . Collectively , our findings suggest an intriguing molecular mechanism that could explain how mHTT expression in HD could compromise genome integrity and neuronal survival .
The construction of plasmids expressing the N-terminal fragment of HTT ( exon1: NT-HTT-Q23 and NT-HTT-Q148 ) and full-length HTT ( FL-HTT-Q23 and FL-HTT-Q148 ) was described previously ( Tanaka et al . , 2006 ) . The N-terminal fragments of wtHTT and mHTT were sub-cloned in pAcGFPC1 ( Clontech , USA ) to construct pGFP-NT-HTT-Q23 and pGFP-NT-HTT-Q97 , respectively . The number of CAG repeats contracted to 97 after propagation in Escherichia coli . The plasmids pGFP-NT-HTT-Q23 and pGFP-NT-HTT-Q97 were digested with NheI and MluI , and the fragments containing GFP-NT-HTT-Q23 and GFP-NT-HTT-Q97 were sub-cloned into the TET-inducible responder plasmid pTRE3G ( Clontech , USA ) using appropriate linkers . The plasmid pTet-ON ( Clontech , USA ) and responder plasmids ( pTRE-GFP-NT-HTT-Q97 or pTRE-GFP-NT-HTT-Q23 ) were transfected into SH-SY5Y cells , and clones were selected with neomycin . The stable inducible clones expressing GFP-NT-HTT-Q97 or GFP-NT-HTT-Q23 were incubated with medium containing doxycycline ( 500 ng/mL ) , and transgene expression was assessed by WB using anti-GFP antibodies . The PNKP cDNA was cloned into pcDNA3 . 1/hygro ( Invitrogen , USA ) to construct pRPS-PNKP , which was transfected into SH-SY5Y cells encoding inducible GFP-NT-HTT-Q23 and GFP-NT-HTT-Q97 . The clones were selected for hygromycin resistance . PNKP expression was examined by WB , and PNKP activity was assessed as described previously ( Chatterjee et al . , 2015 ) . To express PNKP and its functional domains as FLAG-tagged peptides , the full-length cDNA and FHA domain ( 1–300 amino acids ) , kinase domain ( 131–337 amino acids ) , phosphatase domain ( 338–521 amino acids ) , FHA and kinase domain ( 1–337 amino acids ) , and kinase and phosphatase domain ( 131–521 amino acids ) were PCR-amplified using specific primers and cloned into plasmid pCMV-DYKDDDDK ( Clontech , USA ) . Plasmids pBiFC-VN173 ( encoding 1 to 172 N-terminal amino acids of cyan fluorescent protein , CFP ) and pBIFC-VC155 ( encoding 155 to 238 C-terminal amino acids of CFP ) were kindly provided by Dr . Chang-Deng Hu ( Addgene plasmids 22011 and 22010 ) . The N-terminal fragments of HTT cDNA ( encoding 23 or 97 glutamines ) were cloned in-frame with the C-terminal amino acids of CFP in plasmid pBiFC-VC155 to construct pVC-NT-HTT-Q23 and pVC-NT-HTT-Q97 , respectively . Full-length PNKP or its catalytic domain ( phosphatase and kinase domains , 131–521 amino acids ) was cloned in plasmid pBIFC-VN173 to construct pVN-PNKP or pVN- ( PHOS + KIN ) -PNKP , respectively . SH-SY5Y cells ( 2 × 105 cells ) were grown on chamber slides and transfected 24 hr later . Plasmids pVN- ( PHOS + KIN ) -PNKP and pVC-HTT-Q23 or pVN- ( PHOS + KIN ) -PNKP and pVC-HTT-Q97 were cotransfected , and reconstitution of the green/yellow fluorescence of CFP was monitored by fluorescence microscopy . Human neuroblastoma SH-SY5Y cells were purchased from ATCC ( Cat # CRL-2266 ) and cultured in Dulbecco’s minimum essential medium ( DMEM ) containing 15% fetal bovine serum ( FBS ) , and 1% B-27 ( Invitrogen , USA ) . SH-SY5Y cells stably encoding inducible GFP-NT-HTT-Q23 or GFP-NT-HTT-Q97 were cultured in DMEM , and transgene expression was induced by adding doxycycline to the medium to a final concentration of 500 ng/mL . PC12 cells carrying full-length wtHTT-Q23 or mHTT-Q148 were cultured in DMEM containing 15% FBS and doxycycline ( 500 ng/mL ) . HTT expression was induced by withdrawing doxycycline from the media for 5–7 days , and transgene expression was verified by WB . Plasmids expressing the RNAi targeting ATXN3 were from Dharmacon , USA . SH-SY5Y cells were transfected with the ATXN3-RNAi plasmids using Lipofectamine RNAi-MAX reagent ( Invitrogen , USA ) ; stable cells were selected for puromycin resistance and differentiated in DMEM containing 5 µM retinoic acid . All the cell lines were authenticated by short tandem repeat analysis in the UTMB Molecular Genomics Core . The possible mycoplasma contaminations in all the cell lines were tested using GeM Mycoplasma Detection Kit ( SIGMA , Cat# MP0025 ) using a PCR based screening method and cells were found to be free from mycoplasma contamination . Co-IP from NEs: NEs from SH-SY5Y cells were isolated and treated with benzonase to remove DNA and RNA to avoid nucleic acid-mediated Co-IP . Specific target proteins were IP’d , and the IC was washed extensively with cold Tris-buffered saline ( 50 mM Tris-HCl [pH 7 . 5] 200 mM NaCl ) containing 1 mM EDTA , 1% Triton-X100 , and 10% glycerol . The complexes were eluted from the beads with 25 mM Tris-HCl ( pH 7 . 5 ) and 500 mM NaCl and analyzed by WB . Co-IP from tissue: Approximately 250 mg of cortex from freshly sacrificed WT mice was harvested and homogenized with 4 volumes of ice-cold buffer ( 0 . 25 M sucrose , 15 mM Tris-HCl [pH 7 . 9] , 60 mM KCl , 15 mM NaCl , 5 mM EDTA , 1 mM EGTA , 0 . 15 mM spermine , 0 . 5 mM spermidine , 1 mM dithiothreitol , 0 . 1 mM phenylmethylsulfonyl fluoride [PMSF] , and protease inhibitors [Roche Applied Science , Germany] ) with ∼20 strokes to disrupt tissues ( Chakraborty et al . , 2015 ) . Homogenization was continued until a single-cell slurry was obtained , incubated on ice for 15 min , and centrifuged at 1 , 000 × g to obtain the cell pellet . NEs were then prepared from the cell pellet for co-IP analysis . The ICs were analyzed by WB to identify interacting protein partners . Human autopsy specimens were obtained in accordance with local legislation and ethical rules . Control brain samples were collected from age-matched individuals without neurodegenerative disorders . The HD brain tissue samples were obtained from patients with HD who were clinically characterized based on the presence of chorea and motor , mood , and cognitive impairment . The molecular diagnosis of HD was established by analyzing genomic DNA extracted from peripheral blood using a combination of PCR and Southern blotting . HTT CAG repeat lengths were established by sequencing the expansion loci of the mutant allele . All brain autopsies were immediately frozen in liquid nitrogen and stored at −80°C until further analysis . Products purchased from Thermo Fisher Scientific ( Waltham , MA USA ) , unless otherwise specified . Three control: CS25iCTR18n6 , CS14iCTR28n6 , CS83iCTR33n1 and three HD: CS87iHD50n7 , CS03iHD53n3 and CS09iHD109n1 iPSC lines were derived and cultured as previously described on hESC-qualified Matrigel ( HD iPSC HD iPSC Consortium , 2017 ) . Once at 70% confluency , neural induction and further differentiation of neural progenitors with the addition of Activin A ( Peprotech , USA ) , was performed as previously described in Telezhkin et al . ( 2016 ) . Neuronal maturation was performed as previously described ( Telezhkin et al . , 2016 ) on Nunc six well plates . After 3 weeks of maturation , medium was removed and cells were washed once with PBS pH 7 . 4 , without Mg2+ and Ca2+ . Subsequently , cells were washed with 4°C PBS pH 7 . 4 , without Mg2+ and Ca2+ , and scraped using a cell scraper , pipetted into a centrifuge tube and centrifuged at 250 x g for 3 min . PBS was removed and samples were flash frozen in liquid nitrogen . The HD knock-in mouse model zQ175 expresses full-length mHTT from the endogenous mouse Htt promoter ( Menalled et al . , 2012 ) . The N171-82Q transgenic mouse line expresses the truncated N-terminus of human HTT cDNA with a polyQ repeat length of 82 under control of the mouse prion promoter ( Schilling et al . , 1999 ) . Heterozygous transgenic mice and control non-transgenic littermates ( n = 4–5 pools of two animals per genotype ) were sacrificed , and fresh brain tissues were used for enzyme assays , isolating genomic DNA , and obtaining protein for WB analyses . For immunofluorescence assays , transgenic and control littermate mice were deeply anesthetized and transcardially perfused with sterile phosphate-buffered saline ( PBS ) followed by 4% paraformaldehyde in PBS . Brains were post-fixed overnight in fixative solution and embedded in OCT and stored in liquid nitrogen . Slides with 4-μm-thick frozen sections were processed for immunostaining with appropriate antibodies . All procedures involving animals were in accordance with the National Institutes of Health Guide for the care and use of Laboratory Animals , and approved by the Institutional Animal Care and Use Committee of University of California Irivine ( protocol #: AUP-18–155 ) ; and Duke University ( protocol #: A225-17-09 ) . Alkaline comet assays were performed using a Comet Assay Kit ( Trevigen , USA ) . Cells were suspended in 85 μL ice-cold PBS and gently mixed with an equal volume of 1% low-melting-point agarose . The cell suspension was dropped onto an agarose layer and incubated in lysis buffer for 1 hr . After lysis , slides were incubated in buffer containing 0 . 3 M NaOH , 1 mM EDTA ( pH 13 ) for 40 min and electrophoresed for 1 hr . After neutralization , slides were stained and analyzed with a fluorescence microscope . Cell pellets or brain tissues were homogenized , and total protein was isolated using a protein extraction kit ( Millipore , USA ) . The cytosolic and nuclear fractions were isolated from cells/tissue using a NE-PER protein extraction kit ( Thermo Scientific , USA ) . WBs were performed according to the standard procedure , and each experiment was performed at least three times to ensure statistically significant results . The antibodies for p53 ( Cat #9282 ) , p53-S15 ( Cat #9286 ) , p53-S20 ( Cat #9287 ) , p53-S46 ( Cat #2521 ) , Chk2 ( Cat #2662 ) , Chk2-T68 ( Cat #2661 ) , CBP ( Cat #7389 ) and APE1 ( Cat #4128 ) were from Cell Signaling , USA; anti-H2AX ( Cat #ab11175 ) and γH2AX-S139 ( Cat #ab11174 ) were from Abcam , UK; anti-ATM ( Cat #1549–1 ) and ATM-S1981 ( Cat #2152–1 ) were from Epitomics , USA , or anti-ATM from Santa Cruz ( sc-23921 ) , anti-ataxin-3 monoclonal antibody ( Cat #MAB 5360 ) , monoclonal anti-HTT antibody ( MAB 2170 ) and 5TF1-1C2 ( Mab1574 ) were from Millipore , USA . Rabbit polyclonal HAP-1 ( Cat #TA306425 ) was from Origene , USA , and mouse monoclonal HAP-1 ( MA1-46412 ) was from Thermo Scientific , USA . RNA pol II ( sc-899 ) and DNA ligase 3 ( sc-135883 ) were from Santa Cruz Biotechnology , USA . PNKP rabbit polyclonal antibody ( Cat #MBP-1-A7257 ) was from Novus Biologicals , USA , and BioBharati Life Science ( Cat# BB-AB0105 ) , India , and PNKP mouse monoclonal antibody was a kind gift from Dr . Michael Weinfeld ( University of Alberta , Canada ) . SH-SY5Y cells or frozen brain sections were immunostained with anti-PNKP , HTT , CBP , POLR2A , ATXN3 , and anti-polyQ 5T1-1C2 antibodies . Nuclei were stained with DAPI ( Molecular Probe , USA ) and imaged under a confocal microscope . Expression of mHTT or wtHTT was induced in PC12 cells by removing doxycycline from the culture medium for 4–7 days . Induced cells were dissociated with Accutase ( Gibco ) , and collected by centrifugation . Cell toxicity was assayed using a commercially available Annexin-V Cell Toxicity Assay kit ( 4830–01 K , Trevigen , USA ) . 1 × 106 Cells were incubated at room temperature with 1 µl Annexin-V-FITC ( 1 µg/ml ) and 5 ul Propidium Iodide , in the provided binding buffer , for 15 min , before diluting with binding buffer . FITC fluorescence was analyzed by flow cytometry using a Cytoflex ( Beckman Coulter ) , measuring 10 , 000 events per sample . Gating on main cell population was performed by FSC/SSC gating . Positive thresholds determined with unstained negative control , and H2O2 treated positive control samples . Identical thresholds applied to all samples . Data were analyzed using CytExpert software ( Beckman Coulter ) . Images were collected using a Zeiss LSM-510 META confocal microscope with 40 × or 60 × 1 . 2 numerical aperture water immersion objectives . Images were obtained using two excitation wavelengths ( 488 and 543 nm ) by sequential acquisition . Images were collected using 4-frame-Kallman-averaging with a pixel time of 1 . 26 μs , a pixel size of 110 nm , and optical slices of 1 . 0 μm . Z-stack acquisition was performed at 0 . 8 μm steps . Orthogonal views were processed with LSM 510 software . Caspase-3 activities were measured using a Caspase-3 assay kit ( BD Biosciences , USA ) based on hydrolysis of the substrate acetyl-Asp-Glu-Val-Asp p-nitroanilide ( Ac-DEVD-pNA ) , resulting in release of the p-nitroaniline ( pNA ) moiety . Released pNA is detected at 405 nm . Comparison of pNA absorbances from the sample and control allows determination of the fold increase in caspase-3 activity ( relative caspase-3 activity is expressed in arbitrary units ) . SH-SY5Y cells were plated on chamber slides and cultured in DMEM for 24 hr . SH-SY5Y cells or brain sections were fixed with 4% paraformaldehyde , permeabilized with 0 . 2% Tween-20 , washed with 1 × PBS , incubated with primary antibodies for PNKP ( mouse monoclonal ) , HTT ( rabbit polyclonal and mouse monoclonal ) , PNKP ( mouse monoclonal ) , POLR2A ( rabbit polyclonal ) , HAP-1 ( rabbit polyclonal and mouse monoclonal ) , ATXN3 ( rabbit polyclonal and mouse monoclonal ) , and DNA ligase 3 ( rabbit polyclonal ) . These samples were subjected to PLAs using the Duolink PLA kit ( O-Link Biosciences , Sweden ) . Nuclei were stained with DAPI , and PLA signals were visualized under a fluorescence microscope at 20 × magnification . The 3’-phosphatase activity of PNKP in the nuclear extract ( 250–500 ng ) of cells/mouse brains or with purified recombinant His-tagged PNKP ( 25 fmol ) was conducted as we described previously ( Wiederhold et al . , 2004; Mandal et al . , 2012; Chatterjee et al . , 2015 ) . Nuclear extracts for the 3’ phosphatase assay was prepared following standard protocols from cells ( Chakraborty et al . , 2016 ) or mouse brains tissues ( Chakraborty et al . , 2015 ) . A 32P-labeled 3’-phosphate-containing 51-mer oligo substrate with a strand break in the middle ( 5 pmol ) was incubated at 37°C for 15 min in buffer A ( 25 mM Tris-HCl , pH 7 . 5 , 100 mM NaCl , 5 mM MgCl2 , 1 mM DTT , 10% glycerol and 0 . 1 μg/μl acetylated BSA ) with 5 pmol of unlabeled ( cold ) substrate . The reaction was stopped by adding buffer B ( 80% formamide , 10 mM NaOH ) and the reaction products were electrophoresed on a 20% Urea-PAGE to measure the amount of 3’ phosphate release from the radio-labeled substrate . The radioactive bands were visualized in PhosphorImager ( GE Healthcare , USA ) . The data were represented as % of the phosphate release ( % product ) with the total radiolabeled substrate as 100 . Total DNA repair assays were carried out according to the protocol of Wiederhold et al . ( 2004 ) . Briefly , 10 pmol DNA substrate ( a 51-mer DNA-oligo ) annealed to two shorter DNA duplexes , one containing 3'-P and the other with 5'-P with a 4-nt gap in the middle was used to assess total repair activity ( DNA end cleaning +gap filling through polymerization +ligation to fill the ends ) in NEs ( 2 . 5 µg ) from wtHTT- and mHTT-expressing neuronal cells and zQ175 and control mouse brain samples . Total repair activity was also assessed with the same substrate with DNA oligos containing 3'-OH ( clean DNA ends ) . In both cases , the 20 μL reaction mixture contained 1 mM ATP , 50 μM unlabeled dNTPs , and 0 . 5 pmol [α−32P]-dCTP ( the concentration of cold dCTP was lowered to 5 μM ) in BER buffer and incubated for 45 min at 30°C . The reaction products were analyzed with 20% urea-polyacrylamide gel electrophoresis , and the radioactive bands were detected in a Phosphorimager ( GE Life Sciences , USA ) . Freshly dissected brain tissue from transgenic and age-matched control mice was homogenized in TRIzol ( Thermo Scientific , USA ) , and total RNA was extracted using an RNA extraction kit ( Qiagen , USA ) and purified using a DNA-free DNAse Kit ( Ambion , USA ) . Next , 1 μg of total RNA was reverse-transcribed using an RT-PCR kit ( Clontech , USA ) . A cDNA aliquot from each reaction was quantified , and 500 ng of cDNA from each reaction was used for qRT-PCR . 18S rRNA was used as control for the qRT-PCR analysis . The reactions were repeated three times using the following primers . Neurod1: F: AGCCCTGATCTGGTCTCCTT; R: CTGGTGCAGTCAGTTAGGGG Neurod2: F: AAGCCAGTGTCTCTTCGTGG; R: TTGGACAGCTTCTGCGTCTT Neurog1: F: CCAGGACGAAGAGCAGGAAC; R: GGTCAGAGAGTGGTGATGCC Tubb3: F: TGAGGCCTCCTCTCACAAGT; R: ACCACGCTGAAGGTGTTCAT Eno2 γ: F: CCCAGGATGGGGATTTTGCT; R: CCTCCCCTGATCTGCTACCT Pol b: F: TTCCACCGGTAAGACCCAGG; R: GCCAGTAACTCGAGTCAGGA Myod1: F: AGCATAGTGGAGCGCATCTC; R: TTGGGGCTGGATCTAGGACA Myog: F: GAGGAAGTCTGTGTCGGTGG; R: CCACGATGGACGTAAGGGAG Myh2: F: CGAGAGACGAGTGAAGGAGC; R: GAATCACACAGGCGCATGAC Myh4: F: AGCGCAGAGTGAAGGAACTC; R: TCTCCTGTCACCTCTCAACAGA Myh6: F: ATAAAGGGGCTGGAGCACTG; R: TCGAACTTGGGTGGGTTCTG Myh7: F: CCTTACTTGCTACCCTCAGGTG; R: GGCCATGTCCTCGATCTTGT Gapdh F: ATGAGAGAGGCCCAGCTACT; R: TTTGCCGTGAGTGGAGTCAT Bcl2L11: F: TTGGATTCACACCACCTCCG; R: CGGGATTACCTTGCGGTTCT Pmaip 1: F: CTCGCTTGCTTTTGGTTCCC; R: ACGACTGCCCCCATACAATG Bid: F: CCACAACATTGCCAGACATCTCG: R: TCACCTCATCAAGGGCTTTGGC Pidd: F: ACAGAAGAGCCTCGGCAAGTCT: R: GAAAGGCACAGCAGAGGGCTTA Apaf1: F: CACGAGTTCGTGGCATATAGGC: R: GGAAATGGCTGTCGTCCAAGGA ChIP assays were performed using fresh brain tissue of WT mice as previously described ( Chakraborty et al . , 2016; Sailaja et al . , 2012 ) . Briefly , 80–100 mg of freshly harvested CTX was chopped into small pieces and fixed in 1% formaldehyde for 15 min . The samples were centrifuged at 440 × g for 5 min at room temperature , and 0 . 125 M glycine was added to terminate cross-linking . The samples were washed two to three times with ice-cold PBS ( containing protease inhibitors ) and centrifuged each time at 440 × g for 4 min at 4°C . The pellet was resuspended in 1 mL ice-cold lysis buffer ( 10 mM EDTA , 1% [w/v] SDS , 50 mM Tris-HCl [pH 7 . 5] ) with protease inhibitors and PMSF for 15 min and homogenized to produce a single-cell suspension . The samples were then transferred to pre-cooled 1 . 5 mL tubes and centrifuged at 2260 × g for 5 min . The pellet was resuspended in lysis buffer and sonicated to generate ∼500 bp DNA fragments . The samples were centrifuged at 20 , 780 × g for 30 min at 4°C , and supernatants were collected for ChIP . The sheared chromatin was IP’d for 6 hr at 4°C with 10 μg isotype control IgG ( Santa Cruz Biotechnology , USA: sc-2027 ) or anti-HTT antibody . After DNA recovery with proteinase K treatment followed by phenol extraction and ethanol precipitation , 1% of input chromatin and the precipitated DNA were analyzed by qPCR with the following primers . ChIP data are presented as percent binding relative to the input value . Neurod1: F: CTGCAAAGGTTTGTCCCAAGC; R: CTGGTGCAGTCAGTTAGGGG Neurod2: F: CAGGCCCTCCCAAGAGACTT; R: TCGTGTTAGGGTGAAGGCGT Neurog1: F: GCTTGCTCCAGGAAGAACCT; R: AGAGACACCGCTACTAGGCA Tubb3: F: GTGGGGCTCTCCCCTAAAAC; R: TTGGGAGCGCACAGTTAGAG Eno2 γ: F: TAGGGGTGCCTAGTCCTGTC; R: GAGTGCTGGATGTGTGGTCA Myod1: F: ATCTGACACTGGAGTCGCTTT; R: TTAGTCTCAGCTGCTGGTTCC Myog: F: GGCCACCAGAGCTAGAACAG; R: ATGAAGGCTGTGGACTTGGG Myh2: F: TCAGTGAGCAGTGGGAGCTA; R: GTACAAACACGGGGACACCC Myh4: F: AGGTGTACAACTCCGTGGGT; R: GCTCTAGCAAGACCAGTCACG Myh6: F: TCGTGCCTGATGACAAGGAG; R: CTTTCTGGCAAGCGAGCATC Myh7: F: ATTGGTGCCAAGGTGGGTTT; R: CCTGGGGTTCCCAGAATCAC LA-qPCR assays were carried out following an existing protocol ( Santos et al . , 2006 ) . Briefly , tissues were harvested from the cortex ( CTX ) , striatum ( STR ) , and cerebellum ( CRBL ) of control and HD transgenic mice , and genomic DNA was extracted using the genomic-tip 20/G kit ( Qiagen , Germany ) . Genomic DNA was quantified , and gene-specific LA-qPCR analyses were performed using Long Amp Taq DNA polymerase ( NEB , USA ) . Various genomic loci were PCR-amplified from actively transcribing genes in brain ( e . g . , neuronal differentiation factor 1 and 2 [Neurod1 and Neurod2] , neurogenic basic-helix-loop-helix protein neurogenin 1 [Neurog1] , tubulin beta three class III [Tubb3] , neuron-specific enolase 2 [Eno2γ] , and DNA polymerase β [Pol b] ) . Non-transcribing loci ( e . g . , myogenic differentiation factor 1 [Myod1]; myogenic factor 4; myogenin [Myog]; and myosin heavy chain 2 , 4 , 6 , or 7 [Myh2 , Myh4 , Myh6 , or Myh7] ) were amplified using the primers listed below . Loci from genomic DNA isolated from iPSC-derived control and HD primary neurons were PCR-amplified with the primers listed below . The cycle numbers and DNA concentrations were standardized before each final reaction so that the reaction remained within the linear amplification range ( Santos et al . , 2006 ) . The final PCR conditions were optimized at 94°C for 30 s ( 94°C for 30 s , 55–60°C for 30 s depending on the oligo annealing temperature , 65°C for 10 min ) for 25 cycles and 65°C for 10 min . Each reaction used 15 ng of DNA template , and the LA-qPCRs for all studied genes used the same stock of diluted DNA samples to avoid amplification variations due to sample preparation . A small DNA fragment for each gene was amplified to normalize large fragment amplification . The PCR conditions were 94°C for 30 s , 54°C for 20 s , 68°C for 30 s for 25 cycles , and 68°C for 5 min . Short PCR used 15 ng of the template from the same DNA aliquot . The amplified products were visualized on gels and quantified with the ImageJ software based on three independent replicate PCRs . The extent of damage was calculated according to our previously described method ( Chakraborty et al . , 2016 ) . Long: F: CTCGCAGGTGCAATATGAATC; R: GCAACTGCATGGGAGTTTTCT Short: F: CTGCAAAGGTTTGTCCCAAGC; R: CTGGTGCAGTCAGTTAGGGG Long: F: GGCAGTGGTTGGGATGGTAT; R: CTCACTCTGTGCTGTCTGTCTC Short: FP: CAGGCCCTCCCAAGAGACTT; R: TCGTGTTAGGGTGAAGGCGT Long: F: GATGAGCCCCTGAAGACGAG; R: GCCAATCTTGCTTCTTGCGT Short: F: GCTTGCTCCAGGAAGAACCT; R: AGAGACACCGCTACTAGGCA Long: F: GGTACAGGGGATGTGGTTGG; R: GAGTCTCCTGCCTGTCCCTA Short: F: GTGGGGCTCTCCCCTAAAAC; R: TTGGGAGCGCACAGTTAGAG Long: F: CTTGTTCTTCGGGGACCCTC; R: CATCCGTGTGCTTAAGGGGT Short: F: TAGGGGTGCCTAGTCCTGTC; R: GAGTGCTGGATGTGTGGTCA Long: F: TATCTCTCTTCCTCTTCACTT; R: GTGATGCCGCCGTTGAGGGTCTCCTG Short: F: TATGGACCCCCATGAGGAACA; R: AACCGTCGGCTAAAGACGTG Long: F: ATAGACTTGACAGGCCCCGA; R: GGACCGTTTCACCTGCATTG Short: F: ATCTGACACTGGAGTCGCTTT; R: TTAGTCTCAGCTGCTGGTTCC Long: F: ACAAGCCTTTTCCGACCTGA; R: CCATGGCCAAGGCGACTTAT Short: F: GGCCACCAGAGCTAGAACAG; R: ATGAAGGCTGTGGACTTGGG Long: F: ATCTCAGGAGCACCCATCCT; R: GAAAAGGGTGTGCCAAGCAG Short: F: TCAGTGAGCAGTGGGAGCTA; R: GTACAAACACGGGGACACCC Long: F: GACGTGGAACTGTTAGGCCA; R: AAGCCAGAGTCTTCAACCCG Short: F: AGGTGTACAACTCCGTGGGT; R: GCTCTAGCAAGACCAGTCACG Long: F: GACAAGGGGCATTGTAGCCT; R: TCTGCCTACCTTATGGGGCT Short: F: TCGTGCCTGATGACAAGGAG; R: CTTTCTGGCAAGCGAGCATC Long: F: TTTGGGTTGGCCTGTCAGTT; R: ATCCCTAGCTGGGGCTTGTA Short: F: ATTGGTGCCAAGGTGGGTTT; R: CCTGGGGTTCCCAGAATCAC Long: F: TGCTTCTCATGCTTGCTACCAC; R: TCTGTCCCTGTAGGAGGATGT Short: F: CCTGTCCCTTTGTTGGAGGG; R: CGAGGTGGGCTAACAATGGA Long: F: CCGCGCTTAGCATCACTAAC; R: TGGCACTGGTTCTGTGGTATT Short: F: TGCCTCTCCCTTGTTGAATGTAG; R: TTCTTTTTGGGGCCGCGTCT Long: F: CATGTCACCACTGGACTCTGCAC R: CCTGGAGTAGGAACAAAAATTGCT Short: F: AGTGGGCTGGATGTAACCTG R: CCAGTAGATGTGCTGCCAGA Long: F: ACGTGTGCTGCAAGCAATTT; R: CCTGAAACTCCCCTGACACC Short: F: GGTGAGCAATAAGCCAGCCT; R: CAGCTTGTTGCCAGCATGAG Data reported as mean ± SD and the statistical analysis was performed using Sigma Plot ( SYSTAT Software ) . Differences between two experimental groups were analyzed by Student’s t test ( 2-tail , assuming unequal variances ) . When comparing multiple groups , One-way ANOVA was performed followed by Tukey’s post-hoc test to determine significance . In all cases , p<0 . 05 was considered significant . | Our DNA encodes the instructions to make proteins , which then go on to perform many crucial roles in the cell . Breakages and damage to DNA occur over time , and if uncorrected , they can make the instructions illegible or incorrect . A build-up of damages can be harmful – for example , DNA damage from excessive UV light exposure can cause skin cancer . Luckily , cells contain DNA repair complexes , protein machines that surveil DNA and correct errors or breakages . An accumulation of DNA breakages is thought to contribute to the development of Huntington’s disease , a devastating and currently incurable condition where brain cells slowly die . The immediate cause of Huntington’s disease is well known: Huntington’s patients have an abnormal , mutant version of a protein called huntingtin . However , it is still unclear how the mutant huntingtin causes the symptoms of the disease and participates in cell death . Gao et al . carefully studied the proteins that huntingtin physically interacts with . The experiments revealed that huntingtin is part of a newly identified DNA repair complex that fixes breakages in DNA as the molecule is ‘read’ by the cell . The presence of the normal huntingtin protein promoted DNA repair . However , when the healthy huntingtin was replaced with the mutant version found in Huntington’s disease , the activity of the DNA repair complex was greatly reduced . This resulted in a build-up of DNA errors , triggering a series of events that ultimately led to cell death . In addition , in mice engineered to produce the mutant version of huntingtin , the accumulation of DNA damage was particularly important in two brain regions that are severely damaged in patients with Huntington’s disease . There is currently no effective treatment for Huntington’s disease . However , understanding how the mutant huntingtin damages brain cells may provide new targets for future therapies . More broadly , several other brain disorders share similarities with Huntington’s disease , and it remains to be seen whether the same mechanisms could be at work in all these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2019 | Mutant huntingtin impairs PNKP and ATXN3, disrupting DNA repair and transcription |
Enzymes enable life by accelerating reaction rates to biological timescales . Conventional studies have focused on identifying the residues that have a direct involvement in an enzymatic reaction , but these so-called ‘catalytic residues’ are embedded in extensive interaction networks . Although fundamental to our understanding of enzyme function , evolution , and engineering , the properties of these networks have yet to be quantitatively and systematically explored . We dissected an interaction network of five residues in the active site of Escherichia coli alkaline phosphatase . Analysis of the complex catalytic interdependence of specific residues identified three energetically independent but structurally interconnected functional units with distinct modes of cooperativity . From an evolutionary perspective , this network is orders of magnitude more probable to arise than a fully cooperative network . From a functional perspective , new catalytic insights emerge . Further , such comprehensive energetic characterization will be necessary to benchmark the algorithms required to rationally engineer highly efficient enzymes .
Scientists have long marveled at the enormous rate enhancements and exquisite specificities of enzymes . Remarkable progress has been made since catalysis was viewed as a ‘life force’ just over a century ago ( Buchner , 1897; Hein , 1961; Barnett , 2003 ) . Now , the chemical moieties involved in enzymatic transformations can be identified by a combination of structural and functional approaches . The positions of functional groups in X-ray structures can typically be combined with chemical intuition to derive mechanisms for enzymatic reactions ( Benkovic and Bruice , 1966; Walsh , 1979; Sinnott , 1998; Silverman , 2002 ) . Such mechanisms are routinely supported by site-directed mutagenesis experiments in which large deleterious rate effects are observed when putative catalytic residues are mutated . While there remain a subset of reactions that are less understood and whose important and fascinating reaction details are still being worked out ( e . g . , Das et al . , 2011; Weeks et al . , 2012 ) , we can write reasonable chemical mechanisms for the vast majority of enzymes ( Benkovic and Bruice , 1966; Walsh , 1979; Sinnott , 1998; Silverman , 2002 ) . In contrast , our understanding of the energetics that underlie enzymatic catalysis is far less developed . Such understanding is critical for elucidating the pathways that have been followed in molecular evolution and for designing new , highly efficient enzymes . The current dominant mechanistic tools , X-ray crystallography and removal of catalytic residues via site-directed mutagenesis , while powerful , have fundamental limitations . Structures reveal the positions of functional groups in active sites , but these static pictures do not allow reaction probabilities to be determined . Specifically , although energies derived from a given structure can , in principle , reveal potential energies , full sampling of the ensemble states of the enzyme , substrates , and surrounding solvent is needed to obtain free energies . It is difficult to combine such sampling with high-level energy functions and to make nontrivial predictions required for independent assessment of such energetic models . Site-directed mutagenesis has two fundamental limitations . First , site-directed mutagenesis reads out the difference in free energy of the reaction ( ΔΔG ) for the mutant enzyme relative to the wild type ( WT ) , and so does not report an absolute energetic contribution to catalysis ( Kraut et al . , 2003; Herschlag and Natarajan , 2013 ) . As an example , the vastly different assignments of the catalytic contributions of residues in the Ketosteroid Isomerase oxyanion hole , dependent on the type and extent of mutation , provide a particularly clear demonstration of this limitation and underscore the need to clearly and explicitly define the comparison states ( Kraut et al . , 2010; Schwans et al . , 2011 ) . The second limitation of site-directed mutagenesis is that enzymatic residues do not act in isolation ( Narlikar and Herschlag , 1998; Kraut et al . , 2003 ) . The apparent contribution of one residue is also a function of the surrounding residues and the overall structure , as demonstrated by energetic coupling in double mutant cycles and more dramatically by the fact that a denatured protein still contains its catalytic residues but these so-called catalytic residues no longer provide catalysis ( see Kraut et al . , 2003 for discussion ) ( Carter et al . , 1984; Horovitz , 1996 ) . One seemingly important aspect of the interconnectivity of enzymatic residues is highlighted in X-ray structures , which typically reveal or suggest active site hydrogen bond networks and imply a functional connection between the identified catalytic residues and these more extensive ‘network residues’ . Indeed , prior investigations have identified networks of energetically coupled residues in enzymes that contribute synergistically to catalysis . These studies include incisive double mutant cycle analysis demonstrating functional and energetic connections between residues ( Hermes et al . , 1990; Dion et al . , 1993; Horovitz et al . , 1994; Rajagopalan et al . , 2002; Masterson et al . , 2008; Singh et al . , 2014 ) . Larger scale coupled networks have been observed through statistical analysis of co-evolution and characterization of individual residue's relaxation timescales by NMR ( e . g . , Lockless and Ranganathan , 1999; Eisenmesser et al . , 2002 , 2005; McElheny et al . , 2005; Freedman et al . , 2009; Halabi et al . , 2009; Doucet et al . , 2011 ) . A remaining challenge , undertaken in this study , is to link mutant cycle analysis to extended networks . Currently , we lack the ability to ascertain the energetic properties of network residues from structural inspection , first principles , or empirical models . For example , do these networks act as fully cooperative units where disruption of any connection would dissipate the advantage from all of the residues in the network ? In the other extreme , do certain side chains act independently from their network neighbors , positioned for function by their backbone placement and/or packing interactions with other portions of the side chain ? Although either extreme might be presumed unlikely , such expectations are not grounded in data given the current absence of quantitative assessments of these extended networks . We have therefore investigated the functional behaviors of an interaction network hypothesized from available structural data in the Escherichia coli alkaline phosphatase ( AP ) active site ( Figure 1 ) . Our experiments reveal and quantitatively delineate the energetic interconnectivity within this highly proficient active site , the type of active site that will be necessary to create if we are to engineer enzymes that rival the catalytic power of natural enzymes . 10 . 7554/eLife . 06181 . 003Figure 1 . Alkaline phosphatase ( AP ) structure and active site . ( A ) The three-dimensional structure of AP with bound Pi ( PDB 3TG0 ) . Active site residues are depicted as follows: D101 , brown; R166 , black; D153 , red; K328 , green; E322 and Mg2+ ion , blue . ( B ) A close-up of AP active site from two angles . Dashes represent putative hydrogen bonds . Residues colored as in part ( A ) . ( C ) Schematic of AP active site interactions represented with the phosphoryl transfer transition state . Residues colored as in part ( A ) . ( D ) Reaction scheme for phosphomonoester hydrolysis by AP , where ROP represents a phosphate monoester dianion substrate , and E-P represents the covalent seryl-phosphate intermediate ( Coleman , 1992 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06181 . 003
Figure 3A shows the effects from mutating each of the five active site residues depicted in Figure 1C . Each residue has a significant effect ranging from 64 to 88 , 000-fold , with the largest effects coming from R166 mutation and Mg2+ ion removal ( E322Y ) . Previous studies have shown that replacing the E322 side chain with a tyrosine leads to loss of Mg2+ from the active site ( Zalatan et al . , 2008 ) . We confirmed this result and showed that the E322Y mutant reaction is not activated by the presence of Mg2+ ( ‘Materials and methods’ ) . The absence of Mg2+ binding and activation is consistent with the finding that other members of the AP superfamily that lack Mg2+ have a tyrosine at this position ( Zalatan et al . , 2008 ) . Indeed , we chose the E322Y mutation for our studies because it gives the same functional effect as E322A while also creating this steric block to Mg2+ binding ( Zalatan et al . , 2008 ) . While we did not test all possible mutations , several different mutations were tested at each position and shown to give very similar effects , providing no indication of idiosyncratic effects from any of the subtractive mutations made in this study ( Appendix 1 Table 1 ) . 10 . 7554/eLife . 06181 . 006Figure 3 . Single-mutation effects and additivity predictions . Rate effects from removing individual residues from WT AP ( A ) or restoring individual WT residues to AP minimal ( B ) . The symbol ( ± ) indicates which residue is varied . Residues are color-coded as in Figure 1: D101 , brown; R166 , black; D153 , red; K328 , green; and E322 and Mg2+ ion , blue . The following mutations were made: D101A , R166S , D153A , K328A , and E322Y; several alternative mutations gave similar effects ( Appendix 1 Table 1 ) . To the right of the dashed line is the activity of WT AP relative to AP minimal observed ( A , B , grey bars ) and predicted from the effects of removal of each WT residue from the WT background and assuming independent ( energetically additive ) effects ( A , open bar ) or from the effects of addition of each WT residue in the minimal background , assuming independence ( B , open bar ) . DOI: http://dx . doi . org/10 . 7554/eLife . 06181 . 006 We next determined the degree to which these residues were dependent on one another in two simple ways . First , we removed all five residues simultaneously; if independent , the effects of removal would be multiplicative—that is , energetically additive ( Equation 1 ) . However , the cumulative effect assuming additivity gives a predicted activity that is 106-fold lower than the observed activity of the minimal construct ( Equation 1 ) , indicating substantial energetic interdependence between these residues ( Figure 3A ) ;ΔΔG‡=[−RTln ( ( kcat/KM ) obsd/ ( kcat/KM ) predicted ) ]=8 . 3 kcal/mol , ( 1 ) krel , predicted=krel , D101 × krel , R166 × krel , D153 × krel , K328× krel , Mg2+ . We next added each residue back in isolation to the minimal AP construct lacking all five residues , and we compared the rate effect from the added residue to the corresponding rate decrease from removing that residue from WT AP ( Figure 3A , B ) . In each case , there was a larger deleterious effect from removing the residue than the rate enhancement afforded by adding it back . The differential effects ranged from 2 . 6 to >1900-fold , corresponding to differential energetic effects of up to at least 4 . 5 kcal/mol . Indeed , D153 added back in isolation is deleterious by at least 10-fold ( Figure 3B ) but beneficial by 230-fold when added in the otherwise WT background ( Figure 3A ) . Further , the small 2 . 6-fold ( 0 . 57 kcal/mol ) differential between restoring D101 to AP minimal and removing it from WT AP likely arises coincidentally from different mechanistic contributions ( see below ) . Thus , each of the five residues has some energetic connection to at least one residue in this network . Overall , the contribution from adding all five residues back is 580-fold greater than predicted from assuming energetically additive effects from addition of each residue individually ( Equation 1 ) . An empirical or theoretical framework to explain this energetic behavior is lacking . In particular , there is no way to discern from structural inspection of an active site whether all neighbors have substantial energetic interactions , the scale and form of these energetic connections , or whether and how far beyond nearest neighbors functional effects extend . Therefore , we proceeded to define the interconnections between active site residues for this model enzyme system . Figure 2A depicts the set of all 32 possible combinations of mutations of the five catalytic residues , using the color-coding from Figure 1 to represent the presence ( colored ) or absence ( white ) of each residue . In the following sections , we describe the energetic properties and interconnections of these residues . X-ray crystallographic structures reveal connections between residues but cannot define the energetic properties of these connections . Site-directed mutagenesis reveals residues that give large functional effects upon mutation , and double mutant cycles provide powerful tests for energetic dependence vs independence between two residues ( Carter et al . , 1984; Hertel et al . , 1994; Horovitz , 1996; Narlikar et al . , 1999 ) . However , residues never function in isolation , and essentially every residue in a protein structure can be connected , via chains of hydrogen bonding and packing interactions , to every other residue . In cases of allostery , there are energetic interactions over large distances , and in certain cases the residues and conformations involved in the allosteric transition have been mapped ( Monod et al . , 1965; Shulman et al . , 2004; Cui and Karplus , 2008; McLaughlin et al . , 2012 ) . We are particularly interested in active site networks because the residues constituting these networks are critical for the most basic catalytic functions of enzymes , these networks have not been previously mapped , and an understanding of their extent and properties may facilitate engineering of enzymes that rival natural enzymes in catalytic efficiency and specificity . We have interrogated a network of five residues in the active site of the model enzyme AP ( Figure 1B , C; Figure 2A ) . A priori , these residues could have functioned fully cooperatively or fully independently . Although either extreme might have been considered unlikely , there were no data for this system ( or , to our knowledge , other systems ) that would allow us to know or predict the degree and extent of functional interconnectivity . We observed three overlapping functional units , D101/R166/D153 , D153/Mg2+/K328 , and D101/Mg2+ ( Figure 9 ) . Each functional unit exhibited distinct energetic behaviors . The aspartate residues of the D101/R166/D153 unit make nearly independent contributions to the catalytic function of the central arginine residue; D153 and K328 ( and presumably their associated water molecules ) act cooperatively with the Mg2+ ion and increase its contribution to catalysis; and the Mg2+ ion and D101 are energetically anti-cooperative and apparently constitute a redundant functional unit . A mathematical model that describes these energetics quantitatively accounts for all of the AP variants , reproducing the observed catalytic rates for all 28 mutants within a factor of two and predicting the catalytic rates of the remaining four variants that were not tested herein ( Appendix 1 Table 2 ) . Nevertheless , the atomic-level interactions and properties that underlie each of these distinct and complex energetic behaviors are not known . We propose the following models: each aspartate residue of the D101/R166/D153 unit helps to position R166 , thereby lowering the conformational entropy cost for making its interactions with the two phosphoryl oxygen atoms; the network of the D153/Mg2+/K328 functional unit and its two associated water molecules exists in alternative , likely less structured conformations until all three of these groups are present and can position hydrogen bond donors to one of the oxygen atoms of the transferred phosphoryl group; and D101 and the Mg2+ ion can each position the serine oxyanion nucleophile with respect to Zn22+ , rendering their contributions redundant . These models will require future testing and refinement . The interconnections between the functional units ( Figure 9 ) —that is , the residues common to more than one functional unit—may be favored both evolutionarily and functionally . An interconnected active site will have fewer residues and thus require a less expansive search of mutational space . Further , multiple functional groups need to ‘pack’ into a limited space to interact with the substrate and stabilize the transition state . The approach taken in this study introduces a powerful opportunity to deepen the feedback loop between experiment and computation . It is generally recognized that the most powerful tests of computational approaches are true predictions , made ‘blindly’ with pertinent data not yet collected or obtained by others but not released prior to reporting computational results and predictions . Such blind predictions have proven critical in other areas of biochemistry and biophysics to determine whether seemingly predictive and descriptive algorithms were indeed predictive and correct ( Moult et al . , 2004 , 2011; Nielsen et al . , 2011; Cruz et al . , 2012; Dill and MacCallum , 2012 ) . Nevertheless , the vast majority of computational studies of enzyme mechanism make ‘predictions’ for results that are already known and are thus not true independent tests . Further , single active site mutations predominantly give effects that fall within typical and rather narrow ranges , limiting the usefulness of traditional site-directed mutagenesis as a robust test of computational approaches . We suggest that computational predictions of the rate effects from multiple mutations may provide extensive and nontrivial predictions that can be quantitatively tested by experiment and that are needed to effectively advance computational methods and our understanding . Our group is willing to test predictions from individual computational groups or consortia , using AP or other systems where we know robust kinetic measurements are possible; we are willing to carry out such experiments in advance and withhold the results or send them to an independent evaluator to ensure that comparisons between experiment and computation will be possible and made in a timely manner; and we are willing to discuss with computational groups the best systems to carry out such tests . We strongly believe that such synergistic approaches , informal and formal , will be required to unite computational and experimental enzymology and to make the greatest advances in our understanding of catalysis . There has been recent excitement about the ability to design new enzymes , some of which catalyze reactions not seen in nature ( Röthlisberger et al . , 2008; Siegel et al . , 2010; Hilvert , 2013; Bos and Roelfes , 2014 ) . The ability to repurpose and create protein scaffolds and to place functional groups in desired locations is a truly remarkable advance . Nevertheless , designed enzymes to-date have modest rate enhancements relative to naturally occurring enzymes and can lack the stereospecificity observed with naturally occurring enzymes ( Wolfenden and Snide , 2001; Lassila et al . , 2009; Baker , 2010; Bos and Roelfes , 2014 ) . Indeed , enzyme mimics and bovine serum albumin ( BSA ) can catalyze reactions with rate enhancements similar to designed enzymes prior to their improvement by randomization and selection ( Kirkby et al . , 2000; Schmidt et al . , 2013 ) . These observations raise the following questions: what distinguishes naturally occurring enzymes from current designed enzymes , and what is needed to achieve more proficient designs ? The most apparent difference is the absence of extended and extensive hydrogen bond networks in and around active sites of designed enzymes . For example , the most carefully studied designed enzyme , a retroaldolase , has a lysine residue placed within a hydrophobic pocket ( Jiang et al . , 2008 ) . Rate enhancements are achieved by lowering the lysine pKa due to its non-polar environment ( to increase the concentration of the reactive free amine at neutral pH ) and from binding the hydrophobic substrate in this lysine-containing pocket; these mechanisms provide ∼105-fold catalysis for a designed enzyme , with modest additional rate enhancement obtained through subsequent rounds of selection ( Lassila et al . , 2009; Hilvert , 2013 ) . We hope that an empirical understanding of the extent and properties of active site networks will help in future design efforts as well as promote more conceptual and theoretical understanding . Although it would not be appropriate to generalize from the single example of AP , our dissection of the AP active site network does show that optimal positioning of catalytic residues can occur with only a subset of the full network; thus , it may be possible to attempt designs with active site modules that correspond to functional units identified in this study and , we hope , subsequent studies . A recent attempt to incorporate an active site catalytic triad produced designed enzymes that , after several rounds of selection , were able to catalyze a side reaction with efficiencies similar to analogous natural enzymes ( Rajagopalan et al . , 2014 ) . Taking a longer view , the hand-in-hand development and testing of computational approaches , as outlined above ( see ‘Synergy between experimental and computational enzymology’ ) , will ultimately provide foundational models for the efficient design of highly effective and specific new enzymes . The evolution of a fully cooperative network would be exceedingly improbable and more difficult the larger the number of constituent residues , as all of the residues would need to arise through random drift with a selective advantage accruing only once the entire network was in place . A probability model in which each of five residues would arise with a one in twenty probability ( i . e . , one active residue out of the total 20 amino acids ) and all are needed for a selective advantage ( i . e . , full cooperativity ) has a probability of 1/ ( 6 . 8 × 105 ) ( Figure 10A , bottom pathway and Appendix 3 ) . This probability arises because there are five steps ( or positions ) each with 20 possible residues , and there are five chances of ‘choosing’ a WT residue , starting with AP minimal with all five WT residues missing , four chances of choosing the next residue , etc . 10 . 7554/eLife . 06181 . 017Figure 10 . Cooperative and independent models of active site evolution . ( A ) Schematic comparing fully cooperative ( bottom ) and stepwise ( top ) models for a single pathway . In the fully cooperative model , simultaneous acquisition of all five WT residues is required to confer a selective advantage , leading to a mean waiting time of 4 . 7 × 105 ( arbitrary units ) considering all 120 pathways of adding in the residues ( black net rates; for simplicity , only one intermediate of the multiple possible mutant combinations is shown in each step ) . In contrast , the stepwise model , in which acquisition of any WT residue confers a fitness advantage and is thus irreversible ( top , black numbers ) , has a minimum mean waiting time of 32 . If only one of the 120 pathways leads to a stepwise increase in fitness ( top , grey numbers ) then the mean waiting time would be 69 . The model and simplifying assumptions made to highlight the differences arising from the presence or absence of cooperativity are described in Appendix 3 . ( B ) Model of active site evolution showing the 120 possible paths in the AP landscape for introduction of the five residues investigated herein , in an otherwise WT background . A stepwise model in which acquisition of any WT residue is considered irreversible and all paths are possible would result in a mean waiting time of 32 ( all arrows , grey and black , same as part A , top ) . As a subset of mutagenic steps toward WT AP ( 36 of the 80 potential evolutionary steps ) confers a selective advantage ( here defined as a rate increase of >threefold ) and paths containing steps that do not confer such an advantage have much lower probabilities , we consider the 34 of 120 pathways that provide a monotonic fitness increase as all five WT residues are added . This gives a mean waiting time between the mean waiting times for the stepwise models for a single pathway and 120 pathways , 32 and 69 , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 06181 . 017 Conversely , evolution of a hypothetical enzyme in which each active site residue provides an independent rate advantage would be considerably more probable , as addition of those residues would lead to monotonically increasing catalysis ( i . e . , be continually uphill on a fitness landscape and thus selected for after each addition ) . We first consider a simplified model with a single pathway of five successive steps that each lead to increased fitness , each with a one in twenty probability representing a single advantageous residue of the twenty possible residues ( Figure 10A , top path , black numbers ) . The probability of achieving the final state for a multi-step process can be considered in terms of a mean waiting time , akin to a reaction's half-time or the inverse of its rate constant ( scaled by ln 2 ) . In this case , with five successive irreversible steps each with probability 1/20 , the ‘rate constant’ is 1/100 and the mean wait time is 69 ( in arbitrary units; Appendix 3 ) . ( Calculating probabilistic waiting times using a hidden Markov Model , such as is often used for probability calculations , gives the same relative values as the kinetic mean times [Appendix 3] . ) This kinetic mean wait time is about three orders of magnitude lower than that for the fully cooperative process , which is obtained by computing the mean waiting time: 4 . 7 × 105 ( =6 . 8 × 105 × ln 2; Appendix 3 ) . Thus , if it were functionally equally probable to obtain an active site with these different underlying energetic properties ( i . e . , fully cooperative vs stepwise; Figure 10A ) , evolution would favor the non-cooperative solution by greater than 1000-fold . In other words , for every enzyme that evolved five residues with full functional dependence on one another , there would be more than 1000 with active sites containing residues whose stepwise addition each provided a selective advantage . As we observed functional units exhibiting a range of energetic behaviors within the AP active site , we asked the question: where does the energetic behavior of AP place it along this wide range of evolutionary probabilities ? There are 120 possible pathways from AP minimal to WT AP ( Figure 10B ) . To assess the range of possible mean waiting times , we consider two limiting cases ( Figure 10A , top pathway ) : i . The case in which there is a single pathway of the 120 that is favorable; this is the simplified model presented above and gives a mean wait time of 69 . ii . The case in which all 120 potential pathways are favorable—that is , proceed with a selective advantage at each step; in this case the mean waiting time is 32 , even shorter than for case ( i ) because there are more ways to traverse the landscape from AP minimal to WT AP . Because we have rate constants for each AP species ( Table 1; Figure 2 ) , we can determine which pathways would confer a selective advantage . Using an arbitrary minimal cutoff of >threefold increase in kcat/KM , 34 of the 120 pathways confer a fitness advantage; with a cut-off of a fivefold increase in kcat/KM , 28 of these pathways remain . Thus , although the AP active site residues studied herein have varied energetic behaviors , in many cases addition of a WT residue leads to a significant rate increase , providing multiple favorable evolutionary routes . While not all 120 pathways are favorable for AP , many are , and the mean waiting time will thus be within the range of 32–69 , >1000-fold shorter than the expected waiting time to evolve a fully cooperative network . Although a fully cooperative network would probably not have been anticipated , we were unaware of how strong the evolutionary pressure would be for stepwise increases in fitness . Similar conclusions have been drawn for the evolution of unlinked loci where largely additive rather than fully cooperative or synergistic effects have been observed ( Arnegard et al . , 2014 ) . It will be fascinating to explore more broadly the interplay of functional effects and evolutionary probabilities and how this interplay has biased the complement of extant enzymes . Such underlying probabilistic preferences presumably also impact biological solutions at higher levels of function such as gene regulation and neuronal function ( McLean et al . , 2011 ) . Many enzymes have loops or flaps that close over active sites or domains that accomplish analogous closure ( Pai et al . , 1977; Bennett and Steitz , 1978; Alber et al . , 1987 ) . These events exclude solvent , and it is often stated or implied that the exclusion of water provides the underlying driving force to evolve these processes ( Pai et al . , 1977; Bennett and Steitz , 1978; Harris et al . , 1997; Cleland et al . , 1998; Richard and Amyes , 2004 ) . However , the observation that water is excluded does not logically indicate that the exclusion of water causes enhanced catalysis . AP has an open cavity to allow cleavage of a wide range of phosphate esters , consistent with its putative role in scavenging inorganic phosphate; nevertheless , it has one of the largest known rate enhancements , suggesting that full exclusion of water is not required to attain extremely efficient enzymatic catalysis . AP does , however , appear to exclude water from access to the non-bridging phosphoryl oxygen atoms , and we considered whether that exclusion is catalytically important . Removal of R166 ( and D101 ) , which allows solvent access to the phosphoryl oxygen atoms ( O'Brien et al . , 2008 ) does not diminish the catalytic contribution from the Mg2+ ion or the D153/Mg2+/K328 functional unit , which also interact with at least one of these phosphoryl oxygen atoms . Thus , at least for AP , the wholesale exclusion of solvent from the active site does not enhance catalytic contributions . We suggest , most generally , that flap or domain closure allows the establishment of additional specific interactions for optimal transition state interactions while providing a route for the ingress and egress of substrates and products , rather than generic rate enhancements from solvent exclusion ( Wolfenden , 1974 , 1976; Alber et al . , 1987; Herschlag , 1988; Pompliano et al . , 1990; Herschlag , 1991 ) . A deeper investigation into the energetics and energetic interconnections within an active site—that of E . coli AP—has led to new knowledge , new models for catalytic effects , and a quantitative assessment of the evolutionary probability of establishing an active site network . Nevertheless , this is just one study , and it remains to be determined how similar or different the behaviors of active sites of other enzymes are . In addition , while extending beyond traditional studies , our work investigates a miniscule fraction of the possible connections and interrelationships in AP . Of particular interest will be understanding the interplay between the broader scaffold and the active site residues along with their functional roles and energetic connections , and such studies will ultimately rely on the development of methods that are high-throughput and highly quantitative .
WT and mutant AP were purified from a fusion construct containing an N-terminal maltose binding protein ( MBP ) tag and a C-terminal strepII tag with a factor Xa cleavage site between it and the natural C-terminal end of AP . D101 was mutated to alanine , R166 to serine , D153 to alanine , K328 to alanine , and the Mg2+ ion ligand E322 to tyrosine to prevent the Mg2+ from binding in the active site; these mutations were made alone and in combination . To test for idiosyncratic effects , mutations to alternative residues were tested ( Appendix 1 Table 1 ) . E . coli SM547 ( DE3 ) cells were transformed with the MBP-AP-strepII constructs and were grown to an OD600 of 0 . 6 in rich media and glucose ( 10 g of tryptone , 5 g of yeast extract , 5 g of NaCl , and 2 g of glucose per liter ) with 50 µg/ml of carbenicillin at 37°C . IPTG was added to a final concentration of 0 . 3 mM to induce protein expression . Cultures were then grown at 30°C for 16–20 hr . Cells were harvested from 2 l culture by centrifugation at 4400×g for 20 min and lysed with osmotic shock . The cell pellet was resuspended in 800 ml 20% sucrose solution ( 30 mM Tris-HCl , pH 8 . 0 , 1 mM EDTA ) and incubated at room temperature for 10 min on a shaking table . The cells were pelleted by centrifugation at 13 , 000×g for 10 min . The pelleted cells were resuspended in 800 ml ice cold water at 4°C . The cells were incubated on a shaking table at 4°C for 10 min and then pelleted at 13 , 000×g for 20 min . The supernatant was adjusted to 10 mM Tris-HCl , pH 7 . 4 , 200 mM NaCl , and 10 µM ZnCl2 . The sample was then passed over a 10 ml amylose resin ( New England BioLabs , Ipswich , MA ) gravity column . All mutants were purified with fresh amylose resin to prevent inadvertent co-purification with other mutants . The amylose column was washed with 10 column volumes of 10 mM Tris-HCl , pH 7 . 4 , 200 mM NaCl , and 10 µM ZnCl2 and eluted with the same buffer supplemented with 10 mM maltose . Protein-containing fractions were concentrated by centrifugation through a 10 kDa cutoff filter ( Amicon ) and buffer exchanged at least twice into 10 mM sodium MOPS , pH 7 . 0 , 50 mM NaCl , 100 µM ZnCl2 , 1 . 0 mM MgCl2 unless the E322Y mutation was present , in which case MgCl2 was omitted . For all enzymes , purity was determined to be >95% by SDS-PAGE gel electrophoresis based on staining with Coomassie Blue . To further test for a possible contaminant and to determine the reproducibility of the results , nine out of the 28 mutants were re-expressed and re-characterized: K328A , R166S , E322Y , D101A/D153A , D101A/E322Y , R166S/K328A , D101A/D153A/R166S , D153A/R166S/E322Y/K328A , and D101A/D153A/R166S/E322Y/K328A . Independent preparations gave activities within twofold of one another . To most strongly test whether the observed activities might arise from an enzyme contaminant , the construct with all five residues mutated was further mutated by removal of the serine nucleophile ( to give S102G/D101A/D153A/R166S/E322Y/K328A AP ) . This mutant had no measurable activity above background and a reaction rate at least 100-fold lower than that for any of the AP mutants reported herein , suggesting that the observed activities do not arise from a contaminant . To control for potential unintended complications specific to the mutant residue introduced , several additional mutations were tested ( Appendix 1 Table 1 ) . The values of kcat/KM with mutations to different residues were within twofold in all cases . Alternative AP variants investigated from previous papers are also added to the table and have similar efficiencies . Activity measurements were performed in 0 . 1 M MOPS , pH 8 . 0 , 0 . 5 M NaCl , 100 µM ZnCl2 , and 500 µM MgCl2 at 25°C in a UV/Vis Lambda 25 spectrophotometer ( Perkin Elmer , Waltham , MA ) , unless otherwise noted; for mutants containing E322Y , MgCl2 was excluded . The formation of free p-nitrophenolate from hydrolysis of the substrate p-nitrophenol phosphate ( pNPP ) was monitored continuously at 400 nm . Rate constants were determined from initial rates , and the activity of the free enzyme , kcat/KM , was determined . ( Rate measurements were limited to kcat/KM measurements because kcat for WT represents dissociation of product rather than a chemical step and because the rate-limiting step for kcat could vary with mutation . We ensured that the chemical step was rate limiting for the kcat/KM comparisons carried out herein , as described in Table 3 and described in Appendix 1 . ) 10 . 7554/eLife . 06181 . 018Table 3 . Kinetic constants for Me-P hydrolysisDOI: http://dx . doi . org/10 . 7554/eLife . 06181 . 018kcat/KM ( M−1s−1 ) AP mutantpNPPMe-P ( kcat/KM ) pNPP ( kcat/KM ) Me-PiWT* , †3 . 3 × 1071 . 2 × 106283 . 5 × 10580R166S†1 . 0 × 105110910561800E322Y‡7 . 2 × 1031 . 64500D101A9 . 9 × 1062 . 7 × 1033600D153A2 . 8 × 1061 . 1 × 1032500D101A/D153A3 . 3 × 105615400* ( kcat/KM ) obsd; the chemical step is not rate limiting . †Values of kcat/KM of 1 . 2 × 106 M−1s−1 and 110 M−1s−1 for WT and R166S , respectively for Me-P was obtained previously from reference ( O'Brien and Herschlag , 2002; O'Brien et al . , 2008 ) . This difference in value would not affect the conclusions herein . ‡From reference ( Zalatan et al . , 2008 ) . The efficiency of Me-P hydrolysis was measured for mutants with pNPP activities close to the rate of diffusion . The ratios measured for the enzyme with pNPP and Me-P were close to what had been measured previously for R166S and E322Y , two mutants for which chemistry is rate limiting for both substrates . The ratios suggest that chemistry is rate limiting for the pNPP hydrolysis in D101A and D153A . At least two different enzyme concentrations and at least seven different substrate concentrations were used for each enzyme . Enzyme concentrations were varied by at least fivefold , and substrate concentrations were extended to at least fivefold below the KM value for each enzyme based on KM values determined over wider ranges of substrate . Reaction rates were linear in enzyme concentration at each substrate concentration for each enzyme , and no reaction was observed without added enzyme . Values of kcat/KM determined from linear fits to the lowest substrate concentrations were the same , within error , as values determined from full Michaelis–Menten fits , and R2 values were >0 . 98 in all cases . For E322Y AP and D153A/E322Y , because of their very low KM ( ∼0 . 5 µM ) , kcat/KM was determined from rate measurements in the presence of inhibitory Pi and the independently measured inhibition constant of Pi using an alternative , high KM substrate , as described previously ( Zalatan et al . , 2008 ) . Errors were estimated from two independent kinetic measurements , and comparisons with independent enzyme preparations for nine of the AP variants gave the same values and similar error estimates as the same preparation used on separate days . All mutants were incubated at room temperature for at least a week to test for Zn2+ activation . The E322Y mutants exhibited time-dependent Zn2+ activation , as observed previously , and were shown to level to a maximal rate ( Zalatan et al . , 2008 ) . Mutants with E322 not mutated were incubated in storage buffer with Mg2+ added ( 10 mM sodium MOPS , pH 7 . 0 , 50 mM NaCl , 100 µM ZnCl2 , 1 mM MgCl2 ) . To test for kinetic effects arising from decreased metal affinity of the mutants compared to WT , mutants R166S , D101A/R166S/D153A/K328A , D101A/D153A , E322Y , D101A/R166S/D153A/E322Y/K328A , and D101A/R166S were also incubated with varying Zn2+ and Mg2+ concentrations for a week . The following metal ion concentrations were used in the incubation experiment: 100 µM ZnCl2; 1 . 0 mM ZnCl2; 100 µM ZnCl2 and 100 µM MgCl2; 100 µM ZnCl2 and 1 . 0 mM MgCl2; 100 µM ZnCl2 and 10 mM MgCl2; 1 . 0 mM ZnCl2 and 10 mM MgCl2 . Only AP mutants with E322Y exhibited time-dependent activation , as observed previously ( Zalatan et al . , 2008 ) ; the Mg2+ concentration did not affect the Zn2+ activation; and the activities of the E322Y mutants were not dependent on the presence of Mg2+ . The activities of mutants R166S , D101A/R166S/D153A/K328A , D101A/D153A , and D101A/R166S do not increase with higher concentrations of Mg2+ . Phosphate monoester hydrolysis by D101A , D153A , R166S , and D101A/D153A AP was also measured with methyl phosphate ( Me-P ) , using the same reaction buffer and conditions as for pNPP . The formation of the inorganic phosphate ( Pi ) product from Me-P hydrolysis was monitored discontinuously by withdrawing aliquots from ongoing reactions , quenching in 6 M guanidine-HCl , and detecting Pi with a modified Malachite Green assay ( Lanzetta et al . , 1979 ) at eight or more specified times . Rate constants were determined from initial rates , and activity of the free enzyme , kcat/KM , was determined . At least two different enzyme concentrations and at least seven different substrate concentrations were used for each enzyme . Enzyme concentrations were varied by at least fivefold , and substrate concentrations were extended to at least fivefold below the KM value for each enzyme based on KM values determined over wider ranges of substrate . Reaction rates were linear in enzyme concentration at the lowest substrate concentration for each enzyme , and no reaction was observed without added enzyme . R2 values were >0 . 98 in all cases . To test if the most mutated AP mutant , AP minimal ( D101A/R166S/D153A/E322Y/K328A ) , had Mg2+ in the active site and full Zn2+ occupancy , we carried out atomic emission spectroscopy , as previously used for the E322Y single mutant , with the AP minimal mutant ( Zalatan et al . , 2008 ) . The metal ion occupancies were consistent with an active site saturated with Zn2+ and lacking Mg2+ ( Zn2+:protein ratio 2 . 49; Mg2+: protein ratio 0 . 01; Pi: protein ratio 0 . 06 ) . As described above , kinetic experiments were also carried out to test for Mg2+ activation . The MBP tag used for purifying D101A/D153A was cleaved with factor Xa , and the enzyme was separated from the tag over a 5 ml HiTrap Q HP column ( GE Healthcare , Amersham , UK ) . The purified enzyme was buffer exchanged into 10 mM sodium Tris , pH 7 . 0 , 50 mM NaCl , and 100 µM ZnCl2 and concentrated to 5 . 1 mg/ml . Equal volumes of enzyme and precipitant solution ( 22% PEG3350 , 0 . 1 mM Bis-Tris , pH 5 . 0 , 0 . 2 mM ammonium sulfate ) were mixed and placed over a reservoir of 1 ml precipitant solution to crystalize by the hanging drop method . No inorganic phosphate ( Pi ) was added to the precipitant solution , but 0 . 8 mM contaminating Pi was found in the crystallization solution using a Malachite Green assay ( Lanzetta et al . , 1979 ) . Crystals were soaked in a cryoprotectant solution of 30% glycerol , 0 . 1 mM Bis-Tris , pH 5 . 0 , and 0 . 2 mM ammonium sulfate prior to being frozen in liquid nitrogen . Crystallographic data were collected at the Stanford Linear Accelerator at beamline 11-1 . The D101A/D153A mutant of AP crystallized in space group P6322 with one dimer per asymmetric unit . Data were integrated with MOSFLM ( Battye et al . , 2011 ) and scaled and merged with AIMLESS ( Evans and Murshudov , 2013 ) . Five percent of reflections were set aside for calculation of Rfree . Molecular replacement was performed with PHASER ( McCoy et al . , 2007 ) using WT AP ( PDB 3TG0 ) stripped of phosphate and metal ions as a search model . Rounds of alternating manual and automated refinement were performed with COOT and REFMAC5 , respectively ( Emsley et al . , 2010; Murshudov et al . , 2011 ) . Stereochemistry was assessed with MOLPROBITY , and images were generated with PYMOL ( Schrödinger , 2010 ) . The PDB deposition ID is 4YR1 . | Enzymes are biological catalysts that speed up the reactions that are essential for life . As such , enzymes convert ‘reactant’ molecules into other molecules . Reactant molecules bind to part of the enzyme called the active site . Some of the amino acids that make up the active site must directly interact with these molecules to catalyze the reaction . Mutating individual active site amino acids often greatly reduces or destroys the ability of the enzyme to increase reaction rates . These amino acids are known as ‘catalytic residues’ . However , catalytic residues do not work in isolation: instead , they interact with other residues in the enzyme to carry out their function . Therefore , the effects of these interactions need to be characterized in order to fully understand how enzymes work . Sunden et al . explored the interactions within a network of five residues found at the active site of an enzyme , called alkaline phosphatase , which was taken from the bacterial species E . coli . Nearly all of the possible combinations of these five residues were examined . The results of these experiments indicated that even though all five residues are structurally linked , only a subset of the residues affected one another functionally , even though all of them are structurally connected . In particular , three groups—or functional units—of residues were found in the enzyme structure . The residues within each functional unit directly or indirectly cooperate to increase different aspects of the enzyme's catalytic activity . Sunden et al . used this information to develop models that describe how the functional units work together , and suggest that the likelihood of the active site evolving so that its residues are not fully cooperative is high . It remains to be seen whether similar cooperative networks exist in the active sites of other enzymes and how residues further away affect those in and around the active site . Understanding how the residues in the active site work together and being able to model their interactions could help efforts to develop more efficient enzymes for use in biotechnology in the future . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology"
] | 2015 | Extensive site-directed mutagenesis reveals interconnected functional units in the alkaline phosphatase active site |
Genetic hardwiring during brain development provides computational architectures for innate neuronal processing . Thus , the paradigmatic chick retinotectal projection , due to its neighborhood preserving , topographic organization , establishes millions of parallel channels for incremental visual field analysis . Retinal axons receive targeting information from quantitative guidance cue gradients . Surprisingly , novel adaptation assays demonstrate that retinal growth cones robustly adapt towards ephrin-A/EphA forward and reverse signals , which provide the major mapping cues . Computational modeling suggests that topographic accuracy and adaptability , though seemingly incompatible , could be reconciled by a novel mechanism of coupled adaptation of signaling channels . Experimentally , we find such ‘co-adaptation’ in retinal growth cones specifically for ephrin-A/EphA signaling . Co-adaptation involves trafficking of unliganded sensors between the surface membrane and recycling endosomes , and is presumably triggered by changes in the lipid composition of membrane microdomains . We propose that co-adaptative desensitization eventually relies on guidance sensor translocation into cis-signaling endosomes to outbalance repulsive trans-signaling .
The dazzling diversity of functions of the nervous system are brought about by neural networks of definite connectivity , most of which arise during ontogenesis through targeted outgrowth of axons . This is accomplished by growth cones ( GCs ) at the tips of growing axons ( Lowery and Van Vactor , 2009; Vitriol and Zheng , 2012 ) , which sense genetically encoded chemotactic guidance cues ( Dickson , 2002; Kolodkin and Tessier-Lavigne , 2011 ) . Such genetic hardwiring endows the individual at prefunctional stages with computational architectures , which have been fitted for survival through the evolution of the whole species . A well-studied example is the development of the retinotectal projection ( Figure 1 ) , which , in non-mammalian vertebrates , connects the retinal ganglion cells ( RGCs ) of the eye with the midbrain’s optic tectum in a topographic , i . e . , neighborhood-preserving manner ( Feldheim and O'Leary , 2010; Lemke and Reber , 2005; Weth et al . , 2014 ) . Research on this two-dimensional mapping has mainly focused on the anterior-posterior axis , whereby the temporal retina is projected onto the anterior tectum and the nasal retina onto the posterior tectum . The major guidance cues along this axis are EphA receptor tyrosine kinases and their glycosylphosphatidylinositol- ( GPI ) -anchored ephrin-A ligands ( Suetterlin et al . , 2012; Triplett and Feldheim , 2012 ) . EphAs and ephrin-As are counter-graded on both the retina and the tectum ( temporal retina — EphA high , ephrin-A low; anterior tectum — EphA high , ephrin-A low [McLaughlin and O'Leary , 2005] ) . The ephrin-A/EphA system can signal in forward ( EphA acting as receptor ) and in reverse ( ephrin-A as receptor ) directions ( Egea and Klein , 2007; Kania and Klein , 2016; Lisabeth et al . , 2013 ) . Both signaling channels act repulsively on RGC GCs . Receptor/ligand interactions typically occur in trans , between different cells , but , when present on the same cell , as on RGCs , they can additionally happen in cis ( Hornberger et al . , 1999 ) . We have recently proposed a simple but powerful comprehensive computational model ( Gebhardt et al . , 2012 ) that includes all possible ephrin-A/EphA interactions in this system ( fiber–target , fiber–fiber , cis , each forward and reverse ) . The model assumes that axon targeting aims to balance all summed reverse signals against all summed forward signals sensed by the GC . There have been diverse further attempts to explain retinotopic mapping ( Hjorth et al . , 2015; Simpson et al . , 2009 ) . Most of them agree , however , that topographic guidance is based on quantitative signaling , whereby the concentrations of sensors and cues bear the topographic information . This assumption has also been corroborated by experimental evidence ( Baier and Bonhoeffer , 1992; Hansen et al . , 2004; Rosentreter et al . , 1998; von Philipsborn et al . , 2006b ) . 10 . 7554/eLife . 25533 . 003Figure 1 . The retinotectal projection . The topographic projection in the chicken visual system connects RGCs from the retina to the midbrain's optic tectum . The temporal/nasal ( t/n ) axis of the retina is mapped onto the anterior/posterior ( a/p ) axis of the tectum , whereas the retinal dorsal/ventral axis projects onto the lateral/medial axis of the tectum . Retinal GCs are guided to their tectal targets by repulsive signals from counter-gradient distributions of tectal ephrin-As ( red; p>a ) and EphAs ( blue; a>p ) . Detection of these cues is mediated via retinal EphA and ephrin-A sensors , which are expressed in t>n and n>t gradients , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 003 On the other hand , it has been shown repeatedly that GCs can adapt to chemotactic guidance cues . Xenopus spinal GCs , in turning assays , can adapt to attraction by soluble Netrin-1 or brain-derived neurotrophic factor ( BDNF ) ( Ming et al . , 2002 ) . Ligand-specific desensitization was correlated with decreased Ca2+ signaling , whereas re-sensitization required MAP-kinase activation and local protein synthesis . Xenopus retinal axons in collapse assays were shown to adapt towards the collapse-inducing activities of soluble Sema3A and Netrin-1 ( Piper et al . , 2005 ) . Here , ligand-specific desensitization and re-sensitization , depended on endocytosis and protein translation , respectively . Notably , however , none of these guidance cues has been involved in topographic mapping . Indeed , it is difficult to imagine how adaptive signal modulation might be compatible with the need for quantitative signaling . First evidence , however , indicates that retinal GCs might in fact be able to adapt to topographic cues ( Rosentreter et al . , 1998; von Philipsborn et al . , 2006b ) . We therefore decided to investigate this conundrum in more detail . Introducing a novel adaptation assay ( ‘gap assay’ ) , we demonstrate that chick retinal GCs adapt towards both ephrin-A forward and EphA reverse topographic signals . Led by computational modeling , we postulate a distinctive novel type of signal modulation ( ‘co-adaptation’ ) to reconcile adaptation and topographic accuracy . We prove experimentally that such ephrin-A/EphA specific co-adaptation is indeed realized in retinal GCs . Using SNAP-tagged ephrin-A5 to label sensor surface populations , pharmacological inhibition , and co-localization studies with Rab11 , we show that co-adaptation on the cellular level involves trafficking of the guidance sensors between the GC membrane and the recycling endosome . By inducing adaptation by altering the lipid composition of the GC membrane , we provide evidence that repartitioning of guidance sensors between membrane microdomains might drive co-adaptation downstream of ephrin-A/EphA signaling . Eventually we propose , on the basis of our computational model , a hypothetical mechanism of co-adaptation , suggesting that an increase in the efficiency of cis-signaling from endosomes eventually desensitizes GCs against trans-signaling , without altering their topographic identity .
In in-vitro collapse assays ( Cox et al . , 1990; Kapfhammer et al . , 2007 ) typically ~80% of temporal GCs show a collapsed morphology after 20 min incubation with 0 . 25 μg/ml ephrin-A5-Fc , whereas control ( Fc fragment ) -treated GCs mostly remain intact ( Figure 2A , B ) . Surprisingly , however , we observe that temporal GCs recover their morphology , despite the presence of the repulsive cue , after prolonged incubation . After 120 min of incubation with ephrin-A5-Fc , the fraction of collapsed GCs drops from 77 . 1% ( 20min ) to 32 . 9% and , thus , to control levels ( 20 min Fc: 23 . 3%; 120 min: 27 . 2%; Figure 2A ) . This recovery is not caused by a loss of activity of the repulsive cue , as a medium containing 0 . 25 μg/ml ephrin-A5-Fc that has been on a first culture for 120 min triggers full response when reused on a second culture for 20 min ( 81 . 5% collapsed temporal GCs; Figure 2A ) . Therefore , recovery in the presence of ephrin-A indicates desensitization of RGC GCs towards forward signals . 10 . 7554/eLife . 25533 . 004Figure 2 . Growth cone desensitization towards soluble ephrin-A5 and EphA3 . ( A ) Temporal GCs initially collapse upon application of 250 ng/ml ephrin-A5-Fc ( eA5 20 min: 77 . 1% ) , but recover their morphology within 120 min in the presence of the cue ( eA5 120 min: 32 . 9% collapsed GCs ) . Ephrin-A5-Fc is still functional after incubation for 120 min , when re-used on a fresh culture ( eA5 recycled 20 min: 81 . 5% collapsed ) . Fc controls demonstrate the specificity of the effect of ephrin-A5 ( Fc 20 min: 23 . 3%; Fc recycled 20 min: 20 . 7%; Fc 120 min: 27 . 2% ) . ( B ) Representative intact and collapsed GC morphology ( Phalloidin Alexa488 ) . Scale bar: 10 µm . ( C ) Soluble EphA3-Fc triggers a collapse when administered at 15 µg/ml ( EA3 20 min: 64 . 4% collapsed ) , whereas an equimolar concentration of Fc ( 4 . 2 µg/ml ) does not ( Fc 20 min: 24 . 3% ) . GCs desensitize towards EphA3-Fc within 120 min ( EA3 120 min: 22 . 6% ) . Combined data from nasal and temporal GCs . N: number of independent experiments; number of analyzed GCs in brackets . Error bars represent standard deviations . T-test with n . s . : α ≥0 . 05 , ***: α <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 00410 . 7554/eLife . 25533 . 005Figure 2—source data 1 . Original data underlying bar charts of Figure 2A , C . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 00510 . 7554/eLife . 25533 . 006Figure 2—figure supplement 1 . EphA3 collapse assays and ephrin-A5/EphA3 dissociation constants . ( A ) Moderate concentrations of EphA3-Fc ( EA3 ) do not trigger a collapse of nasal or temporal GCs , neither in dimeric nor in an antibody clustered form ( Fc , 1 µg/ml , 20 min: 16 . 1% collapsed; EA3 , 2 µg/ml , 5 min: 15 . 3%; 10 min: 17 . 3%; 30 min: 22 . 4%; EA3 , 5 µg/ml antibody-clustered ( C ) , 20 min: 21 . 7% collapsed ) . Antibody-induced clustering was achieved by preincubation of EphA3 with anti human Alexa Fluor488 goat IgG ( Thermo Fisher Scientific , Waltham , MA , USA ) in a molar ratio of 2:1 ( EphA3:AB ) for 1 hr at room temperature . N: number of independent experiments; number of analyzed GCs in brackets . Error bars represent standard deviations . ( B ) Concentration-dependent binding of ephrin-A5 to EphA3 ( see Materials and methods for details ) . Dissociation constants ( KD ) of ephrin-A5 and EphA3 decrease abruptly with increasing EphA3 concentrations , indicating a sharp increase in binding affinity , which is potentially due to the concentration-dependent formation of EphA3 clusters in solution . KDs ephrin-A5 binding to: EphA3 3 µg/ml: 36 . 68 nM; EphA3 15 µg/ml: 0 . 76 nM; EphA3 33 µg/ml: 1 . 23 nM; EphA3 100 µg/ml: 0 . 55 nM . ( C ) Potential EphA3 clusters are stable for at least one day in solution at a concentration of 11 µg/ml as indicated by a constant KD ( fresh EphA3 , 11 µg/ml: 3 . 37 nM; 1 day old EphA3 , 11 µg/ml: 3 . 40 nM ) . ( B , C ) N: number of independent experiments . Error bars represent standard deviations . T-test with n . s . : α ≥0 . 05; *: α <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 00610 . 7554/eLife . 25533 . 007Figure 2—figure supplement 1—source data 1 . Original data underlying bar charts of Figure 2—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 007 To investigate reverse signaling desensitization , we first established EphA3 collapse assays . At high concentrations ( 15 µg/ml ) , a specific and significant collapse is triggered ( 64 . 4% of GCs; equimolar Fc: 24 . 3% collapsed; Figure 2C ) , but this does not occur at lower concentrations ( Figure 2—figure supplement 1A ) . We suggest that the exclusive response to high EphA concentrations is due to an exclusive activity of EphA3 oligomers ( n > 2 ) , which might form in concentrated solutions only . Supporting evidence comes from measuring ephrin-A5/EphA3 affinity by biolayer interferometry . Remarkably , binding constants depend on the EphA3 concentration and are dramatically increased at high concentrations ( ≥15 µg/ml; Figure 2—figure supplement 1B ) . The assumed oligomers remain stable in solution for at least 1 day ( Figure 2—figure supplement 1C ) . Notably , but similar to the response to ephrin-A5 , GCs completely recover after prolonged incubation with EphA3-Fc ( 22 . 6% collapsed after 120 min; Fc controls after 120 min: 21 . 6%; Figure 2C ) , demonstrating GC desensitization also towards EphA3 reverse signals . Full adaptability comprises desensitization and re-sensitization . To address the latter and to better approximate the in-vivo situation regarding surface-bound cues , we developed a novel in-vitro adaptation assay , the ‘gap assay’ . To this end , we manufactured patterned guidance substrates by protein contact printing ( von Philipsborn et al . , 2006a ) , which consisted of two rectangular fields homogeneously covered with the cue and separated by a permissive gap ( laminin ) of defined width . To ensure the activity of the printed cue , positive controls are performed in which axons grow from homogeneous laminin towards an identical field of the cue . For ephrin-A5-Fc , naїve temporal axons show a robust stop reaction at the boundary to the ephrin field in such controls ( 98% stopping; Figure 3A ) . Negative controls using a corresponding field of human Fc fragment ( 28 . 8% stopping ) demonstrate the specificity of the response ( Figure 3C ) . By contrast , axons starting from a field of homogeneous ephrin-A5-Fc no longer stop and continue growing onto the second ephrin field if the gap is small ( Figure 3A' ) , revealing their desensitization in agreement with the collapse assays . Notably , however , the proportion of GCs that stop in front of the second ephrin field increases with increasing gap sizes . Only 15 . 3% of temporal axons stop in assays with 50 µm wide gaps , but 43 . 0% stop after 75 µm , 59 . 2% after 100 µm and 81 . 0% after 200 µm wide gaps ( Figure 3A'' and C ) , indicating that GCs regain sensitivity towards forward signaling while growing on the permissive gap . 10 . 7554/eLife . 25533 . 008Figure 3 . Growth cone adaptation towards substrate-bound ephrin-A5 and EphA3 . Subfigures in ( A ) and ( B ) each display a cartoon ( left ) illustrating the experimental setup consisting of explant ( thick black strip ) , axons and printed guidance protein ( colored field ( s ) ) , the inverted signal of fluorescent phalloidin-stained axonal actin ( middle; explant not shown ) and the underlying , antibody-labeled substrate ( right ) are shown in a detailed view of a representative microscopic image ( scale: 100 µm ) . ( A ) Naïve temporal axons stop in front of a homogeneous field of ephrin-A5-Fc ( eA5 15 µg/ml; red ) , but do not react to an identical boundary when initially grown on ephrin-A5 after a 65 µm wide gap ( A' ) . Stopping behavior returns in axons having crossed a 215 µm wide , ephrin-free gap ( A'' ) . ( B–B'' ) Nasal RGC axons show very similar behavior on EphA3-Fc ( EA3 15 µg/ml , blue ) gap substrates as described for ephrin-A5 gap assays . ( C ) Quantification of gap assays . Stop reactions were quantified as the percentage of fibers not entering the protein field after the gap . Fc naïve: 28 . 8% stopping . eA5 naïve: 98 . 0% , eA5 50 µm: 15 . 3% , eA5 75 µm: 43 . 0% , eA5 100 µm: 59 . 2% , eA5 200 µm: 81 . 0% stopping . EA3 naïve: 92 . 6% , EA3 65 µm: 25 . 0% , EA3 90 µm: 39 . 9% , EA3 115 µm: 60 . 1% , EA3 215 µm: 75 . 8% stopping . Ephrin-A5 gap assay patterns differ slightly from EphA3 patterns in gap width as a result of a modified stamp geometry . N: number of independent experiments . Error bars represent standard deviations . T-test with *: α <0 . 05 , **: α <0 . 01 , ***: α <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 00810 . 7554/eLife . 25533 . 009Figure 3—source data 1 . Original data underlying bar chart of Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 009 In reverse signaling gap assays , naїve nasal RGC GCs show a clear stop reaction at the boundary to a field of contact printed EphA3 ( 92 . 6% stopping; Figure 3B ) . GCs no longer stop in front of an identical field after a small gap when starting on EphA3 ( 65 µm: 25 . 0% stopping; Figure 3B' ) , indicating desensitization towards reverse signals in accordance with the collapse assays . As in the forward signaling gap assays , they gradually regain sensitivity with increasing gap sizes ( 90 µm: 39 . 9% stopping; 115 µm: 60 . 1% stopping; 215 µm: 75 . 8% stopping; Figure 3B'' and C ) , demonstrating re-sensitization towards reverse signaling . Re-sensitization of forward and reverse signaling in gap assays displays very similar dependencies on the overgrown distance hinting at a potential common mechanism . Together , our results show that topographically mapping retinal GCs robustly adapt towards forward and reverse ephrin-A/EphA signaling . This is puzzling , because topographic guidance is believed to rely on precise quantitative sensing . To conceive how adaptation and topography could be reconciled , we made use of our previous computational model ( Gebhardt et al . , 2012 ) . Briefly , the model ( Figure 4A ) is exclusively based on ephrin/Eph signaling and assumes that GCs do not distinguish the diverse sources of the impinging signals ( target cells , other fibers , GC’s own surface ) . They just sense total reverse and total forward signaling . GC targeting is then considered as a potential minimization process , comparable to a sphere rolling into a well . To describe this process , a ‘guidance potential’ , D , is defined , which is conceived to be proportional to the ratio of total reverse to total forward signaling and to be minimized when both are equal , indicating target arrival . The effective strengths of the signals received by an individual GC depend on its endowment with sensors ( ephrins for reverse and Ephs for forward signaling ) . Thus , the GC’s own ephrin/Eph ratio corresponds to its topographic imprint , determining where on the target area reverse/forward signaling balance will be reached . Therefore , to retain topography in the face of adaptation , which is expected to modulate signal strength , this ratio must remain basically unaltered . In the model , this is achieved by introducing the same multiplicative adaptation factor , a , simultaneously to both reverse and forward signaling . For more mathematical detail see 'Material and methods' . 10 . 7554/eLife . 25533 . 010Figure 4 . Modeling growth cone adaptation and topographic mapping . ( A ) Fiber terminals are modeled as circular discs bearing Gaussian-shaped distributions of EphAs ( RF , blue ) and ephrin-As ( LF , red ) , according to their retinal origin , moving on a rectangular target field of unit squares ( xT , yT ) , carrying target-derived guidance cues ( LT and RT ) . Fiber–target and fiber–fiber ( cis and trans ) interactions and the resulting forward and reverse signals integrate into a guidance potential , D ( equation1 ) , which determines the probability , p , of a change in position ( only fiber–target interactions are illustrated in the inset ) . Additionally , D is used to calculate an adaptation coefficient , a ( equation2 ) , which simultaneously modulates RF and LF ( collectively called S; equation4 ) through 'co-adaptation' . A resetting force , f ( equation3 ) , counteracts a . For mathematical detail see 'Material and methods' . ( B ) Mapping plots of simulations with 200 fiber terminals using different implementations of adaptation . The anterior-posterior ( a-p ) position of a terminal on the target is plotted as a function of naso-temporal ( n-t ) origin ( perfect topography is indicated by all terminals targeting the main diagonal ) . Terminals that are enabled to regulate sensors independently from each other by canonical adaptation are widely scattered across the target field ( white circles ) , whereas co-adapted terminals ( regulating both sensors concomitantly ) find their topographically correct target positions ( gray circles ) . In the basic model , the potential minimum is at a position where fiber and target ephrin/Eph concentrations match . Therefore , canonical adaptation was implemented in terms of a tendency for GCs to independently match their forward and reverse sensor levels to the receptor and ligand levels , respectively , on the current target position . ( C ) After implementation of co-adaptation , naïve terminals stop in front of a field of high ephrin-A or EphA ( LT = 4 , red; RT = 4 , blue ) , respectively , but ignore the same boundary in simulated gap assays with small gap size . In simulations with wider gaps , terminals stop again in front of the second field . Gap size = 20 , 40 , or 100 units respectively; n = 12; i = 2000; target field: 200 × 8 . ( D ) Co-adaptation predicts that ephrin-A adapted terminals will ignore a field of high EphA after a small gap and vice versa . Gap size = 20; other parameters as in ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01010 . 7554/eLife . 25533 . 011Figure 4—figure supplement 1 . Inclusion of co-adaptation does not alter the explanatory power of the computational model . For purposes of comparison , this figure , produced using the updated model that includes co-adaptation presented in this paper is structured in parallel to figure 7 in Gebhardt et al . ( 2012 ) , which did not take adaptation into account . The results are highly similar , demonstrating that the inclusion of co-adaptation does not alter the performance of the original model . ( A ) 200 simulations of single fibers chosen randomly from the retinal field . Consistent with the experimental evidence in zebrafish , single fibers map topographically , as indicated by the accumulation of terminals on the main diagonal of the mapping plot . ( B ) EphA3 knock-in , Rki , in every second fiber ( blue; ki/+: Rki = 2; ki/ki: Rki = 4; axonal ligand on knock-in fibers was reduced reciprocally to the axonal receptor ) . Consistent with the experiment , wild-type fibers ( gray ) are posteriorly displaced , with a more severe effect in homozygotes . ( C ) Map expansion . 100 fibers chosen from the nasal ( n ) or temporal ( t ) half of the retina reach their correct target ( yellow ) after 3000 iterations . After 30 , 000 iterations , fibers stably cover the whole tectum ( gray ) . ( D ) No expansion occurs ( gray ) when a n half retina grows into a field containing the remnants of a previous full innervation ( white ) . ( E ) Map compression . 200 fibers from a whole retina form a compressed topographic projection when growing into an anterior tectum half . ( F ) Mismatch . 100 n fibers form a half-projection , when growing into an anterior , non-matching , tectal half . ( G ) An additional n projection ( gray ) displays polarity reversal when growing into a tectal field still containing properly mapped n fibers ( white ) ( n = 100; fiber–fiber interactions were put into effect earlier than usual , i . e . , j = 2000 in C ( i ) ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 011 After implementing this form of concomitant adaptation of two signaling channels , which we term co-adaptation , the model is still able to form an accurate topographic map ( Figure 4B , gray circles ) . By contrast , uncoupled adaptation by independent negative feedback to each channel completely abolishes map formation ( Figure 4B , white circles ) . The updated model reproduces the experimental results of the gap assays ( Figure 4C ) . This is achieved without changing the model’s global explanatory power , described earlier ( Figure 4—figure supplement 1; Gebhardt et al . , 2012; Weth et al . , 2014 ) . The co-adaptation concept predicts that each channel should be modulated upon its own activation , but counter-intuitively also upon activation of the parallel channel even in the absence of its own ligand . Thus , ephrin-adapted axons should ignore Eph in a ‘double-cue’ gap assay with small gap size . Similarly , Eph-adapted axons should ignore ephrin ( simulations in Figure 4D ) . To test this prediction , we used double-cue gap assays with substrates comprising different cues on either side of the gap . In ephrin-A5/EphA3 double-cue gap assays , ephrin-A5-adapted temporal GCs , after a small gap ( <100 µm ) , ignore the EphA3 field ( 26 . 9% stopping ) , in front of which they naïvely stop ( 93 . 4% stopping; Figure 5A , A' and D ) . As these GCs have not experienced substrate-derived reverse signals before , their desensitization towards these signals can only be explained through co-regulation of reverse and forward signaling upon activation of the forward channel . 10 . 7554/eLife . 25533 . 012Figure 5 . Co-adaptation of retinal growth cones in double-cue gap assays . ( A–C ) Naïve axons stop in front of a homogeneous field of EphA3-Fc ( A; EA3 , blue; nasal axons ) , ephrin-A5-Fc ( B; eA5 , red; temporal axons ) , or Sema3A-Fc ( C; S3A , green; temporal axons ) . Scale: 100 μm . ( A'–C' ) Nasal , eA5-adapted GCs ignore a field of EA3 after a small gap ( A' ) and , vice versa , temporal , EA3-adapted GCs ignore eA5 ( B' ) . By contrast , temporal , eA5-adapted GCs are still strongly repelled by a field of S3A after a small gap , indicating specific co-adaptation of ephrin-A and EphA signals ( C' ) . ( D ) Quantification of combined data of assays with 65 μm and 90 μm wide gaps grouped in ' < 100 μm' bars; 115 μm and 215 μm grouped in ' > 100 μm' bars . eA5-EA3: naïve: 93 . 4% , <100μm: 26 . 9% , >100μm: 76 . 3% stopping . EA3-eA5: naïve: 91 . 6% , <100μm: 40 . 3% , >100μm: 77 . 1% stopping . eA5-S3A: naïve: 93 . 7% , <100μm: 81 . 9% stopping . N: number of independent experiments; error bars indicate standard deviations . T-test with n . s . : α ≥0 . 05 , ***: α <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01210 . 7554/eLife . 25533 . 013Figure 5—source data 1 . Original data underlying bar chart of Figure 5D . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01310 . 7554/eLife . 25533 . 014Figure 5—figure supplement 1 . Growth cone adaptation towards Sema3A . ( A ) Naïve axons stop in front of a homogeneous field of Sema3A-Fc ( S3A , green ) , but predominantly ignore a similar field in gap assays with ( B ) 65 µm , ( C ) 115 µm and ( D ) 215 µm wide gaps . Axons strongly de-fasciculate and begin to branch in the Sema3A-free areas of the gap substrates , but not when growing on the protein fields ( detail shown in C' , white dotted lines indicate the position of the gap ) . The reason for this notable effect is currently unknown . Remarkably , the desensitization towards Sema3A in these assays is so strong that re-sensitization is not reached even after the largest gap-size available with our contact-printing stamps ( 215 µm ) . ( E ) S3A-eA5 double-cue gap assay . Sema3A-adapted GCs do not co-adapt towards ephrin-A5 and stop in front of an ephrin-A5 field after a 65 µm wide gap ( cf . Figure 5B' ) . ( F ) SNAP-ephrin-A5 surface localization on Sema3A-adapted GCs . ( G , G' ) Representative images of SNAP-ephrin-A5 surface-stained GCs growing on laminin or Sema3A , respectively ( outline drawn in the actin channel indicated as a white , dotted line ) . ( H ) Quantification of SNAP-ephrin-A5 surface staining expressed as fold change in relative intensity compared to controls on laminin ( on La: 1; on S3A: 1 . 07 ) . N: number of independent experiments , number of analyzed GCs in brackets; error bars show standard deviations . T-test with n . s . : α ≥ 0 . 05 . Scale ( A–E ) : 100 µm , ( G , G' ) : 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01410 . 7554/eLife . 25533 . 015Figure 5—figure supplement 1—source data 1 . Original data underlying bar chart of Figure 5—figure supplement 1H . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 015 Conversely , in EphA3/ephrin-A5 double-cue gap assays ( <100 µm ) , nasal GCs co-adapt their forward signal by growing on an EphA3 reverse signaling field ( 40 . 3% stopping; naïve: 91 . 6%; Figure 5B , B' and D ) . Consistent with the findings in single-cue gap assays , the degree of co-adaptation abates with increasing gap width ( eA5–EA3 >100µm: 76 . 3% stopping; EA3–eA5 >100µm: 77 . 1% stopping; Figure 5D ) . Co-adaptation is ephrin/Eph-specific , as for example , ephrin-A5-adapted temporal GCs remain sensitive to other repulsive guidance molecules such as Sema3A ( naïve: 93 . 7% stopping , eA5–S3A <100µm: 81 . 9% stopping; Figure 5C , C' and D ) . Although retinal GCs can efficiently desensitize towards Sema3A in gap-assays ( Figure 5—figure supplement 1A–D ) , Sema3A-adapted GCs remain sensitive to ephrin-A5 ( Figure 5—figure supplement 1E ) , further attesting to the orthogonality of these signaling pathways and to the specificity of the co-adaptation mechanism for the ephrin-A/EphA-based topographic mapping system . Together , these findings , being in full agreement with the predictions of our computational model , prove the existence of a novel mechanism of signal modulation ( co-adaptation ) , which allows for topographic mapping in the presence of CG adaptation . Searching for the molecular underpinnings of co-adaptation , we transfected RGCs with a SNAP-tagged ephrin-A5 expression construct ( pSNAP–ephrin-A5–IRES–EGFP; Figure 6—figure supplement 1 ) . The self-labeling SNAP-tag ( Gautier et al . , 2008 ) in combination with a membrane-impermeant fluorescent substrate enabled us to specifically label the surface-bound subpopulation of SNAP–ephrin-A5 in naïve and adapted states . SNAP–ephrin-A5 transfected GCs remain sensitive to substrate-bound ephrin-As ( see below ) . This is in contrast to retrovirally transfected native ephrin-A5 , which rendered chick RGC GCs insensitive to external ephrin-As ( Hornberger et al . , 1999 ) . Seemingly , the amount of cis-signaling added through the expression of SNAP–ephrin-A5 is sufficiently low to prevent significant disturbance of the guidance system , supporting the validity of the localization probe . On GCs growing on an EphA3-Fc surface , SNAP–ephrin-A5 , being the cognate sensor of EphA3 , is dramatically reduced compared to control GCs growing on Fc ( Figure 6A , A' and B ) . Remarkably , SNAP–ephrin-A5 surface levels are also reduced on GCs growing on ephrin-A5-Fc , despite not being detected by this sensor ( Figure 6A'' and B ) . Thus , the reverse signaling sensor is cleared from the surface upon both pure reverse and pure forward signaling , strongly corroborating the co-adaptation concept . By contrast , SNAP–ephrin-A5 surface levels remain unaltered upon adaptation to Sema3A ( Figure 5—figure supplement 1G–H ) , again confirming the specificity of the co-adaptation mechanism to the ephrin-A/EphA system . 10 . 7554/eLife . 25533 . 016Figure 6 . Endocytosis of guidance sensors upon growth cone desensitization . ( A ) SNAP–ephrin-A5 surface signal is strongly reduced on GCs growing on either homogeneous EphA3-Fc ( A' ) or ephrin-A5-Fc ( A'' ) compared to controls ( on Fc ) , indicating clearance of sensors from the surface upon co-adaptation . Outlines of GCs drawn in the GFP ( transfection marker ) channel are indicated as white , dotted lines . Scale: 10 µm . ( A''' ) displays a cartoon of the experiment in ( A–A'' ) with axonal SNAP surface labeled ephrin-A5 in magenta and with the corresponding guidance cues ( Fc , gray; EA3 , blue and eA5 , red ) . ( B ) Quantification of SNAP–ephrin-A5 surface signal intensities . Relative signal intensity within the GC outline normalized to the control situation is given as fold change ( on Fc: 1; on EA3: 0 . 29; on eA5: 0 . 32 ) . ( C ) Effects of Pitstop2 on adaptation towards soluble ephrin-A5 . Pitstop2 ( P; carrier DMSO ) prevents temporal GCs from desensitizing to 0 . 25 µg/ml ephrin-A5-Fc ( eA5+DMSO 120 min: 15 . 8% collapsed; eA5+P 120 min: 74 . 3% collapsed ) . The initial response of GCs towards eA5 is unaltered in the presence of P ( eA5+DMSO 20 min: 85 . 4% collapsed; eA5+P: 83 . 1% collapsed ) . Neither P itself , nor its carrier DMSO induces a collapse ( Fc+DMSO 20 min: 13 . 6% collapsed; Fc+P 20 min: 7 . 5% collapsed ) . ( D ) Effects of Pitstop2 on adaptation towards soluble EphA3 . Soluble EphA3-Fc ( EA3; 15 µg/ml ) triggers a collapse on nasal or temporal GCs ( EA3 20 min: 64 . 4% collapsed; Fc [4 . 2 µg/ml] 20 min: 24 . 3% collapsed ) . GCs desensitize towards EA3 within 120 min ( EA3 120 min: 22 . 6% collapsed ) , but not in the presence of 30 μM Pitstop2 ( EA3+P 120 min: 49 . 8% collapsed ) . P does not impede GCs from recovering their morphology in general ( EA3 20 min , then Fc+P 120 min: 19 . 3% collapsed ) . Combined data from nasal and temporal GCs . ( B , C , D ) N: number of independent experiments , number of analyzed GCs in brackets; error bars represent standard deviations . T-test with n . s . : α ≥0 . 05 , *: α <0 . 05 , ***: α <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01610 . 7554/eLife . 25533 . 017Figure 6—source data 1 . Original data underlying the bar charts of Figure 6B—D . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01710 . 7554/eLife . 25533 . 018Figure 6—figure supplement 1 . SNAP–ephrin-A5 expression constructs pSNAP–ephrin-A5–IRES-FP contains the coding sequence of a fusion protein consisting of the signal sequence of chick ephrin-A5 ( 60 bp ) , the SNAP-tag and full-length chick ephrin-A5 downstream of the CAG enhancer/promoter and upstream of an IRES-FP sequence . The FP is EGFP in pSNAP–ephrin-A5–IRES-EGFP and dTomato in pSNAP–ephrin-A5–IRES–dTom . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 018 To address the mechanism of surface clearance during co-adaptation , we used the clathrin-mediated endocytosis ( CME ) inhibitor Pitstop2 ( von Kleist et al . , 2011 ) . CME is a ubiquitous mechanism for the endocytosis ( McMahon and Boucrot , 2011 ) of numerous receptors including EphAs ( Boissier et al . , 2013; Yoo et al . , 2010 ) . In collapse assays , Pitstop2 does not affect the initial response to the guidance cue ( ephrin-A5+DMSO: 85 . 4%; ephrin-A5+Pitstop2: 83 . 1% collapsed; Figure 6C ) , nor does it induce a collapse by itself ( Fc+DMSO: 13 . 6%; Fc+Pitstop2: 7 . 5% collapsed ) . Strikingly , however , adaptive recovery after prolonged incubation is completely abolished when Pitstop2 is applied together with ephrin-A5 or EphA3 for 120 min ( ephrin-A5+DMSO: 15 . 8%; ephrin-A5+Pitstop2: 74 . 3%; EphA3: 22 . 6%; EphA3+Pitstop2: 49 . 8% collapsed; Figure 6C and D ) , indicating that sensor internalization via CME is required for the desensitization of both channels . We next asked how desensitized GCs replenish their surface sensor pools upon re-sensitization . Previously described forms of GC adaptation involve local translation for re-sensitization ( Ming et al . , 2002; Piper et al . , 2005 ) . Thus , we performed ephrin-A5 gap assays in the presence of anisomycin ( AIM ) to inhibit protein synthesis ( Grollman , 1967 ) . Consistent with previous reports ( Campbell and Holt , 2001 ) , AIM treatment did not affect axon elongation until , after about 4 hr , cells start to die ( data not shown ) . We used time-lapse imaging and administered AIM just before axons left the first ephrin-A5 field . Interestingly , AIM-treated GCs display normal re-sensitization , as indicated by the unaltered recognition of the second ephrin-A5-Fc field after a 200 µm wide gap ( Figure 7A ) . 10 . 7554/eLife . 25533 . 019Figure 7 . Dynamics of SNAP–ephrin-A5 during growth cone re-sensitization . ( A ) Exemplary trajectories of five GCs ( combined from N = 3 independent experiments ) on ephrin-A5 gap patterns ( red ) in the presence of 40 µM AIM . Positions of GCs ( black dots ) were marked at the indicated time points ( in minutes ) after the addition of AIM . GCs stop and pause ( I , III ) or stop and retract ( II , IV , V ) upon reaching the second ephrin field , indicating full re-sensitization despite inhibition of local translation . ( B–B'' ) SNAP–ephrin-A5 dynamics upon GC re-sensitization . Representative images of GCs , stained for formerly intracellular SNAP–ephrin-A5 that has been relocated to the membrane . Outlines of GCs drawn in the actin channel are indicated as white , dotted lines . Surface localization of ephrin-A5 ( anti-fluorescein signal ) is elevated in re-sensitizing GCs ( B'' , off ephrin-A5 ) compared to desensitized GCs ( B' , on ephrin-A5 ) . Scale: 10 μm . ( C ) Experimental setup with retinal explant ( black ) on a field of ephrin-A5 ( red ) and cartoon of the experiment shown in ( B–B'' ) with intracellular SNAP–ephrin-A5 ( green ) , which is detected on the surface ( magenta star ) . ( D ) Quantification of membrane-relocated SNAP–ephrin-A5 signals . Relative anti-fluorescein signal intensity within the GC outline normalized to the control is given as fold change . Control — 1; on eA5 — 0 . 85; off eA5 — 1 . 13 . N: number of independent experiments , number of analyzed GCs in brackets; error bars represent standard deviations . T-test with n . s . : α ≥0 . 05 , *: α <0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 01910 . 7554/eLife . 25533 . 020Figure 7—source data 1 . Original data underlying the bar chart in Figure 7D . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 02010 . 7554/eLife . 25533 . 021Figure 7—figure supplement 1 . Staining procedure for recycled SNAP–ephrin-A5 . Extracellular SNAP–ephrin-A5 was blocked by SNAP surface block ( gray ) , before the cell-permeant SNAP cell fluorescein was applied to the living cells ( green ) . After washing , axons were allowed to grow for another 20–22 hr . Then , anti-fluorescein antibody was added to the living cells for 15 min , washed out and , subsequently , explants were fixed and stained for anti-fluorescein ( magenta stars ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 021 As an alternative to local translation , the GC membrane might be repopulated with sensors from internal stores during re-sensitization . To check for this possibility , we specifically labeled the intracellular population of SNAP–ephrin-A5 in desensitized GCs growing on ephrin-A5 . Subsequently , GCs were given time ( 20–22 hr ) to leave the ephrin-A5 field for re-sensitization ( Figure 7C ) and were then stained for any formerly intracellular label that had appeared on the GC surface in the meantime ( for a detailed staining protocol , see Figure 7—figure supplement 1 and 'Materials and methods' ) . Notably , GCs that have left the ephrin field show a significantly higher surface staining than GCs that are still growing on the field ( fold change in intensity compared to laminin controls: on ephrin-A5 — 0 . 85; off ephrin-A5 — 1 . 13; Figure 7B–B’’ and D ) . This indicates a transport of intracellular sensors to the GC's surface upon re-sensitization . Notably , these results were gained for SNAP–ephrin-A5 on ephrin-A5-Fc substrates . As ephrin-A5 is not a receptor for ephrin-A5-Fc , its regulation has to be assumed to be a result of co-regulation with EphAs , the active sensor in this setting , again supporting the co-adaptation idea . Endocytotic clearance and surface repopulation from internal stores upon co-adaptation suggests the involvement of recycling endosomal compartments ( Goldenring , 2015 ) . To assess whether ephrin-A5 does indeed traffic through recycling endosomes during co-adaptation , we co-transfected RGCs with pSNAP–ephrin-A5–IRES–dTom and pEGFP–Rab11 ( Figure 8A , B , B’ and C , C' ) . In fact , we find an increased colocalization of SNAP–ephrin-A5 and Rab11 in GCs growing on ephrin-A5 when compared to GCs on laminin ( Figure 8B'' , C'' and D ) , suggesting elevated partitioning of these sensors in desensitized GCs into Rab11-positive compartments , from which they can return to the surface upon re-sensitization . 10 . 7554/eLife . 25533 . 022Figure 8 . Colocalization of SNAP–ephrin-A5 and Rab11-positive endosomes . ( A ) eGFP–Rab11 localization in a transfected RGC axon ( outline drawn in the actin channel indicated as white , dotted line ) . Rab11-positive vesicular structures are predominantly observed in the axon shaft and the central domain of GCs and , to a lesser extent , are found to stretch into the more distal GC . ( B–B'' ) Representative images of an eGFP–Rab11 ( B ) and a SNAP–ephrin-A5 ( B' ) co-transfected GC on laminin substrate . Vesicular structures identified by automated image segmentation are indicated in green and red for each channel , respectively . ( B'' ) Colocalization of SNAP–ephrin-A5 and eGFP–Rab11 within the Rab11-positive vesicles is shown as yellow areas . ( C–C'' ) Example images of an eGFP–Rab11- ( C ) and a SNAP–ephrin-A5-transfected GC ( C' ) growing on ephrin-A5-Fc . ( C'' ) Colocalization of SNAP–ephrin-A5 and eGFP–Rab11 within the Rab11-positive vesicles ( yellow areas ) increases upon desensitization . ( D ) Manders' colocalization coefficient ( MCC ) as a measure of colocalization of SNAP–ephrin-A5 and eGFP–Rab11 within the Rab11-positive vesicles in naïve ( on laminin: M2 = 0 . 55 ) and adapted ( on ephrin-A5: M2 = 0 . 72 ) GCs , respectively . N: number of independent experiments , number of analyzed GCs in brackets; error bars indicate standard deviations . T-test with *: α <0 . 05 . Scale: 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 02210 . 7554/eLife . 25533 . 023Figure 8—source data 1 . Original data underlying the bar chart of Figure 8D . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 023 In our model , adaption is adjusted in response to the guidance potential experienced by the GC . A potential regulator of co-adaption should therefore be located downstream of both reverse and forward signaling , which together determine the guidance potential , and it should impact on the membrane activity of the signaling sensors . Baba et al . ( 2009 ) have shown that Fyn-kinase is activated downstream of ephrin-A reverse signaling , inducing an increase of sphingomyelin in the membrane and a corresponding exclusion of ephrin-A from the cell surface . Sphingomyelin , together with cholesterol , is a major component of the lipid rafts of the cell membrane ( Simons and Ikonen , 1997 ) , where ephrin-As have been shown to be localized ( Averaimo et al . , 2016; Gauthier and Robbins , 2003 ) . As Fyn-kinase also acts downstream of EphA receptors ( Ellis et al . , 1996; Wu et al . , 1997 ) , we first asked whether manipulating the sphingomyelin content of the membrane might impact co-adaptation . Pretreatment with sphingomyelinase ( SMase , Neufeld et al . , 1996 ) in collapse assays strongly reduces the primary response of GCs towards ephrin-A5 and EphA3 ( ephrin-A5 — 79 . 6%; ephrin-A5+SMase — 50%; EphA3 — 52 . 9%; and EphA3+SMase — 23 . 8% collapsed; Figure 9—figure supplement 1A and B ) . SMase treatment has been shown to disrupt rafts by replacing them with ceramide-enriched , densely packed solid ordered domains ( Chiantia et al . , 2006; Goñi and Alonso , 2009 ) . Thus , sphingomyelin overproduction or conversion to ceramide might have the same effect of excluding ephrin-As from rafts . To investigate the role of these membrane microdomains in co-adaptation more closely , and to avoid the intricacies of SMase effects , we used methyl-β-cyclodextrin ( MβCD , [Mahammad and Parmryd , 2015] ) to disrupt rafts by extracting cholesterol from the GC membrane in further adaptation assays . In forward signaling collapse assays with 0 . 25 µg/ml ephrin-A5-Fc , co-application of 2 mg/ml MβCD does not significantly alter the primary response after 20 min ( 80 . 3% without , 77 . 4% with MβCD , Figure 9A ) . After 1 hr of incubation , however , significantly more GCs have recovered in the presence of MβCD than without ( 74 . 2% collapse without , 45% with MβCD , Figure 9A ) . In reverse signaling collapse assays with 15 µg/ml EphA3 , a corresponding desensitizing effect is already seen after 20 min ( 78 . 2% collapsed without , 51 . 6% collapsed with MβCD , Figure 9B ) and remains significant after 1 hr ( 28 . 3% collapse without , 14 . 5% with MβCD , Figure 9B ) . In double-cue gap-assays with a gap size actually allowing for substantial re-sensitization ( 115 µm ) , a significantly higher percentage of EphA3- or ephrin-A5-adapted axons are still desensitized when MβCD is present compared to controls; these axons overgrow the boundary and enter the field of ephrin-A5 or EphA3 , respectively , after the gap ( EA3-eA5: 80 . 7% stopping without MβCD , 46 , 3% stopping with MβCD [Figure 9C , C’ and E]; eA5-EA3: 70 . 9% stopping without MβCD , 23 . 4% stopping with MβCD [Figure 9D , D’ and E] ) . Thus , extraction of cholesterol from the GC membrane , which is thought to result in the disintegration of lipid rafts , strongly enhances desensitization in both types of adaptation assays . 10 . 7554/eLife . 25533 . 024Figure 9 . Disintegration of lipid microdomains induces co-adaptive growth cone desensitization . ( A ) Ephrin-A5-Fc and ( B ) EphA3-Fc collapse assays . Methyl-beta-cyclodextrin ( MβCD , 2 mg/ml ) -treated GCs show enhanced desensitization ( reduced collapse rates ) when exposed to soluble ephrin-A5-Fc ( 0 . 25 μg/ml ) or EphA3-Fc ( 15 μg/ml ) compared to controls without MβCD . ( A ) GCs collapsed after 20 min incubation: Fc ( 0 . 25 μg/ml ) , 16 . 7%; Fc+MβCD , 11 . 4%; eA5 , 80 . 3%; and eA5+MβCD , 77 . 4% . GCs collapsed after after 60 min incubation: Fc , 12 . 1%; Fc+MβCD , 13 . 4%; eA5 , 74 . 2%; and eA5+MβCD , 45 . 0% . ( B ) GCs collapsed after 20 min incubation: Fc ( 4 . 2 μg/ml ) , 22 . 8%; Fc+MβCD , 24 . 4%; EA3 , 78 . 2%; and EA3+MβCD , 51 . 6% . GCs collapsed after 60 min incubation: Fc , 14 . 9%; Fc+MβCD , 10 . 4%; EA3 , 28 . 3%; and EA3+MβCD , 14 . 5% . N: number of independent experiments , number of analyzed GCs in brackets; error bars show standard deviations . T-test with n . s . : α ≥0 . 05 , *: α <0 . 05 , ***: α <0 . 001 . ( C ) EA3–eA5 and ( D ) eA5–EA3 double-cue gap assays . After 115 µm wide gaps , untreated GCs predominantly stop , whereas MβCD-treated GCs ( C' , D' ) ignore the protein field after the gap . ( E ) Quantification of double-cue gap assays: EA3–eA5 , 80 . 7%; EA3–eA5+MβCD , 46 . 3% stopping; eA5–EA3 , 70 . 9%; and eA5–EA3+MβCD , 23 . 6% stopping . N: number of independent experiments , error bars show standard deviations , T-test with *: α <0 . 05 , ***: α <0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 02410 . 7554/eLife . 25533 . 025Figure 9—source data 1 . Original data underlying the bar charts in Figure 9A , B , E . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 02510 . 7554/eLife . 25533 . 026Figure 9—figure supplement 1 . Disintegration of lipid microdomains by sphingomyelinase desensitizes growth cones towards soluble ephrin-A5 and EphA3 . ( A ) Ephrin-A5-Fc and ( B ) EphA3-Fc collapse assays . In accordance with MβCD experiments , Sphingomyelinase ( SMase , 400mU/ml ) -treated GCs show reduced collapse rates when exposed to soluble ephrin-A5-Fc ( 0 . 25 μg/ml ) or EphA3-Fc ( 15 μg/ml ) compared to controls without SMase . ( A ) Fc ( 0 . 25 μg/ml ) , 19 . 5%; Fc+SMase , 27 . 5%; eA5 , 79 . 6%; and eA5+SMase , 50% collapsed . ( B ) Fc ( 4 . 2 μg/ml ) , 22 . 5%; Fc+SMase , 27 . 7%; EA3 , 52 . 9%; and EA3+SMase , 23 . 8% collapsed after 20 min incubation . SMase was applied during a 30 min pre-incubation and thereafter together with Fc or ephrin-A5 . N: number of independent experiments , number of analyzed GCs in brackets; error bars show standard deviations . T-test with n . s . : α ≥ 0 . 05 , *: α < 0 . 05 , **: α < 0 . 01 , ***: α < 0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 02610 . 7554/eLife . 25533 . 027Figure 9—figure supplement 1—source data 1 . Original data underlying the bar charts of Figure 9—figure supplement 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 027 We have previously shown that our model , through the inclusion of fiber–fiber in addition to fiber–target interactions , displays a substantial degree of mapping plasticity even in the absence of any adaptation ( Weth et al . , 2014 ) . Thus , in accordance with experimental observations , a half-retinal projection expands to a full tectum , a full retinal projection compresses to a half-tectum , and a half-projection can be forced to map onto a mismatching tectal half ( Goodhill and Richards , 1999 ) . This is due to fiber–fiber interactions , by virtue of which the axons introduce their own mapping apparatus into the target field , eventually overwriting the target cues . Why then should co-adaptation be needed ? Generally speaking , signaling comes at the expense of energy . Adaptation is continuously acting to reduce the guidance potential and , thus , the energetic costs of organizing the connectome , which are presumably limiting to the development of the nervous system as the mature brain is the most highly ordered biological structure we know of . A more specific advantage of the adaptation of retinal GCs can be envisaged from corresponding simulations ( Figure 10 ) . Without adaptation , the majority of axons would be unable to overcome the steep boundary of reverse signaling at the entrance of the target field . By contrast , when we deliberately desensitize the fiber population through the co-adaptation mechanism before entering the target field , they easily overcome the boundary and automatically re-adjust their sensitivities on the target field . To this end , a pre-adapting signal would be required before the target . Fiber–fiber interactions enabled by the intimate contact of the fibers in the optic tract are unlikely to be the preadapting factor , as it has been show that fiber–fiber interactions are required neither for target entry nor for the actual mapping ( Gosse et al . , 2008 ) . This suggests the existence of still elusive biochemical preadapting factors , which should be present in the optic tract , but absent from the target . 10 . 7554/eLife . 25533 . 028Figure 10 . Modeling tectal innervation . Co-adaptation enables fiber terminals to enter a tectal target field initially and allows correct mapping therein ( gray circles ) , whereas non-adapted terminals ( white circles ) fail to enter a tectal target field in simulations with 200 GCs placed in front of the target . The graded distributions of EphAs ( blue ) and ephrin-As ( red ) on the target field are symbolized by the colored wedges . RF and LF of adapted terminals are initially deflected by a factor of 30 . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 028
It has repeatedly been shown before that GCs can adapt to axonal guidance cues that act as global attractants or repellents ( Ming et al . , 2002; Piper et al . , 2005 ) . Adaptation in these cases might be a means of adjusting the dynamic range of the sensors . By contrast , adaptation against topographic cues comes unexpectedly , as these cues have been thought to act by virtue of their concentrations within graded distributions since the first proposition of gradient guidance ( Sperry , 1963 ) . Concentration gradients provide two kinds of information for chemotactic guidance: positional information , via the local concentration , c ( x ) , and directional information , via the derivative , dc ( x ) /dx . Adaptation amounts to deleting the positional information . This is what is found in bacterial chemotaxis ( Parkinson et al . , 2015 ) . At every location in the gradient , the CheR methyltransferase desensitizes chemotaxis receptors so that upon relocation of the bacterium , only the incremental change of the cue concentration is detected . This exclusive readout of the gradient’s directional information enables the bacteria to reach the source or the sink of an attractant or repellent , respectively , but precludes the identification of intermediate targets . The same would hold true for the topographic guidance of GCs . In our previously published computational model of topographic guidance ( Gebhardt et al . , 2012 ) , the topographic identity of the GCs is encoded in the ratio of ephrin-As to EphAs that they express . This is because this ratio determines the ratio of instantaneous reverse to forward signals received by the GC , which , according to the model , has to be balanced upon target approach . Thus , retaining topographic identity in the face of adaptation requires that these sensing channels are co-modulated concomitantly by the same factor . We challenged this idea of co-adaptation experimentally by testing the strong prediction that , through co-adaptive coupling , a signaling channel should be modulated even in the absence of its cognate signal , as long as the coupled channel is activated and modulated . We show that this is what happens in retinal GCs with regard to reverse and forward ephrin-A/EphA signaling . The mechanism is specific to the ephrin/Eph system , and GCs remain sensitive to other guidance signals throughout . To the best of our knowledge , co-adaptation is a novel signaling phenomenon not described previously . Canonical adaptation can be conceived as a negative feedback to channel activation . Co-adaptation , by contrast , also modulates coupled , non-activated signaling channels . It should be noted that the existence of co-adaptation in retinal GCs along the way supports our modeling assumption that topographic identity is encoded in the ephrin-A/EphA ratio . It also corroborates the occasionally disputed role of reverse signaling in topographic mapping . Why should a specific co-adaptation mechanism for the two channels exist , if only one was needed for mapping ? According to our findings , co-adaptation in retinal GCs is based on vesicular trafficking . Desensitization is correlated with surface clearance of ephrin-As and EphAs by CME , whereas re-sensitization implies replenishment of ephrin-As and EphAs from recycling endosomal compartments . In contrast to the transport of ephrin-As , the trafficking of EphA receptors has been described in detail ( Sabet et al . , 2015 ) . Receptors that have been auto-phosphorylated as a result of ligand binding , as well as receptors that are spontaneously auto-phosphorylated in the absence of a ligand , are internalized into Rab5-positive early endosomes . Ligand binding , as opposed to spontaneous auto-phosphorylation , is thought to induce clustering ( Janes et al . , 2012; Nikolov et al . , 2013 ) and additional phosphorylation . The latter recruits ubiquitinylating enzymes , eventually diverting the stable receptor/ligand complexes to the late endosomal/lysosomal compartment for degradation . Unliganded , spontaneously auto-phosphorylated receptors , however , are trafficked to the recycling endosome , bringing them into apposition with tyrosine phosphatase PTP1B on the cytosolic face of the endoplasmic reticulum membrane . After dephosphorylation , the receptors are recycled to the plasma membrane . Despite being triggered by ligand binding , co-adaptation undoubtedly necessitates the trafficking of unbound receptors . It is unlikely , however , that these correspond to the spontaneously auto-phosphorylated subpopulation described above . This is because spontaneous auto-phosphorylation is expected to occur at constant frequency , whereas the degree of co-adaptation must vary with the impinging signal . Thus , we propose that the sensors trafficked during co-adaptation form an independent subpopulation , whose size is adjusted to the required degree of adaptation and which is shuttled on a pathway that is not identical with the routes described above . As co-adaptation requires the shuttling of the proper absolute number of sensors in the topographically appropriate ratio , the question arises as to how this subpopulation might be selected for trafficking . The degree of co-adaptation required in the model is determined by the instantaneous guidance potential , D , measured by a GC , i . e . , downstream of the integration of reverse and forward signals . Sphingomyelin has been shown to increase in the cell membrane upon activation of Fyn-kinase , downstream of ephrin-A/EphA signaling , and to induce a concomitant decrease of membrane ephrin-As ( Baba et al . , 2009 ) . This sphingomyelin appeared to be a promising candidate regulator of co-adaptation . In fact , SMase pretreatment desensitizes retinal GCs to both ephrin-A forward and EphA reverse signaling . Sphingomyelin , together with cholesterol , is thought to be a key organizer of lipid rafts , which are liquid-ordered microdomains of the cell membrane ( Róg and Vattulainen , 2014; Simons and Sampaio , 2011 ) . We therefore went on to test the effects of MβCD , an efficient scavenger of cholesterol , on GC co-adaptation . In fact , we find a strong enhancement of the desensitization of forward and reverse signaling in both collapse and gap co-adaptation assays upon application of MβCD . Thus , manipulation of both major components of lipid rafts yields the same effect , supporting the idea that these membrane microdomains are involved in the observed desensitization . Formally , our experiments cannot rule out the possibility that lipid raft disruption simply impedes primary signaling and is thus unrelated to the co-adaptation mechanism . However , the fact that ephrin-A reverse signaling impacts on the relevant lipid composition of the membrane strongly suggests the idea that raft microdomain alteration is a genuine element of the GC co-adaptation mechanism . Raft disintegration would lead to a release of ephrin-As , which have been shown to be localized in membrane rafts ( Averaimo et al . , 2016; Gauthier and Robbins , 2003 ) , into the liquid disordered domain of the membrane . There is substantial evidence that EphAs might also reside in lipid rafts ( Boscher et al . , 2012; Chakraborty et al . , 2012; Foster et al . , 2003; Gu et al . , 2012; Inder et al . , 2012; Kim et al . , 2012; Yang et al . , 2010; Yi et al . , 2013; Zheng et al . , 2011 ) , potentially driven by the lipid-interactions of their second fibronectin type III domain ( Chavent et al . , 2016 ) , although these rafts seem to be separate from the ephrin-A-containing microdomains ( Kao and Kania , 2011; Marquardt et al . , 2005 ) . Therefore , we suggest that a similar dispersion from microdomains might also happen to EphAs upon variation of the sphingomyelin content of the membrane ( Figure 11 ) . These processes would eventually generate an intermixed subpopulation of dispersed ephrin-As and EphAs in the non-raft membrane . Notably , raft-located proteins are typically internalized by clathrin-independent endocytosis ( Sandvig et al . , 2011 ) . By contrast , we observe CME for both ephrin-As and EphAs during co-adaptation , supporting the idea that the endocytosed subpopulation of molecules might not be raft localized . If , as we propose , alteration of the sphingomyelin content induced the same fraction of ephrin-As and EphAs to re-partition from their respective microdomains to the liquid-disordered phase of the membrane , the dispersed mixture eventually would have the same ratio of ephrin-As and EphAs as the whole membrane . It could thus be endocytotically removed from the surface upon co-adaptation without altering the topographically relevant ephrin-A/EphA ratio . The same would be hardly conceivable if membrane clearance involved only ephrin-A/EphA ligand receptor complexes , as their ratio would be stoichiometric instead of depending on the topographic imprint of the respective cell . The disperse subpopulation would probably not be activated , as activated sensors should stay in their respective microdomains due to increased clustering caused by ligand binding ( Janes et al . , 2012; Nikolov et al . , 2013 ) . Thus , the co-adaptive subpopulation would not be expected to carry any phosphorylation tag to trigger endocytosis . We instead suggest that the disperse distribution by itself might destine for constitutive endocytosis ( Figure 11 ) . 10 . 7554/eLife . 25533 . 029Figure 11 . Proposed mechanism of co-adaptation . ( A ) EphAs ( blue ) and ephrin-As ( red ) , located in membrane lipid microdomains ( dark gray ) , signal in trans due to external cues . Trans forward ( FWD ) and reverse ( REV ) signals integrate into the guidance potential , D , and elicit repulsion when not balanced . Trans signaling also promotes sphingomyelin ( SM ) synthesis , and when incorporated into the membrane , this phospholipid might proportionally solubilize EphAs and ephrin-As from their respective domains . ( B ) Dispersed , unbound sensors are destined for constitutive clathrin-mediated endocytosis ( CME ) and might increase cis signaling upon internalization , as anti-parallel orientation is favored in high-curvature membrane vesicles . Enhanced cis signals desensitize GCs to trans signals . Sensitivity returns with the recycling of unbound sensors , and degraded receptor/ligand complexes should constitutively be replaced through protein synthesis in the long run . DOI: http://dx . doi . org/10 . 7554/eLife . 25533 . 029 Albeit highly intuitive that desensitization and re-sensitization might involve sensor endocytosis and recycling , respectively , our computational modeling suggests unexpected twists on the interpretation of sensor trafficking . Surprisingly , according to the model , the sensor surface concentration is irrelevant to adaptation . This can be seen , for example , when an individual axon is growing on a homogeneous ephrin target . Then , the model equation reduces toD=|ln1LT/LF+1| which includes the ephrin on the fiber terminal , LF , and on the target , LT , but not on the receptor actually detecting the trans signal . In fact , the model suggests that the axonal ligand representing cis-signaling is the relevant parameter determining the guidance potential . As discussed in the previous paper ( Gebhardt et al . , 2012 ) , we assume cis-interactions to be signal transducing instead of signal inactivating . That it is , in fact , cis-signaling that governs co-adaptation in the model can be seen more easily when the model equation for an individual axon is expressed conceptually in terms of signaling channels ( REV: reverse , FWD: forward ) :D=|lnREVtrans+REVcisFWDtrans+FWDcis| Upon introduction of an adaptation factor , a , which was originally meant to equally modulate all sensor concentrations , the model equation becomesD=|lna*REVtrans+a2*REVcisa*FWDtrans+a2*FWDcis| The square in front of the cis-signaling is due to the fact that , in contrast to trans-signaling , cis-signaling involves both axonal sensors . This obviously reduces toD=|lnREVtrans+a*REVcisFWDtrans+a*FWDcis| Thus , effectively , only the efficiency of cis-signaling is modulated during co-adaptation . Figuratively , this compares to a situation in which we listen to music via earphones ( cis-signaling ) to suppress external acoustic signals ( trans-signaling ) . The proposed type of signal transducing cis-interactions has the steric requirement that receptor and ligand be aligned , as in trans-signaling , in an anti-parallel orientation ( Himanen et al . , 2001 ) . This will normally occur only occasionally , when filopodia collide or when membrane ruffles are formed . Thus , the major effect of endocytosis in co-adaptation ( Figure 11 ) might be to provide high membrane curvature ( around the vesicle lumen ) to increase the efficiency of cis-signaling from endosomes ( Schmick and Bastiaens , 2014 ) . In summary , the proposed mechanisms could explain how co-adaptation might reconcile targeting accuracy in topographically mapping GCs with their adaptability . Thus , our study reveals that the primary task of forming a proper topographic map , despite being based on genetic hardwiring , is achieved through an astonishingly flexible mechanism .
Retinae of embryonic day 6 to 7 ( E6–E7 ) chick embryos were dissected in ice-cold Hanks' balanced salt solution ( HBSS ) , sucked onto a nitrocellulose filter and cut into 250–300 μm wide naso-temporal strips . Explants were placed on cover slips coated with 20 µg/ml mouse natural laminin in HBSS ( 1 hr at 37°C ) and grown in F12 medium containing 0 . 4% methylcellulose , 5% fetal calf and 2% chicken serum ( F12-MC ) at 37°C and 4% CO2 for 20–24 hr . The medium on explant cultures was replaced with warm F12-MC containing recombinant human ephrin-A5-Fc , mouse EphA3-Fc ( both RnD Systems , Minneapolis , MN , USA ) or human Fc fragment ( Calbiochem , San Diego , CA , USA ) . For inhibitor experiments , 30 µM Pitstop2 ( Abcam , Cambridge , UK ) or 400 mU/ml sphingomyelinase ( Sigma-Aldrich , St . Louis , MO , USA ) in F12-MC was applied during a 15 or 30 min pre-incubation , respectively , and thereafter together with ephrin-A5-Fc , EphA3-Fc or Fc . Methyl-β-cyclodextrin ( 2 mg/ml; Sigma-Aldrich , St . Louis , MO , USA ) was administered together with ephrin-A5-Fc , EphA3-Fc or Fc . After incubation for 20 or 120 min , cultures were fixed and stained with Alexa488 or Alexa568-phalloidin and the percentage of collapsed GCs was counted . Gap patterns were produced by direct contact printing as described previously ( von Philipsborn et al . , 2006a ) . In short , a PDMS stamp ( Elastosil RT 625; Wacker Chemie , Munich , Germany ) comprising the gap pattern was covered with protein solution for 2 hr at 37°C , rinsed in H2O , dried under a stream of nitrogen gas and then stamped onto an epoxysilanized and poly-L-lysine ( PLL ) -covered glass coverslip ( Ø 18 mm , VWR , Radnor , PA , USA ) for 15 min at 37°C ( cleaned coverslips in 1% [3-glycidoxypropyl]-trimethoxysilane [abcr , Karlsruhe , Germany] in pure EtOH , pH 5 . 5 for 5 min , air dried and covered with 200 μg/ml PLL [Sigma-Aldrich , St . Louis , MO , USA] in PBS overnight ) . The substrate was then covered with laminin , rinsed in H2O , covered with F12 and kept at 37°C until imminent usage . For single-cue gap assays 15 μg/ml ephrin-A5-Fc , 15 μg/ml EphA3-Fc or 25 μg/ml mouse semaphorin3A-Fc in PBS were used . For double-cue gap assays , both fields of the stamp were physically separated using a piece of ParafilmTM and each was covered with a different protein solution ( combinations of 15 μg/ml ephrin-A5-Fc , 15 μg/ml EphA3-Fc and 25 μg/ml mouse semaphorin3A-Fc in PBS ) . The pattern orientation was indicated on the coverslip . Patterns were stained with Alexa-conjugated anti-human goat IgGs . Quantification of gap assays was performed using a custom-written MATLAB ( RRID:SCR_001622 ) code ( Weth , 2017 ) to count fibers in and immediately behind the gap . In short , grayscale images of axons were thresholded and two ROIs were drawn in parallel to the gap ( ROI one in the gap , ROI two directly behind the gap; both ~20 μm wide and over the complete lateral extension of the gap ) . Both ROIs were then evaluated , counting signal peaks in each pixel row , and the percentage of stopping fibers was calculated from mean counts in ROI one and ROI 2 . Methyl-β-cyclodextrin ( 2 mg/ml; Sigma-Aldrich , St . Louis , MO , USA ) was directly added to the culture medium . For experiments with anisomycin ( AIM , Sigma-Aldrich , St . Louis , MO , USA ) , 40 µM AIM was added to the culture medium when axons were just about to leave the first ephrin-A5 field , and GCs were tracked using time-lapse imaging . To construct pSNAP–ephrin-A5–IRES–EGFP ( Figure 6—figure supplement 1 ) , we first reverse transcribed and PCR amplified the message for full-length Gallus gallus ephrin-A5 ( NCBI [RRID:SCR_006472] NM_205184 . 2 ) using primers starting 16 bp upstream of the start and ending 27 bp downstream of the stop codon from total RNA of the E7 chick tectum . Using SdaI and NotI linkers , the amplification product was inserted between the respective sites of the pSNAP-tag ( m ) vector ( New England Biolabs , Ipswich , MA , USA ) . In addition , a synthetic double-stranded DNA encoding the first 20 amino acids of chick ephrin-A5 ( signal peptide ) was inserted into the EcoRV site of the same construct to create a continuous open reading frame for a fusion protein containing the ephrin-A5 signal peptide , the SNAP-tag and full-length ephrin-A5 . This coding sequence was amplified from the construct using the primers starting 8 bp upstream of the start codon and ending 36 bp downstream of the stop codon , and was inserted into the SmaI site of the expression vector pCIG2 , containing a CAG enhancer/promotor upstream and IRES–EGFP downstream of the insert ( Hand et al . , 2005 ) . For pSNAP-ephrin-A5-IRES-dTom , the EGFP coding sequence was excised using MscI from pSNAP-ephrin-A5-IRES-EGFP and replaced by the coding sequence for dTomato amplified from pCAGJC-dTomato ( Lee et al . , 2013 ) using MscI primer linkers starting 4 bp upstream and ending 33 bp downstream of the dTomato coding sequence . pEGFP-Rab11 was a kind gift of Dr . Esther Stoeckli ( University of Zurich ) . It contains the coding sequence of human Rab11 N-terminally fused to EGFP using the pEGFP-C3 vector ( TaKaRa Bio USA , Mountain View , CA , USA ) , which contains the CMV enhancer/promotor ( Alther et al . , 2016 ) . For transgenic SNAP-ephrin-A5 expression , dissected retinae were cut and treated with 100 µl ice-cold Accutase solution for 5 min at room temperature and electroporated using 330 ng/µl pSNAP-ephrin-A5-IRES-EGFP/dTomato plasmid in 0 . 5x PBS and CUY700P20 electrodes ( NepaGene , Ichikawa-City , Japan; 15V , 5 × 50 ms , 950 ms off time ) . Because of higher transfection efficiency , we later switched to whole-eye electroporation as described in Vergara et al . , 2013 using 1 µg/µl pSNAP-ephrin-A5-IRES-EGFP/dTomato and 1 µg/µl pEGFP–Rab11 . All SNAP labeling reagents were purchased from New England Biolabs , Ipswich , MA , USA and administered at 1 µg/ml in warm F12-MC for 40 min at 37°C . For Surface SNAP-ephrin-A5 staining , SNAP Surface488 reagent was followed by washing with warm F12-MC . After fixation , cells were treated with sodium borohydride solution ( 2 × 5 min , about 250 mM in PBS ) to remove background and stained with anti-AlexaFluor488 rabbit IgG ( RRID:AB_221544 ) and anti-rabbit AlexaFluor647 goat IgG ( Jackson ImmunoResearch , Suffolk , UK ) to improve signal/noise . For recycling assays ( Figure 7—figure supplement 1 ) surface SNAP-ephrin-A5 was blocked using SNAP Surface Block . Intracellular SNAP-ephrin-A5 was labeled with SNAP Cell Fluorescein followed by washing with warm F12-MC . Cultures were grown for another 20–22 hr before application of anti-fluorescein mouse IgG in pre-warmed F12-MC to the living cells for 15 min . After washing , cells were fixed and stained with anti-mouse AlexaFluor647 goat IgG for formerly intracellular SNAP-ephrin-A5 now relocated to the cell surface . Images were taken using the ApoTome module on a Zeiss AxioimagerZ1 microscope ( Zeiss , Oberkochen , Germany ) . Colocalization of SNAP–ephrin-A5 with EGFP–Rab11 was evaluated by calculating the Manders' colocalization coefficients M1 and M2 using the Fiji ( RRID:SCR_002285 ) plugin Coloc2 ( https://github . com/fiji/Colocalisation_Analysis/ , V2 . 0 . 2 ) after image segmentation to isolate vesicular structures within GCs . Segmentation was performed using Squassh ( Rizk et al . , 2014 ) in Fiji with parameters: regul . =0 . 05; min . obj . int . =0 . 09; subpix . seg=true; excl . z-edge=true; localint . est . =auto; noise=gauss; PSFstd . dev . xy=0 . 7 , z = 0 . 8; removeint . =0; and removesize = 2 . All simulations were performed using MATLAB 8 . 4 ( RRID:SCR_001622 , The MathWorks , Natrick , MA , USA ) . Except for the introduction of co-adaptation ( see below ) , all modeling conditions were as previously described ( Gebhardt et al . , 2012 ) . Briefly , the projection target , T , a rectangular array of unit squares ( XT , max∗XT , max=50∗8 ) , displays the guidance cues along the x-axis ( exponential counter-distributions of ephrin-As [LT] and EphAs [RT] for simulations of tectal innervation and mapping , step-functions for simulation of gap assays ) . Fiber terminals , F , are modeled as circular discs with a diameter of about seven units , carrying ligands ( LF for terminal under consideration , Lf , for interacting terminals ) , and receptors ( RF and Rf , respectively ) in Gaussian-shaped distributions , according to their origin in the retina , which also bears exponential counter-distributions of LF and RF , and from which the terminals are equally sampled . On the target field , terminals are allowed to overlap freely while performing a random walk , which is biased by the tendency to minimize a guidance potential , D . In every iteration , i , with current terminal center position ( xT∗ , yT∗ ) , DF , i ( xT∗ , yT∗ ) is calculated from total EphA forward and total ephrin-A reverse signals , both comprising fiber–target , fiber–fiber and cis interactions ( assumed to be signal transducing in both directions ) . All signals are calculated from mass action for the corresponding receptors and ligands over all unit increments of the target and of other terminals overlapped by the terminal under consideration . All interactions are weighted equally ( with constants set to one ) , except trans fiber signals , whose influence , C ( i ) , increases with iteration number to conceptually reflect the developmental increase in terminal number and size . Thus , ( 1 ) DF , i ( xT∗ , yT∗ ) =|ln ( ∑xT , yTLF ( xT , yT ) [RT ( xT , yT ) +RF ( xT , yT ) +C ( i ) Rf ( xT , yT ) ]∑xT , yTRF ( xT , yT ) [LT ( xT , yT ) +LF ( xT , yT ) +C ( i ) Lf ( xT , yT ) ] ) | with the nominator representing total reverse and the denominator total forward signals instantaneously impinging on the terminal . A GC has reached its target position , when both signaling channels are balanced ( total reverse/total forward = 1 ) and , therefore , when D is minimized ( abs ( ln ( 1 ) ) =0 ) ) . For more detail , see Gebhardt et al . ( 2012 ) . We updated this model to include co-adaptation by introducing a common adaptation coefficient , a ( i ) , and a Hookian resetting force , f ( i ) , modulating both ligands and receptors ( collectively called sensors , S ) on the fiber terminals at every iteration , i+1 , depending on the recent history ( h ) of D: ( 2 ) S ( i+1 ) =a ( i ) S ( i ) +f ( i ) with ( 3 ) a ( i ) =1+ln ( 1+μ ( ∑k=1hk ( D ( i−h+k ) ) ∑k=1hk ) ) and ( 4 ) f ( i ) =λ ( S ( 0 ) −S ( i ) ) μ and λ are adjusting parameters . Unless explicitly stated , a unique set of parameters was used in all simulations: number of terminals n = 200; iterations i = 30000; μ = 0 . 006; λ = 0 . 0045; and h = 10 . For the simulation of in-vitro assays ( Figure 4 ) , where the area density of terminals is sparse , and for the analysis of tectal entry ( Figure 10 ) , where a role of fiber–fiber interactions has been excluded experimentally , the disproportionate increase of the weight of fiber–fiber interactions was switched off ( C0 = 1 ) . On target fields without guidance cues ( gap assays and in front of the tectum in the analysis of tectal entry ) , the random walk of the terminals was given a slight inherent bias to avoid prolonged dwelling ( qx = 0 . 37 ) . For details on C0 and qx and further parameters of the basic model , see Gebhardt et al . ( 2012 ) . Code is available via GitHub ( Weth , 2017 ) . A copy is archived at https://github . com/elifesciences-publications/RTP_Co-adapt_Model . KDs were measured using the BLItz biolayer interferometer ( Pall ForteBio , Menlo Park , USA ) at different EphA concentrations ( cf . Figure 2—figure supplement 1 ) . Streptavidin-coated biosensors were loaded with 5 µg/ml biotinylated human ephrin-A5-Fc ( RnD Systems , Minneapolis , MN , USA ) and the binding kinetics were measured using concentration series of 0 , 3 , 15 , 33 and 100 µg/ml EphA3-Fc all in BLItz kinetics buffer ( loading — 120 s; baseline — 30 s; association — 120 s; and dissociation —120 s ) . KDs were calculated from on and off rates derived by software from local curve fits corrected for start of association and dissociation . | The human brain contains roughly 100 billion neurons , which are organized into complex networks . But how does the brain establish these networks in the first place ? Neurons have long projections known as axons and , in the developing brain , these axons form structures called growth cones at their tips . The growth cones possess finger-like appendages that probe their surroundings in search of signals displayed on the surface of other cells . These signals guide the growth cones to their targets and move the axon tip into a position where it can form connections with other neurons within a particular network . The signals that growth cones follow are often distributed in concentration gradients so that the levels of a signal may be low at one end of a brain structure and gradually increase to a maximum level at the other end . In the developing visual system , for example , about one million axons from the retina reach their proper targets in visual regions of the brain by reading gradients of signals called ephrins and Ephs . However , when Fiederling et al . studied retinal neurons in a petri dish , they found that the axons became much less sensitive to both signals upon prolonged exposure to them . This unexpected finding raised a new question . If neurons rely upon these gradients for navigation , how do they continue to find their way if they also become less sensitive to those signals over time ? Fiederling et al . used a computer to simulate the events occurring in the developing brain . The simulations were based on the idea that navigating growth cones sense the ratio of ephrins to Ephs , instead of sensing the individual concentrations of these signals . Thus , by keeping the amounts of all involved sensors in strict proportion to each other while continuously re-adjusting them , the axons could still be accurately guided to their targets even though the neurons would become less sensitive to the signals . Experiments in neurons grown in petri dishes confirmed that retinal growth cones do exactly this and regulate the amounts of ephrin and Eph sensors on their outer membranes in a highly coordinated manner using a previously unknown mechanism . Given that signaling requires energy , the brain may have evolved this system to reduce the costs associated with wiring itself up . The system also offers greater flexibility than guidance based on the absolute concentrations of the signals . If other regions of the brain use a similar mechanism to establish their own wiring patterns , then understanding such basic mechanisms might eventually provide insights into diseases of miswiring such as schizophrenia and autism . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Material",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2017 | Ephrin-A/EphA specific co-adaptation as a novel mechanism in topographic axon guidance |
LRRK2 is a kinase expressed in striatal spiny projection neurons ( SPNs ) , cells which lose dopaminergic input in Parkinson’s disease ( PD ) . R1441C and G2019S are the most common pathogenic mutations of LRRK2 . How these mutations alter the structure and function of individual synapses on direct and indirect pathway SPNs is unknown and may reveal pre-clinical changes in dopamine-recipient neurons that predispose toward disease . Here , R1441C and G2019S knock-in mice enabled thorough evaluation of dendritic spines and synapses on pathway-identified SPNs . Biochemical synaptic preparations and super-resolution imaging revealed increased levels and altered organization of glutamatergic AMPA receptors in LRRK2 mutants . Relatedly , decreased frequency of miniature excitatory post-synaptic currents accompanied changes in dendritic spine nano-architecture , and single-synapse currents , evaluated using two-photon glutamate uncaging . Overall , LRRK2 mutations reshaped synaptic structure and function , an effect exaggerated in R1441C dSPNs . These data open the possibility of new neuroprotective therapies aimed at SPN synapse function , prior to disease onset .
Gain-of-function mutations in the leucine-rich repeat kinase 2 ( LRRK2 ) gene represent the most common cause of familial Parkinson’s disease ( PD ) ( Alessi and Sammler , 2018 ) . Carriers are at risk for late onset PD , which is clinically indistinguishable from sporadic PD , consistent with a possibility of common disease mechanisms ( Kluss et al . , 2019; Italian Parkinson's Genetics Network et al . , 2006; Haugarvoll et al . , 2008; Di Maio et al . , 2018 ) . LRRK2 gene product is a large multi-domain protein with two catalytic domains: a GTPase ( ROC-COR ) domain and a serine/threonine-directed protein kinase domain . Pathogenic mutations are found predominantly in these two domains , suggesting that LRRK2 enzymatic activities are involved in PD pathogenesis ( Cookson , 2010 ) , ( Esteves et al . , 2014 ) . Yet , how LRRK2 mutations in the two distinct functional domains contribute to PD pathogenesis and whether they act through a common mechanism is unknown . Enhanced LRRK2 kinase activity conferred by the G2019S ( GS ) mutation in the kinase domain is the most extensively studied property of mutant LRRK2 ( Cookson , 2010 ) , ( Steger et al . , 2016 ) . Meanwhile , the R1441C ( RC ) substitution in the GTPase domain results in impaired GTP hydrolysis , which is thought to indirectly enhance kinase activity through mechanisms that remain to be determined ( Nguyen and Moore , 2017 ) , ( Xiong et al . , 2010 ) . Despite remaining questions , the last decade marks extensive progress in our understanding of LRRK2 function . This kinase is highly expressed in the spiny projection neurons ( SPNs ) of the striatum ( Nguyen and Moore , 2017; West et al . , 2014; Parisiadou et al . , 2014 ) . LRRK2 expression peaks in a developmental time window of extensive glutamatergic excitatory synapse formation , suggesting that LRRK2 may regulate the development or function of excitatory synaptic networks . Consistently , several lines of evidence suggest that loss of LRRK2 alters striatal circuits during postnatal development ( Parisiadou et al . , 2014 ) , and the GS pathogenic mutation increases glutamatergic activity in cultured cortical neurons ( Beccano-Kelly et al . , 2015 ) as well as in acute striatal slices ( Matikainen-Ankney et al . , 2016 ) , ( Volta et al . , 2017 ) . Recent studies assign a critical role of LRRK2 in presynaptic terminal vesicle function ( Pan et al . , 2017 ) . Here , the GS mutation impairs presynaptic glutamatergic release , suggested to underlie changes in glutamatergic activity of striatal neurons ( Volta et al . , 2017 ) . The potential postsynaptic function of LRRK2 remains less well-characterized . We have previously shown that the RC mutation impedes normal striatal protein kinase A ( PKA ) signaling , which in turn results in increased GluA1 phosphorylation in developing SPNs , consistent with a postsynaptic mechanism of action ( Parisiadou et al . , 2014 ) . Similarly , glutamate receptor trafficking perturbations were observed in GS knock-in ( KI ) mice in response to plasticity induction protocols ( Matikainen-Ankney et al . , 2018 ) . Furthermore , although an increase in spontaneous excitatory postsynaptic currents has been reported for the dorsomedial striatum , a recent report failed to show this phenotype for the ventral striatum ( Huntley , 2020 ) , despite LRRK2 expression in that region ( West et al . , 2014 ) , ( Giesert et al . , 2013 ) . These observations suggest that LRRK2 mutations may shape the corticostriatal synaptic function in a synapse- , cell- and area-specific manner . Overall , despite the links between LRRK2 and glutamatergic synapse dysfunction , the field currently lacks a coherent framework for understanding how the two distinct LRRK2 mutations selectively alter synaptic function in specific cell types . Given the complementary role of direct and indirect striatal pathways in behavior ( Kravitz et al . , 2010; Fieblinger et al . , 2014; Kozorovitskiy et al . , 2012 ) the lack of pathway specificity in prior studies limits our understanding of disease related mechanisms associated with LRRK2 mutations . Previous studies did not compare the dysfunction of two LRRK2 mutations which are found in distinct LRRK2 domains and confer divergent biochemical properties to the kinase , choosing instead to focus on the one mutation or the other . Earlier reports have mainly focused on the GS pathogenic mutation , and whether the molecular mechanisms underlying pathology across the two most common mutations remain unknown . In addition , the phenotype of LRRK2 mutant models is subtle and corresponds to a moderate susceptibility for PD . Therefore , refined tools are required to distinguish the early pre-pathology functional abnormalities . We have taken the approach of combining molecular , anatomical and electrophysiological approaches that capture global aspects of LRRK2 function in the striatum , but also single synapse-specific effects , in case abnormalities are associated with specific synapse subtypes . Here , we undertook a systematic synapse function and structure analysis in two different mutant LRRK2 mouse lines to evaluate the contribution of this kinase mutations to SPN synapses across direct and indirect pathways . Using a combination of subcellular fractionation biochemistry , super-resolution imaging , and two-photon laser scanning microscopy with whole-cell physiology approaches , we found a critical role for LRRK2 RC , and to a lesser extent GS mutation , in organizing the structure and function of the SPN excitatory synapses , particularly for dSPNs .
In order to resolve LRRK2 mediated synaptic alterations in identified SPNs in a systematic way , we crossed the pathway-specific reporter BAC transgenic mice ( Tozzi et al . , 2018a ) with two mutant LRRK2 KI mouse lines . Specifically , Drd1-Tomato and Drd2-eGFP reporter mice were crossed with each of the two mutant LRRK2 lines ( RC and GS ) ( Figure 1A ) . This platform enabled pathway-specific interrogation of each LRRK2 mutation side by side , as well as initial experiments described below that were done on aggregated SPN populations . Our previous findings showed that RC KI neurons displayed altered PKARIIβ localization , as compared to wild-type ( +/+ ) neurons , suggesting altered synaptic PKA activity in SPNs ( Parisiadou et al . , 2014 ) . To further explore the synaptic PKA signaling in both KI mouse lines , we employed differential centrifugation , discontinuous sucrose gradient , and detergent extractions to isolate synaptic subcellular fractions from striatal extracts as shown in Figure 1B ( Parisiadou et al . , 2014 ) , ( Bermejo et al . , 2014; Peng et al . , 2004 ) . Consistent with our previous data , PKA activity was found elevated in the P2 crude synaptosomal preparation of +/RC but not +/GS striatal extracts , when phospho-PKA substrate antibody was used to detect phosphorylation of downstream PKA targets . The increased PKA activity in +/RC was further confirmed by an increase in phosphorylation of GluA1 , a key downstream target of PKA . To directly evaluate LRRK2 function , we examined the phosphorylation of two known targets of this kinase , Rab8A and Rab10 . Phosphorylation of both targets was selectively increased in the synaptic fraction ( P2 ) of the RC striatal extracts ( Figure 1D ) . Notably , while Rab8A is a known physiological target of LRRK2 kinase activity ( Steger et al . , 2016 ) , ( Bonet-Ponce and Cookson , 2019 ) , it was unclear whether this is the case in the striatum , since only low levels of Rab8A transcripts are found in dSPNs and iSPNs at the single-cell level ( dropviz . org ) ( Saunders et al . , 2018 ) . Our finding stands in accordance with two recent studies in cell lines , showing that the RC and other mutations found in the GTPase domain of LRRK2 result in greater increase kinase activity , compared to the GS mutation ( Purlyte et al . , 2018; Liu et al . , 2018 ) . We further isolated the post-synaptic density ( PSD ) fraction from the P2 preparation by using a detergent extraction step . The PSD fraction showed enrichment of the postsynaptic marker PSD95 , but not presynaptic protein synaptophysin ( svp38 ) ( Figure 1C ) . Given that PKA mediated phosphorylation of S845 in GluA1 has a significant impact on synaptic trafficking of GluA1 ( Parisiadou et al . , 2014 ) , ( Roche et al . , 1996 ) , we examined the levels of GluA1 in the PSD fractions of KI mice . We observed a significant increase in the GluA1 levels in the PSD fraction from +/RC , but not +/GS mice , compared to wild-type control ( +/RC: 47% increase of control; +/GS: 19% increase relative to control , one-way ANOVA p=0 . 0397 , post-hoc tests as noted , n = 6 per group ) ( Figure 1E , F ) , which was paralleled by elevated PKA activity specifically in the PSD fractions in +/RC ( +/RC , 167% of control , one-way ANOVA p=0 . 0486 , post-hoc tests as noted , n = 6 ) but not +/GS mice ( +/GS , 118% of control , n = 6 ) ( Figure 1E , G ) . Next , we determined the relative levels of PSD-95 protein in the PSD fraction across genotypes , which was used in the ratios in Figure 1F , G . We run subcellular fractions of all genotypes on the same blot ( Figure 1—figure supplement 1A , B ) , and normalized PSD95 to another synaptic protein , Homer1 . PSD95 band intensities across fractions and genotypes showed no difference . Furthermore , PSD95 band intensities in each of S1 , P2 , and PSD fractions were expressed as percentage of the PSD95 intensity of the wild type in each fraction ( Figure 1—figure supplement 1C ) . Similarly , no differences in PSD95 levels in mutant LRRK2 fractions were observed after one-way ANOVA . Overall , our data show increased GluA1 levels in the PSD fractions of +/RC striatal extracts , which parallels enhanced synaptic PKA signaling . To precisely examine the nanoscale organization of GluA1 receptors in the dendritic spines of identified SPNs in mutant LRRK2 mice , we used structured illumination microscopy ( SIM ) . SIM is able to overcome the resolution limits of conventional microscopy , and recently revealed a high degree of organization among scaffold proteins and AMPA receptors , forming nanoscale subsynaptic domains ( Gao et al . , 2018; Crosby et al . , 2019; Smith et al . , 2014 ) . Primary cultured striatal neurons from Drd1-dTomato or Drd2-eGFP mice crossed with either +/RC or +/GS mouse lines ( Figure 2A ) , were immunostained for GluA1 and PSD95 ( Figure 2B , and Figure 3A–B , Figure 2—video 1 ) . We found several lines of evidence demonstrating increased synaptic incorporation of GluA1 in +/RC dSPNs , compared to +/+ ones . Specifically , the distance between GluA1 and PSD95 was smaller in +/RC dSPNs compared to control dSPNs , while no difference was observed between +/GS and +/+ neurons ( +/RC , 63% of control , +/GS 85% of control one-way ANOVA , p=0 . 0208 , post-hoc tests as noted , n = 75–93 dendritic spines/genotype ) ( Figure 2C–D ) . Similarly , the +/RC dendritic spines showed the greatest shift in cumulative distribution for minimum distance between GluA1 and PSD95 nanodomains ( Figure 2E ) . While the overlap area of GluA1 and PSD95 nanodomains was found elevated in the dendritic spines of +/RC dSPNs , no differences were observed in +/GS neurons , compared to +/+ neurons ( +/RC , 137% of control , +/GS , 102% of control one-way ANOVA p=0 . 0215 , post hoc tests as noted , n = 75–93 dendritic spines/genotype ) ( Figure 2F ) . Similarly , the cumulative distribution analysis showed that GluA1-PSD95 distance was shifted towards higher values in the +/RC dSPNs ( Figure 2G ) . Given that the PSD95 area in +/GS dSPNs showed a trend towards decrease ( +/RC , 94% of control , +/GS , 83% of control , one-way ANOVA p=0 . 0904 , post-hoc +/+ vs +/GS p=0 . 075 ) ( Figure 2—figure supplement 1A , B ) , we evaluated the percentage of the GluA1-PSD95 overlap relative to the total PSD95 area across genotypes . 32% of total PSD95 area contained GluA1 in +/RC dSPNs , compared to 23% in +/+ neurons , and 26% in +/GS neurons ( one-way ANOVA p=0 . 0015 , post hoc tests as noted ) ( Figure 2H ) . These data strongly suggest increased levels of GluA1 at the synapses of +/RC dSPNs . Correlation analysis was performed to investigate the association between GluA1-PSD95 overlap nanodomain area and PSD95 area . A positive correlation was observed in all three genotypes ( control , Spearman r = 0 . 5822 , p-value<0 . 0001 , +/RC , r = 0 . 6522 , p<0 . 0001 and +/GS r = 0 . 6569 , p-value<0 . 0001 ) . The correlation was stronger in LRRK2 mutant SPNs ( Figure 2I–K ) . A similar synaptic nanoscale organization analysis was performed for Drd2-eGFP control and mutant LRRK2 SPNs . The neurons were immunostained with GluA1 ( purple ) , and PSD95 ( orange ) ( Figure 3A–B ) and minimum distance summary data ( +/RC , 87% of control , +/GS , 89% of control , one-way ANOVA , p=0 . 6435 n = 72–88 dendritic spines/genotype ) , as well as cumulative distribution of the distance between the GluA1 and PSD95 nanodomains , showed no difference across genotypes ( Figure 3C–D , E ) . We observed that the overlap GluA1 and PSD95 area was increased in +/GS SPNs ( +/RC , 105% of control , +/GS , 155% of control , one-way ANOVA , p<0 . 0001; post hoc tests as noted , p=72–88 dendritic spines/genotype ) ( Figure 3F ) . Further analysis showed an increase in the mean PSD95 area in the +/GS neurons ( +/RC , 123% of control , +/GS , 137% of control , one-way ANOVA , p=0 . 0008 , post hoc tests as noted ) ( Figure 2—figure supplement 1C ) , as well as in cumulative frequency ( Figure 2—figure supplement 1D ) . However , the percentage of the total PSD95 area containing GluA1 nanodomains +/GS iSPNs compared to controls was similar ( one-way ANOVA , p=0 . 469 ) ( Figure 3H ) . Thus , the increased size of GluA1-PSD95 overlapping domains in the +/GS iSPNs reflects the elevated PSD95 area of these neurons ( Figure 2—figure supplement 1C-D ) , and not synaptic GluA1 incorporation , as found in +/RC . We aimed to further investigate the relationship between overlapping nanodomains size and PSD95 area in the iSPNs dendritic spines . In general , PSD95 area correlated with the overlap GluA1-PSD95 domain size in all genotypes ( control , Spearman r = 0 . 7817 , p<0 . 0001 , +/RC , r = 0 . 4040 , p=0 . 0003 and +/GS , r = 0 . 7349 , p<0 . 0001 ) ( Figure 3I–K ) . Based on Figure 2I , we found that for larger PSD95 domains , mutant LRRK2 dSPNs show a tendency towards larger overlap between GluA1 and PSD95 . However , this was not the case for iSPNs of +/RC genotype . In fact , in +/RC iSPNs , some high overlap GluA1-PSD95 domains were characterized by smaller PSD95 areas . This detailed high-resolution imaging analysis points to different GluA1 subsynaptic distribution in the iSPNs of +/RC SPNs , compared to +/+ and +/GS . Taken together , these single synapse imaging findings in pathway identified SPNs demonstrate an increased number of GluA1 receptors in the synapses of +/RC dSPNs , pointing toward cell type and mutant-specific LRRK2 functions . These observations suggest highly specific and likely regulated nanoscale organization of GluA1 receptors in SPN dendritic spines . Since the effects of LRRK2 mutations appear to alter excitatory synapse subunits , next we carried out recordings of pharmacologically isolated miniature excitatory post-synaptic currents ( mEPSCs ) in acute brain slices from mice of the three genotypes . SPN pathway identity was determined based on the presence or absence of Drd2-eGFP ( Kozorovitskiy et al . , 2015 ) . Neurons were held at −70 mV in voltage clamp , and pharmacologically isolated mEPSCs were recorded in the presence of tetrodotoxin ( TTX ) . For dSPNs , the mean frequency of mEPSCs , averaged within each neuron , was decreased relative to controls for both LRRK2 mutations ( GFP- , RC 60% of control , GS , 47% of control; two-way ANOVA , genotype main effect , p<0 . 001 , post-hoc tests as noted ) ( Figure 4A–B ) . For iSPNs , lower mean frequency of mEPSCs , averaged within each neuron , was observed in both RC and GS neurons ( GFP- , RC 60% of control , GS , 52% of control; two-way ANOVA , genotype main effect , p<0 . 001 , post-hoc tests as noted ) . In contrast to the mean frequency of mEPSCs , no significant differences in the mean within-cell response amplitude were observed across all six genotype/cell type combinations on post-hoc comparisons , despite a modest main effect of genotype ( GFP- , RC 107% of control , GS , 92% of control; GFP+ , RC 108% of control , GS 94% of control; two-way ANOVA , genotype main effect , p=0 . 041 , N = 13–14 neurons/group ) ( Figure 4C ) . To evaluate the distribution of individual mEPSC features across genotypes , rather than within-neuron averages , we computed cumulative distributions of inter-event intervals ( IEIs ) and mEPSC amplitudes , along with histograms of response amplitudes ( Figure 4D–E ) . Amplitude data were fitted with a gamma distribution , a non-negative asymmetrical distribution , because the mEPSC amplitudes were not normally or lognormally distributed ( Figure 4 , Figure 4—figure supplement 1A ) , failing D'Agostino and Pearson , Shapiro-Wilk , and KS normality tests , p<0 . 0001 , for raw and log-transformed data . Skew and kurtosis across the 6 datasets were 2 . 19 ± 0 . 16 and 6 . 34 ± 0 . 1 , respectively , appropriate the gamma distribution . Akaike information criterion ( AIC ) values as goodness-of-fit criteria confirmed superior fit of gamma family distribution over Gaussian . To compare the distributions of mEPSC amplitudes , we estimated mEPSC amplitude means across different genotypes for GFP- and GFP+ cell types using generalized linear modeling ( GLM ) with a gamma family distribution . mEPSC amplitudes in RC dSPNs were shifted toward larger events , and smaller events in GS dSPNs , with no differences in iSPNs ( dSPNs , Bonferroni post-hoc comparisons , p<0 . 0001; iSPNs , p>0 . 7 ) ( Figure 4 , Figure 4—figure supplement 1B ) . To investigate whether the observed decrease in the mEPSC frequency was associated with global changes in dendritic spine number and morphology , we relied on identifying pathway-specific SPNs by injecting a Cre-dependent adeno-associated virus ( AAV ) expressing eGFP virus ( AAV8/Flex-GFP ) in the striatum of neonatal pups . Drd1-Cre and Adora2a-Cre mice were crossed with +/RC and +/GS mice , as well as controls ( Figure 5A , Figure 5—video 1 ) . We then analyzed dendritic fragments of d- and iSPNs to determine dendritic spine density , spine head width , and spine length . No differences in any of these parameters were observed for all six genotype/cell type combinations ( two-way ANOVA , n = 9–12 SPNs from 3 to 4 mice/genotype ) ( Figure 5B–E ) . Further dendritic spine classification showed no difference in dendritic spine type in SPNs of either pathway ( two-way ANOVA , n = 9–14 SPNs for 3–4 mice/genotype ) ( Figure 5F ) . Overall , these data suggest that the functional synaptic alterations ( Figure 4 ) observed in +/RC and +/GS mutant SPNs are not reflected in uniform changes in dendritic spine number or morphology . Our results indicate that the abundance of GluA1 subunits is selectively increased in dendritic spines of +/RC dSPNs , along with synaptic AMPAR presence and the prevalence of larger amplitude mEPSCs . This is consistent with a possibility that a specific fraction of +/RC excitatory synapses are functionally stronger . In order to evaluate the physiology of single dendritic spines in pathway-identified SPNs across three genotypes , we used two-photon dual laser glutamate uncaging and imaging , combined with whole-cell electrophysiology in voltage clamp mode ( Figure 6A ) . Drd2-eGFP mice were crossed into G2019S and R1441C KI lines , and maintained on a WT background for control groups . SPNs were filled with a cesium-based internal and Alexa 594 for imaging dendrites and dendritic spines at ~910 nm ( Figure 6B–C ) ; the presence or absence of GFP labeling was used to assign pathway identity . Uncaging evoked EPSCs ( uEPSCs ) were elicited using 0 . 5 ms long pulses of 725 nm laser light , directed near dendritic spines ( ~0 . 5–1 µm away ) , in order to drive focal uncaging of MNI-glutamate . Locations of peak responses , sampled for three sites near each dendritic spine , were chosen for data acquisition for every synapse . Recordings were carried out in the presence of blockers of GABAARs , NMDARs , muscarinic receptors , and sodium channels , for isolation of single synapse AMPAR responses . We found that single spine glutamate uEPSCs were increased in amplitude selectively in dSPNs of RC genotype mice ( GFP- , +/RC 196% of control , GS , 102% of control; one-way ANOVA , p<0 . 05 , post hoc tests , as noted; n = 18–28 dendritic spines/group ) ( Figure 6D–F ) . For Drd2-eGFP+ SPNs ( iSPNs ) , no uEPSC amplitude differences were observed ( GFP+ , RC 70% of control , GS , 80% of control; one-way ANOVA , p=0 . 1682; n = 21–65 dendritic spines/group ) , suggesting a pathway-specific effect on single synapse function in the RC genotype .
In the present study , we characterized synaptic dysfunctions caused by mutant LRRK2 ( RC and GS ) in pathway-identified SPNs . To our knowledge , this is the first comparative approach focusing on more than one PD related LRRK2 mutation and employing both global and single synapse approaches for the study of LRRK2 driven striatal remodeling . By studying RC and GS KI mouse lines side by side , we were able to demonstrate that the two most common LRRK2 gain-of-function pathogenic mutations ( Hernandez et al . , 2016 ) alter the function of SPN excitatory synapses in largely similar ways , with several potentially important differences . The observed changes were frequently stronger for the RC mutations , and more exaggerated in the direct pathway . For example , while whole cell electrophysiology recordings revealed decreases in the mESPCs frequency across both +/RC and +/GS mice , two-photon dual laser glutamate uncaging and imaging , combined with whole-cell electrophysiology , demonstrated larger amplitude in uEPSCs selectively for +/RC dSPNs . This observation paralleled higher synaptic GluA1 receptors levels in the same subtype +/RC SPNs ( Figure 6G ) . Based on the current knowledge , it remains unclear why the RC mutation is associated with somewhat more pronounced effects on sculpting striatal synapses . The GS mutation , which has been the main focus of the current literature , is found in the kinase domain of LRRK2 , whereas the RC mutation is located in the GTPase domain of the protein ( Paisán-Ruiz et al . , 2013 ) , ( Cookson , 2012 ) . It has been previously proposed that all pathogenic mutations lead to increased LRRK2 kinase activity , although the mechanism by which RC does this still remains unclear ( Alessi and Sammler , 2018; Nguyen and Moore , 2017 ) . Moreover , it is not known whether aberrant LRRK2 substrate phosphorylation is the predominant pathogenic mechanism in LRRK2 mediated striatal changes . Recent findings have revealed a subset of Rab family of proteins as the long awaited bona fide LRRK2 substrates ( Steger et al . , 2016 ) . Notably , the RC mutation , mainly in the context of heterologous cell lines , leads to higher increase in Rab phosphorylation compared to the GS mutation , which is localized to the kinase domain ( Liu et al . , 2018 ) , ( Purlyte et al . , 2018 ) . Similarly , the evidence of increased synaptic PKA activities was observed in +/RC and not +/GS striatal synaptic fractions . This increased PKA activity in the RC synapses is expected to positively correlate with elevations in dSPNs signaling ( Zhai et al . , 2019 ) . Specifically , phosphorylation of GluA1 at Ser 845 by PKA facilitates the targeting of the receptors to extrasynaptic membranes and their ‘priming’ for synaptic insertion and this results in elevated GluA1 synaptic incorporation ( Diering and Huganir , 2018 ) . Indeed , our biochemical and SIM imaging data in RC mice support this notion . In contrast to dSPNs , PKA-driven signaling requirements for iSPNs are more complex , due to the involvement of adenosine . This molecule can be released by neurons or glia ( Surmeier et al . , 2007 ) , ( Zhang et al . , 2019 ) , complicating predictions for PKA involvement in mutant LRRK2 function in iSPNs . Several prior reports demonstrate a LRRK2 mediated presynaptic regulation of glutamatergic transmission ( Matikainen-Ankney et al . , 2016; Volta et al . , 2017; Beccano-Kelly et al . , 2015; Piccoli et al . , 2011 ) , while emerging evidence emphasizes a postsynaptic role for LRRK2 ( Parisiadou et al . , 2014; Matikainen-Ankney et al . , 2016 ) . Our previous and current findings showed that LRRK2 contributes to glutamatergic synaptic functions by directing PKA signaling events in SPNs , leading to altered GluA1 synaptic incorporation in the striatum of RC mice ( Parisiadou et al . , 2014 ) . Despite the elevated GluA1 levels in the synapses of +/RC dSPNs , there were no alterations in mean mEPSCs amplitude in these neurons . However , cumulative probability analysis demonstrated that the amplitude shifted toward larger values in the +/RC dSPNs , suggesting an inhomogeneity of synaptic function , where a fraction of synapses exhibit larger GluA1-dependent responses and single-synapse uncaging-evoked EPSCs . This scenario , on top of overall dampening of mEPSC frequency , is expected to bias mutant dSPN responses to be driven by a narrower set of glutamatergic inputs than is normally the case . Overtime , a small bias could be amplified through the recurrent circuitry of basal ganglia loops ( Kozorovitskiy et al . , 2012 ) . While there are some differences in the age of preparation and in recording conditions , our consistent finding of substantial miniature EPSC frequency decreases for both mutations in pathway-identified SPNs are in disagreement with prior recordings of spontaneous events in the GS mutant mice . One recent report showed elevations in spontaneous EPSCs events frequency in GS KI mice at 1–3 months ( Volta et al . , 2017 ) , while another reported changes in spontaneous but not miniature events with the GS mutation in pre-weaning mutants ( Matikainen-Ankney et al . , 2016 ) , although recordings were primarily performed in a mixed population of direct and indirect pathway SPNs , and neither study thoroughly compared the two mutations . Such differences in spontaneous event frequency point to a potential circuit-level compensatory mechanism that may be driven by disturbances in a subset of SPN excitatory synapses . Surprisingly , in the presence of clear functional and nano-architecture abnormalities , aggregate dendritic spine density and morphology , at least at the ages we evaluated , appeared normal . One possible explanation for this mismatch that requires further investigation is the possibility of higher occurrence of nonfunctional or differently functioning spines in LRRK2 mutant SPNs . Given the critical role of LRRK2 in excitatory glutamatergic synapses ( Parisiadou et al . , 2014; Volta et al . , 2017; Matikainen-Ankney et al . , 2016 ) , a more nuanced understanding of LRRK2-mediated synaptic alterations should facilitate the search for more targeted PD therapies . Here , we focused primarily on interrogating cell type specific LRRK2-based responses . Why does this pathway specificity matter ? It is well-established that dopamine loss causes pathway-specific morphological and functional changes in SPNs ( Kravitz et al . , 2010; Fieblinger et al . , 2014; Gertler et al . , 2008 ) . Moreover , dSPNs and iSPNs exhibit opposing PKA pathway signaling properties after dopamine receptor activation ( Gerfen and Surmeier , 2011; Surmeier et al . , 2007 ) . Given the critical role of LRRK2 in synaptic PKA activities ( Parisiadou et al . , 2014; Tozzi et al . , 2018b ) , these mutations are poised to have pathway specific effects and present distinct opportunities for pathway-targeted therapies . Over the past years , a number of transgenic mouse models was generated to study LRRK2 function ( Volta and Melrose , 2017; Xiong et al . , 2017 ) ; however , the results were found to be inconsistent across studies . Gene-targeted mutant LRRK2 KI mice represent the most physiologically relevant model to investigate LRRK2 mediated alterations and unravel disease mechanisms . Here , we attempt to synthesize and further the existing knowledge by employing a comparative approach focusing on two distinct mutations , combined with a powerful suite of techniques allowing for both global and single synapse study of LRRK2 SPNs . The clinical phenotypes that define PD arise after a substantial loss of nigrostriatal dopamine signaling . Thus , the identification of pre-symptomatic dysfunctions offers a potential window of opportunity for the development of neuroprotective therapies in PD . Human asymptomatic LRRK2 mutation carriers represent an appropriate population to define these preclinical symptoms . Indeed , emerging evidence suggests that asymptomatic LRRK2 mutation carriers do show subtle motor and non-motor symptoms , including alterations of corticostriatal circuit organization , in comparison to asymptomatic non-carriers ( PPMI Investigators et al . , 2020; LRRK2 Ashkenazi Jewish Consortium et al . , 2015 ) . Accordingly , cellular and synaptic dysfunctions in etiologically relevant LRRK2 mutant KI mice allow investigations of the early events that precede neuronal death and may be predictive of future dysfunction . Several studies have suggested a central role for LRRK2 in striatal SPNs ( Volta et al . , 2017 ) , ( Parisiadou et al . , 2014; Xenias , 2020 ) . Here , we show that the presence of LRRK2 mutations can influence SPN function in a pathway-specific context , in the absence of dopamine neurodegeneration . Examining PD-relevant cellular functions in the absence of DA depletion represents a move away from the classical ways of studying PD in small rodent models , where basal ganglia cellular and network properties have been examined using several methods for dopamine depletion ( Gerfen and Surmeier , 2011; Kreitzer and Malenka , 2008 ) . While there is evidence that at first iSPNs show the primary structural and functional synaptic and dendritic changes in response to dopamine loss ( Gerfen , 2006; Gertler et al . , 2008; Day et al . , 2006; Villalba et al . , 2009 ) , with time both SPNs types undergo substantial adaptations ( Suarez et al . , 2018 ) , ( Gagnon et al . , 2017 ) . Here , we find that some early changes in LRRK2 mutant dSPNs – including those observed at the single synapse level – are poised to contribute to the fragility of the system to lower levels of DA loss/variation . Because the well-described recurrent circuitry of the basal ganglia ( Kozorovitskiy et al . , 2012 ) , ( Kravitz et al . , 2010; Oldenburg and Sabatini , 2015 ) enables the output of direct and indirect pathways to feed back to regulate cortico-striatal glutamate release , minor synaptic defects in LRRK2 mutant SPNs could be amplified over time , contributing to the life-time risk of the disorder . The presence of early corticostriatal alterations in LRRK2 carriers ( LRRK2 Ashkenazi Jewish Consortium et al . , 2015; Vilas et al . , 2015; the Barcelona LRRK2 Study Group et al . , 2016 ) , along with the evidence for finer scale synaptic dysfunctions reported in this study , open the possibilities for future personalized medicine approaches that take into account the specific LRRK2 mutation and emphasize protecting the health of striatal neurons prior to potential DA losses .
All mouse related experiments followed the guidelines approved by the Northwestern University Animal Care and Use Committee . Young adult male and female mice ( postnatal days 30–50 ) were used in this study . Approximately equal numbers of males and females were used for every experiment . All mice were group-housed , with standard feeding , light-dark cycle , and enrichment procedures . C57BL/6 ( wild-type ) , RC ( Tong et al . , 2009 ) , and GS ( Yue et al . , 2015 ) heterozygous knock in mice were used for subcellular fractionation experiments . For electrophysiological and imaging approaches RC or GS mice were crossed with Drd1-dTomato and/or Drd2-eGFP mice . Heterozygotes for RC and GS allele and hemizygotes for Drd1-dTomato and/or Drd2-eGFP were used in all experiments . For dendritic spine analysis , hemizygous Drd1-Cre , and Adora2a-Cre mice crossed with RC and GS KI mice were used . All animals were backcrossed to C57BL/6 for several generations . Subcellular fractionation of mouse striatum was performed as previously described ( Bermejo et al . , 2014 ) , ( Parisiadou et al . , 2014; Figure 1B ) . Specifically , mouse striata were dissected ( three striata per experiment were pooled ) and rapidly homogenized in four volumes of ice-cold Buffer A ( 0 . 32 M sucrose , 5 mM HEPES , pH7 . 4 , 1 mM MgCl2 , 0 . 5 mM CaCl2 ) supplemented with Halt protease and phosphatase inhibitor cocktail ( Thermo ) using a Teflon homogenizer ( 12 strokes ) . Homogenized brain extract was centrifuged at 1400 g for 10 min . Supernatant ( S1 ) was saved and pellet ( P1 ) was homogenized in buffer A with a Teflon homogenizer ( five strokes ) . After centrifugation at 700 g for 10 min , the supernatant ( S1’ ) was pooled with S1 . Pooled S1 and S1′ were centrifuged at 13 , 800 g for 10 min to the crude synaptosomal pellet ( P2 ) and the supernatant ( S2 ) . P2 was resuspended in Buffer B ( 0 . 32 M sucrose , 6 mM Tris , pH 8 . 0 ) supplemented with protease and phosphatase inhibitors cocktail with Teflon homogenizer ( five strokes ) and was carefully loaded onto a discontinuous sucrose gradient ( 0 . 8 M/1 M/1 . 2 M sucrose solution in 6 mM Tris , pH 8 . 0 ) with a Pasteur pippete , followed by centrifugation in a swinging bucket rotor for 2 hr at 82 , 500 g . The synaptic plasma membrane fraction ( SPM ) in the interphase between 1 M and 1 . 2 M sucrose fractions was collected using a syringe and transferred to clean ultracentrifuge tubes . 6 mM Tris buffer was added to each sample to adjust the sucrose concentration from 1 . 2 M to 0 . 32 M and the samples were centrifuged in a swinging bucket rotor at 200 , 000 g for 30 min . The supernatant was removed and discarded and the SPM pellet was resuspended in 500 μl of 6 mM Tris/2 mM EDTA/0 . 5% Triton X-100 solution and rotated for 30 min at 4°C . In turn , the samples were centrifuged at 32 , 800 g for 20 min . The supernatant contains the Triton-soluble fraction ( TSF ) , whereas the pellet represents the postsynaptic pellet ( PSD ) . S1 , P2 , presynaptic and PSD fractions were separated by 4–12% NuPage Bis-Tris PAGE ( Invitrogen ) and transferred to membranes using the iBlot nitrocellulose membrane Blotting system ( Invitrogen ) by following manufacture protocol . Primary antibodies specific for GluA1 ( Cell Signaling Technology #13185 ) , pGluA1 ( Cell Signaling Technology #8084 ) , phospho-PKA substrates ( Cell Signaling Technologies #9624 ) , phospho-Rab8A ( Abcam , ab188574 ) , total Rab 8A ( Abcam , ab230260 ) , phospho-Rab10 ( Abcam ab230261 ) and total Rab10 ( Cell Signaling Technologies #8127 ) as well as secondary anti-mouse and anti-rabbit ( Thermo Fischer Scientific ) Membranes were incubated with Immobilon ECL Ultra Western HRP Substrate ( Millipore ) for 3 min prior to image acquisition . Chemiluminescent blots were imaged with iBright CL1000 imaging system ( Thermo Fisher Scientific ) . For quantitative analysis , images were analyzed using iBright Analysis Software ( Thermo Fisher Scientific ) . Primary corticostriatal co-cultures were prepared as described previously ( Parisiadou et al . , 2014 ) , ( Tian et al . , 2010 ) . In brief , striatal cultures were prepared from P0 pups of Drd1-Tomato or Drd2-eGFP mice crossed with mutant KI LRRK2 lines ( G2019S and R1441C ) . Tissues were digested by papain ( Worthington Biochemical Corporation ) and the striatal and cortical cells were mixed at a ratio of 1:2 . The neurons were placed on coveslips with plating medium ( Medium I ) containing Basal Eagle Medium ( Sigma-Aldrich ) supplemented with 1 x GlutaMAX ( Gibco ) , 1 × B27 ( Gibco ) , 1 x N-2 ( Gibco ) , 1 x Antibiotic Antimycotic ( Sigma-Aldrich ) , 5% horse serum ( Gibco ) and 5% FBS ( Gibco ) at a density of 4 × 105 for about one hour . After initial plating , medium was changed to Medium I without the horse serum and the FBS supplemented with 2 . 5 μM arabinosylcytosine ( Sigma-Aldrich ) ( Medium II ) . Half of the medium was changed with fresh Medium II every 7 days; experiments were conducted 28–30 days after plating . Multichannel SIM images were obtained with a Nikon Structured Illumination super-resolution microscope using a 100x , 1 . 4 NA objective as previously described ( Smith et al . , 2014 ) . The acquisition was set to 10 MHz , 16 bit depth with EM gain and no binning . Exposure was between 100 and 500 ms and the EM gain multiplier restrained below 1500 . Conversion gain was held at 1x . Laser power was adjusted to keep LUTs within the first quarter of the scale . Single images were processed and analyzed using Nikon Elements software . Reconstruction parameters ( Illumination Modulation Contrast , High Resolution Noise Suppression , and Out of Focus Blur Suppression ) ( 0 . 94 , 0 . 98 , and 0 . 07 ) for GluA1/PSD95 and ( 1 . 2 , 3 . 0 , and 0 . 07 ) for GFP/tdTomato were kept consistent across all acquisitions and experiments . Single spine analyses were carried out on 72–95 spines across 9–12 neurons per genotype . Images of dendritic fragments were collected from secondary to tertiary dendrites across genotypes . All the spines on the fragments were analyzed . GFP/tdTomato signal ( Thermo Fischer Scientific , #A10262/#M11217 ) was used to generate a mask to identify the spine shape , subsequent analysis were done with the nanodomains of GluA1 ( NeuroMab #75327 ) and PSD95 ( Thermo Fischer Scientific , #51–6900 ) within the mask . 3D reconstructions in a single plane used nine images captured with 2D SIM and reconstructed with 3D SIM utilizes 15 to increase xy and z resolution were generated by Nikon Element and the illumination modulation contrast was set automatically by the software . Nikon Elements software ( general analysis ) was used for quantification of colocalization of GluA1 and PSD95 proteins . The images were thresholded for each channel and kept constant across experiments . The regions of interest ( puncta ) were outlined , and total immunofluorescence number and binary area for each region per channel were measured automatically . Regions in one channel were overlayed on the other channel . The minimum distances between puncta ( e . g . PSD95 to GluA1 ) were measured from surface to surface automatically by the software . Intracranial injections were performed as previously described ( Kozorovitskiy et al . , 2012 ) . Briefly , P4-5-day-old pups were placed into a stereotaxic frame under cryoanesthesia . 200 nl of AAV8/Flex-GFP virus ( 6 . 2*1012 used at 1:3 dilution , UNC Vector Core , Chapel Hill , NC ) were delivered into the dorsal striatum at a rate of 100 nl min−1 using microprocessor-based controller , Micro4 ( WPI ) . In order to ensure targeting of the dorsal striatum the needle was placed 1 mm anterior to midpoint between ear and eye , 1 . 5 mm from midline and 1 . 8 mm ventral to brain surface . Confocal images of fixed 80-μm-thick brain sections of P30 pups injected with the AAV8/Flex-GFP virus were obtained with the Nikon A1R scope . Fluorescence projection images of dendrites and the corresponding spines were acquired with a 60x oil immersion objective ( NA = 1 . 4 ) at 0 . 1 μm intervals with 1024 × 1024 pixel resolution . For each genotype , 2–4 segments per neuron , 9–14 neurons from 3 to 4 animals were used to generate z-stacks . Fragments between secondary to tertiary dendrites without overlap with other neurons or discontinuous were chosen for analysis . Dendritic spine density and morphology was performed using Imaris 9 . 21 software ( Bitplane , Concord , USA ) . Images of dendritic fragments were collected from secondary to tertiary dendrites , for tracing of the dendritic fragments the autopath mode of the filament tracer and default settings were selected . The following settings were used for spine detection: 0 . 5 μm minimal head size , 1 . 8 μm maximum length , seed point threshold approx . 10 , no branched spines were allowed . Spine detection was manually corrected if necessary . Classification of spines into stubby , mushroom-like and filopodia was performed using the Imaris XTension classify spines with following definitions: stubby: spine length <0 . 75 μm; mushroom: spine length <3 . 5 μm , spine head width >0 . 35 μm and spine head width >spine neck width; filopodia: when it did not fit the criteria mentioned above ( Schier et al . , 2017 ) . Two hundred and fifty µm thick acute coronal brain slices were prepared as described previously ( Kozorovitskiy et al . , 2015 ) , ( Xiao et al . , 2017 ) and incubated in artificial cerebral spinal fluid ( ACSF ) containing ( in mM ) 127 NaCl , 2 . 5 KCl , 25 NaHCO3 , 1 . 25 NaH2PO4 , 2 . 0 CaCl2 , 1 . 0 MgCl2 , and 25 glucose ( osmolarity ~310 mOsm/L ) . Slices were recovered at 34°C for 15 min , followed by 30 min at RT , and transferred to a recording chamber perfused with oxygenated ACSF at a flow rate of 2–3 mL/min . Whole-cell recordings were obtained from dorsolateral striatal neurons visualized under infrared Dodt or DIC contrast video microscopy using patch pipettes of ~4–6 MΩ resistance . Drd2 BAC GFP signal visualized under epifluorescence was used to target iSPNs , while Drd2 GFP-negative neurons with spiny dendrites and electrophysiological properties of SPNs ( holding current , input resistance ) were considered dSPNs ( Gagnon et al . , 2017 ) , ( Gittis et al . , 2010 ) . Both fluorescent and non-fluorescent neurons were targeted for recording . Internal solution consisted of ( in mM ) : 120 CsMeSO4 , 15 CsCl , 10 HEPES , 2 QX-314 Chloride , 2 Mg-ATP , 0 . 3 Na-GTP , 1 EGTA ( pH ~7 . 2 , ~295 mOsm ) . Morphology was confirmed using 20 µM Alexa 594 in the recording pipette . Recordings were made using a Multiclamp 700B amplifier ( Molecular Devices ) . Data were sampled at 10 kHz and filtered at 4 kHz , acquired in MATLAB ( MathWorks ) . Series resistance , measured with a 5 mV hyperpolarizing pulse in voltage clamp , averaged under 20 MΩ and was left uncompensated . Miniature EPSCs were recorded from voltage clamped SPNs held at −70 mV in the absence of stimulation . Over 2 min of recording per neuron was used for analyses . For recordings of miniature EPSCs , 50 µM gabazine , 10 µM scopolamine , 10 µM CPP , and 1 µM TTX were added to the ACSF . Two-photon laser-scanning microscopy and two-photon laser photoactivation were accomplished on a modified Scientifica microscope with a 60X ( 1 . 0 NA ) objective . Two mode-locked Ti:Sapphire lasers ( Mai-Tai eHP Deep See and Mai-Tai eHP; Spectra Physics ) were separately tuned , with beam power controlled by independent Pockels cells ( ConOptics ) . The beams were separately shuttered , recombined using a polarization-sensitive beam-splitting cube ( Thorlabs ) , and guided into the same galvanometer scanhead ( Cambridge ) . The Mai Tai eHP Deep See was tuned to ~910 nm for excitation of genetically encoded GFP and Alexa 594 , and the Mai Tai eHP was tuned to 725 nm for photoactivation of recirculated caged MNI-L-glutamate ( 2 mM , Tocris ) ( Xiao et al . , 2018; Banala et al . , 2018 ) . Neurons in the dorsolateral striatum were targeted for recording , as for other electrophysiology assays , with secondary and tertiary dendrites targeted for uncaging . All 2P uncaging voltage clamp recordings were made at a holding potential of −70 mV . Internal solution contained ( in mM ) 115 K-gluconate , 20 KCl , 4 MgCl2 , 10 HEPES , 4 Mg-ATP , 0 . 3 Na-GTP , 7 phosphocreatine , 0 . 1 EGTA , ( pH 7 . 2 , 290 mOSm ) . Alexa Fluor 594 ( 10–20 μM ) was added to the internal solution to visualize cell morphology for uncaging/imaging with physiology experiments . Uncaging evoked EPSCs were elicited by 0 . 5 ms pulses of 725 nm laser light ( ~20 mW at the focal plane ) . Up to three locations in a single field of view were stimulated ( 1 s apart ) in a single sweep . Stimulation of a single location occurred with a minimum 10 s ISI . A spot diameter of ≤0 . 8 μm , based on measurements of 0 . 5 μm beads ( 17152–10; Polysciences Inc ) was used for all two-photon laser flash photolysis experiments . Two GaAsP photosensors ( Hamamatsu , H7422 ) with 520/28 nm band pass filters ( Semrock ) , mounted above and below the sample , were used for imaging fluorescence signals . A modified version of ScanImage was used for data acquisition ( Pologruto et al . , 2003 ) , ( Kozorovitskiy et al . , 2015 ) . Group statistical analyses were done using GraphPad Prism seven software ( GraphPad , LaJolla , CA ) . For n sizes , both the number of trials recorded and the number of animals are provided . All data are expressed as mean + SEM , or individual plots . For two-group comparisons , statistical significance was determined by two-tailed Student’s t-tests . For multiple group comparisons , one-way analysis of variance ( ANOVA ) tests were used for normally distributed data , followed by post-hoc analyses . For non-normally distributed data , non-parametric tests for the appropriate group numbers were used , such as the Mann-Whitney test . For the analysis of mEPSC amplitude distributions , data were fit with a gamma distribution , a non-negative asymmetrical distribution , because the mEPSC amplitude data were not normally or lognormally distributed but displayed appropriate skew and kurtosis to match the gamma distribution . Generalized linear modeling ( GLM ) with a gamma family distribution was used to estimate and compare mEPSC amplitudes , with Bonferroni corrections for multiple posthoc comparisons . All mEPSC distribution analyses were done in R , using fitdistrplus library and the emmeans package . Spearman correlation was used to detect correlation between two groups of data . p<0 . 05 was considered statistically significant . | Parkinson’s disease is caused by progressive damage to regions of the brain that regulate movement . This leads to a loss in nerve cells that produce a signaling molecule called dopamine , and causes patients to experience shakiness , slow movement and stiffness . When dopamine is released , it travels to a part of the brain known as the striatum , where it is received by cells called spiny projection neurons ( SPNs ) , which are rich in a protein called LRRK2 . Mutations in this protein have been shown to cause the motor impairments associated with Parkinson’s disease . SPNs send signals to other regions of the brain either via a ‘direct’ route , which promotes movement , or an ‘indirect’ route , which suppresses movement . Previous studies suggest that mutations in the gene for LRRK2 influence the activity of these pathways even before dopamine signaling has been lost . Yet , it remained unclear how different mutations independently affected each pathway . To investigate this further , Chen et al . studied two of the mutations most commonly found in the human gene for LRRK2 , known as G2019S and R1441C . This involved introducing one of these mutations in to the genetic code of mice , and using fluorescent proteins to mark single SPNs in either the direct or indirect pathway . The experiments showed that both mutations disrupted the connections between SPNs in the direct and indirect pathway , which altered the activity of nerve cells in the striatum . Chen et al . found that individual connections were more strongly affected by the R1441C mutation . Further experiments showed that this was caused by the re-organization of a receptor protein in the nerve cells of the direct pathway , which increased how SPNs responded to inputs from other nerve cells . These findings suggest that LRRK2 mutations disrupt neural activity in the striatum before dopamine levels become depleted . This discovery could help researchers identify new therapies for treating the early stages of Parkinson’s disease before the symptoms of dopamine loss arise . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2020 | Pathway-specific dysregulation of striatal excitatory synapses by LRRK2 mutations |
The mating-type switching endonuclease HO plays a central role in the natural life cycle of Saccharomyces cerevisiae , but its evolutionary origin is unknown . HO is a recent addition to yeast genomes , present in only a few genera close to Saccharomyces . Here we show that HO is structurally and phylogenetically related to a family of unorthodox homing genetic elements found in Torulaspora and Lachancea yeasts . These WHO elements home into the aldolase gene FBA1 , replacing its 3' end each time they integrate . They resemble inteins but they operate by a different mechanism that does not require protein splicing . We show that a WHO protein cleaves Torulaspora delbrueckii FBA1 efficiently and in an allele-specific manner , leading to DNA repair by gene conversion or NHEJ . The DNA rearrangement steps during WHO element homing are very similar to those during mating-type switching , and indicate that HO is a domesticated WHO-like element .
Mating-type switching is an important process in the natural life cycle of many budding yeast species . If an uninhabitable environment improves and becomes habitable , any yeast spores that germinate earlier than others will have a competitive advantage , provided that they have a backup mechanism to prevent death if they germinate too early . Mating-type switching provides this backup ( Hanson and Wolfe , 2017 ) . Yeast spores are survival capsules , and mating-type switching enables the microcolony formed from a newly-germinated spore to sporulate again after just a few cell divisions if necessary , preventing its extinction ( Herskowitz , 1988 ) . Across the phylogenetic tree of budding yeasts , mating-type switching has arisen independently at least 11 times , indicating strong natural selection in favor of switching ( Krassowski et al . , 2019 ) . In Saccharomyces cerevisiae , HO is the central gene in the mating-type switching process . It was one of the first yeast genes ever discovered , because haploid strains with a functional HO gene can switch their mating type and hence auto-diploidize and form visible spores , whereas ho mutants cannot ( Winge and Roberts , 1949; Oshima , 1993 ) . HO codes for an endonuclease that makes a double-strand DNA break at the mating type ( MAT ) locus , and is essential for efficient switching ( Kostriken et al . , 1983; Russell et al . , 1986 ) . Much of our knowledge about how eukaryotes repair double-strand DNA breaks in their chromosomes comes from studies that used HO as a model system ( Haber , 2016 ) . But despite the comprehensive functional and genetic characterization of HO , its evolutionary origin remains mysterious ( Keeling and Roger , 1995; Haber and Wolfe , 2005; Koufopanou and Burt , 2005; Muller et al . , 2007 ) . The HO gene is a relatively recent evolutionary addition into the yeast genome , because it is found only in a few genera closely related to Saccharomyces ( Butler et al . , 2004; Hanson and Wolfe , 2017 ) . The ‘three-locus’ system for mating-type switching , involving an active MAT locus and silent HML and HMR loci , pre-dates the origin of HO . An outgroup species , Kluyveromyces lactis , also has a three-locus system but has no HO gene , and employs alternative mechanisms to create a double-strand break at the MAT locus to initiate switching ( Barsoum et al . , 2010; Rajaei et al . , 2014 ) . Some more distantly related budding yeasts switch mating types using ‘two-locus’ flip/flop inversion systems , and again do not have an HO gene ( Hanson and Wolfe , 2017; Krassowski et al . , 2019 ) . As well as being a recent evolutionary innovation , HO endonuclease also has an unusual protein domain structure that begs the question of where it came from . It resembles inteins , but is not an intein itself . Inteins are mobile genetic elements that are completely protein-coding and occur as in-frame fusions within a host gene ( Novikova et al . , 2014 ) . After the host gene is transcribed and translated , the intein is excised post-translationally and the host protein is assembled by protein splicing , making a peptide bond between its N- and C-terminal parts ( exteins ) . HO has highest sequence similarity to the VDE intein of budding yeasts , which is the only intein in S . cerevisiae ( Koufopanou and Burt , 2005; Green et al . , 2018 ) . The host gene for VDE is VMA1 , which codes for a subunit of vacuolar H+-ATPase ( Gimble and Thorner , 1992; Anraku et al . , 2005 ) . The excised VDE intein has endonuclease activity and can cleave empty ( inteinless ) alleles of VMA1 , enabling the intein to spread through the population by homing – a selfish , super-Mendelian mode of inheritance ( Burt and Koufopanou , 2004; Burt and Trivers , 2006 ) . The VMA1 genes of several species in the budding yeast family Saccharomycetaceae are polymorphic for the presence/absence of VDE , due to active homing and interspecies spread of the intein ( Koufopanou et al . , 2002; Okuda et al . , 2003 ) . Homing of VDE into empty alleles of VMA1 occurs during meiosis in diploids that are heterozygotes for intein-containing and empty alleles of VMA1 ( Gimble and Thorner , 1992 ) . The VDE protein has two domains ( Moure et al . , 2002 ) : a protein splicing domain that enables the host protein Vma1 to be made , and a homing endonuclease domain that enables the VDE DNA sequence to home into empty alleles of VMA1 . Most inteins are found in bacteria and archaea , not yeasts ( Poulter et al . , 2007; Novikova et al . , 2014; Green et al . , 2018 ) . In phylogenetic and other sequence similarity analyses , the two yeast proteins HO and VDE were found to be each other’s closest relatives ( Dalgaard et al . , 1997; Koufopanou and Burt , 2005; Green et al . , 2018 ) . Although HO is related to inteins , and more distantly related to other homing endonucleases in the LAGLIDADG superfamily ( Chevalier and Stoddard , 2001 ) , it is an independently expressed standalone gene , whereas inteins and other homing endonucleases are self-splicing entities embedded within their host genes ( Belfort et al . , 2005; Belfort , 2017 ) . HO does not undergo protein splicing and has no exteins . HO also has a unique zinc finger domain at its C-terminus that is not present in other intein-like proteins or homing endonucleases . In S . cerevisiae HO , amino acid residues essential for cleavage of the MAT locus are located both in the zinc finger and in the endonuclease domain of the intein-like region ( Meiron et al . , 1995; Bakhrat et al . , 2004; Bakhrat et al . , 2006 ) , and the endonuclease has a stringent requirement for zinc ions ( Jin et al . , 1997 ) . A key feature differentiating HO from true homing endonucleases is that it does not propagate its own DNA sequence – in other words , it does not home . HO has become a normal cellular gene , and is an example of a domesticated mobile genetic element ( Volff , 2006; Rusche and Rine , 2010 ) . However , until now it has been unclear what type of mobile element HO originated from . Here , we show that HO is related to a large and diverse family of intein-zinc finger fusion proteins ( WHO proteins ) that occur mostly in the yeast genus Torulaspora . WHO proteins are encoded by a newly discovered homing genetic element , whose genomic organization is different from all other known homing elements , and whose host is the aldolase gene FBA1 . The similarities between WHO and HO proteins , and between the DNA rearrangement steps that occur during WHO element homing and mating-type switching , show how the HO-catalyzed system of mating-type switching originated .
In the genome sequence of the type strain of Torulaspora delbrueckii ( CBS1146; Gordon et al . , 2011 ) , we identified a cluster of five genes ( TDEL0B06670 to TDEL0B06710 ) spanning 14 kb that have sequence similarity to HO . We renamed these genes WHO1 to WHO5 , for ‘weird HO’ ( Figure 1A ) . Two of them are pseudogenes , with a single frameshift in WHO1 and more extensive damage in WHO5 . The WHO gene cluster is located downstream of the FBA1 gene encoding fructose-1 , 6-bisphosphate aldolase , an enzyme that functions bidirectionally in glycolysis and gluconeogenesis ( Schwelberger et al . , 1989 ) . Amino acid sequence identity among the inferred WHO proteins is unusually low for a tandem gene cluster , ranging from 55% ( Who2 vs . Who4 ) down to 24% ( Who3 vs . Who4 ) . T . delbrueckii also has an HO gene ( TDEL0A00850 ) elsewhere in its genome , orthologous and syntenic with the HO gene of S . cerevisiae . The five inferred WHO proteins have only 22–25% identity to T . delbrueckii HO ( BLASTP E-values in the range 1e-6 to 3e-30 ) . The length and content of the WHO gene cluster is polymorphic among natural isolates of T . delbrueckii . This species has a primarily haploid ( haplontic ) life cycle ( Kurtzman , 2011 ) , so there is only one allele per strain . We found six different allelic WHO cluster arrangements among 15 strains we examined , with content ranging from 2 to 9 WHO genes and pseudogenes ( Figure 1A; Supplementary file 1 ) . The largest cluster ( 18 kb ) is in strain L09 , with three intact genes ( WHO6 , WHO2 , WHO3 ) and six WHO pseudogenes . As well as WHO genes and pseudogenes , some of the T . delbrueckii clusters contain one or two duplicated fragments of the 3’ end of FBA1 , interspersed with WHO genes ( Figure 1A ) . The FBA1 fragments are about 400 bp long ( full-length FBA1 is 1080 bp ) . Some FBA1 fragments contain frameshifts or internal stop codons , and others are intact . The WHO proteins consist of an intein-like region followed by a zinc finger domain ( Figure 1B ) . This structure is similar to HO , but distinct from VDE which has no zinc finger . WHO and HO are the only intein-zinc finger fusion proteins known in any organism . WHO and HO both have no exteins , and their intein domains both lack an amino acid motif that is normally found at the C-terminal end ( motif G; Pietrokovski , 1994 ) . Similar to HO , the zinc finger domains of the WHO proteins contain variable numbers ( 3-5 ) of a Cys-X-X-Cys motif . By BLAST searches , we found that the most similar zinc finger domains in other yeast proteins occur in orthologs of S . cerevisiae Ash1 ( a regulator of HO transcription ) , which is an atypical type of GATA zinc finger domain ( Scazzocchio , 2000; Münchow et al . , 2002 ) , though the level of sequence identity is low ( maximally 38% over 76 residues ) . WHO gene/pseudogene clusters are present downstream of FBA1 in all five other species of the genus Torulaspora that we examined , and they again show within-species polymorphism in their gene content . We found three allelic WHO cluster arrangements among 9 T . pretoriensis strains , and two allelic arrangements in 2 T . globosa strains ( Figure 1A ) . A pair of WHO genes is also present downstream of FBA1 in Zygotorulaspora mrakii , which is closely related to Torulaspora . There are no WHO genes in the next most closely related genus , Zygosaccharomyces . By database searches , we found that WHO genes also occur in the genus Lachancea , but they are absent from almost every other sequenced budding yeast genome . Eleven of the 12 sequenced Lachancea species have clusters of WHO genes or pseudogenes downstream of FBA1 , and five of these species have intact WHO genes ( Figure 1—figure supplement 1 ) . The WHO clusters in four Lachancea species also contain duplicated fragments of the 3’ end of FBA1 . These structures in Lachancea are remarkably similar to the ones seen in Torulaspora , especially considering that these two genera are not closely related to each other , and there are no WHO genes in most other yeast genera . In addition , a few Lachancea genomes contain other WHO genes or pseudogenes at loci separate from FBA1 ( Figure 1—figure supplement 1 ) . In summary , WHO genes resemble homing endonuclease genes ( Gimble , 2000 ) . Like HO , but unlike any other homing endonucleases , the endonuclease domain is fused to a zinc finger domain . The structure of WHO genes , and their occurrence in clusters that are mixtures of intact genes and pseudogenes , suggests that they are part of a mobile genetic element . We hypothesized that WHO proteins are homing endonucleases whose target is the FBA1 gene . This hypothesis was motivated by two observations . First , WHO genes with very diverse sequences occur in tandem clusters downstream of FBA1 , in both Torulaspora and Lachancea , suggestive of repeated integrations of different members of a mobile element family into the same target locus . Second , fragments of FBA1 are present within the WHO gene clusters in both Torulaspora and Lachancea ( Figure 1 and Figure 1—figure supplement 1 ) . All the FBA1 fragments consist of only the 3’ end of the gene , and many of them begin at approximately the same position ( base 670–680 in the gene sequence ) , which we hypothesized could indicate a possible endonuclease cleavage site in FBA1 . To test the hypothesis that WHO proteins target the FBA1 gene , we carried out experiments in S . cerevisiae because few tools exist for genetic manipulation of Torulaspora or Lachancea . We chose T . delbrueckii WHO6 for these experiments because it is intact , present in only a minority of the T . delbrueckii isolates we examined ( 3 of 15 ) , and it is located at the end of the cluster closest to FBA1 ( Figure 1A ) . Together , these features suggested that WHO6 could be the most recently-inserted WHO gene in the cluster in the strains that contain it , and therefore that WHO6 is a good candidate for a homing element that is currently active and spreading through the T . delbrueckii population . FBA1 is an essential gene in most growth conditions ( Lobo , 1984; Schwelberger et al . , 1989; Boles and Zimmermann , 1993 ) , so we reasoned that if it is the natural target of WHO endonuclease cleavage , then strains of T . delbrueckii that contain a WHO6 gene should contain alleles of FBA1 that are resistant to cleavage by Who6 endonuclease , whereas other T . delbrueckii strains might contain alleles that are sensitive to Who6 . We constructed haploid strains of S . cerevisiae that contain the open reading frame ( ORF ) of T . delbrueckii FBA1 ( TdFBA1 ) integrated into the ADE2 gene on chromosome XV . These strains also have the native S . cerevisiae FBA1 gene on chromosome XI . We used two different alleles of TdFBA1: one from a T . delbrueckii isolate ( strain L09 ) that has a WHO6 gene downstream of it , and one from an isolate ( CBS1146 ) that has no WHO6 gene ( Figure 1A ) . We then introduced a high copy-number panARS plasmid ( pWHO6-HA ) on which WHO6 was expressed from the constitutive T . delbrueckii TDH3 promoter ( Figure 2A ) , with a 3xHA epitope tag at its 3’ end . As described below , we found that this plasmid induces cleavage of the allele from strain CBS1146 , but not of the allele from strain L09 . Hence we designated the CBS1146 allele TdFBA1-S ( for sensitivity to cleavage by Who6 ) , and the L09 allele TdFBA1-R ( for resistance ) . We also found that the plasmid does not induce cleavage of S . cerevisiae FBA1 . Our experiment is similar to one carried out by Moore and Haber ( 1996 ) who overexpressed HO so that it continually cleaved the MAT locus in haploid S . cerevisiae cells that had no HML/HMR loci . They found that the only cells that survived HO overexpression were ones in which inaccurate DNA repair ligated the chromosome back together but modified the target site sequence in such a way that HO could no longer cleave it , because chromosomes with accurate repairs were re-cleaved by HO . Similarly , in our experiment , the only haploid S . cerevisiae cells containing TdFBA1-S that survived overexpression of Who6 were ones in which the TdFBA1-S sequence became modified , either by gene conversion or by imprecise non-homologous end joining ( NHEJ ) ( Figure 2A ) . By genome sequencing , we found that in two independent experiments where pWHO6-HA was introduced into S . cerevisiae strains containing TdFBA1-S , this T . delbrueckii FBA1 allele was modified by gene conversion with the native S . cerevisiae FBA1 gene . In contrast , no gene conversion was seen when pWHO6-HA was introduced into S . cerevisiae strains containing the TdFBA1-R allele ( two independent transformants ) , nor in control transformations in which a similar plasmid expressing 3xHA-tagged Green Fluorescent Protein ( pGFP-HA ) was transformed into S . cerevisiae strains containing TdFBA1-S or TdFBA1-R . Apart from the gene conversions at the TdFBA1-S transgene , no other nucleotide changes were detected in the genomes of the strains transformed with pWHO6-HA , and the native S . cerevisiae FBA1 gene remained unchanged in this experiment . To verify that the gene conversions were not caused by the 3xHA tag , we carried out additional experiments in which S . cerevisiae strains containing either TdFBA1-S or TdFBA1-R were each independently transformed 10 times with either pWHO6 ( a plasmid expressing Who6 with no 3xHA tag ) or pGFP-HA as a control . A single colony was chosen from each transformation , and its ade2::TdFBA1 locus was amplified by PCR and sequenced . Approximately 100-fold fewer colonies were obtained from the transformations of pWHO6 into the TdFBA1-S strain than in any of the other three combinations of plasmid ( pWHO6 or pGFP-HA ) and allele ( TdFBA1-S or TdFBA1-R ) . Evidence of cleavage of the T . delbrueckii FBA1 ORF was observed in all 10 independent transformants of pWHO6 into the TdFBA1-S strain: five transformants showed gene conversion between TdFBA1-S and the native S . cerevisiae FBA1 gene , and the other five had single-nucleotide insertions or deletions in TdFBA1-S consistent with cleavage and repair by imprecise NHEJ ( Figure 2B ) . In contrast , no sequence changes at ade2::TdFBA1 were detected in the 10 independent transformants of pWHO6 into the TdFBA1-R strain , nor ( as expected ) in any of the 20 transformants with pGFP-HA . From these experiments we conclude that the T . delbrueckii isolate ( L09 ) that contains the WHO6 gene also contains , 588 bp upstream , an allele of FBA1 ( TdFBA1-R ) that is resistant to cleavage by the Who6 endonuclease . It can therefore stably maintain this endonuclease gene in its genome . In contrast , an isolate ( CBS1146 ) that has no WHO6 gene contains an FBA1 allele ( TdFBA1-S ) that is sensitive to cleavage by Who6 . Similar to the HO overexpression survivors in Moore and Haber ( 1996 ) , our transformants that contained the sensitive TdFBA1-S allele of T . delbrueckii FBA1 and survived overexpression of Who6 have acquired mutations that can be inferred to have damaged the Who6 recognition or cleavage sites , making the cells resistant to Who6 . These mutations enable us to identify the approximate location of the Who6 recognition and cleavage sites in T . delbrueckii FBA1 . The five transformants showing evidence of imprecise NHEJ ( survivors 1–5 in Figure 2B ) each sustained a single 1 bp insertion or deletion in the TdFBA1-S sequence , at the same site in the gene ( positions 667–669: TTT→TTTT , or TTT→TT ) . Like other LAGLIDADG endonucleases , HO and VDE both make a staggered double-strand break with 4 bp 3’ overhangs when they cleave DNA , and they have large ( ~24 bp ) degenerate recognition sequences that span the cleavage site ( Nickoloff et al . , 1990; Gimble and Thorner , 1993; Taylor et al . , 2012 ) . We therefore infer that the overhang made by Who6 must include some or all of the TTT sequence centered on position 668 ( Figure 2B ) . The HO recognition site at the MAT locus is moderately conserved between S . cerevisiae and other budding yeasts , and has at its core the sequence CGCAACA , where the 4 bp overhang is underlined . By analogy , we suggest that the core of the Who6 recognition site in the sensitive TdFBA1-S allele of T . delbrueckii FBA1 is CGCATTT ( positions 663–669 ) . In the resistant TdFBA1-R allele , 3 of these seven bases are different ( CagcTTT ) . In S . cerevisiae FBA1 , which is also resistant to cleavage by Who6 , 3 of 7 bases are different ( tGCtTTc ) ( Figure 2B ) . The seven transformants showing evidence of gene conversion ( survivors 6–12 , including two from pWHO6-HA transformations and five from pWHO6 transformations ) each replaced a section of the T . delbrueckii FBA1-S sequence with the corresponding section of the S . cerevisiae FBA1 gene from chromosome XI ( Figure 2B ) . The T . delbrueckii and S . cerevisiae genes have 84% nucleotide sequence identity overall . The gene conversion tracts are asymmetrical: they extend rightwards ( towards the stop codon of FBA1 ) from the cleavage site for 106–306 bp , whereas they extend leftwards for only 5–14 bp . This asymmetry suggests that one side of the WHO-induced double-strand break is more active in recombination than the other . Similarly , during mating-type switching in S . cerevisiae , the DNA on one side of the HO-induced break ( the Z-side ) participates in exchange with HML/HMR , whereas DNA on the Y-side remains inert until it is eventually clipped off ( Lee and Haber , 2015 ) . In summary , the pattern of NHEJ events and gene conversions in S . cerevisiae cells that contain TdFBA1-S and survive continuous expression of a WHO protein is very similar to the pattern in cells that contain MAT and survive continuous expression of HO ( Moore and Haber , 1996 ) . We infer that WHO endonucleases cleave FBA1 genes at approximately base 668 , which is slightly upstream of the 5’ ends of many of the FBA1 fragments seen in Torulaspora and Lachancea species . The core of the putative recognition site of a WHO endonuclease is also similar to that of HO , with the sequence 5’-CGC-3’ adjacent to the overhang . The TdFBA1-S and TdFBA1-R alleles have only 85% nucleotide sequence identity downstream of base 668 , which is remarkably low for two alleles from the same species . In contrast , they have 99% identity upstream of this position . More generally , among the full-length FBA1 genes of the 15 T . delbrueckii isolates we analyzed , nucleotide sequence diversity is much higher in the 3’ part of the gene ( Figure 3A ) . Moreover , phylogenetic trees constructed from the 5’ and 3’ parts of TdFBA1 ( upstream and downstream of position 668 ) have contradictory topologies ( Figure 3B ) . The heterogeneous evolution of the two ends of FBA1 in T . delbrueckii , and the presence of 3’ FBA1 fragments in its genome , suggest a mechanism for how the homing genetic element containing WHO genes operates . We propose that WHO elements home into the FBA1 locus by using a mechanism that involves replacing the 3’ end of FBA1 , thereby converting a sensitive FBA1 allele into one that is resistant to the particular WHO protein encoded by the element ( Figure 4A , left column ) . During meiosis in a heterozygous diploid cell , the WHO protein encoded by the donor allele cleaves the sensitive full-length FBA1 gene of the recipient allele at position 668 . The double-strand break is then repaired by using the WHO-containing chromosome as a template . The 3’ region of the cleaved full-length FBA1 gene interacts with the FBA1 fragment in the template , resulting in incorporation of the WHO gene into the previously empty allele . After homing , the recipient chromosome contains a WHO gene located between a resistant full-length FBA1 gene ( a chimera of the recipient’s previous 5’ end and a copy of the donor’s 3’ end ) and a new FBA1 fragment formed from the recipient’s previous 3’ end . In this model , the FBA1 fragment downstream of the donor’s WHO gene is an essential part of the WHO element because it provides a region of homology that acts as a recombination site ( Figure 4A ) . The fragment is not part of the expressed FBA1 gene so it does not need to maintain an open reading frame . While the modified FBA1 is now resistant to the newly acquired WHO gene , it may still be sensitive to other WHO genes . Repeated homing of multiple different WHO elements into the same chromosome will build tandem clusters of WHO genes and FBA1 fragments , with the most recent elements being located closest to the full-length FBA1 gene ( Figure 4B ) . Each homing event replaces the 3’ end of the full-length FBA1 gene with the sequence from a different allele , causing its rapid evolution and discordant phylogenies . Over time , the WHO genes and FBA1 fragments can decay into pseudogenes or become deleted , because they are not required for the aldolase function of FBA1 . The functional unit of a WHO element can be defined as a WHO gene and the resistance-conferring FBA1 3’ region upstream of it ( Figure 4B ) . In summary , T . delbrueckii WHO genes are part of a homing genetic element that targets FBA1 . Our model for its mechanism of action explains how WHO clusters and FBA1 fragments are formed , and the unusual chimeric mode of evolution of FBA1 . We investigated the phylogenetic relationship among WHO , HO , and VDE genes , using amino acid sequences inferred from intact genes and from some of the less-damaged WHO pseudogenes . In view of the high divergence among the sequences , the tree topology may not be fully accurate , but it permits identification of approximately 14 families of WHO genes ( Figure 5 ) . The WHO families form a monophyletic group , separate from HO and VDE . Most of the families are either Torulaspora-specific or Lachancea-specific , indicative of recent gene duplications within each genus . The WHO2 , WHO4 , WHO5 and WHO11 families are specific to T . delbrueckii so they must be young . Overall , the tree indicates a dynamic history of extensive WHO gene duplication and frequent formation of pseudogenes , consistent with the ‘cycle of degeneration’ expected for a homing genetic element ( Burt and Koufopanou , 2004 ) . Although most WHO genes and pseudogenes are located downstream of FBA1 genes ( magenta branches in Figure 5 ) , a few of them are not . It is striking that these non-FBA1-associated genes fall into a small number of clades ( blue branches in Figure 5 ) . The WHO genes in these clades seem to have lost their target specificity for FBA1 and transposed to other places in the genome , and several of these WHO genes are intact . Most notably , the WHO10 family includes five intact genes from T . globosa that are located at five different places in the genomes of the two isolates we sequenced ( Figure 6 ) . There is a WHO10 pseudogene beside FBA1 in one T . globosa isolate and the sister species T . maleeae ( Figure 1A ) , indicating that WHO10 was originally associated with FBA1 . Another WHO clade unlinked to FBA1 is WHO8 , which is present only in two species of Lachancea ( Figure 5 ) . We detected only a few WHO sequences in species other than Torulaspora ( or Zygotorulaspora ) and Lachancea in BLAST searches against the NCBI database , which includes genome sequences from hundreds of yeast species including members of almost every genus in the family Saccharomycetaceae ( Shen et al . , 2018 ) . Thus the WHO family has a very limited phylogenetic distribution , and occurs mostly in two genera that are not sisters of each other ( Shen et al . , 2018 ) . The few WHO sequences outside Torulaspora and Lachancea all lie in the WHO13 and WHO12 families , which are outgroups to the other WHO families ( Figure 5 ) , and most of them are pseudogenes . WHO13 is FBA1-associated but WHO12 is not . The WHO13 sequences are all pseudogenes and were detected only in a small clade of Kazachstania species , downstream of FBA1 . Interestingly , these species have an intron in FBA1 at precisely the inferred WHO endonuclease cleavage site ( Figure 5—figure supplement 1 ) , and this is the only intron in any budding yeast FBA1 gene . In other eukaryotes , evolutionarily novel introns are gained at sites of double-strand DNA breakage ( Li et al . , 2009 ) . The position of WHO12 as an outgroup to all the other WHO families ( Figure 5 ) raises the possibility that the common ancestor of all the WHO families might have used a different gene as its original target , before changing target to FBA1 after the WHO12 family separated from the others . The only intact WHO12 gene occurs in Naumovozyma castellii , where it is located beside GLK1 ( glucokinase ) . We sequenced the genomes of four strains of this species and found two with intact WHO12 , and two with frameshifted WHO12 pseudogenes . There was no structural polymorphism or high sequence divergence at this locus in N . castellii , and no GLK1 gene fragments , so no evidence of active homing . The other WHO12 sequences are pseudogenes in K . lactis ( Fabre et al . , 2005 ) and two Lachancea species ( Figure 1—figure supplement 1 ) , beside a CAO copper amine oxidase gene in all three cases . The K . lactis WHO12 pseudogene lies between full-length CAO and a damaged fragment of CAO . It is therefore possible that WHO12 is an active element targeting CAO , and that CAO pre-dates FBA1 as the target of the whole WHO superfamily , but given the rarity of WHO12 it seems more likely that FBA1 is the ancestral target and WHO12 is a family that lost its specificity for FBA1 more recently , similar to WHO10 . In summary , the phylogenetic tree indicates that WHO genes have been located beside FBA1 throughout most of their diversification . The WHO genes that are not now located beside FBA1 are descended from FBA1-linked ancestors . The patchy taxonomic distribution of WHO genes suggests that they are native to the genus Torulaspora and/or Lachancea and have probably been transmitted between these genera by horizontal gene transfer . Horizontal transfer among budding yeast species has previously been inferred for the VDE intein ( Koufopanou et al . , 2002; Okuda et al . , 2003 ) . The split between the zinc finger-containing proteins HO and WHO pre-dates the diversification of the WHO families . The WHO10 family has become amplified in T . globosa . In the two strains that we sequenced , strain-specific insertions of intact WHO10 genes are present at five loci not linked to FBA1 ( Figure 6 ) . The five Who10 proteins have only 72–80% amino acid sequence identity to one another . The T . globosa WHO10 genes are located within regions of inserted DNA 2 . 8–6 . 6 kb long . In two cases , the inserted DNA includes a duplicated fragment of the 3’ end of a host gene ( BXI1 and E1850 ) at one end , so that the host gene was not disrupted ( Figure 6A , B ) . These duplications resemble the FBA1 fragments seen downstream of FBA1 in many species ( Figure 1A ) . In the other three cases , the WHO10-containing insertion interrupts a host gene without forming a duplication , probably inactivating it ( Figure 6C , D , E ) . All five T . globosa insertions include an FBA1 fragment immediately downstream of the WHO10 gene , even though the insertions are not near the FBA1 locus . Some insertions also contain fragments of various other genes , mostly from their 3’ ends ( Figure 6 ) . The structure and variable location of the DNA insertions indicate that WHO10 genes are part of a mobile genetic element that is active in T . globosa . The mobile element consists of WHO10 and the gene fragments , which may be molecular fossils of previous host sites into which the element inserted . The genomic evidence indicates that most WHO genes function as part of a homing genetic element that targets the FBA1 locus ( Figure 4 ) . However , in T . globosa the WHO10 family has lost its specificity for FBA1 and become a more general mobile element rather than a homing element , resulting in the proliferation of intact WHO10 genes to multiple other sites in the genome . How the WHO10 genes become integrated into the non-FBA1 sites in T . globosa is not clear , because there are no homologous flanking sequences to guide integration of WHO10 into a double-strand break .
We have shown that WHO elements are homing genetic elements in the budding yeast genera Torulaspora and Lachancea , that primarily target the aldolase gene FBA1 . They have diversified into a large family with very divergent endonuclease genes . Our model proposes that WHO elements home into sensitive alleles of FBA1 by using a duplicated fragment of the 3’ end of FBA1 as a second region of homology downstream of the WHO gene ( Figure 4A , left column ) . Homing replaces the 3’ end of FBA1 , making it resistant to cleavage by the element’s WHO endonuclease . The DNA manipulation steps in WHO’s homing mechanism are identical to those that occur during VDE intein homing into VMA1 , but the gene organization of WHO elements and their relationship to the host gene differ substantially from VDE ( Figure 4A , first two columns ) . Resistance to endonuclease cleavage in FBA1 comes from allelic sequence differences , whereas resistance in VMA1 comes directly from interruption of the cleavage site by the VDE element ( Gimble and Thorner , 1992 ) . FBA1 can be described as the host gene for the WHO element , even though the element lies downstream of FBA1 rather than interrupting it . This structural organization makes WHO elements different from the two currently recognized classes of homing genetic elements , which are inteins and intron-encoded homing endonucleases ( Belfort et al . , 2005; Belfort , 2017 ) . In both of these other classes the homing element is a self-splicing entity , transcribed as an internal part of the host gene , that must be removed ( by mRNA or protein splicing ) in order to express the mature host protein . In contrast , WHO genes are transcribed independently of FBA1 ( some of them are in the opposite orientation to FBA1; Figure 1A ) , and FBA1 will remain functional even if the WHO gene becomes a pseudogene . WHO elements therefore constitute a third structural class of homing element , and the only one with a propensity to form clusters . The mechanism of action of WHO elements has altered the evolutionary trajectory of their host gene FBA1 , disrupting the normal vertical inheritance of this gene and leading to a chimeric mode of evolution in which the two ends of the gene have different histories ( Figure 3 ) . The reason why WHO elements chose FBA1 as their host gene is probably that aldolase is absolutely required for spore formation , due to its role in gluconeogenesis ( Dickinson and Williams , 1986 ) . Meiosis and sporulation require cells to be grown on a non-fermentable carbon source such as acetate , and in these conditions gluconeogenesis is necessary to make the glucose monomers used for synthesis of the polysaccharide layers of the spore wall , a late stage in the meiosis-sporulation pathway ( Neiman , 2005; Walther et al . , 2014 ) . S . cerevisiae fba1 mutants cannot make spores ( Lobo , 1984; Dickinson and Williams , 1986 ) but they should not be blocked in meiosis , which is when WHO element homing is expected to occur . It is unlikely that the FBA1 genes in either the donor or the recipient chromosome can be transcribed at the same time as DNA cleavage and recombination is occurring during homing . By temporarily inactivating FBA1 , the WHO element may be able to delay the cell from progressing from meiosis into sporulation until homing has finished . Homing is likely to be a slow process , because mating-type switching takes more than an hour ( Lee and Haber , 2015 ) . During mating-type switching ( Figure 4A , right column ) , HO initiates a series of DNA manipulation steps that closely resemble the steps that occur during homing of WHO elements and the VDE intein . Together with the sequence similarity among the three proteins , this similarity of the molecular mechanisms indicates a shared evolutionary origin of the three processes . While the mechanisms of WHO and VDE homing are essentially identical , the mechanism of HO action at the MAT locus has diverged from them in two critical ways . First , the HO gene is not part of the template used for DNA repair . Second , switching occurs in haploids , whereas homing occurs in diploids during meiosis . There is no homologous chromosome for the cleaved MAT locus to interact with , so instead it interacts with HML or HMR . Our results finally illuminate the origin of HO endonuclease . Based on the fact that WHO and HO share features that are otherwise unique – the presence of a zinc finger domain and the absence of exteins – we propose the following evolutionary model ( Figure 7 ) . ( 1 ) An intein from a bacterial source invaded the VMA1 gene of an early budding yeast species to become VDE . ( 2 ) VDE subsequently duplicated and mis-homed into a zinc finger protein gene ( possibly a paralog of ASH1 ) , close to the 5’ end , to make a fusion gene that was the common ancestor of WHO and HO . The zinc finger directed the endonuclease to new target gene ( s ) in the genome . ( 3 ) The fusion gene became located between the target gene and a duplicated fragment of the target gene , forming a proto-WHO element . This step resembles some of the WHO10 insertion sites seen in T . globosa . The target gene may have been FBA1 , or possibly a different , unknown , gene . To function as a homing element , the proto-WHO element must have had a meiosis-specific promoter . ( 4 ) The proto-WHO element diversified and spread through yeast populations and into additional species , with FBA1 as its main target . Occasional mis-homing events spread the element into new targets such as the WHO10 locations in T . globosa . ( 5 ) At an early stage of diversification , a WHO endonuclease developed an ability to cleave MATα1 , in a species that already contained a three-locus MAT/HML/HMR mating-type switching system , and became domesticated as HO . During domestication , the transcriptional regulation of the gene must have changed from meiosis-specific expression to haploid-specific expression , as well as gaining cell lineage and cell cycle constraints ( Stillman , 2013 ) . The boundary between the Y and Z regions of the MAT locus , which was previously variable among species , became permanently fixed at the site where the endonuclease cleaved MATα1 ( Figure 4A; Hanson and Wolfe , 2017 ) . Many examples are known of mobile genetic element genes that have been domesticated to take on a new role in the cell ( Volff , 2006 ) . In some of these examples , a domesticated endonuclease gene has retained its nucleolytic activity and functions in a programmed genome rearrangement process , such as the RAG1 gene in V ( D ) J recombination in the immune system of jawed vertebrates ( Huang et al . , 2016 ) , and the PiggyMac gene in elimination of germline sequences during development of the macronucleus in Paramecium ( Baudry et al . , 2009 ) . In other examples the ability to cleave DNA has been lost , such as in the bacterial DUF199/WhiA family which originated as a LAGLIDADG endonuclease but is now a regulator of transcription ( Kaiser et al . , 2009 ) . HO has retained its endonuclease activity , and its origin from a homing element may help explain some unusual properties of this protein in vitro , such as its extreme catalytic inefficiency and its ability to attach to both ends of linear DNA molecules , forming loops visible by electron microscopy ( Jin et al . , 1997 ) . The domestication of a WHO element to become HO is similar to the domestication of the transposon-derived genes KAT1 and α3 to act as generators of double-strand breaks at the MAT locus during mating-type switching in K . lactis ( Barsoum et al . , 2010; Rusche and Rine , 2010; Rajaei et al . , 2014 ) . In all three cases , a mobile element gene was domesticated in a genome that already had a three-locus MAT/HML/HMR arrangement and probably switched mating types by a passive process based on homologous recombination without a specific mechanism for making a double-strand break at MAT . Why were mobile genetic elements repeatedly recruited into these switching systems ? If we consider that , in any population of haploid cells , ( 1 ) mating-type switching can only increase a cell’s probability of mating , ( 2 ) mating leads to the formation of a diploid and inevitably to meiosis , even if many vegetative generations later , and ( 3 ) homing genetic elements can only home during meiosis , it logically follows that it is in a homing genetic element’s self-interest to increase the frequency of mating-type switching ( Hanson and Wolfe , 2017 ) . Thus , a WHO element could increase its rate of spread into empty FBA1 alleles in a population , if its WHO protein developed a secondary activity of cleaving the MAT locus as well as cleaving FBA1 . Importantly , haplontic species such as Torulaspora and Lachancea require a nutritional signal to mate , so mating-type switching is unlikely to be followed by immediate mating of switched cells with their clonal relatives as occurs in S . cerevisiae . Instead , we expect that a switched cell in a haplontic species could go through many cycles of mitotic division before mating , which increases the opportunity for switched cell lineages to disperse and outbreed , and therefore increases the opportunity for homing elements to spread . Since frequent and accurate switching are probably favored by natural selection ( Hanson et al . , 2014 ) , the subsequent steps that domesticated a WHO element to form the non-mobile and exquisitely regulated gene HO ( Stillman , 2013 ) would also have been advantageous .
Genomes analyzed in this study are listed in Supplementary files 1 and 2 . New genome sequences were obtained as follows . T . delbrueckii strains L09–L20 were from the strain collection of Lallemand Inc ( Montréal , Canada ) , generously provided by Dr . Caroline Wilde . They were sequenced at the University of Missouri core facility ( Illumina , SE 1 × 50 bp ) . T . delbrueckii strain NCYC696 data were downloaded from opendata . ifr . ac . uk/NCYC on 23-Feb-2017 as unassembled Illumina sequence reads ( PE 2 × 100 bp ) . T . globosa strains CBS764T and CBS2947 were purchased from the Westerdijk Institute ( Netherlands ) and sequenced using both Illumina ( PE 2 × 150 bp; BGI Tech Solutions , Hong Kong ) and Pacific Biosciences Sequel technologies ( 1 SMRT cell; Earlham Institute , Norwich , UK ) . T . pretoriensis strain CBS2187T ( Illumina , PE 2 × 100 bp + 6 kb library MP 2 × 100 bp ) and T . franciscae strain CBS2926T ( Illumina , PE 2 × 100 bp ) were sequenced and assembled at INRAE Montpellier . Eight other T . pretoriensis strains ( CBS2785 , CBS5080 , CBS9333 , CBS11100 , CBS11121 , CBS11123 , CBS11124 from the Westerdijk Institute , and UWOPS 83–1046 . 2 from M . A . Lachance , University of Western Ontario ) were sequenced at the Earlham Institute using their proprietary LITE protocol for the Illumina platform . Zygotorulaspora mrakii strain NRRL Y-6702T was obtained from the USDA Agricultural Research Service ( Peoria , IL , USA ) and sequenced at the Earlham Institute using both Pacific Biosciences RSII ( 4 SMRT cells ) and Illumina LITE methods . Naumovozyma castellii strains Y056 , Y174 , Y287 and Y668 were gifts from Prof . Jure Piškur ( Lund University , Sweden ) and were sequenced at the Earlham Institute using the Illumina LITE method . Cultures were grown under standard rich-medium conditions . DNA for Illumina sequencing was harvested from stationary-phase cultures by homogenization with glass beads followed by phenol-chloroform extraction and ethanol precipitation . Purified DNA was concentrated with the Genomic DNA Clean and Concentrator-10 ( Zymo Research , catalog D4010 ) . DNA for PacBio sequencing was prepared as in Ortiz-Merino et al . ( 2017 ) . Illumina data were assembled using SPAdes version v3 . 11 . 1 ( Bankevich et al . , 2012 ) . PacBio data were assembled using HGAP3 ( Chin et al . , 2013 ) . Other genome sequences used in this study were taken from the NCBI database . The previously published genome sequences for Torulaspora species are from Gordon et al . ( 2011 ) , Gomez-Angulo et al . ( 2015 ) , Tondini et al . , 2018 , Galeote et al . ( 2018 ) and Shen et al . ( 2018 ) ; Lachancea species are from Souciet et al . ( 2009 ) , Sarilar et al . ( 2015 ) , Vakirlis et al . ( 2016 ) , Freel et al . ( 2016 ) and Kellis et al . ( 2004 ) ; and Kluyveromyces species are from Dujon et al . ( 2004 ) and Varela et al . ( 2019 ) . S . cerevisiae strains in which the coding region ( ORF ) of T . delbrueckii FBA1 was integrated into the S . cerevisiae ADE2 gene , in opposite orientation to ADE2 so that it is not functional , were constructed using CRISPR-Cas9 as follows . The ADE2-targeting sgRNA ADE2 . Y from DiCarlo et al . ( 2013 ) was synthesized as a gene fragment by Integrated DNA Technologies , and inserted into the sgRNA plasmid pMEL13 from Mans et al . ( 2015 ) by restriction digestion and ligation . This plasmid was then transformed into S . cerevisiae strain IMX585 expressing Cas9 ( Mans et al . , 2015 ) , together with a repair template containing the T . delbrueckii FBA1 ORF ( TdFBA1-S or TdFBA1-R allele ) flanked with homology to S . cerevisiae ADE2 in reverse orientation ( bases 564456 . . 564832 and 565952 . . 566366 of S . cerevisiae chromosome XV ) . Sequences of the ade2::TdFBA1-S and ade2::TdFBA1-R constructs are given in Supplementary file 3 . Transformants were selected on YPAD ( YPD ( Formedium ) supplemented with 40 μg/ml adenine sulfate ( Sigma ) ) containing 200 μg/ml G418 . ADE2 knockouts were identified by formation of red colonies . Successful integrants were confirmed by PCR amplification of the ade2::TdFBA1 locus and Sanger sequencing ( Eurofins ) . Two replicate ade2::TdFBA1-S strains were designated C1 and C4 , and two replicate ade2::TdFBA1-R strains were designated L1 and L3 . Replicating plasmids constitutively expressing Who6 , Who6-HA , or GFP-HA were constructed in the panARS replicating vector pIL75 ( Liachko and Dunham , 2014 ) containing a KanMX marker . The nucleotide sequences of these plasmids ( pWHO6 , pWHO6-HA , pGFP-HA ) are given in Supplementary file 3 . In pWHO6 , the WHO6 gene from T . delbrueckii strain L09 was placed under the control of the promoter and terminator of the T . delbrueckii glyceraldehyde-3-phosphate dehydrogenase gene TDH3 ( TDEL0E04750 ) . These regions were amplified by PCR from T . delbrueckii genomic DNA using high fidelity polymerase ( New England Biolabs , M0492S ) and inserted into pIL75 by restriction digestion and ligation . Plasmid pWHO6-HA is identical to pWHO6 except that its WHO6 gene is fused to a C-terminal 3xHA tag . A similar control plasmid ( pGFP-HA ) was made containing a GFP gene fused to 3xHA tag ( GBlock made by Integrated DNA Technologies and inserted into pIL75 by restriction digestion and ligation ) , under the control of the T . delbrueckii TDH3 promoter and terminator . The S . cerevisiae ade2::TdFBA1 strains ( C1 , C4 , L1 , L3 ) as described above were transformed with the plasmids pWHO6 , pWHO6-HA , or pGFP-HA . Multiple independent transformations were made to ensure that all gene conversion or NHEJ events recovered were independent . First , pWHO6-HA and pGFP-HA were each transformed into strains C1 , C4 , L1 and L3 , and the whole genomes of these eight strains were sequenced , resulting in the unexpected discovery of gene conversion at the ade2::TdFBA1 locus when C1 and C4 were transformed with pWHO6-HA ( called survivors 11 and 10 , respectively , in Figure 2B ) . Genome sequencing was done by BGI Tech Solutions using a BGISEQ instrument with 50 bp single-end reads . Second , pWHO6 ( no 3xHA tag ) and pGFP-HA were each transformed 10 times into strains C4 and L1 . The ade2::TdFBA1 locus from each transformant was amplified by PCR with a high-fidelity polymerase ( New England Biolabs , M0492S ) and Sanger sequenced . Primers for amplification were TGACCACGTTAATGGCTCC and CACCAGCTCCAGCGATAATTG . For transformation , S . cerevisiae cells were grown overnight in liquid cultures of YPAD . Cultures were reinoculated in 50 ml and grown to mid-log phase . Cells were incubated with 1M LiAc , 50% PEG , salmon sperm DNA and plasmid DNA for 30 min , and then heat shocked at 42°C for 15 min . Transformants were selected on YPAD containing 200 μg/ml G418 . Only one colony was picked from each transformation plate . The number of colonies obtained on the plates with the combination C4 + pWHO6 was dramatically lower ( approximately 100-fold ) than on the three other combinations . Individual colonies were grown in liquid YPAD overnight , before genomic DNA was harvested using a QIAamp DNA Mini kit ( Qiagen ) . | In the same way as a sperm from a male and an egg from a female join together to form an embryo in most animals , yeast cells have two sexes that coordinate how they reproduce . These are called “mating types” and , rather than male or female , an individual yeast cell can either be mating type “a” or “alpha” . Every yeast cell contains the genes for both mating types , and each cell’s mating type is determined by which of those genes it has active . Only one mating type gene can be ‘on’ at a time , but some yeast species can swap mating type on demand by switching the corresponding genes ‘on’ or ‘off’ . This switch is unusual . Rather than simply activate one of the genes it already has , the yeast cell keeps an inactive version of each mating type gene tucked away , makes a copy of the gene it wants to be active and pastes that copy into a different location in its genome . To do all of this yeast need another gene called HO . This gene codes for an enzyme that cuts the DNA at the location of the active mating type gene . This makes an opening that allows the cell to replace the ‘a’ gene with the ‘alpha’ gene , or vice versa . This system allows yeast cells to continue mating even if all the cells in a colony start off as the same mating type . But , cutting into the DNA is risky , and can damage the health of the cell . So , why did yeast cells evolve a system that could cause them harm ? To find out where the HO gene came from , Coughlan et al . searched through all the available genomes from yeast species for other genes with similar sequences and identified a cluster which they nicknamed “weird HO” genes , or WHO genes for short . Testing these genes revealed that they also code for enzymes that make cuts in the yeast genome , but the way the cell repairs the cuts is different . The WHO genes are jumping genes . When the enzyme encoded by a WHO gene makes a cut in the genome , the yeast cell copies the gene into the gap , allowing the gene to ‘jump’ from one part of the genome to another . It is possible that this was the starting point for the evolution of the HO gene . Changes to a WHO gene could have allowed it to cut into the mating type region of the yeast genome , giving the yeast an opportunity to ‘domesticate’ it . Over time , the yeast cell stopped the WHO gene from jumping into the gap and started using the cut to change its mating type . Understanding how cells adapt genes for different purposes is a key question in evolutionary biology . There are many other examples of domesticated jumping genes in other organisms , including in the human immune system . Understanding the evolution of HO not only sheds light on how yeast mating type switching evolved , but on how other species might harness and adapt their genes . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology",
"genetics",
"and",
"genomics"
] | 2020 | The yeast mating-type switching endonuclease HO is a domesticated member of an unorthodox homing genetic element family |
Exercise induces beneficial responses in the brain , which is accompanied by an increase in BDNF , a trophic factor associated with cognitive improvement and the alleviation of depression and anxiety . However , the exact mechanisms whereby physical exercise produces an induction in brain Bdnf gene expression are not well understood . While pharmacological doses of HDAC inhibitors exert positive effects on Bdnf gene transcription , the inhibitors represent small molecules that do not occur in vivo . Here , we report that an endogenous molecule released after exercise is capable of inducing key promoters of the Mus musculus Bdnf gene . The metabolite β-hydroxybutyrate , which increases after prolonged exercise , induces the activities of Bdnf promoters , particularly promoter I , which is activity-dependent . We have discovered that the action of β-hydroxybutyrate is specifically upon HDAC2 and HDAC3 , which act upon selective Bdnf promoters . Moreover , the effects upon hippocampal Bdnf expression were observed after direct ventricular application of β-hydroxybutyrate . Electrophysiological measurements indicate that β-hydroxybutyrate causes an increase in neurotransmitter release , which is dependent upon the TrkB receptor . These results reveal an endogenous mechanism to explain how physical exercise leads to the induction of BDNF .
It has been traditionally thought that physical activity and the processes of learning and memory formation are independent and carried out by different organ systems . However , from an evolutionary perspective , these processes needed to be tightly intertwined to ensure the survival of animal species . Indeed , physical effort usually occurred in response to an imminent danger . Responding to that danger not only required running , but also necessitated better functioning of the brain through increased plasticity in order to adapt to new sources of stress , to learn to avoid dangers or better respond to them , and to map surroundings and learn the locations of hazards ( Noakes and Spedding , 2012 ) All of these responses require improved memory . Physical exercise produces many benefits in the brain that enhance cognitive function , blood flow and resistance to injury . One mechanism to account for the changes in brain plasticity is through the action of growth factors ( Cotman et al 2007 ) . A major contributor to the processes of learning and memory formation involves brain derived neurotrophic factor ( BDNF ) signaling pathways . It has been known for over two decades that physical activity or neuronal activity markedly enhances Bdnf gene expression in the brain ( Isackson et al . , 1991; Neeper et al . , 1995 ) and that this increase in BDNF protein leads to activation of signaling pathways that result in exercise-dependent enhanced learning and memory formation ( Vaynman et al . , 2004 ) . Though these results are widely recognized , it is important to note that very little is known about the molecular mechanisms that link exercise and Bdnf expression . Regulation of Bdnf expression occurs by many means , but how exercise influences the expression of trophic factors is not understood . In this paper , we are interested in understanding how physical exercise induces Bdnf gene expression . This is a significant question , since cognitive ability and synaptic plasticity are influenced by the levels of BDNF ( Lu et al . , 2013; Park and Poo , 2013; Vaynman et al . , 2004 ) and BDNF signaling is reduced in many neurodegenerative and psychiatric diseases ( Autry and Monteggia , 2012; Zuccato and Cattaneo , 2009 ) . During development , BDNF is required for the survival of specific neuronal populations and it participates in axonal and dendritic growth and synaptogenesis ( Alsina et al . , 2001; Bibel and Barde , 2000 ) . A number of studies have indicated that decreased levels of BDNF are associated with depression and become enhanced following antidepressant treatment ( Duman and Monteggia , 2006; Martinowich et al . , 2007 ) . Moreover , exercise frequently leads to an increase in BDNF in the central nervous system to promote improvement in cognitive ability and depressive-like behavior ( Marais et al . , 2009; Russo-Neustadt et al . , 2000 ) . Indeed , physical activity has been shown to have anti-depressant effects and to improve outcomes in animal models and for patients with neurodegenerative diseases such as Parkinson’s Disease ( Frazzitta et al . , 2014 ) or Alzheimer’s disease ( Smith et al . , 2014 ) . As a result , by understanding the molecular mechanisms by which exercise induces Bdnf expression , we aim to harness the therapeutic potential of physical exercise and eventually identify novel therapeutic targets for both psychiatric and neurodegenerative diseases . In animal models , exercise induces Bdnf mRNA expression in multiple brain regions ( Cotman et al . , 2007 ) , most prominently in the hippocampus . BDNF production provides trophic support and increases in synaptogenesis and dendritic and axonal branching and spine turnover . Blocking BDNF signaling attenuates the exercise-induced improvement of spatial learning tasks ( Vaynman et al . , 2004 ) , as well as the exercise-induced expression of synaptic proteins ( Vaynman et al . , 2006 ) . However , how BDNF is selectively increased after physical activity-dependent changes in the nervous system is not well understood . One mechanism that has been proposed is that exercise induces Bdnf expression through the induction of expression of Fndc5 ( Wrann et al . , 2013 ) , a PGC-1α-dependent myokine . This hypothesis proposes that the FNDC5 protein is cleaved into a small circulating protein called irisin , which has been associated with the browning of fat ( Bostrom et al . , 2012 ) . However , there are contradictory reports about whether Fndc5 is translated and expressed at high levels after exercise and whether irisin is produced and found in blood ( Albrecht et al . , 2015; Jedrychowski et al . , 2015 ) . This raises questions about how and whether a myokine can be responsible for the induction of Bdnf gene regulation . A second hypothesis is that exercise may induce BDNF levels by altering the epigenetic landmarks of the Bdnf promoters ( Guan et al . , 2009; Koppel and Timmusk , 2013 ) . Because exercise induces metabolic changes and because epigenetics lies at the interfaces between the environment and changes in gene expression , it is conceivable that an endogenous molecule is produced after exercise , which can serve as a metabolite as well as a regulator of Bdnf transcription . In this paper , we provide a mechanism demonstrating that exercise induces the accumulation of a ketone body ( D-β-hydroxybutyrate or DBHB ) in the hippocampus , where it serves both as an energy source and an inhibitor of class I histone deacetylases ( HDACs ) to specifically induce BDNF expression .
To assess how exercise enhances Bdnf gene expression , we established a voluntary running protocol for mice ( 4 weeks of age ) , which has been previously shown to mediate increases in BDNF ( Marlatt et al . , 2012 ) . Mice were individually housed in cages and divided into two groups: control and exercise . Each mouse in the exercise group was provided with a running wheel , whereas the control group mice were not . Mice were not forced to run on the wheels , but rather were allowed to run voluntarily . Long term running for up to 30 days was monitored ( Figure 1A ) for distance and time . We chose this exercise model because of its voluntary aspect . Stress has been reported to decrease BDNF levels ( Murakami et al . , 2005 ) . For that reason , we avoided any handling or stress induction in the mice . Animals were sacrificed and tissues were collected for analysis . To determine the effects of voluntary running in the hippocampus , we initially focused our study on selective promoters from the Bdnf gene , which has a complex structure containing multiple promoters that generate many transcripts with a common coding exon ( Pruunsild et al . , 2011 ) . Promoter I ( pI ) , a neuronal activity dependent promoter ( Tabuchi et al . , 2002 ) and other nearby promoters ( pIIA , pIIB , pIV ) ( Pruunsild et al . , 2011 ) , were also assessed for mRNA expression . Our results showed that voluntary exercise for 4 weeks significantly induced BDNF promoter I and II expression in the hippocampus ( Figure 1B and data not shown ) . Western blot analysis of the hippocampus of animals after voluntary running showed a significant increase of mature BDNF protein levels , as compared to control mice ( Figure 1C and D ) . 10 . 7554/eLife . 15092 . 003Figure 1 . The experimental design and time course of the exercise regime is shown . Exercise induces changes in brain BDNF levels in a voluntary exercise protocol ( 16 ) . ( A ) Experimental design for the Voluntary exercise model . ( B ) Voluntary exercise for 4 weeks significantly induces Bdnf promoter I expression in the hippocampus as measured by real-time RTPCR . The number of animal used for each group ( control and exercise ) is 10 . *p<0 . 05 as measured by unpaired t-test . ( C ) Western blot analysis depicting the increase in mature BDNF protein levels in the hippocampus of exercise animals as compare to wild type . In this representative image , the BDNF levels from 2 control hippocampal lysates and 3 exercise hippocampal lysates are depicted . This experiment was replicated from additional 3 different animals in each group . ( D ) Quantification of the BDNF western blot . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 003 One epigenetic mechanism that has been proposed for inducing Bdnf gene expression is histone acetylation ( Koppel and Timmusk , 2013 ) . A number of studies have reported that blocking HDACs enhances memory formation and the expression of synaptic plasticity genes , such as Creb , Bdnf and CamkII ( Guan et al . , 2009; Koppel and Timmusk , 2013 ) . HDACs catalyze the deacetylation of histones and are divided into four classes . Class I ( HDAC1 , HDAC2 , HDAC3 and HDAC8 ) , II and IV are zinc dependent enzymes and are represented by at least 11 different proteins , whereas class III includes the sirtuins ( Haberland et al . , 2009 ) . We have found that the general class I/IIb HDAC inhibitor , SAHA ( vorinostat ) , a clinically approved anticancer agent and memory enhancer ( Guan et al . , 2009 ) and TSA ( trichostatin A ) , a broad-spectrum inhibitor , were effective in inducing Bdnf mRNA in primary cortical neuron cultures ( Figure 2A ) . More specific HDAC inhibitors , such as CI-994 and MS-275 , that target class I HDAC members were also capable of elevating Bdnf transcripts ( Figure 2B ) , whereas a non-active analog of CI-994 , BRD4097 ( manuscript submitted ) , did not give a response . While these results implicate class I HDAC inhibition in the regulation of synaptic plasticity genes , such as Bdnf , it is not clear how exercise is translated into increases in BDNF . Interestingly , we found that exercise reduces Hdac2 and Hdac3 mRNA levels , but not Hdac1 levels in the hippocampus ( Figure 3A ) . This is consistent with the observation that HDAC2 , but not HDAC1 , is more important in binding to promoters of activity-related genes ( Guan et al . , 2009 ) . These results suggest that exercise may modulate Bdnf gene expression in the hippocampus by inducing changes to the epigenetic landscape of its promoter . To test this possibility , we performed chromatin immunoprecipitation experiments and showed that the binding of both HDAC2 and HDAC3 to the Bdnf promoter in hippocampi of exercise animals was decreased , in comparison to sedentary animals ( Figure 3B and data not shown ) . Even though some variation was detected in the basal binding of HDAC3 to the Bdnf pI promoter , the overall data is consistent with exercise inducing a significant decrease in HDAC3 binding to the Bdnf promoter . As a result , these results focused our search for endogenous metabolites that are produced in response to physical activity and that may also act as an epigenetic modulator . 10 . 7554/eLife . 15092 . 004Figure 2 . HDAC inhibitors induce Bdnf expression . ( A ) Broad spectrum HDAC inhibitors such as TSA ( 0 . 67 μM ) and SAHA ( 5 μM ) induce coding and pI Bdnf expression levels as measured by real-time RTPCR . For the coding promoter , the n number for controls , TSA and SAHA treatments are 6 , 5 and 4 respectively . For the pI promoter , the n number for controls , TSA and SAHA treatments are 5 , 5 and 4 respectively . Each replicate consisted of primary neurons obtained from different set cultures and treated with fresh dilutions of the compounds . Significance was measured by 1way anova **p< 0 . 01 . and ***p<0 . 001 . ( B ) Treatment with class I HDAC inhibitors such as CI-994 ( 10 μM ) and MS-275 ( 10 μM ) induce coding and pI Bdnf levels , whereas the negative control analog for CI-994 , BRD4097 does not as measured by real-time RTPCR . For the coding promoter , the n number for controls , CI-994 , BRD4097 and MS-275 treatments are 6 , 6 , 5 and 5 respectively . For the pI promoter , the n number for controls , CI-994 , BRD4097 and MS-275 are 5 , 6 , 5 and 5 respectively . Each replicate consisted of primary neurons obtained from different cultures and treated with fresh dilutions of the compounds . Significance was measured by 1way anova **p<0 . 01 . and ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 00410 . 7554/eLife . 15092 . 005Figure 3 . Exercise affects Hdac expression and binding to the hippocampal BDNF promoter . ( A ) Exercise induces significant decreases in both Hdac2 and Hdac3 expression in the hippocampus without affecting Hdac1 expression as measured by real-time RTPCR . The expression was analyzed from the hippocampi of 4 different control and exercise animals . Unpaired t-tests were used to measure statistical significance *p<0 . 05 . ( B ) Exercise induces decreases in HDAC3 binding to the Bdnf promoter as measured by chromatin immunoprecipitation followed by real-time RTPCR . The experiment was conducted by using tissues from 7 different control and exercise animals . Unpaired t-tests were used to measure statistical significance *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 005 Exercise is accompanied with increases in energy requirements leading to mitochondrial biogenesis , as well as higher metabolism and oxygen consumption . For example , exercise is often associated with increases in ketone body production . The ketone body D-β-hydroxybutyrate ( DBHB ) is a major energy metabolite that is increased in the liver after prolonged exercise . DBHB levels are frequently increased after caloric restriction , fasting and ketogenic diets and DBHB is believed to serve as a signaling molecule in response to metabolic changes ( Newman and Verdin , 2014 ) . DBHB is synthesized from acetyl-CoA generated from the β-oxidation of fatty acids in the liver . As DBHB production is sensitive to environmental factors , nutritional states and energy levels , it has been hypothesized that ketone bodies may serve as an intermediary to regulate gene expression and chromatin structure . Interestingly , DBHB is transported in the blood stream to the brain where it can serve as an energy source . In addition , DBHB also has been shown to act as a class I HDAC inhibitor in non-neuronal tissues ( Shimazu et al . , 2013 ) . To assess the effects of voluntary running , we measured DBHB levels in the hippocampus following exercise . We found that an increase in DBHB levels occurred in mice after exercise , compared to mice not subjected to voluntary running ( Figure 4A ) . 10 . 7554/eLife . 15092 . 006Figure 4 . Exercise increases DBHB levels in the hippocampus . DBHB in turn can induce Bdnf expression in vitro in neurons and in vivo in the hippocampus ( A ) Exercise induced DBHB levels in the hippocampus . The number for controls and exercise hippocampi is 10 and 11 respectively . Statistical significance was analyzed by the unpaired t-test *p<0 . 05 . DBHB amounts are expressed as millimole per gram of total hippocampal protein ( B ) Overnight treatment of DIV6 primary cortical neurons with 5 mM of DBHB significantly induces coding and pI Bdnf expression levels as measured by real-time RTPCR . For both promoters , the n number for controls and DBHB are 4 and 4 respectively . Each replicate consisted of primary neurons obtained from different cultures and treated with fresh dilutions of the compounds . Significance was measured by unpaired t-test *p<0 . 05 and **p<0 . 01 . ( C ) Three hour treatment of hippocampal slices with DBHB ( 0 . 8 mM ) induces coding and pI Bdnf levels as measured by real-time RTPCR . For both promoters , the n number for controls and DBHB are 5 and 5 respectively . Each replicate consisted of slices obtained from different animals and treated with fresh dilutions of the compounds . Significance was measured by unpaired t-test *p<0 . 05 and **p<0 . 01 . ( D ) Intraventricular delivery of DBHB ( 2 mM ) significantly induces coding and pI Bdnf levels as measured by real-time RTPCR . The n number for controls and DBHB are 3 and 5 respectively . Each replicate consisted of hippocampi from different animals that were subjected to the surgical procedure . Significance was measured by unpaired t-test *p<0 . 05 . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 006 To evaluate if DBHB had an effect upon Bdnf expression , we treated cortical neurons , which represent an abundant source of activity-dependent BDNF . Overnight treatment with DBHB significantly induced the coding and pI-driven Bdnf transcripts ( Figure 4B ) , consistent with the effects of exercise . Treatment of hippocampal slices with DBHB also gave similar results ( Figure 4C ) . To test the ability of DBHB to induce Bdnf in an in vivo setting , we administered exogenous DBHB intraventricularly in mice . Gene expression analysis indicated an increase specifically in Bdnf promoter I activity in the hippocampi of mice receiving the intraventricular injection of DBHB as compared to saline injections ( Figure 4D ) . One possible explanation for the effects of exercise is that it induces DBHB accumulation in the hippocampus , which in turn inhibits class I HDACs such as HDAC2 and HDAC3 and leads to induction of Bdnf expression . To determine whether DBHB induction of Bdnf represents an HDAC-dependent mechanism , we tested whether DBHB treatment affected HDAC2 and HDAC3 occupancy at the Bdnf promoters . We find that treatment of primary neurons with DBHB led to decreased HDAC2 and HDAC3 binding to the Bdnf promoters ( Figure 5A ) . This was consistent with an increase in levels of acetylated histone H3 after DBHB treatment ( Figure 5B ) . Because DBHB is a potent inhibitor of HDAC3 ( Shimazu et al . , 2013 ) and HDAC2 has been previously shown to repress BDNF gene expression ( Guan et al . , 2009 ) , we focused our attention on HDAC3 . Inactivation of HDAC3 either by treating cortical neurons with a HDAC3 selective HDAC inhibitor BRD3308 ( Barton et al . , 2014 ) Wagner et al . , 2016 ( Figure 5C ) or through knockdown using shRNA targeting HDAC3 ( Figure 5D , E , & F , E ) both led to increases in Bdnf expression . Taken together , these results suggest that inhibition of HDAC2 and HDAC3 in the hippocampus induces Bdnf expression . 10 . 7554/eLife . 15092 . 007Figure 5 . DBHB induces Bdnf expression by inhibiting HDAC2 and HDAC3 . ( A ) DBHB ( 2 mM or 5 mM ) treatment decreased the binding of both HDAC2 and HDAC3 on the pI promoter of Bdnf as measured by chromatin immunoprecipitation followed by real-time RTPCR . The number of chromatin immunoprecipitations for each HDAC for each treatment were 3 ( control ) , 3 ( DBHB 2 mM ) and 2 ( DBHB 5 mM ) . Statistical significance was measured by 1way anova ****p<0 . 001 . ( B ) HISTONE H3 acetylation is increased in neurons upon treatment with different doses of DBHB . Representative blot shown in the figure . ( C ) The HDAC3 selective inhibitor BRD3308 significantly induces Bdnf pI expression as measured by real-time RTPCR . . For the coding promoter , the n number for controls and BRD3308 treatments are 6 and 3 respectively . For the pI promoter , the n number for controls and BRD3308 treatments are 5 and 4 respectively . Each replicate consisted of primary neurons obtained from different cultures and treated with fresh dilutions of the compounds . Significance was measured by unpaired t-test **p<0 . 01 . ( D ) Hdac3 knockdown is verified by real-time RTPCR . N = 5 represents independent times knockdown was achieved in different primary culture using nucleofection . Significance was measured by unpaired t-test **p<0 . 01 . ( E ) : HDAC3 knockdown is verified by western blotting . ( F ) : HDAC3 knockdown significantly induces Bdnf coding gene expression as verified by real-time RTPCR . N = 5 and significance was measured by unpaired t-test *p<0 . 05 . The Hdac3 shRNA sequence 2 that significantly reduced HDAC3 protein levels induced Bdnf expression . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 007 While brain cells catabolize glucose for energy under normal physiological conditions , ketone bodies are utilized when blood glucose levels decrease , as observed during exercise , fasting or caloric restriction . In order to mimic the increases in DBHB in the brain observed after exercise , we injected mice with 2-deoxy-d-glucose ( 2-DG ) , a structural analog of glucose that inhibits glycolysis and increases the brain’s capacity to utilize ketone bodies as fuel . 2-DG is transported by glucose transporters into the cell where it binds to , but cannot be phosphorylated by hexokinase , resulting in the inhibition of the first step of glycolysis ( Ralser et al . , 2008 ) . This inhibition leads to the activation of a compensatory mechanism resulting in the production of ketone bodies by the liver . By activating this alternative energy pathway , 2-DG treatment induces DBHB levels in the brain . Interestingly , we found that 2-DG injection , like exercise , also induces both DBHB levels and Bdnf expression in the hippocampus to a similar extent ( Figure 6A and B ) . Interestingly , co treatments of a DBHB transporter inhibitor ( AR-C155858 ) with 2-DG attenuated the 2-DG-induced DBHB levels and Bdnf expression . Hence , these findings are consistent with a model in which exercise induces DBHB accumulation in the hippocampus ( Figure 6A and B ) . DBHB stimulates histone acetylation at the Bdnf promoters through reduced HDAC2 and HDAC3 occupancy . This results in an increase in Bdnf gene transcription which reflects an epigenetic mechanism after exercise . 10 . 7554/eLife . 15092 . 008Figure 6 . DBHB can serve as an exercise factor linking metabolic changes in response to exercise to changes in gene expression in the brain . ( A ) 2-DG treatment known to induce ketone bodies in the brain induces Bdnf pI expression in the hippocampus as measured by real-time RTPCR . This effect was blocked by a DBHB transporter inhibitor . N = 10 ( Ctrl ) , N = 9 ( 2-DG injected ) and N = 5 ( 2-DG and AR-C155858 ) . Significance was measured by unpaired t-test *p<0 . 05 . ( B ) 2-DG treatment induced DBHB levels in the hippocampus and this effect is blocked by a DBHB transporter inhibitor . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 008 To investigate the consequences of an increase of DBHB in the hippocampus , we conducted electrophysiological measurements in hippocampal slices to evaluate DBHB’s effects on synaptic transmission . We incubated hippocampal slices with DBHB ( 0 . 8 mM ) and then carried out field recording of post-synaptic potentials ( fEPSPs ) in CA1 evoked by stimulating the Schaffer collateral fibers ( Yano et al . , 2006 ) . Paired pulsed facilitation ( PPF ) was measured as the ratio of fEPSP slopes in response to two stimuli delivered to the Schaffer collateral inputs . PPF was followed after interstimulus intervals of 20–40 milliseconds . Incubation of hippocampal slices with DBHB ( 0 . 8 mM , 3 hrs ) increased the fEPSP slope which was blocked by K252a ( 200 nM ) . Consistent with these results , DBHB decreased paired pulse ratio in a K252a-dependent manner suggesting a pre-synaptic modulation mediated by TrKB signaling ( Figure 7A and B ) . The actions of BDNF upon synaptic transmission are known to result in an increase in the frequency of synaptic currents , indicating a presynaptic role of BDNF ( Park and Poo , 2013 ) . In fact , BDNF rapidly increases spontaneous neurotransmitter release in hippocampal neurons ( Jovanovic et al . , 2000; Li et al . , 1998b ) , which require TrkB receptors localized at pre-synaptic sites ( Li et al . , 1998 ) . These measurements indicated that DBHB treatment enhanced glutamatergic transmission at CA3-CA1 synapses ( Figure 7 ) and that this DBHB-induced effect was blocked by K252a , a frequently used inhibitor of TrkB receptors in synaptic transmission experiments ( Chen et al . , 2015; Li et al . , 1998b; Takei et al . , 1998 ) . K-252a at 200 mM concentration specifically inhibits Trk receptors and not other tyrosine kinase receptors ( Berg et al . , 1992 ) . Hence the presynaptic enhancement by DBHB is dependent upon the BDNF TrkB receptors . These findings are consistent with an increase of BDNF by DBHB in the hippocampus and furthermore indicate there are additional physiological outcomes mediated by DBHB that involve an increase in neurotransmitter release . 10 . 7554/eLife . 15092 . 009Figure 7 . DBHB increases glutamatergic transmission at the CA3-CA1 synapses in a TrkB-sensitive manner . ( A ) Average fEPSP slope in control ( 6 slices/6 mice ) , DBHB ( 6 slices/6 mice ) , K252a ( 6 slices/3 mice ) , and K252a+DBHB ( 6 slices/3 mice ) groups . Two-way repeated measures ANOVA showed significant difference between groups ( F3 , 20 = 10 . 13 , p<0 . 001 ) . Upper panel shows examples fEPSP traces . ( B ) Average paired pulse ratio in control ( 6 slices/6 mice ) , DBHB ( 6 slices/6 mice ) , K252a ( 6 slices/3 mice ) , and K252a+DBHB ( 6 slices/3 mice ) groups . Two-way repeated measures ANOVA showed significant difference between groups ( F3 , 20 = 4 . 03 , p = 0 . 022 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 009
These results provide a link between running exercise , the ketone body DBHB and Bdnf gene expression . Previous work with DBHB showed it was an effective neuroprotective agent in Huntington’s disease ( Lim et al . , 2011 ) and Parkinson’s disease ( Kashiwaya et al . , 2000; Tieu et al . , 2003 ) , affecting striatal and dopaminergic neurons , respectively . It is highly conceivable that DBHB might act to increase the levels of BDNF , which can rescue neurons that are vulnerable in Huntington’s and Parkinson’s disorders ( Autry and Monteggia , 2012 ) . Indeed , treatment of Alzheimer’s disease 3xTgAD mice with 2-DG to induce ketone bodies delayed the progression of bioenergetic deficits in the brain and the associated β-amyloid burden . Similar to our experiments , 2-DG was capable of inducing BDNF in 3xTgAD mice ( Yao et al . , 2011 ) . Brain is a major source of the secretion of BDNF after exercise ( Rasmussen et al . , 2009 ) . Moreover , changes in the levels of BNDF have been suggested in the pathophysiology of schizophrenia , addiction and eating disorders ( Autry and Monteggia , 2012 ) . A number of studies have shown that exercise can also improve depressive-like behavior through increased levels of hippocampal BDNF , which can enhance plasticity and synaptogenesis and reduce neurodegeneration ( Aguiar et al . , 2011; Cotman et al . , 2007; Duman and Monteggia , 2006; Lu et al . , 2013; Marais et al . , 2009; Martinowich et al . , 2007; Russo-Neustadt et al . , 2000 ) . Given that histone acetylation in the hippocampus and cortex is associated with effects on learning and memory , the ketone body DBHB serves as a metabolic signal to link environment changes to epigenetic effects on the transcription of neurotrophic factors , such as BDNF . Ketone bodies are widely distributed from the liver to the heart , muscle and the brain after fasting , dieting and intense exercise . When glucose levels are reduced , ketone bodies , produced in the liver from fatty acids in the form of DBHB and acetoacetate , serve as an energy source . In the brain , the levels of ketone bodies can reach very high levels ( 1–5 mM ) ( Mitchell et al . , 1995; Robinson and Williamson , 1980 ) . A number of studies have demonstrated neuroprotective effects of ketone bodies in neurodegenerative and neuronal activity , such as under epileptic conditions ( Maalouf et al . , 2009 ) . It is probable there are multiple mechanisms that are involved in the effects of exercise in increasing trophic factor expression . For example , it has been reported that exercise induces Bdnf expression through the induction of hippocampal expression of Fndc5 ( Wrann et al . , 2013 ) , a PGC-1α and ERRα -dependent myokine . Once BDNF protein levels increase , TrkB signaling in turn inhibits Fndc5 expression in a negative feedback mechanism . We found prolonged exercise decreased Fndc5 expression levels consistent with elevated BDNF protein in the hippocampus ( Wrann et al . , 2013 ) . However , DBHB did not induce Pgc-1α , Fndc5 , or Erra mRNA in primary neurons ( data not shown ) , suggesting that there are alternative ways of affecting BDNF levels . The complexity of the BDNF gene and its alternative promoters and splicing events ( Haberland et al . , 2009 ) indicates that several regulatory mechanisms must exist to explain its many functions in synaptic transmission and neuronal survival ( Park and Poo , 2013 ) . Interestingly , exercise induces Bdnf expression in the hippocampus and not in all the other regions of the brain . How this specificity is achieved is not clear . One possible explanation suggests that exercise may regulate the levels or functions of ketone body transporters in different regions of the brain . Indeed , acute exercise induces Mct2 transporter expression in the hippocampus immediately after and up to 10 hr post-exercise . This exercise-induced upregulation of Mct2 was associated with increases in BDNF and TrkB levels ( Takimoto and Hamada , 2014 ) . Our finding of an endogenous metabolite , DBHB , which upregulates Bdnf transcription in the hippocampus , provides one explanation linking the effects of exercise and peripheral metabolism to changes in epigenetic control and gene expression in the brain . There are multiple links between metabolic changes resulting from exercise and epigenetic modulation . For example , it has been reported that the blood levels of some Kreb’s cycle intermediates such as α-ketoglutarate are increased after exercise ( Leibowitz et al . , 2012 ) . Interestingly , α-ketoglutarate is an essential co-substrate for jumanji histone demethylases as well as for the tet proteins ( DNA demethylases ) . Another molecule that may play a critical role in linking metabolic changes induced by exercise to epigenetic modulation is acetyl CoA which serves as the source of the acetyl group for all acetylation reactions catalyzed in the body including those catalyzed by histone acetyltransferases . Finally , coenzymes such as NAD+ , or FAD produced during metabolic reactions are also required by sirtuins and the histone demethylase LSD1 , respectively . In this paper , we provide evidence that an endogenous molecule , DBHB , that crosses the blood brain barrier , is increased by physical exercise to enhance the expression of a fundamental trophic factor in the brain and in turn affect synaptic transmission ( Figure 8 ) . Further studies aiming at identifying molecules that can also serve the dual purpose of an energy fuel and epigenetic modulator will help us accumulate additional members of the “exercise pill” . The identification of these molecules is of great interest as many people afflicted with depression or with neurodegenerative diseases are likely to benefit from the ability of exercise to stimulate BDNF through small metabolites , such as DBHB . The involvement of ketone bodies in many other syndromes , such as glucose utilization , diabetes and epilepsy , suggests they represent vital molecules with broad metabolic effects upon chromatin and gene expression . 10 . 7554/eLife . 15092 . 010Figure 8 . A proposed model by which exercise induces Bdnf expression in the hippocampus . Exercise induces DBHB synthesis in the liver . DBHB is transported through the circulation to peripheral organ including the brain . In the hippocampus , DBHB induces Bdnf expression through a mechanism involving HDAC inhibition . This induction in turn mediates exercise’s positive effects on memory , cognition and synaptic transmission . DOI: http://dx . doi . org/10 . 7554/eLife . 15092 . 010
Male mice were individually housed with food and water ad libitum , lights on: 6 AM and lights off 6 PM . They were divided into 2 groups ( Control and Running , n = 10 each ) ; Control mice and running animals were housed with free access to a running wheel . Running distance and time were monitored ( Coulborn Instruments ) . Animals were sacrificed after 30 days and hippocampus and cortex were collected and immediately frozen on dry ice . Animal care and use was in accordance with the guidelines set by the National Institutes of Health and the New York State Department of Health Immature primary cortical neurons were obtained from C57BL/6 mice [embryonic day 17 ( E17 ) ] as previously described ( Ratan et al . , 1994 ) . Mature cortical neurons were maintained in Neurobasal media ( Invitrogen ) supplemented with MACS NeuroBrew- 21 ( Miltenyi Biotec ) , and Glutamax ( Invitrogen ) . Primary neurons wereisolated as described above and 0 . 5 million cells were plated in six well plates . On Day 6 , cells were treated with different concentrations of DBHB overnight or HDAC inhibitors for 5 hr . DBHB was prepared as 1 M stock in PBS and used at a final concentration of 5 mM . suberoylanilide hydroxamic acid ( SAHA ) , Acetyldinaline ( CI-994 ) , Entinostat ( MS-275 ) , BRD3308 and BRD4097 were prepared as 10 mM stocks in DMSO and used at a final concentration of 10 μM . Trichostatin A ( TSA ) was prepared as a 670 μM stock and used at a final concentration of 0 . 67 μM . Total RNA was prepared from primary cortical neurons or hippocampi usingthe NucleoSpin RNA II Kit ( Clontech ) according to the manufacturer’s protocol . Real-time PCRs wereperformed using standard PCR protocol . Details of the Primers used are provided below: Bdnf coding ( Fwd ) : GCGGCAGATAAAAAGACTGC Bdnf coding ( Rev ) : GCAGCCTTCCTTGGTGTAAC Bdnf Var ( Rev ) : GCCTTCATGCAACCGAAGTA Bdnf Var I ( Fwd ) : CAGGACAGCAAAGCCACAAT Gapdh ( Fwd ) : CTCTCTGCTCCTCCCTGTTC Gapdh ( REV ) : CCGACCTTCACCATTTTGTC Hdac3 ( Fwd ) : ATGCCTTCAACGTGGGTGAT Hdac3 ( Rev ) : CCTGTGTAACGGGAGCAGAAC DBHB levels were measured using a DBHB Assay kit ( MAK041 , Sigma ) according to manufacturer’s protocol . To determine HDAC3 , acetyl HISTONE H3 , and HISTONE H3 protein levels , total cell proteins were prepared by lysing cells in RIPA-B ( 1% Triton X-100 , 1% SDS , 50 mM Tris-Cl , pH 7 . 4 , 500 mM NaCl and 1 mM EDTA ) in the presence of protease inhibitors ( Sigma ) , the proteasome inhibitor MG-132 ( Sigma ) and phosphatase inhibitors ( Sigma ) followed by benzonase nuclease ( Sigma ) digestion for 15 min . Samples were boiled in Laemmlibuffer and electrophoresed under reducing conditions on NuPAGE Novex 4–12% for Bis-Tris Gel polyacrylamidegels ( Invitrogen ) . Proteins were transferred to a nitrocellulose membrane ( Bio-Rad ) by electroblotting . Nonspecific binding was inhibitedby incubation in Odyssey blocking buffer ( LI-COR Biosciences ) . Antibodies against HISTONE H3 ( Cell signaling ) and acetylated HISTONE H3 ( Millipore ) , HDAC3 ( ab16047 , Abcam ) and β-ACTIN ( AC-74; Sigma-Aldrich ) were diluted 1:1000 , 1:1000 , 1:1000; and 1:10 , 000 , respectively , in odyssey blocking buffer and the membranes were incubated overnight at 4°C . Fluorophore-conjugated Odyssey IRDye-680 or IRDye-800 secondary antibody ( LI-COR Biosciences ) was used at 1:10 , 000 dilution followed by incubation for 1 hr at room temperature . Finally , proteins were detected using an Odyssey infrared imaging system ( LI-COR Biosciences ) . To determine BDNF protein levels , total hippocampal proteins were prepared by homogenizing tissues in RIPA-B in the presence of protease inhibitors , the proteasome inhibitor MG-132 and phosphatase inhibitors followed by benzonase nuclease digestion for 15 min . Samples ( 100μg ) were boiled in Laemmlibuffer and electrophoresed under reducing conditions on NuPAGE® Novex 12% Bis-Tris Gel polyacrylamidegels . Proteins were transferred to a nitrocellulose membrane ( Bio-Rad ) by electroblotting . The membranes were washed twice with 1X PBS followed by 30 min incubation with 2 . 5% glutaraldehyde/PBS . The membranes were washed twice again with PBS and then with TBS-T . Nonspecific binding was inhibitedby incubation in 5% milk/TBS-T followed by incubation with the BDNF antibody ( N-20 , Santa Cruz ) for 2 hr at room temperature . A peroxidase conjugated goat anti-rabbit or anti mouse was used for BDNF or actin detection , respectively . The membranes were detected by chemiluminescence . Animal care and use was in accordance with the guidelines set by the National Institutes of Health and the New York State Department of Health . 5-week old male mice were anesthetized and decapitated . The brain was dissected and placed in 4°C cutting buffer ( 126 mM sucrose , 5mM KCl , 2 CaCl2 , 2 mM MgSO4 , 26 mM NaHCO3 , 1 . 25 mM NaH2PO4 , and 10 mM D-glucose , pH 7 . 4 ) . The hippocampus was dissected and submerged in ice-cold cutting buffer and cut horizontally into 300-μm sections , which were immediately placed in recovery solution {50% cutting solution/ 50% artificial cerebrospinal fluid ( ACSF ) } and oxygenated ( 95% CO2-5% O2 ) for 20 min . The slices were then transferred in to ACSF oxygenated chambers and treated with DBHB ( 0 . 8 mM concentration ) for 3 hr after which RNA was extracted for Real-Time RT PCR analysis . All surgeries were performed in accordance to IACUC rules . Mice were anesthetized under isoflurane and placed on a stereotaxic frame ( David Kopf instruments , CA ) . The body temperature of the animals was maintained at 37°C using a homeothermic blanket . DBHB ( 2 mM and 5 mM; 5 µl ) was delivered by a Hamilton syringe at a flow rate of 0 . 5 µl/min using a nanomite syringe pump ( Harvard apparatus , MA ) . The stereotaxic coordinates relative to bregma were as follows: AP , -0 . 46 mm; L , -1 . 20 mm; DV , 2 . 20 mm . ( Paxinos and Watson , 1998 ) . In sham control animals , 5 µl of saline ( vehicle ) were infused . Proper post-operative care was taken until the animals recovered completely . Mice were sacrificed after 6 hr following the injection and the different brain regions were dissected . Hdac3 ( NM_010411 ) short hairpin RNA ( shRNA ) clone ( TRCN0000039391 , 5’ CCGGGTGTTGAATATGTCAAGAGTTCTCGAGAACTCTTGACATATTCAACACTTTTTG 3′; Sigma ) and Non-Target shRNA Control Vector ( Sigma ) were introduced into immature primary cortical neurons ( E17 ) using the Amaxa mouse Neuron Nucleofector kit as directed by the manufacturer ( Lonza ) . On Day 6 , HDAC3 knockdown was confirmed by Real-time RTPCR and whole-cell lysate Western blots . The Ez-Magna ChIP assay kit was used as directed by the manufacturer ( Millipore ) . Briefly , primary cortical cells were crosslinked with 1% formaldehyde at 37°C for 7 min . Cells were then sonicated using the Bioruptor ( Diagenode ) and immunoprecipitated with primary antibodies ( 10 µg ) . The crosslinking was reversed , and the DNA was isolated on the columns provided by the kit . Shearing size was determined to be between 150 and 1000 bp . Real-time PCR was conducted with primers targeted to the BDNF pI promoter ( TGATCATCACTCACGACCACG and CAGCCTCTCTGAGCCAGTTACG ) and SYBR Green PCR Master mix ( Applied Biosystems ) . Each experiment was conducted at least three times by crosslinking cells from different primary cortical neuron preparations . mice received intraperitoneal injections of 10mg/Kg of 2-DG or saline . They were left to recover for 6 hrs after which the animals were sacrificed and the tissue was harvested . 2–4 months old C57BL6 male mice were anesthetized by pentobarbital anesthesia . After decapitation , hippocampi were isolated in ice-cold artificial cerebrospinal fluid ( ACSF ) containing the following ( in mM ) , NaCl ( 118 ) , KCl ( 4 . 5 ) , glucose ( 10 ) , NaH2PO4 ( 1 ) , CaCl2 ( 2 ) and MgCl2 ( 2 ) ( aerated by 95%O2/5% CO2 , pH adjusted to 7 . 4 with NaHCO3 ) . Hippocampal slices ( 300 μm ) were prepared on a vibratome ( Campden Instruments ) and maintained at room temperature for 1 hr in a brain slice keeper . Brain slices were incubated in vehicle ( 0 . 00007% DMSO ) , K252a ( 200 nM ) , DBHB ( 0 . 8 mM ) or combination of K252a and DBHB for 3 hrs at room temperature before recording CA1 field excitatory post-synaptic potentials ( fEPSPs ) in a recording chamber perfused with ACSF at 32°C . fEPSPs were evoked by stimulation of the Schaffer collateral fibers using a concentric bipolar electrode ( FHC ) . Input-output data were generated by plotting fEPSP slope in response to 100 , 200 , 300 , 400 and 500 μA stimulation . Paired pulse ratio was measured as ratio of second fEPSP slope to first EPSP slope at 20 , 40 , 80 , 120 and 160 ms inter-stimulus interval . A response approximately 35% of the maximum evoked response was used for studying paired pulse ratio . Data were acquired using pCLAMP 10 program and MultiClamp 700B amplifier ( Molecular Devices ) . Data analysis was carried out using Clampfit program . unpaired t-test , 1way or 2way ANOVA followed by the Dunnett or Bonferroni post tests respectively were used to measure statistical significance . p<0 . 05 was considered to be statistically significant . | Exercise is not only good for our physical health but it benefits our mental health and abilities too . Physical exercise can affect how much of certain proteins are made in the brain . In particular , the levels of a protein called brain derived neurotrophic factor ( or BDNF for short ) increase after exercise . BDNF has already been shown to enhance mental abilities at the same time as acting against anxiety and depression in mice , and might act in similar way in humans . Nevertheless , it is currently not clear how exercise increases the production of BDNF by cells in the brain . Sleiman et al . have now investigated this question by comparing mice that were allowed to use a running wheel for 30 days with control mice that did not exercise . The comparison showed that the exercising mice had higher levels of BDNF in their brains than the control mice , which confirms the results of previous studies . Next , biochemical experiments showed that this change occurred when enzymes known as histone deacetylases stopped inhibiting the production of BDNF . Therefore Sleiman et al . hypothesised that exercise might produce a chemical that itself inhibits the histone deacetylases . Indeed , the exercising mice produced more of a molecule called β-hydroxybutyrate in their livers , which travels through the blood into the brain where it could inhibit histone deacetylases . Further experiments showed that injecting β-hydroxybutyrate directly into the brains of mice led to increase in BDNF . These new findings reveal with molecular detail one way in which exercise can affect the expression of proteins in the brain . This new understanding may provide ideas for new therapies to treat psychiatric diseases , such as depression , and neurodegenerative disorders , such as Alzheimer’s disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2016 | Exercise promotes the expression of brain derived neurotrophic factor (BDNF) through the action of the ketone body β-hydroxybutyrate |
CpG dinucleotides are suppressed in most vertebrate RNA viruses , including HIV-1 , and introducing CpGs into RNA virus genomes inhibits their replication . The zinc finger antiviral protein ( ZAP ) binds regions of viral RNA containing CpGs and targets them for degradation . ZAP does not have enzymatic activity and recruits other cellular proteins to inhibit viral replication . We found that KHNYN , a protein with no previously known function , interacts with ZAP . KHNYN overexpression selectively inhibits HIV-1 containing clustered CpG dinucleotides and this requires ZAP and its cofactor TRIM25 . KHNYN requires both its KH-like domain and NYN endonuclease domain for antiviral activity . Crucially , depletion of KHNYN eliminated the deleterious effect of CpG dinucleotides on HIV-1 RNA abundance and infectious virus production and also enhanced the production of murine leukemia virus . Overall , we have identified KHNYN as a novel cofactor for ZAP to target CpG-containing retroviral RNA for degradation .
A major component of the innate immune system are cell intrinsic antiviral proteins . These act at multiple steps in viral replication cycles and some are induced by type I interferons ( Schneider et al . , 2014 ) . Many viruses have evolved mechanisms to evade inhibition by these proteins . First , viruses can encode proteins that counteract specific antiviral factors . Examples of this mechanism in HIV-1 are the accessory proteins Vif and Vpu that counteract APOBEC3 cytosine deaminases and Tetherin , respectively ( Malim and Bieniasz , 2012 ) . Second , viral protein or nucleic acid sequences can evolve to prevent recognition by antiviral factors . The abundance of CpG dinucleotides is suppressed in many vertebrate RNA virus genomes and when CpGs are experimentally introduced into picornaviruses or influenza A virus , replication is inhibited ( Atkinson et al . , 2014; Burns et al . , 2009; Gaunt et al . , 2016; Karlin et al . , 1994; Tulloch et al . , 2014 ) . This shows that CpG suppression in diverse RNA viruses is required for efficient replication . CpG dinucleotides are also suppressed in the HIV-1 genome and multiple studies have shown that they are deleterious for replication ( Antzin-Anduetza et al . , 2017; Kypr et al . , 1989; Shpaer and Mullins , 1990; Takata et al . , 2017; Theys et al . , 2018; Wasson et al . , 2017 ) . Recently , the cellular antiviral protein ZAP was shown to bind regions of HIV-1 RNA with high CpG abundance and target them for degradation , which at least partly explains why this dinucleotide inhibits viral replication ( Takata et al . , 2017 ) . ZAP ( also known as ZC3HAV1 ) is a component of the innate immune response targeting viral RNAs in the cytoplasm to prevent viral protein synthesis ( Li et al . , 2015 ) . ZAP inhibits the replication of a diverse range of viruses including retroviruses , alphaviruses , filoviruses , hepatitis B virus and Japanese encephalitis virus as well as retroelements ( Bick et al . , 2003; Chiu et al . , 2018; Gao et al . , 2002; Goodier et al . , 2015; Mao et al . , 2013; Moldovan and Moran , 2015; Müller et al . , 2007; Takata et al . , 2017; Zhu et al . , 2011 ) . There are two human ZAP isoforms , ZAP-L and ZAP-S ( Kerns et al . , 2008 ) . Both isoforms contain a N-terminal RNA binding domain containing four CCCH-type zinc finger motifs but ZAP-L also contains a catalytically inactive C-terminal poly ( ADP-ribose ) polymerase ( PARP ) -like domain ( Chen et al . , 2012; Guo et al . , 2004; Kerns et al . , 2008 ) . Importantly , neither isoform of ZAP has nuclease activity and it likely recruits other cellular proteins to degrade viral RNAs . Identifying and characterizing these cofactors for ZAP is essential to understand how it restricts viral replication . ZAP requires the E3 ubiquitin ligase TRIM25 for its antiviral activity against Sindbis virus and HIV-1 with clustered CpGs ( Li et al . , 2017; Takata et al . , 2017; Zheng et al . , 2017 ) . While ZAP has been reported to interact with several components of the 5’−3’ and 3’−5’ RNA degradation pathways , depletion of these proteins did not substantially increase infectious virus production for HIV-1 containing clustered CpG dinucleotides ( Goodier et al . , 2015; Guo et al . , 2007; Takata et al . , 2017; Zhu et al . , 2011 ) . This suggests that additional proteins may be required for ZAP to inhibit viral replication . Herein , we identify KHNYN as a cytoplasmic protein that interacts with ZAP and is necessary for CpG dinucleotides to inhibit HIV-1 RNA and protein abundance .
To identify candidate interaction partners for ZAP , a yeast two-hybrid screen was performed for full-length ZAP-S and ZAP-L using prey fragments from a mixed Pam3CSK4-induced and IFNβ-induced human macrophage cDNA library . Candidate interacting proteins were assigned a Predicted Biological Score ( PBS ) of A to F ( Formstecher et al . , 2005 ) : A = very high confidence in the interaction , B = high confidence in the interaction and C = good confidence in the interaction . Scores of D to F are low confidence interactions , non-specific interactions or proven technical artifacts . For ZAP-S , 11 clones were obtained from 60 . 4 million tested interactions . 10 of these contained a prey fragment encoding KHNYN ( Figure 1 ) and had a PBS = A . One clone had an insert encoding MARK3 but this was in the antisense orientation and therefore did not receive a score . For ZAP-L , two positive clones were analyzed from 104 million tested interactions . Both of these had an insert encoding KHNYN and had a PBS = C . KHNYN has two isoforms ( KHNYN-1 and KHNYN-2 ) that contain a N-terminal KH-like domain and a C-terminal NYN endoribonuclease domain ( Figure 1 ) ( Anantharaman and Aravind , 2006 ) . The selected interaction domain , which is the amino acid sequence shared by all prey fragments matching KHNYN , comprised amino acids 572–719 of KHNYN-2 for the clones identified in both screens . A yeast two-hybrid screen was then performed using the same library with full length KHNYN-2 as the bait . Nine clones were isolated that encode ZAP and these had a PBS = A . The selected interaction domain was amino acids 4–352 , which is present in both isoforms ( Figure 1 ) . Supporting the reproducibility of this interaction , KHNYN has also been identified as a ZAP-interacting factor in large-scale affinity purification–mass spectrometry and in vivo proximity-dependent biotinylation ( BioID ) screens ( Huttlin et al . , 2017; Youn et al . , 2018 ) . We first confirmed the interaction between ZAP and KHNYN by co-immunoprecipitation and found both KHNYN isoforms interacted with both isoforms of ZAP ( Figure 2A and B ) . This interaction was RNase insensitive ( Figure 2C ) . Since ZAP mediates degradation of HIV-1 RNAs with clustered CpG dinucleotides in the cytoplasm ( Takata et al . , 2017 ) , its cofactors are likely to be localized in this compartment . Therefore , we analyzed the subcellular localization of KHNYN and observed that it localizes to the cytoplasm similar to ZAP ( Figure 2D ) . Its localization was not affected when ZAP was knocked out using CRISPR-Cas9-mediated genome editing . The mechanisms that allow a virus to escape the innate immune response often have to be inactivated to study the effect of antiviral proteins . For example , HIV-1 Vpu or Vif have to be mutated to allow Tetherin or APOBEC3 antiviral activity to be analyzed ( Malim and Bieniasz , 2012 ) . Since CpG dinucleotides are suppressed in HIV-1 , endogenous ZAP does not target the wild-type virus ( Takata et al . , 2017 ) . However , a ZAP-sensitive HIV-1 can be created by introducing CpGs through synonymous mutations into the env open-reading frame in the viral genome . This makes HIV-1 an excellent system to study the mechanism of action of this antiviral protein because isogenic viruses can be analyzed that differ only in their CpG abundance and therefore ZAP-sensitivity ( Takata et al . , 2017 ) . To determine if KHNYN overexpression inhibited wild-type HIV-1 or HIV-1 with 36 CpG dinucleotides introduced into env nucleotides 86–561 ( HIV-1EnvCpG86-561 ) ( Figure 2—figure supplement 1 ) , each isoform was overexpressed in the context of a single cycle replication assay . As expected , transfection of the HIV-1EnvCpG86-561 provirus into HeLa cells yielded substantially less infectious virus than wild-type HIV-1 , which was accounted for by reduced expression of Gag and Env proteins ( Figure 2E and F ) . While KHNYN-1 or KHNYN-2 overexpression decreased wild-type HIV-1 infectivity by ~5 fold , they decreased HIV-1EnvCpG86-561 infectivity by ~400 fold ( Figure 2E ) . The inhibition of infectivity by KHNYN-1 or KHNYN-2 correlated with decreases in Gag expression , Env expression , and virion production ( Figure 2F ) . Overall , KHNYN appeared to selectively inhibit HIV-1EnvCpG86-561 infectious virus production . We then determined whether ZAP is necessary for KHNYN to inhibit HIV-1 with clustered CpG dinucleotides . Control or ZAP knockout cells ( Figure 3A ) were co-transfected with pHIV-1 or pHIV-1EnvCpG86-561 and increasing amounts of pKHNYN-1 . Wild-type HIV-1 infectious virus production was not affected by ZAP depletion and HIV-1EnvCpG86-561 infectivity was restored in ZAP knockout cells ( Figures 3B , 0 ng of KHNYN-1 ) , confirming that ZAP is necessary to inhibit HIV-1 with CpGs introduced in env ( Takata et al . , 2017 ) . At low levels of KHNYN-1 overexpression ( such as 62 . 5 ng ) , there was no substantial decrease in infectivity for wild-type HIV-1 while HIV-1EnvCpG86-561 infectivity was inhibited in a ZAP-dependent manner ( Figures 3B and 4A ) . The decrease in infectivity for HIV-1EnvCpG86-561 in control cells transfected with pKHNYN-1 correlated with decreases in Gag expression , Env expression and virion production ( Figure 3C ) . Next , we analyzed how ZAP and KHNYN regulate HIV-1 genomic RNA abundance in cell lysates and media . As expected ( Takata et al . , 2017 ) , HIV-1EnvCpG86-561 genomic RNA abundance was decreased in control cells but was similar to wild-type HIV-1 in ZAP knockout cells ( Figure 4B–C , compare GFP samples ) . In control cells , 62 . 5 ng of KHNYN-1 or KHNYN-2 inhibited HIV-1EnvCpG86-561 genomic RNA abundance compared to the GFP control ( Figure 4B–C ) . Importantly , KHNYN-1 and KHNYN-2 did not affect wild-type HIV-1 genomic RNA levels and did not substantially inhibit HIV-1EnvCpG86-561 genomic RNA abundance in ZAP knockout cells . This demonstrates that KHNYN targets HIV-1 RNA containing clustered CpG dinucleotides in a ZAP-dependent manner . TRIM25 is required for ZAP’s antiviral activity , although the mechanism by which it regulates ZAP is unclear ( Li et al . , 2017; Zheng et al . , 2017 ) . To determine if TRIM25 is necessary for the antiviral activity of KHNYN , 62 . 5 ng of pKHNYN-1 or pKHNYN-2 was co-transfected with pHIV-1 or pHIV-1EnvCpG86-561 in control and TRIM25 knockout cells . Both isoforms of KHNYN inhibited HIV-1EnvCpG86-561 much less potently in TRIM25 knockout cells than control cells and had no effect on wild-type HIV-1 in either cell line ( Figure 5A–B ) . One possible reason that TRIM25 is necessary for KHNYN antiviral activity could be that it regulates the interaction between ZAP and KHNYN . We pulled down KHNYN-FLAG and western blotted for ZAP in control and TRIM25 knockdown cells ( Figure 5C ) . Both isoforms of KHNYN pulled down ZAP in both cell lines , indicating that TRIM25 is not required for the interaction between these proteins . Interestingly , KHNYN also pulled down TRIM25 in control cells ( Figure 5C ) . Therefore , we analyzed whether KHNYN interacted with TRIM25 in control and ZAP knockout cells and observed that both isoforms of KHNYN-FLAG immunoprecipitated TRIM25 in the presence and absence of ZAP ( Figure 5D ) . In sum , KHNYN requires TRIM25 to inhibit HIV-1 containing clustered CpG dinucleotides , but TRIM25 is not necessary for the interaction between ZAP and KHNYN . Furthermore , ZAP , KHNYN and TRIM25 appear to be in a complex together . As its name implies , KHNYN contains a KH-like domain and a NYN domain ( Figure 6A ) . The KH-like domain differs from canonical KH domains due to a potential small metal chelating module containing two cysteines and a histidine inserted into the central region of the domain ( Anantharaman and Aravind , 2006 ) . Since this has diverged substantially from a standard KH domain , it has also been called a CGIN1 domain and is only known to be present in two other proteins ( Marco and Marín , 2009 ) . While most KH domains bind nucleic acids ( Nicastro et al . , 2015 ) , the insertion in the KH-like domain in KHNYN may disrupt RNA binding and indicate that it has a different function . To analyze the functional importance of the KH-like domain , we deleted it and found that KHNYN-1∆KH and KHNYN-2∆KH had reduced antiviral activity compared to the wild-type protein ( Figure 6B–C ) . These mutant proteins localized to the cytoplasm and formed foci that were not present for wild-type KHNYN-1 or KHNYN-2 ( Figure 6—figure supplement 1 ) . NYN domains have endonuclease activity and belong to the PIN nuclease domain superfamily . There are at least eight human proteins with a potentially active NYN domain and they have been structurally characterized in several proteins including ZC3H12A and MARF1 ( Matelska et al . , 2017; Matsushita et al . , 2009; Nishimura et al . , 2018; Xu et al . , 2012; Yao et al . , 2018 ) . These domains contain a negatively charged active site with four aspartic acid residues coordinating a magnesium ion , which activates a water molecule for nucleophilic attack of the phosphodiester group on the target RNA . Mutation of these acidic residues inhibits nuclease activity by disrupting the bonds that directly or indirectly interact with the magnesium ion . ZC3H12A ( also known as MCPIP1 and Regnase ) is a RNA binding protein that , similar to ZAP , contains a CCCH zinc finger domain and degrades cellular and viral RNAs ( Takeuchi , 2018 ) . The NYN domain in ZC3H12A has 56% identity to the NYN domain in KHNYN ( Figure 6—figure supplement 2A ) . ZC3H12A containing a D141N mutation in the NYN domain had decreased endonuclease activity and did not degrade RNA containing the IL-6 3’ UTR ( Matsushita et al . , 2009 ) . MARF1 is required for meiosis and retrotransposon silencing in oocytes and a D426A/D427A mutation inhibited its endoribonuclease activity ( Nishimura et al . , 2018; Su et al . , 2012 ) . We made these equivalent mutations in KHNYN ( Figure 6A and Figure 6—figure supplement 2B ) and tested their ability to inhibit HIV-1EnvCpG86-561 gene expression and infectious virus production ( Figure 6B–C ) . KHNYN-1 D443N and KHNYN-2 D484N had substantially decreased activity against HIV-1EnvCpG86-561 . Strikingly , KHNYN-1 D524A/D525A and KHNYN-2 D565A/D566A had no antiviral activity even though they were expressed at similar levels to the wild-type KHNYN . KHNYN proteins with mutations in the NYN domain localized to the cytoplasm , indicating that disruption of this domain’s activity did not substantially affect their subcellular localization ( Figure 6—figure supplement 1 ) . To determine if KHNYN is required for CpG dinucleotides to inhibit infectious HIV-1 production , we depleted it using CRISPR-Cas9-mediated genome editing using single-guide RNAs ( sgRNAs ) targeting two independent sequences in KHNYN ( Figure 7A-C = sgRNA 1 , Figure 7—figure supplement 1 = sgRNA 2 ) . We were unable to identify an antibody that detected endogenous KHNYN . However , FLAG-tagged KHNYN-1 and KHNYN-2 were depleted in the bulk population of cells expressing each sgRNA as well as two clonal cell lines from each bulk population ( Figure 7A and Figure 7—figure supplement 1A ) . Importantly , when the CRISPR PAM sequence was mutated in the KHNYN plasmids , KHNYN-1 and KHNYN-2 were no longer depleted . The relevant region of KHNYN was also sequenced to identify the insertion or deletion in the bulk populations as well as the clones and genetic alterations that inactive the protein were identified in each ( Figure 7—figure supplement 2 ) . These KHNYN CRISPR cells were then transfected with pHIV-1 or pHIV-1EnvCpG86-561 . In the KHNYN CRISPR cells , the CpG dinucleotides in HIV-1EnvCpG86-561 no longer inhibited infectious virus production , Gag expression or Env expression ( Figure 7B–C , Figure 7—figure supplement 1B–C ) . Wild-type HIV-1 Gag expression , Env expression and infectious virus production was not altered in the KHNYN CRISPR cells compared to the control cells . Overall , the KHNYN CRISPR cells phenocopied the ZAP CRISPR cells ( Figure 7B–C ) . We also analyzed the effect of KHNYN depletion on murine leukemia virus ( MLV ) . While most retroviruses are suppressed in CpG abundance , the degree of this suppression varies between the different genera ( Berkhout et al . , 2002 ) . HIV-1NL4-3 is highly suppressed ( 9 CpGs/kb; 0 . 2 observed/expected ) , which is conserved in HIV-1 ( Berkhout et al . , 2002; Kypr et al . , 1989; Shpaer and Mullins , 1990 ) . However , the CpG abundance in MLV is much less suppressed ( 35 CpGs/kb; 0 . 5 observed/expected ) and ZAP was initially identified as an antiviral protein based on its ability to bind MLV RNA and target it for degradation ( Gao et al . , 2002; Guo et al . , 2004; Guo et al . , 2007 ) . To determine if KHNYN inhibits MLV , control , ZAP and KHNYN CRISPR cells were co-transfected with pMLV , p2 . 87 Vpu ( which encodes a highly active HIV-1 Vpu to counteract endogenous Tetherin expression in these cells [Neil et al . , 2008; Pickering et al . , 2014] ) and pGFP . MLV Gag expression and virion production were measured by immunoblotting . Since ZAP is a type I interferon-stimulated gene ( Shaw et al . , 2017 ) , MLV Gag expression and virion production were also analyzed after type I interferon treatment . There was a small but reproducible increase in MLV Gag expression and virion production in the ZAP and KHNYN CRISPR cells in the absence of type I interferon ( Figure 7D ) . However , after type I interferon treatment , MLV virion production was decreased to almost undetectable levels in the control CRISPR cells but was substantially higher in the ZAP and KHNYN CRISPR cells ( Figure 7E ) . The Sindbis virus genome is not substantially depleted in CpG dinucleotides ( 58 CpGs/kb; 0 . 9 observed/expected ) and is restricted by ZAP ( Bick et al . , 2003 ) . However , unlike retroviruses , the predominant ZAP antiviral activity for alphaviruses is to inhibit viral RNA translation , although there may be an additional effect on RNA stability ( Bick et al . , 2003; Kozaki et al . , 2015 ) . As expected ( Bick et al . , 2003; Kozaki et al . , 2015; Li et al . , 2017; Zheng et al . , 2017 ) , depletion of TRIM25 and ZAP substantially increased Sindbis virus replication ( Figure 7—figure supplement 3 ) . In contrast , there was no substantial increase in Sindbis virus replication in the KHNYN CRISPR cells . Thus , KHNYN appears to be required for the restriction of retroviral genomes , but not all ZAP-sensitive RNA viruses . We then analyzed HIV-1 genomic RNA abundance in the KHNYN CRISPR cells . Similar to HIV-1 protein expression and infectivity , HIV-1EnvCpG86-561 genomic RNA abundance was similar in the cell lysate and media to wild-type HIV-1 ( Figure 8A and C ) , indicating that the CpG dinucleotides no longer inhibited RNA abundance . As expected , nef mRNA abundance was not affected by the introduction of CpG dinucleotides in env or by KHNYN depletion since it is a fully spliced mRNA that does not contain the introduced CpGs ( Figure 8B ) . The wild-type HIV-1 genomic RNA abundance was not altered in the KHNYN CRISPR cells compared to the control cells , further showing the specific effect of KHNYN for viral RNA containing CpG dinucleotides . To determine the specificity of the KHNYN knockdown , we titrated CRISPR-resistant pKHNYN-1 or pKHNYN-2 into the KHNYN CRISPR cells . Even very low levels of KHNYN-1 or KHNYN-2 restored selective inhibition of HIV-1EnvCpG86-561 in these cells ( Figure 9A–B ) and KHNYN-1 was consistently slightly more active than KHNYN-2 . This shows that both isoforms are capable of inhibiting infectious virus production of HIV-1 containing clustered CpG dinucleotides . We also analyzed whether KHNYN with the KH-like domain deleted or the putative catalytic mutations in the NYN domain could inhibit HIV-1EnvCpG86-561 infectious virus production in the CRISPR cells . Expression of 31 . 25 ng of KHNYN-1 in the KHNYN CRISPR cells inhibited HIV-1EnvCpG86-561 ( Figure 9C–D ) and all of the mutations substantially reduced KHNYN antiviral activity . In sum , endogenous KHNYN is required for CpG dinucleotides to inhibit HIV-1 infectious virus production .
Several members of the CCCH zinc finger domain protein family target viral and/or cellular mRNAs for degradation ( Fu and Blackshear , 2017 ) . For example , ZC3H12A degrades pro-inflammatory cytokine mRNAs and also inhibits the replication of several viruses , including HIV-1 and hepatitis C virus ( Lin et al . , 2013; Lin et al . , 2014; Liu et al . , 2013; Matsushita et al . , 2009 ) . It contains a CCCH zinc finger domain as well as a NYN endonuclease domain , which allows it to degrade specific RNAs ( Matsushita et al . , 2009; Xu et al . , 2012 ) . ZAP has four CCCH zinc finger domains and specifically interacts with CpG dinucleotides in RNA ( Gao et al . , 2002; Guo et al . , 2004; Takata et al . , 2017 ) . However , it does not contain nuclease activity . While ZAP has been reported to directly or indirectly interact with components of the 5’−3’ and 3’−5’ degradation pathways including DCP1-DCP2 , XRN1 , PARN and the exosome , knockdown of several proteins in these pathways did not substantially rescue infectious virus production of HIV-1 containing clustered CpG dinucleotides ( Goodier et al . , 2015; Guo et al . , 2007; Takata et al . , 2017; Zhu et al . , 2011 ) . Therefore , we hypothesized that ZAP may interact with additional unidentified proteins that regulate viral RNA degradation . Herein , we have identified that KHNYN is an essential ZAP cofactor that inhibits HIV-1 gene expression and infectious virus production when the viral RNA contains clustered CpG dinucleotides . KHNYN overexpression inhibits genomic RNA abundance , Gag expression , Env expression and infectious virus production for HIV-1 containing clustered CpG dinucleotides . This activity requires ZAP and TRIM25 . Furthermore , depletion of KHNYN using CRISPR-Cas9 specifically increased genomic RNA abundance , Gag expression , Env expression and infectious virus production for HIV-1 containing clustered CpG dinucleotides . This indicates that KHNYN is essential for CpG dinucleotides to inhibit infectious virus production . Similarly , KHNYN depletion increased MLV Gag expression and virion production . However , Sindbis virus replication was not substantially increased in the KHNYN knockout cells . The difference between the requirement for KHNYN to inhibit retroviruses versus Sindbis virus may be because the antiviral effect for retroviruses is predominantly at the level of RNA stability and for alphaviruses it is predominately at the level of translation ( Bick et al . , 2003; Gao et al . , 2002; Guo et al . , 2004; Guo et al . , 2007; MacDonald et al . , 2007 ) . A mechanistic explanation for why the major antiviral effect of ZAP appears to be promoting RNA degradation for some viruses and inhibiting translation for other viruses remains unclear , although ZAP has been reported to inhibit translation initiation by interfering with the interaction between eIF4A and eIF4G ( Zhu et al . , 2012 ) . Therefore , KHNYN may not be required for ZAP to inhibit translation . We hypothesize that a complex containing ZAP and KHNYN binds HIV-1 CpG-containing RNA . ZAP and KHNYN could directly interact to form a heterodimer or there could be other factors mediating this interaction . The interaction between ZAP and KHNYN has been detected using several different assays including yeast-two-hybrid , co-immunoprecipitation , affinity purification–mass spectrometry ( Huttlin et al . , 2017 ) and BioID ( Youn et al . , 2018 ) . If there is an unknown factor mediating this interaction , it would have to present in the yeast-two-hybrid assay . It remains unclear how TRIM25 regulates ZAP , but it is not required for ZAP and KHNYN to interact . Interestingly , TRIM25 co-immunoprecipitates with KHNYN and the ZAP antiviral complex may simultaneously consist of all three proteins . ZAP and TRIM25 are interferon-stimulated genes while KHNYN is not induced by interferon in human cells ( Shaw et al . , 2017 ) . Whether KHNYN is regulated by type I interferons or viral infection in a different way , such as post-translational modification , is not known . The zinc finger RNA binding domains in ZAP could target KHNYN to CpG regions in viral RNA . This would allow the endonuclease domain in KHNYN to cleave this RNA , thereby inhibiting viral RNA abundance . Conceptually , the ZAP-KHNYN complex could function similarly to ZC3H12A , but with the RNA binding and endonuclease domains divided between the two proteins . The NYN domain in KHNYN could cleave HIV-1 RNA containing CpG dinucleotides similar to how ZC3H12A cleaves a specific site in the 3’ UTR of the IL-6 mRNA ( Matsushita et al . , 2009 ) . While we do not yet have evidence that the NYN domain in KHNYN is an active endonuclease domain , it is highly conserved with the active NYN domain in ZC3H12A and is required for KHNYN antiviral activity . Strikingly , mutation of two conserved aspartic acid residues in the NYN domain predicted to coordinate a magnesium ion necessary for nucleophilic attack of the target RNA eliminated KHNYN antiviral activity . However , biochemical and structural studies will be necessary to determine the specific nature of the interaction between ZAP , KHNYN , TRIM25 and RNA and how these interactions promote viral RNA degradation . An increasingly common theme for RNA decay is that endonucleic and exonucleic degradation pathways work together to fully degrade RNAs . For example , nonsense-mediated decay ( NMD ) targets mRNAs that do not efficiently terminate translation at the stop codon and uses up to four mechanisms to degrade these mRNAs: endonucleic cleavage , deadenylation , decapping and exonucleic degradation ( Lykke-Andersen and Jensen , 2015 ) . In this pathway , the endonuclease SMG6 interacts with the core regulatory protein UPF1 and cleaves mRNA near a premature termination codon . The 5’ and 3’ cleavage fragments are then degraded by the 5’−3’ exonuclease XRN1 and the 3’−5’ exonuclease exosome complex . The CCR4–NOT deadenylase complex and DCP1-DCP2 decapping complex are recruited by proteins in the NMD complex including UPF1 , SMG5 , SMG7 and PNRC2 . Similarly , the 5’−3’ and 3’−5’ degradation pathway components previously shown to interact with ZAP could work in conjunction with KHNYN-mediated endonucleic decay ( Goodier et al . , 2015; Guo et al . , 2007; Zhu et al . , 2011 ) . In this model , KHNYN would initiate cleavage of the viral RNA and the exonucleic pathways would then degrade the resulting RNA fragments . Identifying the full complement of ZAP-interacting factors and characterizing how these target viral RNAs for degradation will be an exciting area of future investigation . Another important area of future research will be to determine how KHNYN and other cellular proteins that contain an NYN endonuclease domain inhibit the replication of different viruses in different cell types with and without interferon treatment . The interferon-stimulated gene N4BP1 , which is a KHNYN paralog ( Anantharaman and Aravind , 2006 ) , was recently identified to genetically interact with ZAP in a CRISPR-based screen to identify interferon-induced antiviral proteins targeting HIV-1 ( OhAinle et al . , 2018 ) . In the monocytic THP-1 cell line , depletion of N4BP1 led to a small increase in wild-type HIV-1 replication . However , N4BP1 depletion did not inhibit replication of the alphavirus Semliki Forest virus , indicating that it may have a virus specific effect . While ZAP inhibits a range of viruses in different cell types , it remains unknown whether its cofactor requirements are cell type dependent . In this study , we have analyzed the antiviral activity of ZAP and KHNYN on HIV-1 and MLV in HeLa cells , but the role of NYN domain-containing proteins in targeting viral RNAs for degradation may be an important component of the antiviral innate immune response in a variety of cell types . It will also be interesting to determine if proteins containing an endonuclease domain other than KHNYN interact with CCCH zinc finger proteins to mediate antiviral activity . There are 57 human CCCH zinc finger proteins ( Fu and Blackshear , 2017 ) . At least 15 of these proteins are known to promote RNA decay and , including ZAP , six human CCCH zinc finger proteins are antiviral ( Fu and Blackshear , 2017 ) . Identifying the full complement of CCCH zinc finger proteins that inhibit viral replication and determining whether they require proteins containing endonuclease domains such as KHNYN or N4BP1 for this activity will increase our understanding of antiviral responses targeting viral RNA .
All primers were ordered from Eurofins . Polymerase chain reactions ( PCR ) for cloning steps were performed with Phusion High Fidelity polymerase ( New England Biolabs ) . KHNYN-2 ( NM_001290256 ) was synthesized by GenScript . KHNYN-1 was cloned by amplifying the nucleotides 123–2157 from KHNYN-2 and sub-cloning the PCR product into the pcDNA3 . 1 ( + ) backbone using the HindIII site in the vector and SbfI site in the KHNYN open reading frame . KHNYN-1 ΔKH , D443N , D524A/D525A , -CRG1 , -CRG2 and KHNYN-2 ΔKH , D484N , D565A/D566A , -CRG1 , -CRG2 were generated via overlap extension PCR and subsequently sub-cloning the PCR product into the pcDNA3 . 1 ( + ) backbone as described above . pGFP-FLAG was cloned by amplifying GFP from pcDNA3 . 1-GFP ( Swanson et al . , 2010 ) and cloning it into pcDNA3 . 1 . Diagnostic restriction enzyme digestion and DNA sequencing ( Eurofins , Genewiz ) was used to ensure the correct identity of modified sequences inserted into plasmids . pHIV-1NL4-3 contains the HIV-1 provirus in the pGL4 vector ( Antzin-Anduetza et al . , 2017 ) . To generate HIV-1EnvCpG86-561 , we synthesized a HIV-1NL4-3 EcoRI/StuI DNA fragment with synonymous mutations that inserted 36 CpG dinucleotides into env ( Figure 2—figure supplement 1 ) . This DNA fragment was digested with EcoRI and StuI and inserted into the corresponding sites of pHIV-1NL4-3 . pMLV is pNCS , which contains the Moloney murine leukemia virus provirus DNA ( Yueh and Goff , 2003 ) . pCR3 . 1 2 . 87 Vpu contains the HA-tagged codon-optimized HIV-1 Vpu 2 . 87 ( Pickering et al . , 2014 ) . HeLa and HEK293T cells were obtained from the ATCC and were maintained in high glucose DMEM supplemented with GlutaMAX , 10% fetal bovine serum , 20 µg/mL gentamicin or 100 U/ml penicillin and 100 μg/ml streptomycin and incubated with 5% CO2 at 37°C . BHK-21 cells were obtained from ATCC and were maintained in GMEM supplemented with 10% fetal bovine serum , 10% tryptose phosphate broth , and penicillin/streptomycin . Their identity has not been authenticated and they are routinely tested for mycoplasma contamination with all tests negative . CRISPR guides targeting the firefly luciferase gene , lacZ gene , human TRIM25 , ZAP ( also known as ZC3HAV1 ) , and KHNYN genes were cloned into BsmBI restriction enzyme sites in the lentiviral vector genome plasmid lentiCRISPRv2 ( Sanjana et al . , 2014 ) . The CRISPR guide sequences are: LacZ-G1: 5’- CGA TTA AGT TGG GTA ACG CC −3’ , Luciferase-G1: 5’-CTT TAC CGA CGC ACA TAT CG-3’ , TRIM25-G1: 5’-GAG CCG GTC ACC ACT CCG TG −3’ , ZAP-G1: 5’- ACT TCC ATC TGC CTT ACC GG −3’ , KHNYN-G1: 5’-GGG GGT GAG CGT CCT TCC GA-3’ , KHNYN-G2: 5’-CAG ACA CCG CAA AGC GAT CT-3’ . LentiCRISPR vectors encoding guide RNAs targeting KHNYN or LacZ were produced in HEK293T cells seeded in a 10 cm dish and transfected using 100 ul PEI with 8 µg of lentiCRISPRv2-Guide , 8 µg of pCRV1-HIV-Gag Pol ( Neil et al . , 2008 ) and 4 µg of pCMV-VSV-G ( Neil et al . , 2008 ) . Lentiviral vectors encoding guide RNAs targeting ZAP or TRIM25 were produced by transfecting HEK293T cells seeded in a six-well plate using 10 µl PEI with 0 . 5 µg pVSV-G ( Fouchier et al . , 1997 ) , 1 . 0 µg pCMV∆R8 . 91 ( Zufferey et al . , 1997 ) , and 1 . 0 µg LentiCRISPRv2-Guide . Virus containing supernatant was harvested 48 hr after transfection , rendered cell-free via filtration through 0 . 45 µM filters ( Millipore ) and used to transduce HeLa or HEK293T cells followed by selection in puromycin . The yeast two-hybrid screen was performed by Hybrigenics Services , S . A . S . , Paris , France ( http://www . hybrigenics-services . com ) . Full length ZAP-S , ZAP-L and KHNYN-2 were PCR-amplified and cloned into pB27 as a C-terminal fusion to LexA . These constructs were used as a bait to screen a random-primed Induced Macrophages cDNA library constructed into pP6 . pB27 and pP6 derive from the original pBTM116 ( Vojtek and Hollenberg , 1995 ) and pGADGH ( Bartel and Fields , 1995 ) plasmids , respectively . Clones were screened using a mating approach with YHGX13 ( Y187 ade2-101::loxP-kanMX-loxP , matα ) and L40∆Gal4 ( mata ) yeast strains as previously described ( Fromont-Racine et al . , 1997 ) . His + colonies were selected on a medium lacking tryptophan , leucine and histidine . Because KHNYN-2 had some autoactivating activity , the selection medium was supplemented with 10 mM 3-Aminotriazol . The prey fragments of the positive clones were amplified by PCR and sequenced at their 5’ and 3’ junctions . The resulting sequences were used to identify the corresponding interacting proteins in the GenBank database ( NCBI ) using a fully automated procedure . A confidence score ( PBS , for Predicted Biological Score ) was attributed to each interaction as previously described ( Formstecher et al . , 2005 ) . The PBS relies on two different levels of analysis . First , a local score takes into account the redundancy and independency of prey fragments , as well as the distribution of reading frames and stop codons in overlapping fragments . Second , a global score takes into account the interactions found in all the screens performed at Hybrigenics using the same library . This global score represents the probability of an interaction being nonspecific . The scores were divided into six categories: A ( highest confidence ) to D ( lowest confidence ) plus category E that demarcates interactions involving highly connected prey domains previously found several times in screens performed on libraries derived from the same organism and category F that indicates highly connected domains that have been confirmed as false-positives . The PBS scores have been shown to positively correlate with the biological significance of interactions ( Rain et al . , 2001; Wojcik et al . , 2002 ) . HeLa and HEK293T cells were grown to 70% confluence in six-well plates . HeLa cells were transfected according to the manufacturer’s instructions using TransIT-LT1 ( Mirus ) at the ratio of 3 µL TransIT- LT1 to 1 µg DNA . HEK293T cells were transfected according to the manufacturer’s instructions using PEI ( 1 mg/mL ) ( Sigma-Aldrich ) at the ratio of 4 µL PEI to1 µg DNA . For the HIV-1 experiments , 0 . 5 µg pHIV-1 and the designated amount of pKHNYN-FLAG , pGFP-FLAG or pGFP ( Swanson et al . , 2010 ) were transfected for a total of 1 µg DNA . For the MLV experiments , 0 . 65 µg pMLV , 0 . 25 µg pCR3 . 1 2 . 87 Vpu and 0 . 10 µg pGFP were transfected . The transfection medium was replaced with fresh medium after a 6 hr incubation ( HEK293T ) or 24 hr incubation ( HeLa ) . HeLa cells were seeded on six-well plates and transfected the following day with as described above . Approximately 48 hr post-transfection , HeLa cells were lysed in radioimmunoprecipitation assay ( RIPA ) buffer ( 150 mM NaCl , 1 mM EDTA , 1% Triton X-100 , 1% sodium deoxycholate , 0 . 1% SDS , 10 mM Tris–HCl pH 7 . 5 ) supplemented with complete protease inhibitor ( Roche ) and sheared using a 0 . 25G needle . Media was filtered through a 0 . 45 µM filter and virions were pelleted for 2 hr at 20 , 000 x g through a 20% sucrose cushion in phosphate-buffered saline ( PBS ) solution . The pellet was resuspended in 2X loading buffer ( 60 mM Tris–HCl pH 6 . 8 , 2% sodium dodecyl sulfate ( SDS ) , 10% glycerol , 10% β-mercaptoethanol , 0 . 1% bromophenol blue ) . Cell lysates and virions were resolved by 10% SDS-polyacrylamide gel electrophoresis ( PAGE ) , transferred to a nitrocellulose membrane ( GE Healthcare ) and blocked in 1% non-fat milk in PBS with 0 . 1% Tween 20 ( Fischer Bioreagents ) . Primary antibodies were incubated for 2 hr at room temperature . After washing in PBS , blots were incubated for 1 hr with the appropriate secondary antibody . Bound antibodies were visualized on the LI-COR ( Odyssey Fc ) measuring the immunofluorescence or using Amersham ECL Prime Western Blotting Detection reagent ( GE Lifesciences ) for HRP-linked antibodies using an ImageQuant ( LAS4000 Mini ) . For the co-immunoprecipitation experiments , lysates were resolved using precast Mini-PROTEAN TGX gels 8–16% gradient gels ( Bio-Rad ) and transferred to nitrocellulose membranes ( Bio-Rad ) . Antibodies used in study were 1:50 HIV- one anti-p24Gag ( Chesebro et al . , 1992 ) , 1:3000 anti-HIV-1 gp160/120 Rabbit ( ADP421; Centralized Facility for AIDS Reagents ( CFAR ) ) , 1:10 , 0000 anti-HSP90 ( sc7947 , Santa Cruz Biotechnology ) , 1:5000 anti-HSP90 Rabbit ( GeneTex , GTX109753 ) , 1:4000 anti-HSP90 Mouse ( SantaCruz , sc-515081 ) , 1:5000 anti-ZAP ( Abcam , ab154680 ) , 1:1000 anti-β-Actin Mouse ( Sigma , A2228 ) , 1:2500 anti-DYKDDDDK ( Rabbit ) ( Cell Signaling , 14793 ) , 1:2500 anti-FLAG ( Mouse ) ( Sigma , F1804 ) , 1:2500 anti-FLAG ( Rabbit ) ( Sigma , F7425 ) , 1:10000 anti-TRIM25 ( Abcam , ab167154 ) , 1:10000 anti-MLV p30 ( Rat ) ( ATCC CRL-1912 ) ( Chesebro et al . , 1983 ) , 1:10 , 000 Dylight 800- conjugated secondary antibodies ( Cell Signaling Technology , 5151S and 5257S ) , 1:5000 anti-rabbit HRP ( Cell Signaling Technology , 7074 ) , 1:5000 anti-mouse HRP ( Cell Signaling Technology , 7076 ) , 1:4000 anti-rabbit IRDye 800CW ( LI-COR , 926–32211 ) or 1:4000 anti-mouse IRDye 680RD ( LI-COR , 926–68070 ) . Media was recovered approximately 48 hr post-transfection and cell-free virus stocks were generated by filtering the media through 0 . 45 µM filters ( Millipore ) . The TZM-bl indicator cell line was used to quantify the amount of infectious virus ( Derdeyn et al . , 2000; Platt et al . , 1998; Wei et al . , 2002 ) . TZM-bl cells were seeded at 70% confluency in 24-well plates and infected by overnight incubation with virus stocks . 48 hr post infection , the cells were lysed and infectivity was measured by analyzing β-galactosidase expression using the Galacto-Star System following manufacturer’s instructions ( Applied Biosystems ) . β-galactosidase activity was quantified as relative light units per second using a PerkinElmner Luminometer . HEK293T cells in six-well plates were transfected with 800 ng of pKHNYN-1-FLAG , pKHNYN-2-FLAG , pGFP-FLAG as a control using 3 µL TransIT-LT1 per 1 µg of DNA added . For the experiments in which lysates were treated with Ribonuclease A ( RNase A ) , 500 ng of pHA-ZAP-L ( Kerns et al . , 2008 ) was also added . The cells were lysed on ice in lysis buffer ( 0 . 5% NP-40 , 150 mM KCl , 10 mM HEPES pH 7 . 5 , 3 mM MgCl2 ) supplemented with complete EDTA-free Protease inhibitor cocktail tablets ( Sigma-Aldrich ) , 10 mM N-Ethylmaleimide ( Sigma-Aldrich ) and PhosSTOP tablets ( Sigma-Aldrich ) . The lysates were sonicated and then centrifugated at 20 , 000 x g for 5 min at 4°C . 50 μl of the post-nuclear supernatants was saved as the input lysate and 450 μl were incubated with either 18 μg of anti-Flag antibody ( Sigma-Aldrich , F7425 ) or 4 . 275 μg of anti-ZAP antibody ( Abcam ) for one hour at 4°C with rotation . Protein G Dynabeads ( Invitrogen ) were then added and incubated for 3 hr at 4°C with rotation . The lysates were then washed four times with wash buffer ( 0 . 05% NP-40 , 150 mM KCl , 10 mM HEPES pH 7 . 5 , 3 mM MgCl2 ) before bound proteins were eluted with Laemmli buffer and boiled for 10 min . When indicated , RNase A ( Sigma-Aldrich ) was added to the post-nuclear supernatant and incubated for 30 min at 37°C . Protein expression was analyzed via western blot as described above . Total RNA was isolated from transfected HeLa cells using a QIAGEN RNeasy kit accordingly with the manufacturer’s instructions . Viral RNA was extracted from cell supernatants using a QIAGEN QIAamp Viral mini kit accordingly with the manufacturer’s instructions . 500 ng of purified cellular RNA was reverse transcribed using random hexamer primers and a High-Capacity cDNA Reverse Transcription kit ( Applied Biosystems ) . Quantitative PCR was performed using a QuantiStudio 5 System ( Thermo Fisher ) . For genomic RNA and nef mRNA in the cell lysate , the HIV-1 RNA abundance was normalized to GAPDH levels using the GAPDH Taqman Assay ( Applied Biosystems , Cat# Hs99999905_m1 ) . For the genomic RNA in the media , absolute quantification was determined using a standard curve of the HIV-1 provirus DNA plasmid . The genomic RNA primers were GGCCAGGGAATTTTCTTCAGA/TTGTCTCTTCCCCAAACCTGA ( forward/reverse ) and the probe was FAM-ACCAGAGCCAACAGCCCCACCAGA-TAMRA . The nef mRNA primers were GGCGGCGACTGGAAGAAGC/GATTGGGAGGTGGGTTGCTTTG-3’ ( forward/reverse ) ( Jablonski and Caputi , 2009 ) . Cells were seeded on 24-well plates on coverslips pre-treated with poly-lysine . HEK293T cells expressing a control guide RNA targeting the LacZ gene or a guide RNA targeting ZAP were transfected with 250 ng of pKHNYN-FLAG . 24 hr post-transfection , the cells were fixed with 4% paraformaldehyde for 15 min at room temperature , washed with PBS , and then washed with 10 mM glycine . The cells were then permeabilized for 15 min with 1% BSA and 0 . 1% Triton-X in PBS . Mouse anti-FLAG ( 1:500 ) and rabbit anti-ZAP ( 1:500 ) antibodies were diluted in PBS/0 . 01% Triton-X and the cells were stained for 1 hr at room temperature . The cells were then washed three times in PBS/0 . 01% Triton-X and incubated with Alexa Fluor 594 anti-mouse and Alexa Fluor 488 anti-rabbit antibodies ( Molecular Probes , 1:500 in PBS/0 . 01% Triton-X ) for 45 min in the dark . Finally , the coverslips were washed three times with PBS/0 . 01% Triton-X and then mounted on slides with Prolong Diamond Antifade Mountant with DAPI ( Invitrogen ) . Imaging was performed on a Nikon Eclipse Ti Inverted Microscope , equipped with a Yokogawa CSU/X1-spinning disk unit , under 60-100x objectives and laser wavelengths of 405 nm , 488 nm and 561 nm . Image processing and co-localization analysis was performed with NIS Elements Viewer and Image J ( Fiji ) software . The ‘analyze base composition’ tool in MacVector was used to calculate the CpG frequencies for the HIV-1NL4-3 ( NCBI accession number M19921 ) genomic RNA , MLV genomic RNA ( J02255 ) and Sindbis virus ( NC_001547 ) . The CpG frequencies were calculated using the following formula: number of CpG occurrences / ( frequency of C * frequency of G ) where frequency of the base is the number of occurrences of the base/total number of bases in sequence . Sindbis virus ( SINV ) , a kind gift from Penny Powell ( University of East Anglia ) , was expanded and titrated in BHK-21 cells ( Mazzon et al . , 2018 ) . Control , ZAP , TRIM25 or KHNYN HeLa cells were plated at 100 , 000 cells/well in 12-well plates . The following day , the cells were infected with Sindbis virus at a multiplicity of infection of 0 . 005 pfu/cell . After 90 mins , the infectious media was removed , the cells were washed once with PBS and then incubated with 1 ml of media . The media from the infected cells was harvested at 8 , 16 , 24 and 32 hr post-infection . 100 μl of serial diluted media from the cells ( from 10−1 to 10−8 ) were added onto BHK-21 cells in 96 well plates ( 8 , 000 cells/well plated the previous day ) . After 90 min , the media was removed and replaced with fresh media . An MTT assay was carried out on each plate 24 hr later . Briefly , 20 μl of 50 mg/ml Thiazolyl Blue Tetrazolium Bromide in PBS were added onto the cell media for 2 hr at 37°C , after which the supernatant is removed and replaced with 40 μl of a 1:1 solution of isopropanol and DMSO . 20 min later , 35 μl of the supernatant are transferred onto a 96 well plate and signal read at 570 nm . Values from this assay were used to determine the TCID50 and pfu/ml . Statistical significance was determined using unpaired two-tailed t tests calculated using Microsoft excel software . Data are represented as mean ± SD . Significance was ascribed to p values p < 0 . 05 . | Like many viruses , the genetic information of the human immunodeficiency virus ( or HIV for short ) is formed of molecules of RNA , which are sequences of building blocks called nucleotides . Once the virus is inside human cells , a protein called ZAP can identify viral RNAs by binding to a precise motif , a combination of two nucleotides called CpG . This allows the cell to destroy the viral RNA , thus preventing the virus from multiplying . However , HIV and other viruses that infect mammals are often able to ‘hide’ from ZAP because their genetic codes have many fewer CpG nucleotides than what would be expected by chance . ZAP by itself does not appear to be able to cut up RNA , so it is thought that it recruits other , as yet unidentified , proteins to destroy the genome of viruses . Here , Ficarelli et al . used genetic techniques to identify a new human protein called KHNYN that interacts with ZAP . First , a new version of the RNA genome of HIV was engineered , which contained higher numbers of CpGs: this CpG-enriched virus could be inhibited by ZAP in human cells . The experiments showed that increasing the amount of KHNYN protein led to lower levels of HIV genomes enriched in CpG . However , increasing the levels of KHNYN protein in mutant cells without ZAP had no effect on how well CpG-enriched HIV multiplied . CpG-enriched HIV and another related virus with many CpG nucleotides were able to multiply more successfully in mutant cells lacking the KHNYN protein than in normal cells . Further experiments also suggested that mutating a region of KHNYN which is likely to cut RNA prevented it from inhibiting HIV enriched with CpGs . Artificially manipulating the CpG nucleotide content of viral sequences could help create viruses useful for human health . For instance , weakened viruses could be designed for use in vaccines . Some human tumors have decreased levels of ZAP , and it could therefore be possible to build viruses that healthy cells can destroy , but which could multiply in and kill cancer cells . However , before these approaches can be developed , exactly how ZAP and KHNYN degrade strands of viral RNA needs to be characterized . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease",
"immunology",
"and",
"inflammation"
] | 2019 | KHNYN is essential for the zinc finger antiviral protein (ZAP) to restrict HIV-1 containing clustered CpG dinucleotides |
Mechanisms underlying the central regulation of food intake and fat accumulation are not fully understood . We found that neurosecretory protein GL ( NPGL ) , a newly-identified neuropeptide , increased food intake and white adipose tissue ( WAT ) in rats . NPGL-precursor gene overexpression in the hypothalamus caused increases in food intake , WAT , body mass , and circulating insulin when fed a high calorie diet . Intracerebroventricular administration of NPGL induced de novo lipogenesis in WAT , increased insulin , and it selectively induced carbohydrate intake . Neutralizing antibody administration decreased the size of lipid droplets in WAT . Npgl mRNA expression was upregulated by fasting and low insulin levels . Additionally , NPGL-producing cells were responsive to insulin . These results point to NPGL as a novel neuronal regulator that drives food intake and fat deposition through de novo lipogenesis and acts to maintain steady-state fat level in concert with insulin . Dysregulation of NPGL may be a root cause of obesity .
Dysregulated energy balance can result in obesity and lead to serious health problems such as diabetes and cardiovascular disease ( Steinberger and Daniels , 2003; Hill et al . , 2012; Ahima and Lazar , 2013 ) . Thus , it is important for human health to gain insight into the physiological mechanisms underlying the regulation of obesity . As obesity results mainly from overfeeding , most research to date has focused on hypothalamic regulation of feeding and satiety . Several hypothalamic neuropeptides and peripheral factors influence food intake and body mass ( Schwartz and Porte , 2005; Morton et al . , 2006 , 2014 ) . For instance , the arcuate nucleus ( Arc ) of the hypothalamus produces neuropeptide Y ( NPY ) /agouti-related peptide ( AgRP ) and α-melanocyte-stimulating hormone ( α-MSH ) , which are potent orexigenic and anorexigenic factors , respectively ( Schwartz and Porte , 2005; Morton et al . , 2006 , 2014 ) . In addition , leptin and insulin derived from peripheral tissues act on the hypothalamus to influence energy homeostasis ( Baskin et al . , 1999; Varela and Horvath , 2012 ) . Precise hypothalamic control of energy balance affects puberty , thermoregulation , energy storage , survival , and other critical processes during different life-history stages ( Schneider et al . , 2013 ) . Prior to the last decade , several bioactive peptides in the brain were discovered: neuromedin S , TLQP-21 , nesfatin-1 , and neuroendocrine regulatory peptide ( NERP ) ( Mori et al . , 2005; Bartolomucci et al . , 2006; Oh-I et al . , 2006; Yamaguchi et al . , 2007 ) . These neuropeptides are also involved in energy homeostasis and feeding behavior ( Ida et al . , 2005; Bartolomucci et al . , 2006; Oh-I et al . , 2006; Toshinai et al . , 2010 ) . Recently , nonadecaneuropeptide derived from Acyl-CoA binding domain-containing seven was found to be a novel anorexigenic factor in the mouse hypothalamus ( Lanfray et al . , 2016 ) . Despite considerable progress in understanding the regulation of energy homeostasis over the last several decades , the neural control of hyperphagia or obesity is not completely understood . To further understand the mechanism regulating energy intake and/or storage , we sought to identify previously unknown bioactive substances in the hypothalamus that regulate energy metabolism . As part of our search for novel neuropeptides and/or peptide hormone precursors in the hypothalamus , we identified a novel cDNA in the chicken hypothalamus and deduced a precursor protein including a secretory protein of 80 amino acids ( Ukena et al . , 2014 ) . This small protein has Gly-Leu-NH2 at its C-terminal and was named neurosecretory protein GL ( NPGL ) . In chickens , subcutaneous infusion of NPGL increased body mass gain , suggesting that NPGL may be involved in growth processes , including energy homeostasis ( Ukena et al . , 2014 ) . Subsequently , we found homologous Npgl genes in mammals , including human , rat , and mouse; the primary structure of NPGL is highly conserved among mammals and avian species ( Figure 1—figure supplement 1A ) . Rat NPGL assumes a circular structure , although the mature structure has not been determined ( Figure 1A ) . Given the effects of NPGL administration observed in chickens , along with the highly conserved nature of this gene across species , we hypothesized that NPGL and its precursor serve a prominent , unexplored role in energy homeostasis in mammals . More recently , we found that NPGL could induce food intake in mice ( Matsuura et al . , 2017 ) . However , the physiological significance of NPGL in metabolic control in mammals remains to be elucidated . 10 . 7554/eLife . 28527 . 003Figure 1 . Structure of NPGL and expression of NPGL in rats . ( A ) The amino acid structure of NPGL is shown schematically . The bold line between cysteine residues indicates a disulfide bond . ( B ) Expression levels of the NPGL-precursor mRNA in the entire brain and different brain regions , including the telencephalon , diencephalon , mesencephalon , cerebellum , and mediobasal hypothalamus ( n = 4 ) . ( C ) Western blot analysis of mature NPGL in the hypothalamus . Synthetic NPGL served as a reference marker ( 1 ) . The extract of the hypothalami from five rats ( 2 ) . ( D ) Schematic representation of the localization of NPGL-immunoreactive fibers ( blue dots ) and cells ( red dots ) in the mediobasal hypothalamus . Abbreviations; Arc: the arcuate nucleus , ArcLP: lateroposterior part of the Arc , ArcMP: medial posterior part of the Arc , DMH: doromedial hypothalamus , f: fornix , LH: lateral hypothalamus , PMD: dorsal premammillary nucleus , PMV: ventral premammillary nucleus , Te: terete hypothalamic nucleus , VMH: ventromedial hypothalamus , and VTM: ventral tuberomammillary nucleus . ( E–G ) Photomicrographs of the cells containing NPGL-precursor mRNA in the mediobasal hypothalamus . The squares including the ArcLP and VTM are shown magnified in ( F ) and ( G ) , respectively . Arrowheads in ( G ) indicate signals . Scale bar = 100 μm . ( H and I ) Photomicrographs of NPGL-immunoreactive cells in the ArcLP ( H ) and VTM ( I ) . Arrowheads in ( I ) indicate signals . Scale bar = 100 μm . ( J ) Photomicrograph of NPGL-immunoreactive fibers between the DMH and VMH . Scale bar = 100 μm . Mean ± s . e . m . ( one-way ANOVA with Tukey’s test as a post-hoc test: ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 00310 . 7554/eLife . 28527 . 004Figure 1—figure supplement 1 . Amino acid sequences and expression site of NPGL . ( A ) Alignment of NPGL-precursor proteins deduced from human , rat , mouse , and chicken cDNA sequences . Black and gray boxes highlight fully conserved and highly conserved amino acids , respectively . The predicted mature sequences of NPGL are underlined . The two conserved Cys ( C ) residues , which are involved in the intramolecular disulfide bond formation , are indicated by asterisks . Gaps , indicated by hyphens , were inserted to optimize the sequence alignment . ( B–D ) Expression level of NPGL-precursor mRNA in various tissues of human ( B ) , rat ( C ) and mouse ( D ) measured by real-time RT-PCR . Each value for the NPGL-precursor mRNA represents the mean from duplicate analyses . Data were normalized to β-actin ( ACTB ) mRNA and expressed relative to NPGL-precursor mRNA expression in the whole brain . ( E ) Alignment of mature NPGL , its analogs NPGL-Gly and NPGL ( 32-80 ) , and NPGM . Gray boxes highlight conserved amino acids . The two conserved Cys ( C ) residues , which are involved in the intramolecular disulfide bond formation , are indicated by asterisks . ( F ) Dot blot analysis of synthetic NPGL , NPGM and NPGL ( 32-80 ) using an antibody raised in rabbit ( left panel ) or guinea pig ( right panel ) . ( G–L ) Photomicrographs of NPGL-immunoreactive cells in the ArcLP ( G ) and VTM ( J ) . Specificity control was obtained by preadsorbing the working dilution of the antibody with a saturating concentration ( 10 μg/ml ) of NPGL or NPGM in the ArcLP ( H and I ) and VTM ( K and L ) . Scale bar = 100 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 004 The present investigation sought to characterize whether NPGL impacts food intake and energy metabolism using a rat model . To accomplish this goal , we first examined the pattern of expression in the brain and peripheral tissues , along with the specific localization and distribution of NPGL-producing cells in the brain . Subsequently , we investigated the biological action of NPGL and its precursor by overexpression of the precursor gene for Npgl in the hypothalamus , and intracerebroventricular ( i . c . v . ) infusion of NPGL or a specific antibody directed against this protein . We further examined the effects of NPGL on food intake , blood chemistry , and body composition when animals were fed normal chow , high calorie diet , or macronutrient diet . Finally , we determined that NPGL plays a role in monitoring energetic status and appropriate adjustment of feeding and energy metabolism .
First , we examined the expression of Npgl mRNA in the brain and various peripheral tissues in human , rat , and mouse . The results revealed that Npgl mRNA expression is high in the brain and testis of human and rat , but in mouse it only exhibits high expression in the brain ( Figure 1—figure supplement 1B–D ) . To gain insight into the functional role of NPGL , we assessed the neuroanatomical localization of Npgl mRNA and its mature protein in the rat brain . Npgl mRNA was exclusively expressed in the mediobasal hypothalamus ( Figure 1B ) . Western blot analysis demonstrated the presence of the mature form of NPGL in the hypothalamus ( Figure 1C ) . Histological analyses showed that both Npgl mRNA and its mature protein were localized to the lateroposterior division of the Arc ( ArcLP ) and the ventral tuberomammillary nucleus ( VTM ) ( Figure 1D–I ) . Immunoreactive cells were detected only after rats were treated with colchicine to prevent axonal transport of NPGL . NPGL-immunoreactive fibers were observed only in the hypothalamus ( Figure 1D and J ) . This latter result suggests that NPGL’s physiological role is mostly limited to regulation of the hypothalamus . Because NPGL is mainly localized to the Arc , an important feeding and energy metabolic center , we speculated that NPGL produced in the ArcLP might be involved in the regulation of feeding behavior and body mass . Therefore , we investigated the effects of chronic exposure to NPGL on food intake , blood chemistry , and body composition in the following experiments . The analysis of in vivo translation of the NPGL-precursor gene was conducted to survey the more long-term effect on phenotype , including body mass . To explore the effects of chronic NPGL-precursor gene ( Npgl ) overexpression in the mediobasal hypothalamus , we prepared an adeno-associated virus ( AAV ) that would allow chronic expression of the NPGL precursor protein ( Figure 2—figure supplement 1A ) and injected it into the hypothalamus of rats ( Figure 2—figure supplement 1B and C ) . We then monitored food intake ( high-fat/high-sucrose diet; high calorie diet , and normal chow ) for 6 weeks ( Figure 2 ) . When we investigated the effect of Npgl overexpression in rats fed with high calorie diet , cumulative food intake , body mass , blood insulin , and WAT mass also markedly increased ( Figure 2A–D and Table 1A ) . 10 . 7554/eLife . 28527 . 005Figure 2 . The effects of Npgl overexpression . The panels show the data obtained by the injection of AAV-based control vector ( AAV-CTL ) or AAV-based NPGL-precursor gene vector ( AAV-NPGL ) in high calorie diet ( A–D ) and normal chow ( E–H ) . ( A ) Cumulative food intake ( n = 8 ) . ( B ) The body mass ( n = 8 ) . ( C ) Serum insulin levels ( n = 6–7 ) . ( D ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 6–7 ) . ( E ) Cumulative food intake ( n = 6–7 ) . ( F ) The body mass ( n = 6–7 ) . ( G ) Serum insulin levels ( n = 6–7 ) . ( H ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 6–7 ) and representative photographs of retroperitoneal WAT . Scale bar = 5 cm . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 00510 . 7554/eLife . 28527 . 006Figure 2—figure supplement 1 . Construction of AAV-based vectors and verification of overexpression . ( A ) Schematic structures of AAV-based control vector ( AAV-CTL ) and AAV-based NPGL-precursor gene vector ( AAV-NPGL ) . Abbreviations: L- and R-ITR: left and right inverted terminal repeat , CMV: cytomegalovirus promoter , IRES: internal ribosome entry site , GFP: green fluorescence protein , pA: polyadenylation signal sequence . ( B ) Representative micrographs of mediobasal hypothalamus at 14 days after injection of the AAV-CTL or AAV-NPGL . Scale bar = 500 µm . ( C ) Western blot analysis of NPGL in the hypothalamus . Synthetic NPGL served as a positive control ( PC ) . The extract of the hypothalamus from one animal was separated on HPLC , and the fractions were electrophoresed on 15% polyacrylamide gel . A single band of mature NPGL was only detected in the hypothalamic extract at 14 days after the injection of AAV-NPGL . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 00610 . 7554/eLife . 28527 . 007Figure 2—figure supplement 2 . The effects of Npgl overexpression in normal chow . ( A ) Changes in daily food intake ( n = 6–7 ) . ( B ) The frequency of various adipocyte sizes measured in 1000 μm2 areas and representative photographs in sections of retroperitoneal WAT ( n = 5 ) . Scale bar = 100 μm . ( C ) Ratio of the interscapular BAT mass/body mass ( n = 6–7 ) . ( D ) Ratios of liver , heart , and kidney mass/body mass ( n = 6–7 ) . ( E ) Ratios of soleus and gastrocnemius muscle mass/body mass ( n = 6–7 ) . ( F ) Masses of soleus and gastrocnemius muscles ( n = 6–7 ) . ( G ) Body length ( nose to anus ) ( n = 6–7 ) . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 00710 . 7554/eLife . 28527 . 008Table 1 . Blood chemistry during Npgl overexpression . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 008A . Blood chemistry during Npgl overexpression under high calorie diet . AAV-CTLAAV-NPGLGlucose ( mg/dl ) 144 ± 6 . 1156 ± 6 . 2Free Fatty Acid ( mEq/l ) 0 . 462 ± 0 . 0380 . 460 ± 0 . 033Triglyceride ( mg/dl ) 192 ± 12 . 7184 ± 13 . 7Cholesterol ( mg/dl ) 149 ± 9 . 6127 ± 6 . 1†Insulin ( ng/ml ) 3 . 56 ± 0 . 776 . 71 ± 0 . 90*B . Blood chemistry during Npgl overexpression under normal chow . AAV-CTLAAV-NPGLGlucose ( mg/dl ) 105 ± 2 . 6108 ± 2 . 6Free Fatty Acid ( mEq/l ) 0 . 703 ± 0 . 0280 . 618 ± 0 . 031†Triglyceride ( mg/dl ) 202 ± 13 . 4253 ± 23 . 9†Cholesterol ( mg/dl ) 80 . 8 ± 4 . 085 . 4 ± 2 . 1Insulin ( ng/ml ) 3 . 52 ± 0 . 213 . 77 ± 0 . 16Leptin ( ng/ml ) 12 . 0 ± 0 . 5918 . 5 ± 1 . 60***Corticosterone ( ng/ml ) 511 ± 2 . 1512 ± 2 . 1†p <0 . 1 , *p<0 . 05 , ***p<0 . 005 . In rats fed normal chow , daily food intake did not change until approximately 3 weeks after the AAV injection , but then slightly increased thereafter ( Figure 2—figure supplement 2A ) . Overall , cumulative food intake significantly increased ( Figure 2E ) . No overall effect on body mass was observed ( Figure 2F ) . Despite no effect on body mass , Npgl overexpression caused a significant increase in the mass of WAT and the size of adipocytes ( Figure 2H and Figure 2—figure supplement 2B ) . The masses of interscapular brown adipose tissue ( BAT ) , liver , heart , and kidney remained unchanged ( Figure 2—figure supplement 2C and D ) , while the masses of soleus and gastrocnemius muscles did not increase as much as controls , as did body length ( Figure 2—figure supplement 2E–G ) . Blood leptin increased , but other blood measures , including insulin , did not change ( Table 1B and Figure 2G ) . We measured the O2 consumption ( VO2 ) and CO2 production ( VCO2 ) at four weeks after AAV injection in normal chow-fed rats . There were no significant differences in the VO2 and VCO2 of treatment and control animals ( Figure 3A and B ) . Although overall energy expenditure did not change , the respiratory quotient ( RQ ) was significantly elevated in the Npgl overexpression group ( Figure 3C ) . The locomotor activity of the two groups did not differ ( Figure 3D ) . Therefore , it is likely that the elevation of RQ value after Npgl overexpression is caused either by an upregulation of lipogenesis or downregulation of lipolysis in some tissues . To elucidate these possibilities , we analyzed the expression levels of lipogenic and lipolytic enzyme mRNAs in retroperitoneal WAT ( rWAT ) and liver . Specifically , we chose acetyl-CoA carboxylase ( ACC ) , fatty acid synthase ( FAS ) , stearoyl-CoA desaturase 1 ( SCD1 ) , glycerol-3-phosphate acyltransferase 1 ( GPAT1 ) , and adiponutrin ( ADPN ) as lipogenic enzymes , and carnitine palmitoyltransferase 1a ( CPT1a ) , adipose triglyceride lipase ( ATGL ) , and hormone-sensitive lipase ( HSL ) as lipolytic enzymes ( Shi and Burn , 2004 ) . We found that mRNA expression levels of Acc , Fas , Adpn , and Atgl in rWAT significantly increased after Npgl overexpression , but no differences were detected in liver ( Figure 3E ) . The protein level of FAS in rWAT also increased after Npgl overexpression , but the amount of phosphorylation of HSL was not different ( Figure 3F–H ) . The fatty acid ratio of 16:1/16:0 significantly increased in rWAT of Npgl overexpression rats , but not 18:1/18:0 , showing the activation of enzymatic activity of SCD1 ( Figure 3I and J ) . These results indicate that the activation of NPGL-induced de novo lipogenesis occurs in WAT but not in liver . 10 . 7554/eLife . 28527 . 009Figure 3 . The effects of Npgl overexpression on O2/CO2 metabolism and lipogenic and lipolytic enzymes . All panels show data obtained by the injection of AAV-based control vector ( AAV-CTL ) or AAV-based NPGL-precursor gene vector ( AAV-NPGL ) in normal chow . ( A ) O2 consumption ( VO2 ) measured in the metabolic cage ( n = 6–7 ) . ( B ) CO2 production ( VCO2 ) measured in the metabolic cage ( n = 6–7 ) . ( C ) The respiratory quotient ( RQ ) measured in the metabolic cage ( n = 6–7 ) . ( D ) The spontaneous locomotor activity measured by infrared ray passive sensor ( n = 6–7 ) . ( E ) mRNA expression levels for lipogenic ( Acc , Fas , Scd1 , Gpat1 , and Adpn ) and lipolytic ( Cpt1a , Atgl , and Hsl ) enzymes in retroperitoneal WAT ( rWAT ) and liver ( n = 5–7 ) . Representative photographs ( F ) of the western blot and protein expression levels of FAS ( G ) and phosphorylated HSL ( pHSL ) ( H ) in retroperitoneal WAT ( n = 6 ) . ( I ) Ratio of fatty acids ( 16:1 and 16:0 ) in retroperitoneal WAT ( n = 6–7 ) . ( J ) Ratio of fatty acids ( 18:1 and 18:0 ) in retroperitoneal WAT ( n = 6–7 ) . Mean ± s . e . m . ( two-way ANOVA followed by Bonferroni’s test or Student’s t-test: *p<0 . 05 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 009 In addition to the analysis in vivo translation of Npgl overexpression , we performed a 13 day chronic i . c . v . infusion of mature small protein , NPGL ( 15 nmol/day ) using osmotic pumps in rats fed with high calorie diet and normal chow . When we investigated the effects of chronic i . c . v . infusion of NPGL in rats fed with high calorie diet , food intake in the light period , blood insulin , cholesterol , and leptin all increased without changes in overall body mass ( Figure 4A–C and Table 2A ) . Increases in the masses of inguinal WAT ( iWAT ) and perirenal WAT were also observed under this diet ( Figure 4D ) . 10 . 7554/eLife . 28527 . 010Figure 4 . The effects of chronic i . c . v . infusion of NPGL or antibody against NPGL . All panels show data obtained by the infusion of vehicle ( CTL ) or NPGL ( A–I ) in high calorie diet ( A–D ) and normal chow ( E–I ) , and the control IgG ( CTL ) or the antibody against NPGL ( NPGL-Ab ) in high calorie diet ( J ) . ( A ) The average of food intake during the light period , dark period , or over 24 hr ( n = 8 ) . ( B ) Serum insulin levels ( n = 8 ) . ( C ) Body mass ( n = 8 ) . ( D ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 8 ) . ( E ) The average of food intake during the light period , dark period , or over 24 hr ( n = 7 ) . ( F ) Serum insulin levels ( n = 7 ) . ( G ) Body mass ( n = 7 ) . ( H ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 7 ) . ( I ) The frequency of various adipocyte sizes measured in 1000 μm2 areas and representative photographs in sections of retroperitoneal WAT after the infusion of vehicle ( CTL ) or NPGL in normal chow ( n = 4–5 ) . Scale bar = 100 μm . ( J ) The frequency of various adipocyte sizes measured in 2000 μm2 areas and representative photographs in sections of retroperitoneal WAT after the infusion of the control IgG ( CTL ) or antibody against NPGL ( NPGL-Ab ) in high calorie diet ( n = 4 ) . Scale bar = 100 μm . Mean ± s . e . m . ( Student’s t-test or one-way ANOVA with Tukey’s test as a post-hoc test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01010 . 7554/eLife . 28527 . 011Figure 4—figure supplement 1 . The effects of chronic i . c . v . infusion of NPGL on lipogenic and lipolytic enzymes . All panels show data obtained by the infusion of vehicle ( CTL ) or NPGL in normal chow . ( A ) mRNA expression levels for lipogenic ( Acc , Fas , Scd1 , Gpat1 , and Adpn ) and lipolytic ( Cpt1a , Atgl , and Hsl ) enzymes in retroperitoneal WAT ( rWAT ) and liver ( n = 5–7 ) . Representative photographs ( B ) of the western blot and protein expression levels of FAS ( C ) and phosphorylated HSL ( pHSL ) ( D ) in retroperitoneal WAT ( n = 7 ) . ( E ) Ratio of fatty acids ( 16:1 and 16:0 ) in retroperitoneal WAT ( n = 6–7 ) . ( F ) Ratio of fatty acids ( 18:1 and 18:0 ) in retroperitoneal WAT ( n = 6–7 ) . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01110 . 7554/eLife . 28527 . 012Figure 4—figure supplement 2 . The effects of acute i . c . v . injection of NPGL on lipogenic and lipolytic enzymes . All panels show data obtained after 5 hr of i . c . v . injection of vehicle ( CTL ) or NPGL in normal chow . mRNA expression levels for lipogenic ( Acc , Fas , Scd1 , and Gpat1 ) and lipolytic ( Cpt1a , Atgl , and Hsl ) enzymes in retroperitoneal WAT ( rWAT ) and liver ( n = 5 ) . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01210 . 7554/eLife . 28527 . 013Figure 4—figure supplement 3 . The effects of chronic i . c . v . infusion of synthetic analogs in normal chow . ( A ) Sequences of mature NPGL , and its analogs NPGL-Gly and NPGL ( 32-80 ) . The two conserved Cys ( C ) residues , which are involved in the intramolecular disulfide bond formation , are indicated by bold . ( B ) Ratios of inguinal , epididymal , and retroperitoneal WAT mass/body mass after the infusion of vehicle ( CTL ) or NPGL-Gly ( n = 7 ) . ( C ) Ratios of inguinal , epididymal , and retroperitoneal WAT mass/body mass after the infusion of vehicle ( CTL ) or NPGL ( 32-80 ) ( n = 4–7 ) . Mean ± s . e . m . ( Student’s t-test: ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01310 . 7554/eLife . 28527 . 014Figure 4—figure supplement 4 . The effects of chronic i . c . v . infusion of antibody against NPGL in high calorie diet . ( A ) The average intake of a high calorie diet during infusion of control IgG ( CTL ) or antibody against NPGL ( NPGL-Ab ) during the light period , dark period , or over 24 hr ( n = 8 ) . ( B ) Body mass ( n = 8 ) . ( C ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 8 ) . Mean ± s . e . m . ( Student’s t-test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01410 . 7554/eLife . 28527 . 015Table 2 . Blood chemistry during chronic i . c . v . infusion of NPGL or the antibody against NPGL . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 015A . Blood chemistry during chronic i . c . v . infusion of NPGL under high calorie diet . CTLNPGLGlucose ( mg/dl ) 138 ± 5 . 7135 ± 2 . 5Free Fatty Acid ( mEq/l ) 0 . 526 ± 0 . 0430 . 56 ± 0 . 066Triglyceride ( mg/dl ) 101 ± 13 . 2149 ± 20 . 9†Cholesterol ( mg/dl ) 99 . 0 ± 5 . 1120 . 4 ± 5 . 3*Insulin ( ng/ml ) 2 . 53 ± 0 . 456 . 03 ± 1 . 22*Leptin ( ng/ml ) 8 . 76 ± 1 . 017 . 07 ± 1 . 6***B . Blood chemistry during chronic i . c . v . infusion of NPGL under normal chow . CTLNPGLGlucose ( mg/dl ) 123 ± 7 . 0128 ± 7 . 2Free Fatty Acid ( mEq/l ) 0 . 276 ± 0 . 0170 . 289 ± 0 . 017Triglyceride ( mg/dl ) 73 . 0 ± 4 . 5101 . 7 ± 11 . 6†Cholesterol ( mg/dl ) 66 . 5 ± 2 . 167 . 5 ± 3 . 4Insulin ( ng/ml ) 4 . 51 ± 0 . 264 . 87 ± 0 . 33Leptin ( ng/ml ) 8 . 84 ± 1 . 47 . 93 ± 0 . 70Corticosterone ( ng/ml ) 523 ± 3 . 6519 ± 1 . 4C . Blood chemistry during chronic i . c . v . infusion of the antibody against NPGL under high calorie diet . CTLNPGL-AbGlucose ( mg/dl ) 134 ± 2 . 38137 ± 1 . 84Free Fatty Acid ( mEq/l ) 0 . 529 ± 0 . 0780 . 479 ± 0 . 048Triglyceride ( mg/dl ) 129 ± 14 . 6118 ± 18 . 8Cholesterol ( mg/dl ) 111 . 9 ± 3 . 798 . 5 ± 4 . 7*Insulin ( ng/ml ) 2 . 23 ± 0 . 292 . 53 ± 0 . 33Leptin ( ng/ml ) 9 . 14 ± 0 . 857 . 24 ± 0 . 51†D . Blood chemistry during chronic i . c . v . infusion of NPGL under macronutrient diet . CTLNPGLGlucose ( mg/dl ) 116 ± 4 . 4124 ± 4 . 4Free Fatty Acid ( mEq/l ) 0 . 359 ± 0 . 0340 . 416 ± 0 . 042Triglyceride ( mg/dl ) 74 . 4 ± 6 . 497 . 7 ± 7 . 8*Cholesterol ( mg/dl ) 99 . 7 ± 3 . 590 . 8 ± 5 . 5Insulin ( ng/ml ) 1 . 94 ± 0 . 163 . 05 ± 0 . 68Leptin ( ng/ml ) 3 . 92 ± 0 . 326 . 24 ± 0 . 33***†p <0 . 1 , *p<0 . 05 , ***p<0 . 005 . Under normal chow , although no changes in food intake , blood chemistry , including insulin , or body mass were observed ( Figure 4E–G and Table 2B ) , the masses of rWAT and perirenal WAT significantly increased ( Figure 4H ) . The induction of de novo lipogenesis in rWAT but not liver was observed ( Figure 4—figure supplement 1 ) . In addition , acute i . c . v . injection of NPGL also induced de novo lipogenesis in rWAT ( Figure 4—figure supplement 2 ) . Analysis of the structure-activity relationship using a C-terminal Gly extended form of NPGL ( NPGL-Gly; Figure 4—figure supplement 3A ) and an N-terminal deletion form of NPGL [NPGL ( 32-80 ) ; Figure 4—figure supplement 3A] revealed that the longer form containing a disulfide bond is functional and C-terminal amidation is not necessary for induction of adiposity ( Figure 4—figure supplement 3B and C ) . Furthermore , we measured the size of adipocytes in rWAT in rats fed normal chow . Infusion of 7 . 5 and 15 nmol/day NPGL increased the frequency of medium ( 3001–4000 μm2 ) and large ( >5001 μm2 ) adipocytes , respectively , relative to vehicle controls , indicating a dose-dependent effect of NPGL ( Figure 4I ) . To suppress the activity of endogenous NPGL , we performed a 13 day i . c . v . infusion of an antibody directed against NPGL in rats fed high calorie diet . Although food intake and body mass did not change ( Figure 4—figure supplement 4A and B ) , blood cholesterol significantly decreased and the mass of iWAT and blood leptin tended to decrease ( Table 2C and Figure 4—figure supplement 4C ) . The lipid droplets in rWAT were significantly smaller after antibody infusion than those of the controls ( Figure 4J ) . This result suggests that endogenous NPGL promotes adiposity , a finding in agreement with the results observed with NPGL infusion and Npgl overexpression . As mentioned above , we observed that NPGL induced more potent food intake in animals fed a high calorie diet than those fed normal chow after Npgl overexpression and i . c . v . infusion of NPGL . Blood insulin increased only under a high calorie diet in both studies of Npgl overexpression and NPGL infusion . This high calorie diet includes high levels of sucrose as the carbohydrate source . Therefore , we hypothesized that NPGL may induce carbohydrate intake for the purpose of de novo lipogenesis and increase blood insulin levels . We tested this hypothesis using ad libitum selective feeding of macronutrient diets , protein , carbohydrate , and fat diets in the next series of experiments . Rats infused with NPGL increased their intake of carbohydrate diet during the light period ( Figure 5A and Figure 5—figure supplement 1 ) . This carbohydrate feeding may result in increased total calorie intake over 24 hr ( Figure 5A and Figure 5—figure supplement 1 ) . Blood triglyceride and leptin increased without changes in insulin and overall body mass ( Figure 5B and C , and Table 2D ) . The masses of WAT also significantly increased in this experiment ( Figure 5D ) . Although the masses of BAT , liver , heart , and kidney remained unchanged , the mass of the soleus muscle was lower than in controls ( Figure 5E–H ) . In addition , mRNA expression of lipogenic enzymes increased in rWAT , but not liver ( Figure 5I ) . Furthermore , mRNA expression of the fatty acid transporter ( Cd36 ) , glucose transporters ( Slc2a2 or Slc2a4 ) , and a carbohydrate metabolism gene ( glyceraldehyde-3-phosphate dehydrogenase; Gapdh ) were upregulated in rWAT or liver ( Figure 5I ) . 10 . 7554/eLife . 28527 . 016Figure 5 . Feeding behavior of macronutrient diets during chronic i . c . v . infusion of NPGL . ( A ) The calorie intake of macronutrient diets ( total , protein , carbohydrate , and fat ) by the infusion of vehicle ( CTL ) or NPGL during the light period , dark period , or over 24 hr ( n = 8 ) . ( B ) Serum insulin levels ( n = 8 ) . ( C ) Body mass ( n = 8 ) . ( D ) Ratios of inguinal , epididymal , retroperitoneal , and perirenal WAT mass/body mass ( n = 8 ) . ( E ) Ratio of the interscapular BAT mass/body mass ( n = 8 ) . ( F ) Ratios of liver , heart , and kidney mass/body mass ( n = 8 ) . ( G ) Ratios of soleus and gastrocnemius muscle mass/body mass ( n = 8 ) . ( H ) Masses of soleus and gastrocnemius muscles ( n = 8 ) . ( I ) mRNA expression levels for lipogenic ( Acc , Fas , Scd1 , Gpat1 , and Adpn ) and lipolytic ( Cpt1a , Atgl , and Hsl ) enzymes , sterol regulatory element binding protein 1 ( Srebp1 ) , lipoprotein lipase ( Lpl ) , Cd36 , fatty acid transport protein 1 ( Fatp1 ) , Slc2a4 , Gapdh , Slc2a2 , glucokinase ( Gk ) , and liver pyruvate kinase ( L–Pk ) in retroperitoneal WAT ( rWAT ) and liver ( n = 7–8 ) . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 01610 . 7554/eLife . 28527 . 017Figure 5—figure supplement 1 . The effects of chronic i . c . v . infusion of NPGL under macronutrient diets . ( A ) Changes in calorie intake of a macronutrient diet ( total , protein , carbohydrate , and fat ) during the light period ( n = 8 ) . ( B ) The changes in calorie intake of macronutrient diets ( total , protein , carbohydrate , and fat ) during the dark period ( n = 8 ) . ( C ) Changes in calorie intake of a macronutrient diet ( total , protein , carbohydrate , and fat ) over 24 hr ( n = 8 ) . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 017 To explore the role of NPGL in monitoring energetic state , the expression of Npgl mRNA was examined during a negative energy balance created through fasting . The expression of Npgl mRNA increased in rats fasted for 48 hr , with animals showing low blood glucose and insulin levels ( Figure 6A ) . In addition , to manipulate blood glucose and insulin levels , we examined the expression of Npgl mRNA in a diabetic rat model in which animals possess high blood glucose and low insulin following i . p . streptozotocin ( STZ; toxic to the beta cells of pancreas ) administration . The expression of Npgl mRNA increased in STZ-treated rats ( Figure 6B ) . In contrast , i . p . injection of insulin reduced the expression of Npgl mRNA ( Figure 6B ) . These results indicate that Npgl mRNA expression is upregulated by fasting and low insulin and downregulated by high insulin . 10 . 7554/eLife . 28527 . 018Figure 6 . The changes of Npgl mRNA by fasting , streptozotocin or insulin and responsibility of NPGL-producing cells to insulin . ( A ) The effects of fasting . Npgl mRNA levels in the mediobasal hypothalamus of fasted rats ( n = 8 ) . ( B ) The effects of streptozotocin ( STZ ) or insulin . Npgl mRNA levels in the mediobasal hypothalamus after 7 days of i . p . injection of STZ ( n = 7 ) or after 7 hr of i . p . injection of insulin ( n = 6–7 ) . ( C ) Induction of phosphorylated Akt ( red ) in NPGL-immunoreactive cells ( green ) after 45 min of i . c . v . injection of insulin in overnight fasted rats . Arrowheads indicate double-labeled cells ( co-localization of phospho-Akt and NPGL ) . Scale bar = 100 μm . Mean ± s . e . m . ( Student’s t-test: *p<0 . 05 , **p<0 . 01 , ***p<0 . 005 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 018 In the final experiment , whether NPGL-producing cells respond to insulin was investigated . I . c . v . injection of insulin immediately induced phosphorylation of Akt , an insulin-response kinase , in NPGL cells ( Figure 6C ) .
In the present study , we detected high expression of Npgl mRNA in the human , rat , and mouse brain , particularly in the mediobasal hypothalamus . Furthermore , we found that mature NPGL is produced in restricted hypothalamic nuclei and specifically plays a major role in food intake and fat accumulation in rats . Obesity is a burgeoning health issue worldwide , and yet how the brain regulates hyperphagia or adiposity is unclear . Under a high calorie diet including high sucrose , NPGL stimulates feeding behavior and increases blood insulin . At this time , blood glucose levels are not different in control and NPGL-treated groups , suggesting that insulin resistance is not induced by NPGL . Therefore , it seems that NPGL stimulates carbohydrate intake and induces de novo lipogenesis because these phenomena occur readily under high carbohydrate conditions ( Figure 7B ) . Subsequently , NPGL stimulates more carbohydrate usage to provide the glucose substrate for lipogenesis , and finally accelerated carbohydrate intake increases blood insulin levels as an indirect action of NPGL ( Figure 7B ) . This step can recur to achieve fat accumulation in WAT efficiently ( Figure 7B ) . It is well known that insulin is an anabolic hormone , including its role in fat deposition in adipocytes ( Dimitriadis et al . , 2011 ) . However , as revealed in the present experiments , administration of insulin inhibits the expression of Npgl mRNA ( Figure 6B ) . This result suggests that excess de novo lipogenesis by NPGL may be inhibited through negative feedback action of insulin ( Figure 7B ) . Together , the present findings point to NPGL as a novel neuronal lipogenic factor that maintains steady levels of fat storage in harmony with insulin . However , the excess and prolonged action of NPGL under high calorie diet finally leads to obesity ( Figure 7B ) . In developing or developed countries , abdominal adiposity easily leads to obesity , which is associated with an increased risk of metabolic syndrome and cardiovascular disease ( Oliveros et al . , 2014 ) . The relationship between the action of human NPGL and its pathophysiological status has yet to be assessed , but the conserved nature of NPGL indicates that it is likely to play an important metabolic role in humans . The present findings indicate that NPGL induces adiposity in WAT without remarkable changes in overall body mass in the experiments except for Npgl overexpression under high calorie diet . This increased fat accumulation , combined with no increase in muscle mass and body length , appears to explain the overall lack of body mass difference between the control and NPGL-treatment groups ( Figure 2—figure supplement 2E–G and Figure 5G and H ) . Furthermore , this study shows that NPGL-mediated fat accumulation principally results from de novo lipogenesis in WAT but not in liver even when animals are fed normal chow and increases in food intake and blood insulin levels do not occur ( Figure 7A ) . Indeed , the data from acute i . c . v . injection of NPGL show that NPGL can induce de novo lipogenesis in WAT rapidly ( Figure 4—figure supplement 2 ) . To our knowledge , our study is the first report on the presence of an endogenous hypothalamic mediator whose function is to control peripheral de novo lipogenesis independently of feeding behavior in animals . De novo lipogenesis is the biochemical process of converting non-lipid precursors into fatty acids for storage as energy ( Moore et al . , 2014 ) . Recent evidence suggests that de novo lipogenesis in WAT plays an important role in maintaining metabolic homeostasis , although de novo lipogenesis in non-adipose tissues including liver leads to ectopic lipid accumulation , lipotoxicity , and metabolic stress ( Lodhi et al . , 2011; Solinas et al . , 2015 ) . In adipose tissues , glucose taken up via SLC2A4 ( referred to as GLUT4 ) is converted into citrate via the glycolysis system and TCA cycle ( Rui , 2014; TeSlaa and Teitell , 2014 ) . Citrate is converted to palmitoleic acid by lipogenic enzymes including ACC , FAS and SCD1 , and eventually stored as triglycerides via GPAT1 and ADPN ( Shi and Burn , 2004; Rui , 2014; Jeong et al . , 2011 ) . In addition to lipogenic enzymes , the mRNA expression of Atgl , a lipolytic enzyme , also increased in NPGL-treated rats in the cases of three different experiments of this study ( Figures 3E and 5I , and Figure 4—figure supplement 1A ) . Furthermore , the mRNA expression of Cd36 , a fatty acid transporter , increased in WAT of NPGL-infused rats under macronutrient diets ( Figure 5I ) . Analyses using adipose tissue specific Atgl knockout mice and whole body Cd36 knockout mice suggest that the fat oxidation system and fatty acid transport are required for de novo lipogenesis in adipose tissues and liver ( Mottillo et al . , 2014; Clugston et al . , 2014 ) . This result suggests that the activation of de novo lipogenesis by NPGL may be coupled with lipolysis to accelerate fat accumulation in WAT . 10 . 7554/eLife . 28527 . 019Figure 7 . Potential role of NPGL in food intake and fat accumulation . Downward blue and upward red arrows indicate ‘inhibition’ and ‘stimulation’ , respectively and other-colored arrows demonstrate downstream events . ( A ) During the ordinary feeding state , NPGL stimulates de novo lipogenesis and promotes fat accumulation of WAT . ( B ) In the restricted feeding state like fasting , a decrease in circulating insulin induces the expression of Npgl mRNA and NPGL production in the hypothalamus . When food is readily available , NPGL stimulates carbohydrate intake and induces de novo lipogenesis using available carbohydrate . Subsequently , NPGL stimulates more carbohydrate usage to provide the glucose substrate for lipogenesis . Finally carbohydrate intake increases blood insulin levels . This step can be repeated to achieve fat accumulation in WAT efficiently . Excess de novo lipogenesis by NPGL is inhibited through negative feedback actions of insulin to maintain steady-state fat storage in WAT . Under high calorie diet , NPGL induces overweight and finally obesity . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 019 In contrast with its action during a positive energy state , NPGL may act to prepare animals for a negative energy state during food restriction , as fasting induced Npgl mRNA expression ( Figure 6A ) . It is well known that the expression of anabolic neuropeptides such as NPY and AgRP are upregulated under negative energy balance ( Morton et al . , 2006 ) . NPGL selectively stimulates carbohydrate intake during the light period when rats are not typically active or feeding ( Figure 5A ) . Circulating insulin levels are lower at this time than during the dark period ( Marcheva et al . , 2010 ) . Taken together , our data suggest that a decrease in circulating insulin during a restricted feeding state such as fasting may induce the expression of Npgl mRNA ( Figure 7B ) . This mechanism of NPGL induction may serve as a hunger signal to prepare for fat accumulation during a food available state . NPGL-producing cells are located in the ArcLP and the VTM in the posterior hypothalamus . In addition , we detected NPGL-immunoreactive fibers in the anterior Arc where NPY/AgRP- or α-MSH-producing cells are located . More recently , we found that NPGL innervated α-MSH-producing cells and the acute i . c . v . injection of NPGL stimulated feeding behavior in mice fed with normal chow ( Matsuura et al . , 2017 ) . Taken together , NPGL may regulate the transcription or activity of these well-known orexigenic or anorexigenic factors to modulate energy homeostasis involved in food intake and fat accumulation in rats . In the present study , chronic i . c . v . infusion of NPGL did not increase food intake in rats fed normal chow ( Figure 4E ) , whereas a single injection of a lower dose of NPGL has been shown to increase food intake in mice ( Matsuura et al . , 2017 ) . The reason for this discrepancy is unclear and future studies in which rats and mice are investigated under identical conditions are needed to clarify whether differences exist between species in the role played by this neuropeptide . Furthermore , the VTM is known as the origin of histaminergic neurons , and neuronal histamine is also an anorexigenic and anti-obesity factor ( Haas and Panula , 2003 ) . Thus , it is possible that NPGL-expressing cells interact with histamine neurons to regulate fat accumulation . NPGL-immunoreactive fibers were also detected around the dorsomedial hypothalamic nucleus ( DMH ) and ventromedial hypothalamic nucleus ( VMH ) ( Figure 1D and J ) . These regions also participate in energy homeostasis ( Schwartz and Porte , 2005; Morton et al . , 2006 ) . In addition , it has been demonstrated that lipogenesis and lipolysis in WAT are regulated in the mediobasal hypothalamus through the sympathetic nervous system ( Buettner et al . , 2008; Scherer et al . , 2011 ) . Therefore , it is possible that NPGL may participate in the sympathetic regulation of adiposity . Future studies are necessary to identify the target sites and neuronal networks concerning NPGL-producing neurons . In conclusion , the present data reveal novel molecular and functional relationships among the hypothalamus , insulin action , and peripheral adiposity and elucidate a previously unknown role for NPGL in energy homeostasis . Together , these findings reveal a novel neurochemical gateway where signals from the brain and periphery converge to monitor energetic status and adjust feeding and metabolism appropriately . The cognate receptor for NPGL has not yet been identified . The identification of the receptor will help to elucidate the mechanisms of NPGL activity in the mammalian brain . However , the present data bring to light an important role for NPGL in regulation of food intake , body mass , and onset of obesity .
Male Wistar rats ( 7 weeks old ) were purchased from a commercial company ( Kyudo , Saga , Japan ) and housed for one week at 23 ± 2°C under a 12:12 hr light–dark cycle with ad libitum access to tap water and normal chow ( CE-2; CLEA Japan , Tokyo , Japan ) . In several studies , a high-fat/high-sucrose diet ( HFSD; 32% of calories from fat/20% of calories from sucrose , D14050401; Research Diets , New Brunswick , NJ ) was also used as a high calorie diet . In a food choice experiment , protein diet ( 98 . 5% of calories from casein/1 . 5% of calories from L-cystine , D14082901; Research Diets ) , carbohydrate diet ( 43 . 6% of calories from corn starch/40 . 6% of calories from sucrose/15 . 8% calories from maltodextrin 10 , D14082902; Research Diets ) , and fat diet ( 92 . 2% of calories from lard/7 . 8% of calories from soy bean oil , D14082903; Research Diets ) were used as macronutrient diets . All animal procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals prepared by Hiroshima University ( permit numbers: C11-2 , C13-12 , and C13-17 ) . Rat NPGL of 80 amino acid residues , its analogs NPGL-Gly and NPGL ( 32-80 ) , and NPGM ( Figure 1—figure supplement 1E ) were synthesized by microwave-assisted solid-phase peptide synthesis using an automated peptide synthesizer ( Syro Wave; Biotage , Uppsala , Sweden ) as previously described ( Masuda et al . , 2015 ) . Rabbit antisera were produced following our published method ( Ukena et al . , 2010 ) using synthetic NPGL as the antigen . Antigen solution was mixed with Freund’s complete adjuvant and injected into rabbits . After a booster injection , blood was collected from each rabbit and the optimal serum with high titer was selected by a dot-blot analysis ( Figure 1—figure supplement 1F ) . The rabbit antibody against NPGL ( RRID: AB_2636993 ) was purified on an NPGL-conjugated sepharose 6B column . The guinea pig antibody against NPGL ( RRID: AB_2636992 ) was similarly produced . We followed a previously reported method ( Chen et al . , 2007 ) . The full-length open reading frame of rat NPGL was amplified from cDNA of the MBH and inserted into pAAV-IRES-GFP expression vector ( Cell Biolabs , San Diego , CA ) . The primers for rat NPGL were 5′-CGCGGATCCACCATGACTGATCCTGGGA-3′ for sense primer and 5′-CGGAATTCTTAGCTTCGATTTCTCTTTATT-3′ for antisense primer . AAV-based vectors AAV-DJ/8-NPGL-IRES-GFP for NPGL ( AAV-NPGL ) and AAV-DJ/8-IRES-GFP for control ( AAV-CTL ) were produced in 293AAV cells ( Cat# AAV-100; Cell Biolabs ) using AAV-DJ/8 Helper Free Packaging System containing pAAV-DJ/8 and pHelper plasmids ( Cell Biolabs ) . The triple plasmids ( AAV-DJ/8-NPGL-IRES-GFP or AAV-DJ/8-IRES-GFP , pAAV-DJ/8 , and pHelper ) were mixed with the Polyethylenimine MAX transfection reagent ( PEI-MAX; Polysciences , Warrington , PA ) ; the mixture was diluted with Opti-MEM I medium ( Life Technologies , Carlsbad , CA ) and added to 293AAV cells in 150 mm dishes . Transfected cells were cultured in DMEM containing 10% fetal bovine serum . For the purification of AAV-based vectors , three days after transfection , the cells and supernatants were harvested and purified using chloroform and condensed using Amicon Ultra-4 Centrifugal Filter Devices ( 100K MWCO; Merck Millipore , Billerica , MA ) . For AAV titration , 1 μl of AAV solution was treated with RQ1 DNase ( Promega , Madison , WI ) according to manufacturer’s directions . Virus titers were determined by quantitative PCR with EGFP primer pairs . The primers for EGFP were 5′-ACCACTACCTGAGCACCCAGTC-3′ for sense primer and 5′-GTCCATGCCGAGAGTGATCC-3′ for antisense primer . After titration , the AAV-based vectors were prepared at a concentration of 1 × 109 particles/µl and stored at −80°C until use . The stereotaxic coordinates used in the surgery were plotted according to the Rat Brain Atlas of Paxinos and Watson ( Paxinos and Watson , 2007 ) . Eight-week-old rats weighing 270–310 g were placed in a stereotaxic frame ( model 963; David Kopf Instruments , Tujunga , CA ) under isoflurane anesthesia . To survey Npgl mRNA expression in human , rat , and mouse tissues , multiple tissue cDNA panels were employed according to the manufacturer’s directions ( Clontech Laboratories , Mountain View , CA ) . In addition , the whole brain , telencephalon , diencephalon , mesencephalon , cerebellum , MBH , rWAT , and liver were dissected from rats and snap frozen in liquid nitrogen for RNA processing at the endpoint of NPGL infusion and Npgl overexpression . RNA was extracted using TRIzol reagent for brain tissues and liver ( Life Technologies ) or QIAzol lysis reagent for adipose tissues ( QIAGEN , Venlo , Netherlands ) following the manufacturer’s instructions . First-strand cDNA was synthesized from total RNA using a ReverTra Ace kit ( TOYOBO , Osaka , Japan ) . The sequences of primers used in this study are listed in Table 3 . PCR amplifications were conducted with THUNDERBIRD SYBR qPCR Mix ( TOYOBO ) using the following conditions: 95°C for 20 s , followed by 40 cycles of 95°C for 3 s , and 60°C for 30 s . The PCR products in each cycle were monitored using a StepOne Real-Time Thermal Cycler ( Life Technologies ) . Relative quantification of each gene was determined by the 2−ΔΔCt method using β-actin ( Actb ) for brain tissues and liver or ribosomal protein S18 ( Rps18 ) for rWAT as internal controls . 10 . 7554/eLife . 28527 . 020Table 3 . Sequences of oligonucleotide primers for real-time RT-PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 28527 . 020A . Sequences of oligonucleotide primers for real-time RT-PCR in human tissues . GeneForward primerReverse primerAccession no . NPGLGGAACCATGGCTTAGGAAGGCCTTAGGAGCTGAGAATATGTANM_001102659 . 1ACTBGGCACCACACCTTCTACAATAGGTCTCAAACATGATCTGGNM_001101 . 3B . Sequences of oligonucleotide primers for real-time RT-PCR in rat tissues . GeneForward primerReverse primerAccession no . NpglGGAACCATGGCTTAGGAAGGTCTAAGGAGCTGAGAATATGCALC003309AccGAGGTGGATCAGAGATTTCATTCAGCTCTAACTGGAAAGCNM_022193 . 1FasAGGATGTCAACAAGCCCAAGACAGAGGAGAAGGCCACAAANM_017332 . 1Scd1TGAAAGCTGAGAAGCTGGTGCAGTGTGGGCAGGATGAAGNM_139192 . 2Gpat1CAGCGTGATTGCTACCTGAACGGAAGGTGTGGACAAAGATNM_017274 . 1AdpnGGGCTACGCTATGTCTGAGCGAGACTGCACACGAAGGTGANM_001282324 . 1Cpt1aGGATGGCATGTGGGTAAAAGTACTGACACAGGCAGCCAAANM_031559 . 2AtglGCTGCAAGTGGGTTTTTGATGTGAACGGTAAGGCACAGGTNM_001108509 . 2HslGAGACGGAGGACCATTTTGACGGAGGTCTCTGAGGAACAGNM_012859 . 1Srebp1TCACAGATCCAGCAGGTCCCCGGTCCCTCCACTCACCAGGGTNM_001276707 . 1LplTCTCCTGATGATGCGGATTTCAACATGCCCTACTGGTTTCNM_012598 . 2Cd36GAGGTCCTTACACATACAGAGTTCGTTACAGACAGTGAAGGCTCAAAGATGNM_031561 . 2Fatp1GCGGCGTTCGGTGTGTACGCACGCGGATCAGAACAGANM_053580 . 2Slc2a2GACATCGGTGTGATCAATGCTGTCGTATGTGCTGGTGTGANM_012879Slc2a4CCTCCAGGATGAAGGAAACAGGGAGAAAAGCCCATCTAGGNM_012751GapdhCGGCAAGTTCAACGGCACAGACTCCACGACATACTCAGCACNM_017008 . 4GkTTGAGACCCGTTTCGTGTCAAGGGTCGAAGCCCCAGAGTNM_001270850 . 1L-PkTGATGATTGGACGCTGCAAGAGTTGGTCGAGCCTTAGTGATCNM_012624 . 3NpyTATCCCTGCTCGTGTTGATTGATGTAGTGTCGCAGANM_012614 . 2AgrpGCAGACCGAGCAGAAGATGTGACTCGTGCAGCCTTACACANM_033650 . 1PomcTAAGAGAGGCCACTGAACATGTCTATGGAGGTCTGAAGCANM_139326 . 2ActbGGCACCACACTTTCTACAATAGGTCTCAAACATGATCTGGNM_031144 . 3Rps18AAGTTTCAGCACATCCTGCGAGTATTGGTGAGGTCAATGTCTGCTTTCNM_213557 . 1C . Sequences of oligonucleotide primers for real-time RT-PCR in mouse tissues . GeneForward primerReverse primerAccession no . NpglGGAACCATGGCTTAGGAAGGTCTAAGGAGCTGAGAATATGCALC088498ActbGGCACCACACCTTCTACAATAGGTCTCAAACATGATCTGGNM_007393 . 4 Npgl mRNA levels in several conditions were measured . For food deprivation , MBH was harvested from forty-eight hours fasted rats and frozen down immediately in liquid nitrogen . The control rats were fed with ad libitum . For induction of experimental diabetes , rats were injected i . p . with streptozotocin ( 50 mg/kg; Sigma–Aldrich , St . Louis , MO ) . One week after injection when blood glucose levels were higher in the streptozotocin-injected rats , MBH was collected . For injection of insulin , rats were injected i . p . with regular human insulin ( 3 U/kg , humulin R ) after overnight fasting . Seven hours after injection , MBH were harvested . Blood glucose levels were lower in the insulin-injected rats at this time . Npgl mRNA expression levels in MBH were measured by real-time RT-PCR . For the detection of mature NPGL in the hypothalamus , the hypothalami from five rats were boiled and homogenized in 5% acetic acid . The homogenate was centrifuged at 10 , 000 × g for 20 min at 4°C and the precipitate was collected . After extraction with dimethyl sulfoxide , the supernatant was passed through a disposable C18 cartridge column ( Sep-Pak Vac; Waters , Milford , MA ) . The retained material was subjected to reversed-phase HPLC using a C4 column ( Protein C4-300 , 4 . 6 × 150 mm; Tosoh , Tokyo , Japan ) with a linear gradient of 35–75% acetonitrile containing 0 . 1% trifluoroacetic acid for more than 80 min at a flow rate of 0 . 5 ml/min . The fractions were evaporated , dissolved in SDS sample buffer , and subjected to 15% SDS-PAGE . After transfer onto polyvinylidene fluoride membrane ( Immobilon-P; Merck Millipore ) , the blot was probed with the rabbit antibody against NPGL ( RRID: AB_2636993 , 1:1000 dilution ) and incubated with horseradish peroxidase-labeled donkey anti-rabbit IgG ( RRID: AB_772206 , 1:1000 dilution , Cat# NA934; GE Healthcare , Little Chalfont , England ) . The protein bands were detected by ECL Prime ( GE Healthcare ) or ECL Pro ( PerkinElmer , Waltham , MA ) western blotting detection reagents . For the detections of FAS , phospho-HSL ( pHSL ) and total HSL in rWAT , the tissue was homogenized in ice-cold RIPA buffer ( Thermo Fisher Scientific , Waltham , MA ) , Halt Protease Inhibitor Cocktail , EDTA-Free ( Thermo Fisher Scientific ) and centrifuged at 10 , 000 × g for 20 min to remove the lipids . Protein content in the aqueous solution was measured using the Pierce BCA Protein Assay Kit ( Thermo Fisher Scientific ) with BSA as a standard , 30 μg-aliquots were subjected to 6% SDS-PAGE . The western blot procedure was similar to the one described above . The antibodies against FAS , pHSL , total HSL , and α-tubulin as an internal control were employed for FAS ( RRID: AB_2100801 , 1:1000 dilution , Cat# 10624–2-AP; Proteintech , Chicago , IL ) , pHSL ( RRID: AB_490997 , 1:500 dilution , Cat# 4126; Cell Signaling Technology , Danvers , MA ) , total HSL ( RRID: AB_2296900 , 1:500 dilution , Cat# 4107; Cell Signaling Technology ) , and for α-tubulin ( RRID: AB_10598496 , 1:2000 dilution , Cat# PM054; Medical and Biological Laboratories , Nagoya , Japan ) . Four weeks after injecting AAV-based vectors , indirect calorimetry was performed using an O2/CO2 metabolism-measuring system for small animals ( MK-5000RQ; Muromachi Kikai , Tokyo , Japan ) . The system monitored VO2 ( ml/min ) and VCO2 ( ml/min ) at 3 min intervals and calculated the respiratory quotient ( RQ ) ratio ( VCO2/VO2 ) . Locomotor activity was simultaneously measured using the SuperMex infrared ray passive sensor system ( Muromachi Kikai ) . Measurements were collected hourly over a 23 hr period ( light period: 10:00–21:00 , dark period: 21:00–9:00 ) after habituation for 30 min . Energy expenditure was calculated using the following equation: energy expenditure ( cal/kg/hr ) = VO2 ( ml/kg/hr ) × [3 . 815 + ( 1 . 232 × RQ ) ] ( Lusk , 1928 ) . For the analysis of endogenous SCD1 activity in rWAT , the lipids were extracted according to the previous method ( Folch et al . , 1957 ) . WAT ( 100 mg ) was extracted with 1 ml of chloroform: methanol ( 2:1 ) using beads crusher ( μT-12; TAITEC , Saitama , Japan ) and 0 . 25 ml of distilled water was added and mixed by inversion . After incubation for 30 min , the sample was centrifuged at 3000 × g and the lower organic phase was collected and evaporated . Extracted fatty acids were methylated using Fatty Acid Methylation Kit ( nacalai tesque , Kyoto , Japan ) and purified using Fatty Acid Methyl Ester Purification Kit ( nacalai tesque ) . The eluted solution was evaporated to dryness and kept at –20°C . The residues were resolved into hexane and fatty acids were identified by GC-MS ( JMS-T100 GCV; JEOL , Tokyo , Japan ) . The SCD1 activity was estimated as oleate to stearate ratio ( 18:1/18:0 ) and palmitoleate to palmitate ratio ( 16:1/16:0 ) from individual fatty acids . The 16:1/16:0 ratio seems to be a better indicator of endogenous SCD1 activity than the 18:1/18:0 ratio ( Sampath and Ntambi , 2008 ) . Serum levels of glucose , lipids , and hormones were measured using appropriate equipment , reagents , and kits . The GLUCOCARD G+ meter was used to measure glucose content ( Arkray , Kyoto , Japan ) . The NEFA C-Test Wako ( Wako Pure Chemical Industries , Osaka , Japan ) was used to measure free fatty acid levels . The Triglyceride E-Test Wako ( Wako Pure Chemical Industries ) was used to measure triglyceride levels and the Cholesterol E-Test Wako ( Wako Pure Chemical Industries ) for cholesterol content . The Rebis Insulin-rat T ELISA kit ( Shibayagi , Gunma , Japan ) was used to measure insulin levels , the Leptin ELISA kit ( Morinaga Institute of Biological Science , Yokohama , Japan ) for leptin , and the Corticosterone ELISA kit ( Assaypro , St . Charles , MO ) for corticosterone . Blood chemistry is listed in Tables 1 and 2 . Data were analyzed using the student’s t-test or two-way analysis of variance ( ANOVA ) followed by Bonferroni’s test for two groups . In addition , data from three or six groups were statistically analyzed using one-way ANOVA with Tukey’s post-hoc test as appropriate . Statistical significance was set at p<0 . 05 . All results are presented as the mean ± standard error of the mean ( ± s . e . m . ) . | Throughout history , our ancestors needed to accumulate fat to survive during times when food sources were scarce . However , for most people in the modern age , food is abundant and eating too much is a major cause of weight gain , obesity and diseases affecting the metabolism . Obesity in particular , can lead to diseases such as diabetes and heart disease . Hunger and appetite are regulated by proteins and other chemicals that act as messengers , for example insulin , and a region of the brain called the hypothalamus . However , the full mechanisms that regulate these sensations remain unclear . Only recently , a protein called NPGL was discovered in a part of the hypothalamus of birds and mammals . However , it was not known if NPGL plays a role in regulating eating habits and weight gain . Iwakoshi-Ukena et al . have now discovered that NPGL is found in the hypothalamus of rats and is regulated by diet and insulin . When the gene for NPGL was manipulated to produce too much of the protein , rats fed a high calorie diet started to eat more , and gained more weight and body fat . Adding additional NPGL to their brains had the same effect . When the animals were fed a normal diet , NPGL only moderately affected how much they ate , but it substantially increased how much fat they produced . Iwakoshi-Ukena et al . also observed that when animals were starved and insulin levels were low , the rats started to produce more NPGL . These results suggest that NPGL plays a role in fat storage when energy sources are limited , and can contribute to obesity when too much NPGL is produced in animals on a high calorie diet . These findings indicate that NPGL could be an additional brain chemical that regulates hunger and fat storage in mammals . A next step will be to reveal the specific mechanisms by which NPGL regulates overeating and fat accumulation . These findings will further advance the study and treatment of obesity and obesity-related diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Neurosecretory protein GL stimulates food intake, de novo lipogenesis, and onset of obesity |
In eukaryotes , DNA replication requires the origin recognition complex ( ORC ) , a six-subunit assembly that promotes replisome formation on chromosomal origins . Despite extant homology between certain subunits , the degree of structural and organizational overlap between budding yeast and metazoan ORC has been unclear . Using 3D electron microscopy , we determined the subunit organization of metazoan ORC , revealing that it adopts a global architecture very similar to the budding yeast complex . Bioinformatic analysis extends this conservation to Orc6 , a subunit of somewhat enigmatic function . Unexpectedly , a mutation in the Orc6 C-terminus linked to Meier-Gorlin syndrome , a dwarfism disorder , impedes proper recruitment of Orc6 into ORC; biochemical studies reveal that this region of Orc6 associates with a previously uncharacterized domain of Orc3 and is required for ORC function and MCM2–7 loading in vivo . Together , our results suggest that Meier-Gorlin syndrome mutations in Orc6 impair the formation of ORC hexamers , interfering with appropriate ORC functions .
The faithful inheritance of genetic information depends on the precise replication of DNA; errors such as under- or over-replication of DNA can lead to genetic instabilities , malignant transformation , or cell death ( Klingseisen and Jackson , 2011; Abbas et al . , 2013 ) . In eukaryotes , replication initiation machineries are loaded onto DNA during the G1-phase of the cell cycle , ‘licensing’ origins of replication for subsequent replisome assembly in S-phase ( Masai et al . , 2010; Abbas et al . , 2013 ) . The eukaryotic initiator , ORC ( origin recognition complex ) , performs a central role in this licensing reaction ( Bell and Stillman , 1992 ) , working with the loading factors Cdc6 and Cdt1 to recruit the MCM2–7 ( minichromosome maintenance ) replicative helicase into a pre-replicative complex ( pre-RC ) at replication origins ( Liang et al . , 1995; Cocker et al . , 1996; Tanaka et al . , 1997; Maiorano et al . , 2000; Nishitani et al . , 2000 ) . The MCM complex is loaded onto double-stranded DNA as an inactive double-hexamer ( Evrin et al . , 2009; Remus et al . , 2009; Gambus et al . , 2011 ) , and is activated only after origin firing in the subsequent S-phase to commence DNA replication ( Vijayraghavan and Schwacha , 2012 ) . How ORC accurately executes the replication initiation program , and how disregulation of ORC function is linked to chromosomal abnormalities and disease , constitute long-standing questions in the field . ORC is a heterohexameric protein complex containing the subunits Orc1–Orc6 ( Bell and Stillman , 1992; Gavin et al . , 1995; Gossen et al . , 1995; Rowles et al . , 1996; Moon et al . , 1999; Vashee et al . , 2001 ) . The Orc1–Orc5 subunits all contain AAA+ ( ATPases associated with a variety of cellular activities ) or AAA+-like domains ( Neuwald et al . , 1999; Iyer et al . , 2004; Speck et al . , 2005; Clarey et al . , 2006 ) , a subset of which use ATP binding and hydrolysis to support replicative helicase loading , DNA replication , and cell viability ( Chesnokov et al . , 2001; Bowers et al . , 2004; Giordano-Coltart et al . , 2005; Randell et al . , 2006 ) . How individual ORC subunits co-assemble with each other and the nature of the structural transitions that occur within ORC and other pre-RC components during nucleotide-dependent initiation steps are ill-defined , in part due to a paucity of structures for key ORC reaction states . Electron microscopy ( EM ) studies have provided medium-to-low resolution pictures of the general architecture of Saccharomyces cerevisiae and Drosophila ORC , and subunit positions within ORC further have been systematically mapped for the yeast complex ( Speck et al . , 2005; Clarey et al . , 2006; Chen et al . , 2008; Sun et al . , 2012 , 2013 ) ; however , poor similarity between the 3D reconstructions obtained for the different species , combined with conflicting biochemical data on ORC subunit interactions , have led to different models of subunit arrangement and DNA engagement for the two eukaryotic initiators ( Vashee et al . , 2001; Kneissl et al . , 2003; Chen et al . , 2007; Matsuda et al . , 2007; Siddiqui and Stillman , 2007; Clarey et al . , 2008; Li and Stillman , 2012; Sun et al . , 2012 ) . Whether these differences reflect true differences in structure , or potentially distinct functional states , remains unclear . Recently , mutations in genes encoding the pre-RC components Orc1 , Orc4 , Orc6 , Cdc6 , and Cdt1 have been linked to Meier-Gorlin syndrome ( MGS ) , a rare genetic disorder characterized by primordial dwarfism , small ears and aplastic or hypoplastic patella ( Guernsey et al . , 2011; Bicknell et al . , 2011a , b ) . Based on the fact that MGS mutations cluster in pre-RC components , it has been suggested that the clinical phenotype is caused by defects in DNA replication initiation ( Guernsey et al . , 2011; Bicknell et al . , 2011a , b ) . Consistent with this interpretation , cells derived from MGS patients with mutations in Orc1 show reduced recruitment of ORC subunits and replicative helicase components to chromatin , and are compromised in their ability to replicate plasmid DNA ( Bicknell et al . , 2011b ) ; recent work has also shown a link between origin licensing factors and non-replicative cellular processes , such as centrosome duplication and cilia formation , in MGS patient cells and in cells depleted of licensing proteins ( Hossain and Stillman , 2012; Stiff et al . , 2013 ) . Most MGS mutations in Orc1 lie in the N-terminal domain , which is not part of the central core of ORC , and have been implicated interfering with Orc1’s Cyclin E-CDK2 inhibitory function of centrosome duplication ( Hossain and Stillman , 2012 ) . By contrast , the means by which certain mutations in Orc4 and Orc6 result in an MGS phenotype has yet to be determined . Orc6 is particularly enigmatic in this regard: unlike Orc1–Orc5 , Orc6 is poorly conserved between budding yeast and metazoans , both in sequence and in aspects of its function ( Lee and Bell , 1997; Chesnokov et al . , 2001; Prasanth et al . , 2002; Chesnokov et al . , 2003; Semple et al . , 2006; Balasov et al . , 2007; Chen et al . , 2007; Duncker et al . , 2009; Bernal and Venkitaraman , 2011; Chen and Bell , 2011; Liu et al . , 2011; Takara and Bell , 2011 ) . Here we investigate the structural organization of Drosophila ORC ( DmORC ) and how Orc6 is recruited into this complex in metazoans . Using EM we find that apo-DmORC is conformationally heterogeneous , but that the non-hydrolysable nucleotide analog ATPγS locks the complex in a specific conformation that permitted the determination of its 3D structure to 22 Å resolution . Experimental localization of all subunits reveals that the subunit architecture of DmORC is preserved compared to S . cerevisiae ORC ( ScORC ) , reconciling different models proposed previously for the two systems . Interestingly , sequence analyses show that this similarity extends to the domain structure of Orc6 , suggesting analogous functions for Orc6 from yeast and metazoans . Consistent with this notion , we find that recruitment of Orc6 into ORC across multiple species is aided by the interaction of its C-terminus with a novel domain that resides between the AAA+-like and winged-helix domains of Orc3 . Strikingly , a mutation found in patients with Meier-Gorlin syndrome maps to a highly-conserved C-terminal segment in Orc6 , destabilizing the interaction of Drosophila and human Orc6 with both Orc3 alone and with the core Orc1–5 subcomplex . Orc6 transgenes containing mutations in the conserved C-terminal tail fail to incorporate into DmORC in vivo , and do not complement an orc6-null allele when introduced into flies . Taken together , these results suggest that the MGS mutation in Orc6 directly affects the integrity of ORC , and indicate that destabilization of ORC contributes to the pathogenesis of Meier-Gorlin syndrome caused by mutations in Orc6 , ultimately interfering with pre-RC assembly and origin licensing .
EM reconstructions of S . cerevisiae ORC and Drosophila ORC have been determined previously ( Speck et al . , 2005; Clarey et al . , 2006; Sun et al . , 2012 , 2013 ) , and have allowed individual subunits to be localized in budding yeast ORC . Unfortunately , the modest resolution of these structures , together with the poor resemblance between the Drosophila and S . cerevisiae ORC models , have made it difficult to extrapolate the yeast subunit order into metazoan ORC ( Speck et al . , 2005; Clarey et al . , 2006 , 2008; Chen et al . , 2008; Sun et al . , 2012 ) . To overcome this gap , we set out to improve the resolution of the Drosophila ORC model by negative-stain electron microscopy and to further map the complex’s higher-order organization by subunit tagging and visualization . To obtain samples for EM , we purified recombinant Drosophila ORC from insect cells ( Figure 1—figure supplement 1A ) . As previously observed , Drosophila ORC forms a stable , hexameric complex containing subunits Orc1–Orc6 ( Figure 1A , Figure 1—figure supplement 1B ) ( Chesnokov et al . , 1999 , 2001 ) . Examination of ORC by EM after negative staining showed a monodisperse population of particles , with no indication of complex instability at the low-nanomolar concentrations used for sample deposition onto EM grids ( Figure 1—figure supplement 1C ) . 10 . 7554/eLife . 00882 . 003Figure 1 . 3D structure of metazoan ORC . ( A ) Coomassie-stained SDS-PAGE gel of purified recombinant Drosophila ORC used for EM studies . Molecular weight markers and ORC subunits are indicated . ( B ) Nucleotides have different effects on the conformation of ORC . Drosophila ORC was incubated with 1 mM ADP , ATP , AMPPNP or ATPγS and analyzed by electron microscopy and compared to apo-ORC . Representative 2D class averages are shown . ( C ) Single-particle EM reconstruction of Drosophila ORC corrected for the absolute hand of ORC . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 00310 . 7554/eLife . 00882 . 004Figure 1—figure supplement 1 . Purification of Drosophila ORC . ( A ) Overview of recombinant expression and purification strategy used for ORC . ( B ) Analytical gel filtration chromatography of purified Drosophila ORC on a Superose 6 column demonstrates that Drosophila ORC elutes as a hexamer with the predicted molecular weight of ∼390 kDa . Elution volume of the molecular weight markers thyroglobulin ( 670 kDa ) , γ-globulin ( 158 kDa ) and ovalbumin ( 44 kDa ) are indicated above the UV trace . ( C ) Electron micrograph of negatively stained Drosophila ORC showing a monodisperse population of particles . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 00410 . 7554/eLife . 00882 . 005Figure 1—figure supplement 2 . Single particle 3D EM reconstruction of Drosophila ORC . ( A ) Previously-determined EM structure of Drosophila ORC ( EMDB-1252 ) ( Clarey et al . , 2006 ) . ( B ) Initial starting model used for projection-matching refinement obtained by low-pass filtration of the EM structure in ( A ) using a Butterworth filter with pass band at 200 Å and stop band at 100 Å . ( C ) Refined Drosophila ORC EM structure determined in this study prior to hand determination and flipping ( Figure 1C , Figure 1—figure supplement 3 ) . ( D ) Fourier shell correlation curve for reconstructed EM volume . The resolution at 0 . 5 FSC is ∼22 Å . ( E ) Euler angle distribution of particle images used for 3D reconstruction . The points in the hemisphere of the Euler space correspond to particle images assigned to a given reference image during projection-matching refinement , with the size of the point being proportional to the number of images assigned to a given reference . ( F ) Comparison of experimental 2D class averages with forward projection of final 3D EM volume demonstrates good matches between the experimental data and the 3D reconstruction . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 00510 . 7554/eLife . 00882 . 006Figure 1—figure supplement 3 . Determination of the absolute hand of ORC . The final refined Drosophila ORC EM volume shown in Figure 1—figure supplement 2C was used in the freehand test using 0° and +50° tilt-pair images . The cross-correlation coefficients for the tilt transformations are plotted . The highest correlation is observed at a tilt angle of −50° and not +50° , indicating the absolute hand of the reconstructed ORC volume in Figure 1—figure supplement 2C is incorrect . Flipping the hand corrects the handedness of the model ( Figure 1C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 006 Nucleotide binding and hydrolysis often generate significant conformational changes in oligomeric assemblies of AAA+ ATPases ( Neuwald et al . , 1999; Iyer et al . , 2004 ) , a behavior seen previously by EM for Drosophila ORC in the absence and presence of the non-hydrolysable ATP analog ATPγS ( Clarey et al . , 2006 ) . However , the maximal obtainable resolution for these early EM reconstructions was relatively low ( 34 Å and 33 Å resolution , respectively ) , suggesting that conformational heterogeneity still existed within the ORC sample . To improve the resolution , we therefore tested whether different nucleotide analogs or increased concentrations of ATPγS could improve conformational uniformity ( Figure 1B ) . Similar to prior observations ( Clarey et al . , 2006 ) , 2D EM class averages of apo-ORC yielded crescent-shaped particles . Inclusion of 1 mM ATP , ADP or the non-hydrolysable ATP analog AMPPNP appeared similar to apo-ORC in 2D class averages ( Figure 1B ) . However , the addition of 1 mM ATPγS revealed hitherto unobserved detailed features of ORC , indicating that ATPγS stabilizes interactions between subunits and traps ORC in a better defined conformational state . The ATPγS-dependent stabilization seen here proved more pronounced than that seen previously ( Clarey et al . , 2006 ) , likely due to the higher concentration of nucleotide ( 1 mM vs 5 µM ) used to ensure saturation of ORC with ATPγS . Such an effect of ATPγS on Drosophila ORC is somewhat similar to the nucleotide requirement observed for human ORC complex formation and stability ( Ranjan and Gossen , 2006; Siddiqui and Stillman , 2007 ) ; based on this improved behavior , all subsequent EM studies of Drosophila ORC were performed in the presence of 1 mM ATPγS . The more detailed features observed in 2D class averages of ATPγS-ORC suggested that it should be possible to obtain a 3D EM reconstruction of Drosophila ORC at a significantly improved resolution than was possible in earlier efforts . We therefore calculated a 3D EM reconstruction , using a low-pass filtered volume of the previously published Drosophila ORC EM structure as an initial starting model ( Figure 1—figure supplements 2A–C ) . The structure was refined by iterative projection-matching to a resolution of 22 Å ( as judged by the 0 . 5 Fourier shell correlation criterion ) ( Figure 1—figure supplement 2D ) , yielding forward projections of the 3D structure in excellent agreement with our reference-free class averages ( Figure 1—figure supplement 2F ) . The resultant model contains contributions from a near-complete coverage of angular space ( Figure 1—figure supplement 2E ) , and showed significantly more detail than the previous Drosophila ORC EM structure ( compare Figure 1—figure supplement 2A , C ) . Since the handedness of ORC was not previously determined experimentally , we next collected tilt-pairs of ORC and subjected them to the Freehand test ( Rosenthal and Henderson , 2003; Henderson et al . , 2011 ) . This analysis showed that the initially calculated ORC structure ( shown in Figure 1—figure supplement 2C ) was of the incorrect absolute hand ( Figure 1—figure supplement 3 ) , leading us to flip the hand of the EM volume for further analysis ( Figure 1C ) . Taken together , our results indicate that nucleotide binding induces conformational changes within ORC that are propagated throughout the entire complex to stabilize a specific structural state . While the higher resolution features of our new Drospohila EM model allowed us to delineate likely subunit borders , it still remained unknown which subunits might occupy different regions of the density . We therefore set out to localize individual subunits experimentally by EM , using purified recombinant ORC containing N-terminal MBP- and C-terminal GFP-tags on individual ORC subunits , respectively . Fusion proteins of subunits Orc1–Orc5 were able to form stable hexameric ORC complexes ( Figure 2—figure supplement 1 ) , allowing for the collection of electron micrographs , 2D classification , and 3D EM reconstructions . Comparison of class averages and 3D volumes of tagged ORC with those of untagged ORC revealed extra densities corresponding to the MBP- or GFP-tags , thus marking the positions of the tagged subunits ( Figure 2A ) . We did not observe extra density for MBP when fused to the N-terminus of full-length Orc1 or Orc2 , nor did we see GFP fused to the C-terminus of Orc2 and Orc4 ( data not shown ) , suggesting these protein termini are flexible . However , after deletion of the N-terminal domains of Orc1 and Orc2 ( which are predicted to contain disordered regions ) , density became visible for MBP ( Figure 2A ) . During the subunit mapping process , we were unable to definitively distinguish Orc2 and Orc3 from each other , suggesting that the two subunits lie in close proximity within ORC . Collectively , our data indicate that the AAA+-like subunits Orc1–Orc5 comprise the crescent-shaped core of ORC , in the order Orc1-Orc4-Orc5-Orc2/Orc3 ( when starting at the lower left of the EM structure ) ( Figure 2B ) . 10 . 7554/eLife . 00882 . 007Figure 2 . Arrangement of subunits within metazoan ORC . ( A ) Localization of AAA+ subunits Orc1–Orc5 within ORC by EM . 2D class averages of ORC containing N-terminally MBP-tagged or C-terminally GFP-tagged Orc1–5 subunits are compared to representative averages of untagged ORC . Additional densities observed that can be attributed to MBP- or GFP-tags are noted by arrowheads and mark the position of the respective subunit within ORC . 3D EM reconstructions of MBP- or GFP-tagged ORCs further pinpoint the localization of Orc1–Orc5 within ORC . With the exception of MBP–Orc2ΔN , MBP- or GFP-densities ( highlighted in color ) were also observed in 3D EM reconstructions of MBP- or GFP-tagged ORCs . No additional density was observed in 2D class averages for ORC containing N-terminally MBP-tagged full-length Orc1 and Orc2 nor for C-terminally GFP-tagged Orc2 and Orc4 , indicating conformational flexibility of the tag ( data not shown ) . NA—this view was not observed in 2D class averages . ( B ) Surface coloring of the DmORC EM volume based on subunit localization . The asterisks marks density that likely contains a domain inserted between the AAA+-like domain and the winged helix domain in Orc3 ( Figure 5 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 00710 . 7554/eLife . 00882 . 008Figure 2—figure supplement 1 . Purification of ORC with tagged Orc1–5 subunits for EM localization . The AAA+ subunits Orc1–Orc5 were tagged at the N-terminus with MBP ( M ) or at the C-terminus with GFP ( G ) and the respective tagged ORC complexes purified . A Coomassie-stained SDS-PAGE gel of the complexes is shown . For each tagged complex , the tagged subunit is marked by an asterisk to the left of the lane . The N-terminal extensions in Drosophila Orc1 and Orc2 were deleted in MBP–Orc1 and MBP–Orc2 . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 008 To localize Orc6 , we initially compared class averages of ORC lacking Orc6 ( referred to as ORC1–5 ) with class averages of ORC containing Orc6 ( simply referred to as ORC ) . Despite the reasonable molecular weight of Orc6 ( 29 kDa ) and its stoichiometric presence in ORC purifications ( Figure 3—figure supplement 1 ) , 2D EM class averages of ORC and ORC1–5 looked virtually identical , suggesting that Orc6 might be flexibly attached to the rest of the complex ( Figure 3A ) . To investigate the position of Orc6 further , we fused MBP or GFP to the N- or C-terminus of Orc6 , respectively . In class averages of ORC containing C-terminally GFP-tagged Orc6 , extra density appeared near the upper left tip of ORC , close to the position of Orc2 and Orc3 ( Figure 3A ) . We did not observe density with an N-terminal MBP–Orc6 fusion , nor did we observe GFP density in 3D reconstructions ( Figure 3A and data not shown ) . These data indicate that Orc6 is flexibly tethered to ORC , likely via an interaction with Orc2 , Orc3 , or both . 10 . 7554/eLife . 00882 . 009Figure 3 . Orc6 localizes near Orc2 and Orc3 within Drosophila ORC but is conformationally flexible . ( A ) ORC containing N-terminally MBP-tagged Orc6 or C-terminally GFP-tagged Orc6 was purified and subjected to EM analysis . Representative 2D class averages are shown . Drosophila ORC with untagged Orc6 is shown for comparison . Additional density corresponding to the tag ( arrowheads ) is only observed for the Orc6-GFP and not for MBP–Orc6 . Class averages of ORC lacking Orc6 subunit reveal no difference density compared to ORC containing Orc6 , indicating that Orc6 is flexibly tethered to ORC . ( B ) Drosophila Orc6 interacts with Orc3 but not with Orc2 . MBP–Orc2 , MBP–Orc3 and MBP–Orc6 fusions were co-expressed in High5 cells with 6 × His-tagged Orc2 , Orc3 and Orc6 and subjected to pull-downs ( PD ) using amylose resin . Orc3 interacts with both Orc2 and Orc6 , whereas no interaction is observed between Orc2 and Orc6 as assessed by SDS-PAGE and Coomassie staining . In addition , no interaction is observed between MBP alone and any of the ORC subunits . The asterisk denotes a degradation product of MBP–Orc2 . Western blot ( WB ) analysis of whole cell extracts for 6 × His tagged subunits ( lower panels ) demonstrates that they are expressed as expected . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 00910 . 7554/eLife . 00882 . 010Figure 3—figure supplement 1 . Purification of ORC complexes for EM localization of the Orc6 subunit . ORC containing N-terminally MBP-tagged ( M ) Orc6 , C-terminally GFP-tagged ( G ) Orc6 , or the N-terminally MBP-tagged C-terminal domain of Orc6 ( amino acids 187–257 ) and ORC lacking the Orc6 subunit were purified . A Coomassie-stained SDS-PAGE gel of respective complexes is shown . Tagged Orc6 is marked by an asterisk to the left of the lane . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 010 The localization pattern of Orc6 led us to hypothesize that we might be able to use it to determine the order of Orc2 and Orc3 in our EM volume by identifying Orc6’s binding partner . Interactions between human and yeast Orc2 , Orc3 , and Orc6 have previously been investigated; however , conflicting results of subunit interactions have been reported ( Vashee et al . , 2001; Kneissl et al . , 2003; Chen et al . , 2007; Matsuda et al . , 2007; Siddiqui and Stillman , 2007; Sun et al . , 2012 ) . Moreover , the interactions between the Drosophila proteins have not been tested . To identify the subunit that recruits Drosophila Orc6 into ORC , we performed pull-down assays from insect cells co-expressing pairwise combinations of Drosophila Orc2 , Orc3 , and Orc6 , with one subunit containing an MBP-tag used as purification bait . As previously observed for both human and yeast proteins ( Dhar et al . , 2001; Vashee et al . , 2001; Ranjan and Gossen , 2006; Matsuda et al . , 2007 ) , Drosophila Orc2 interacted with Orc3 ( Figure 3B ) . This finding is consistent with our EM localization , which positioned these subunits within similar regions of ORC . Orc6 , on the other hand , interacted only with Orc3 and not with Orc2 ( Figure 3B ) . Since Orc6 appears to contact the subunit at the upper left end in the EM class averages , the Orc3–Orc6 interaction allowed us to refine the subunit order of the core AAA+ subunits to Orc1-Orc4-Orc5-Orc2-Orc3 for Drosophila ORC ( Figure 2B ) . Metazoan Orc6 is composed of three domains: an N-terminal and middle domain , both of which have homology to the cyclin-box folds of transcription initiation factor IIB ( referred to as TFIIB domain A and B ) , and a C-terminal region of unknown structure ( Figure 4A , Figure 4—figure supplement 1 ) ( Chesnokov et al . , 2003; Liu et al . , 2011 ) . The C-terminal domain of metazoan Orc6 has been thought to be mainly important for non-replicative functions of Orc6 in cytokinesis and daughter cell abscission , and in the case of Drosophila Orc6 , the C-terminal domain also has been shown to interact with the septin Pnut ( Prasanth et al . , 2002; Chesnokov et al . , 2003; Bernal and Venkitaraman , 2011 ) . Surprisingly , however , we found that the C-terminal domain of Orc6 proved to be both essential and sufficient for Orc3 binding when co-expressed in insect cells ( Figure 4B , Figure 4—figure supplement 2A for controls ) . The necessity of the C-terminal domain was not limited to binary protein interactions , but was also observed either when all ORC subunits were co-expressed in insect cells ( Figure 4—figure supplement 2B ) , or when purified full-length or C-terminal truncated Orc6 proteins were mixed with ORC1–5 in vitro and subjected to pull-down assays ( Figure 4C ) . 10 . 7554/eLife . 00882 . 011Figure 4 . Drosophila Orc6 is recruited into ORC via interaction of its C-terminal domain with Orc3 . ( A ) Schematic of domain organization of Orc6 and of the N- and C-terminal deletion constructs of Orc6 used in this study . ( B ) The C-terminal domain ( CTD ) of Orc6 is essential and sufficient for the interaction of Orc6 with Orc3 . His-tagged full-length Orc6 ( 1–257 ) and Orc6 deletion constructs were co-expressed with MBP-tagged Orc3 in High5 cells and analyzed for interactions by pull-downs using amylose resin . Bound proteins were analyzed by SDS-PAGE and Coomassie-staining . Western blot analysis of whole cell extracts ( lower panel ) shows that C-terminally truncated Orc6 proteins are expressed at similar levels to full-length wild type Orc6 . Despite the clear pull-down results for the N-terminally truncated Orc6 proteins , we were not able to detect two of the proteins ( Orc6 147–257 and 187–257 ) in whole cell extracts , possibly due to expression levels too low to be detected with our Western blot conditions . ( C and D ) The C-terminal domain of Orc6 is essential and sufficient for recruitment of Orc6 into ORC . Purified ORC1–5 ( lacking Orc6 subunit ) was incubated in vitro with purified full-length Orc6 , Orc6 lacking C-terminal residues ( Orc6 1–232 , Orc6 1–220 and Orc6 1–200 ) , and the isolated Orc6 C-terminal domain ( Orc6 187–257 ) . Formation of ORC containing all six subunits was tested in pull-downs using ( C ) MBP-tagged Orc4 as bait and by ( D ) fluorescence anisotropy after N-terminally labeling Orc6 proteins with Alexa Fluor-488 . C-terminal deletions of Orc6 abolish recruitment of Orc6 into ORC in vitro , whereas the C-terminal domain of Orc6 binds to ORC1–5 with nanomolar affinity . Binding data of full-length Orc6 and the Orc6 C-terminal domain ( Orc6 187–257 ) were fit to the quadratic equation describing single-site binding under ligand depletion . ( E ) The localization of the C-terminal domain of Orc6 is indistinguishable from the localization of full-length Orc6 within ORC . Purified ORC containing full-length Orc6-GFP or MBP–Orc6 187–257 was analyzed by negative stain EM . 2D image analysis revealed extra densities ( arrow heads ) in similar positions that correspond to GFP and MBP tags on the respective Orc6 constructs . Density corresponding to the affinity tag was observed in the 3D volume of ORC containing MBP–Orc6 187–257 , further pinpointing the localization of Orc6 within ORC . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 01110 . 7554/eLife . 00882 . 012Figure 4—figure supplement 1 . The Orc6 TFIIB-like domains as well as residues in the C-terminal domain containing the Meier–Gorlin syndrome mutation in human Orc6 are conserved across eukaryotes , including fungi . Orc6 protein sequences from representative eukaryotes and human transcription factor TFIIB ( PDB code 1vol , chain A ) ( Nikolov et al . , 1995 ) were aligned using MAFFT ( Katoh et al . , 2005; Katoh and Toh , 2008 ) . Conserved domains and secondary structure elements as observed in the crystal structure of human TFIIB are outlined below the alignment . Amino acid residues in Drosophila melanogaster ( Dm ) Orc6 that demarcate truncations as well as the residue mutated in Meier-Gorlin syndrome ( MGS ) are indicated above the alignment . Accession numbers of Orc6 protein sequences used in the alignment can be found in the ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 01210 . 7554/eLife . 00882 . 013Figure 4—figure supplement 2 . C-terminal truncations of Drosophila Orc6 do not associate with ORC1–5 when all subunits are co-expressed in insect cells . ( A ) Orc6 proteins do not interact with MBP nor do they bind unspecifically to amylose beads . MBP and different Orc6 constructs were co-expressed in High5 cells and pull-downs ( PD ) were performed using amylose beads . Only MBP is recovered from beads . Western blots ( WB ) show that Orc6 proteins are expressed in insect cells . We were not able to detect Orc6 147–257 and Orc6 187–257 in Western blots of whole-cell lysates; however , the two proteins efficiently co-purified with MBP–Orc3 ( Figure 4B ) , MBP–Orc3INS ( Figure 5C ) , and ORC1–5 ( panel B in this figure ) , indicating they are expressed . ( B ) The C-terminal domain of Orc6 is essential for hexameric ORC formation in insect cells . ORC subunits and Orc6 truncated proteins were co-expressed in insect cells and extracts were subjected to amylose pull-downs ( PD ) targeting MBP-tagged Orc1 . Whereas N-terminal deletions of Orc6 co-purify with ORC1–5 , C-terminal deletions of Orc6 abrogate the interaction with ORC1–5 . The asterisk marks a protein contaminant . Coomassie-stained SDS-PAGE gels are shown in ( A and B ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 013 To rule out the possibility that our pull-downs might miss a potential weak interaction between ORC1–5 and Orc6 lacking the C-terminal domain , we performed equilibrium binding experiments using fluorescence anisotropy and fluorescently-labeled Orc6 . Full-length Orc6 bound to ORC1–5 with apparent low-nanomolar affinity ( Kd , app = 5 . 0 ± 0 . 7 nM ) ( Figure 4D ) . Consistent with the pull-down results , C-terminally truncated Orc6 proteins did not bind ORC1–5 up to concentrations of 1 µM ORC1–5 ( Figure 4D ) . In contrast , the C-terminal 71 amino acid residues of Orc6 bound to ORC1–5 with an affinity similar to that of full-length Orc6 ( Kd , app = 4 . 8 ± 0 . 6 nM ) . In addition , the Orc6 C-terminal region co-purified with ORC1–5 during several chromatography steps ( Figure 3—figure supplement 1 , lane 5 ) , and when examined by EM , an MBP-fusion of the C-terminal domain of Orc6 localized to the same region as full-length Orc6 in 2D class averages of ORC , indicating that Orc6 did not dissociate from ORC at concentrations used for EM ( 30 nM ORC ) ( Figure 4E ) . Collectively , these results demonstrate that Orc6 is recruited into ORC via interaction of its C-terminal domain with Orc3 , and not through its TFIIB-like domains . We next asked which region of Orc3 interacts with Orc6 . In comparing Orc3 protein sequences from various eukaryotes with archaeal Orc1/Cdc6 proteins , we noticed that Orc3 contains an additional domain inserted between its AAA+-like and winged-helix domains that is not found in archaeal homologs ( Figure 5A , B , Figure 5—figure supplement 1 ) . This insertion is conserved across eukaryotic Orc3 orthologs , varies in length from 70 amino acids in S . cerevisiae to 190 amino acids in Drosophila , and is not found in Orc1 , Orc2 , Orc4 , or Orc5 ( Figure 5—figure supplement 1 ) . Since this extra domain is specific to Orc3 , we hypothesized that it might interact with Orc6 , specifically with the Orc6 C-terminus . To test this premise , we co-expressed the Orc3 insertion as an MBP fusion protein with full-length or truncated Orc6 and performed pull-down experiments . The Orc3 insertion was able to interact with both full-length Orc6 and all N-terminal Orc6 truncations ( Figure 5C ) . Conversely , truncating the C-terminus of Orc6 abolished the interaction with the Orc3 insertion . These results mirror our observations with both full-length Orc3 and with ORC1–5 , and further demonstrate that the C-terminus of Orc6 helps to tether Orc6 to ORC . Moreover , whereas full-length Orc3 was able to interact with Orc2 in addition to Orc6 , the Orc3 insertion was unable to do so , suggesting the Orc2–Orc3 interaction is mediated through the two subunits’ AAA+-like and/or winged-helix domains ( Figure 5D ) . 10 . 7554/eLife . 00882 . 014Figure 5 . A conserved domain inserted between the AAA+-like domain and the winged helix ( WH ) domain of Orc3 binds the C-terminal domain of Orc6 and recruits Orc6 into ORC . ( A ) Schematic domain organization of Orc3 . ( B ) The position of the Orc3 insertion maps between the AAA+-like domain and the winged helix domain in the crystal structure of archaeal Orc1–1 ( PDB code 2qby chain A ) ( Dueber et al . , 2007 ) . ( C ) The Orc3 insertion is sufficient for Orc6 binding . Drosophila Orc3 containing residues 371–559 ( Orc3INS ) was co-expressed as an N-terminal MBP-fusion protein with His-tagged full-length and truncated versions of Drosophila Orc6 in High5 cells and their interaction probed by pull-downs using amylose resin . ( D ) The insertion in Orc3 does not interact with Orc2 . N-terminal MBP-fusion proteins of Drosophila full-length Orc3 ( Orc3FL ) , the Orc3 insertion ( Orc3INS ) or MBP alone were co-expressed with 6 × His-tagged Drosophila Orc2 in High5 cells and subjected to amylose pull-downs . Coomassie-stained SDS-PAGE gels are shown in ( C and D ) . Western blot analysis of whole-cell extract using an anti-His antibody ( lower panels in C and D ) demonstrates that lack of Orc2 or Orc6 pull-down is not due to lack of their expression . As in Figure 4B , we were not able to detect Orc6 147-257 and Orc6 187-257 in cell extracts despite being clearly visible in pull-downs , likely because of low expression levels . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 01410 . 7554/eLife . 00882 . 015Figure 5—figure supplement 1 . Eukaryotic Orc3 contains a domain inserted between its AAA+-like and winged helix domains that is not present in archaeal Orc1/Cdc6 , nor in eukaryotic Orc1 , Orc2 , Orc4 or Orc5 . Protein sequences of archaeal Orc1/Cdc6 , various eukaryotic Orc3 proteins , and Drosophila melanogaster Orc1 , Orc2 , Orc4 , and Orc5 were aligned using MAFFT ( Katoh et al . , 2005; Katoh and Toh , 2008 ) . Secondary structural elements as seen in the crystal structure of archaeal Orc1–1 ( PDB code 2qby , chain A ) ( Dueber et al . , 2007 ) are shown below the alignment . Accession numbers for protein sequences can be found in the ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 015 Although the overall function of Orc6 has remained somewhat debatable , the subunit is one of the pre-RC components recently shown to be mutated in patients with Meier-Gorlin syndrome ( MGS ) ( Bicknell et al . , 2011a ) . Intriguingly , the MGS mutation in human Orc6 is a missense alteration that maps to the protein’s C-terminal domain , changing a tyrosine at position 232 to serine . The tyrosine is located in a highly conserved , predicted helical element , and is itself highly conserved among eukaryotic Orc6 orthologs , including budding yeast ( Figure 6A , Figure 4—figure supplement 1 ) . Since our findings show that the C-terminal domain of Orc6 is essential for the recruitment of Orc6 into Drosophila ORC , we hypothesized that the MGS mutation may interfere with this function . To test this assumption , we introduced the corresponding Y225S mutation into Drosophila Orc6 and asked whether it could affect either the binary Orc3–Orc6 interaction or hexameric ORC formation . In addition , we included another mutant , W228A/K229A , which substitutes an invariant tryptophan close to Y225 and a highly conserved basic amino acid for alanine . Orc6-W228A/K229A is not able to rescue an Orc6 deletion in Drosophila to support viability in vivo , and cells bearing the two mutations showed reduced BrdU incorporation in brain neuroblasts , suggesting defects in DNA synthesis ( Balasov et al . , 2009 ) . When co-expressed in insect cells with MBP–Orc3 , drastically reduced levels of Orc6-Y225S and no Orc6-W228A/K229A were pulled-down ( Figure 6B ) . Similar results were obtained when the Orc3 insertion was used instead of full-length Orc3 ( Figure 6—figure supplement 1A ) . Interestingly , the Orc3–Orc6 interaction was not stabilized by additional co-expression of Orc2 , suggesting that Orc2 binding to Orc3 has no effect on the Orc3–Orc6 interaction in Drosophila ( Figure 6—figure supplement 1B ) . 10 . 7554/eLife . 00882 . 016Figure 6 . A Drosophila Orc6 mutation corresponding to the Meier-Gorlin-Syndrome mutation in human Orc6 weakens the interaction of Orc6 with Orc3 and diminishes its recruitment into ORC . ( A ) The tyrosine mutated in Meier-Gorlin syndrome ( MGS ) is a well conserved residue and maps to the C-terminal domain of Orc6 . A sequence LOGO ( Workman et al . , 2005 ) of the Orc6 C-terminus is shown . The amino acid residue corresponding to the human MGS mutation was mutated in Drosophila Orc6 ( Y225S ) . ( B ) Mutations in the C-terminus of Drosophila Orc6 , including the MGS mutation , reduce the affinity of Orc6 for Orc3 . Wild-type ( WT ) Orc6 , the Y225S mutant or the W228A/K229A double-mutant Orc6 were co-expressed with MBP–Orc3 in High5 cells and tested for interaction with Orc3 by pull-down analysis using amylose beads and MBP–Orc3 as bait . Coomassie-stained SDS-PAGE gels are shown for pull-downs and a Western blot ( lower panel ) is shown to confirm Orc6 expression in insect cells . ( C and D ) C-terminal mutations in Drosophila Orc6 interfere with recruitment of Orc6 into ORC in vitro . Wild-type Orc6 , Orc6-Y225S and Orc6-W228A/K229A were purified and tested for association with ORC1–5 into heterohexameric ORC using pull-downs targeting MBP–Orc4 ( C ) or by fluorescence anisotropy using N-terminally Alexa Fluor-488 labeled Orc6 ( D ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 01610 . 7554/eLife . 00882 . 017Figure 6—figure supplement 1 . Mutations in the C-terminal domain of Drosophila Orc6 , including the MGS-like mutation , abrogate binding of Orc6 to the Orc2–Orc3 subcomplex , to the Orc3 insertion ( Orc3INS ) and to ORC1–5 . ( A and B ) Amylose pull-downs ( PD ) from High5 cells co-expressing His-tagged Orc6 wild type ( WT ) or Orc6 containing the MGS-like ( Y225S ) or WK ( W228A/K229A ) mutation with either ( A ) MBP–Orc3INS or ( B ) MBP–Orc3 and Orc2 . The results show that both mutations in the Orc6 C-terminus interfere with Orc6 binding to the insertion in Orc3 and with recruitment of Orc6 into the Orc2-Orc3-Orc6 subcomplex . Expression of prey proteins was verified in Western blots ( WB ) of whole cell extract ( lower panels ) . ( C ) The MGS-like mutation in the context of the Orc6 C-terminal domain ( CTD ) alone significantly reduces the affinity of the Orc6-CTD for ORC1–5 . The results of fluorescence anisotropy experiments with fluorescently labeled wild-type and Y225S Orc6-CTD are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 017 To address whether the MGS-like mutation and the WK mutant affected hexameric ORC formation , we next tested whether purified wild-type or mutant Drosophila Orc6 could bind to purified ORC1–5 in vitro . Neither Orc6-Y225S nor Orc6-W228A/K229A bound to ORC1–5 in pull-down experiments ( Figure 6C ) . Orc6-Y225S incorporation into ORC was also reduced when it was co-expressed with Orc1–Orc5 subunits in insect cells ( Figure 4—figure supplement 2B , lane 8 ) . In equilibrium-binding experiments using fluorescence anisotropy , Orc6-W228A/K229A showed no binding to ORC1–5 ( Figure 6D ) . Orc6-Y225S was able to bind weakly to Orc1–5; however , the affinity was reduced at least 100-fold compared to wild type Orc6 , and similar results were obtained for the MGS-like mutant when only the C-terminal domain of Orc6 was used in binding experiments ( Figure 6—figure supplement 1C ) . Together , these results show that highly conserved amino acid residues in the C-terminal domain of Drosophila Orc6 , including the tyrosine mutated in human Orc6 in Meier-Gorlin syndrome , mediate the recruitment of Orc6 into ORC . In flies , the Orc6 C-terminus is essential for viability , as GFP–Orc6 transgenes containing mutations in the C-terminus are ( in contrast to wild type GFP–Orc6 ) unable to rescue an orc6-null allele ( Balasov et al . , 2009 ) . Neuroblasts of such Orc6 mutant flies incorporate BrdU less efficiently than cells expressing wild-type Orc6 , suggesting that they suffer from defects in DNA replication ( Balasov et al . , 2009 ) . We re-investigated flies carrying these transgenes to test whether the previously observed phenotypes are due to defects in hexameric ORC formation in vivo , and to determine the physiologic consequences of our in vitro results . We expressed GFP–Orc6 wild type , GFP–Orc6 D224A/Y225A or GFP–Orc6 W228A/K229A transgenes in flies and asked whether they associate with other ORC subunits in vivo . ORC was immunoprecipitated with an anti-Orc2 antibody from ovary extract and , as expected , immunoprecipitates were found to contain Orc3 , Orc5 , and GFP–Orc6 wild type ( Figure 7A ) . In contrast , the amounts of GFP–Orc6 mutants that were immunoprecipitated were greatly reduced compared to wild type GFP–Orc6 , both in Western blots and when GFP absorbance was measured directly ( Figure 7A , B ) . Thus , mutations in the conserved C-terminus of Orc6 result in a reduction of ORC hexamer formed both in vitro and in vivo . 10 . 7554/eLife . 00882 . 018Figure 7 . Mutations of conserved amino acids in the C-terminal domain of Drosophila Orc6 result in a loss of its association with ORC and in reduced MCM chromatin association in vivo in flies . ( A ) Western immunoblotting analysis of ORC complexes precipitated using anti-Orc2 antibody from extracts isolated from fly ovaries expressing different GFP-tagged Orc6 mutant proteins: GFP–Orc6 D224A/Y225A , GFP–Orc6 W228A/K229A and GFP–Orc6 WT ( wild type ) . GFP–Orc6 fusion proteins were detected with anti-GFP monoclonal antibody , while Orc3 and Orc5 subunits were detected using anti-Orc3 or anti-Orc5 antibodies . Ovary extract ( In ) , supernatant after immunoprecipitations ( S ) and immunoprecipitated material ( IP ) are shown for each transgene . ( B ) Quantitation of immunoprecipitated GFP from ovary extracts containing different Orc6 mutants . The Y-axis shows fluorescence emission intensity . Mean and standard deviation from three independent experiments are plotted ( p≤0 . 05 ) . ( C ) Chromatin association of Mcm4 , a member of the MCM2–7 complex , is reduced in orc6-null fly larvae expressing Orc6 mutants . Brains of homozygous , orc6-null ( orc635/orc635 ) or heterozygous ( orc635/Cy ) Drosophila larvae expressing GFP–Orc6 transgenes were isolated and subjected to salt extraction to solubilize cellular proteins . Insoluble and chromatin associated proteins were pelleted by centrifugation and analyzed by Western blotting using Mcm4 polyclonal antibodies . Pnut was used as a loading control . Total brain extracts were also analyzed to ensure equal Mcm4 expression in larvae expressing Orc6 transgenes . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 018 The reduced levels of hexameric ORC may be expected to lead to impaired pre-RC formation . To test this premise , we analyzed chromatin association of a member of the MCM2–7 complex , Mcm4 , in Drosophila larval brains expressing either wild-type or mutant GFP–Orc6 transgenes . After salt extraction of larval brain extracts , the amount of Mcm4 in the pellet fraction was greatly reduced when mutant transgenes were expressed as the sole source of Orc6 ( orc635/orc635 ) compared to larval brains expressing the wild-type GFP–Orc6 transgenes ( Figure 7C ) . As expected , heterozygous larval brains , which express wild-type , endogenous Orc6 , did not show a Mcm4 chromatin recruitment defect . Thus , conserved C-terminal residues of Orc6 are not only required for hexameric ORC formation but also for MCM loading onto chromatin , explaining the reduced DNA synthesis previously observed in mutant flies ( Balasov et al . , 2009 ) . Although Orc6 , including the C-terminal helix , is well conserved among metazoans , human Orc6 does not appear to tightly associate with human ORC1–5 ( Dhar and Dutta , 2000; Dhar et al . , 2001; Vashee et al . , 2001; Giordano-Coltart et al . , 2005; Ranjan and Gossen , 2006; Siddiqui and Stillman , 2007 ) . We therefore set out to test whether the C-terminus of human Orc6 is also important for assembly of the subunit with the rest of human ORC and whether the Orc6 MGS mutation might affect this interaction . We purified recombinant human ORC1–5 from insect cells , incubated it with purified human Orc6 proteins in vitro , and assayed ORC hexamer formation by pull-downs targeting MBP-tagged Orc4 within ORC1–5 . Although weaker than seen for Drosophila , we nonetheless clearly observed an interaction between wild-type Orc6 and human ORC1–5 , particularly when an excess of Orc6 was used ( Figure 8A ) . In contrast , human Orc6 containing the MGS mutation ( Orc6-Y232S ) did not bind appreciably to ORC1–5 ( Figure 8A ) . Similar results were obtained when human ORC subunits were co-expressed in insect cells ( Figure 8B ) . As seen here with the Drosophila protein , human Orc6 has been reported to interact with Orc3 ( Figure 8C ) ( Vashee et al . , 2001; Siddiqui and Stillman , 2007 ) . Importantly , human Orc6 containing the MGS mutation could not bind Orc3 in either the absence or presence of Orc2 ( Figure 8D ) . These data indicate that the Orc6 C-terminus is functionally conserved in metazoans , and that the amino acid mutated in MGS patients is critical for formation of a stable hexameric initiator complex . 10 . 7554/eLife . 00882 . 019Figure 8 . The Orc6–Orc3 interaction is conserved in humans and is affected by a mutation in human Orc6 that is found in MGS patients . ( A ) Human wild-type ( WT ) , but not MGS mutant ( Y232S ) , Orc6 binds to human ORC1–5 in vitro . Purified Orc6 was incubated with recombinant ORC1–5 at either equimolar ratios or with a 10 × excess of Orc6 . Orc6 binding to ORC1–5 was analyzed by pull-downs using amylose beads that bind MBP–Orc4 . Contrary to wild-type Orc6 , MGS-mutant Orc6 does not associate with ORC1–5 , even when a 10-fold excess of Orc6 is used in binding reactions . ( B ) Human Orc6 containing the MGS mutation does not co-purify with human ORC1–5 when co-expressed in insect cells . High5 cells expressing human ORC1–5 and either Orc6-WT or Orc6-Y232S were used in pull-downs targeting MPB-Orc1 . As a negative control , Orc6-WT or Orc6-Y232S were co-expressed with MBP . Only Orc6-WT is recovered from beads , and only when Orc1–Orc5 are co-expressed . The presence of Orc6 was detected by Western blotting in both pull-downs ( PD ) and whole-cell extracts ( lower panel ) . The asterisk marks degradation products of MBP–Orc1 . ( C ) Subunit interactions within the human Orc2-Orc3–Orc6 subcomplex . MBP-tagged human ORC subunits were co-expressed in High5 cells with 6 × His-tagged subunits in a pairwise manner and probed for interaction by amylose pull-downs . Interactions are observed between human Orc3 and Orc2 as well as between Orc3 and Orc6 . As a control , 6 × His-tagged subunits were co-expressed with MBP alone , and no interaction is observed with MBP . Western blots ( lower panels ) verified that all His-tagged subunits are present in whole cell extracts . ( D ) The MGS mutation in human Orc6 decreases the affinity of human Orc6 for human Orc3 and reduces Orc6 recruitment into the Orc2-Orc3-Orc6 subcomplex . Human ORC subunits were co-expressed in High5 cells as indicated and analyzed by amylose pull-downs for binding . The asterisk marks a degradation product of MBP–Orc3 . Western blot of cell lysate shows that wild-type and mutant Orc6 are expressed in insect cells at similar levels ( lower panel ) . Coomassie-stained gels are shown for pull-downs in ( A–D ) . Orc6-Y232S always migrated slightly slower than wild-type Orc6 in SDS-PAGE gels . Mass spectrometry confirmed that the molecular weight of wild type and Y232 Orc6 are as expected based on their amino acid sequence . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 019 Metazoan and Schizosaccharomyces pombe Orc6 have been reported to share little sequence homology with S . cerevisiae Orc6 , and unique functions have been reported for metazoan and budding yeast Orc6 ( Chesnokov et al . , 1999; Moon et al . , 1999; Dhar and Dutta , 2000; Duncker et al . , 2009 ) . For example , metazoan but not budding yeast Orc6 binds DNA , while yeast Orc6 has been reported to bind Cdt1 ( Balasov et al . , 2007; Chen et al . , 2007; Liu et al . , 2011 ) . This lack of similarity is surprising considering the generally high conservation of other licensing factors and the central role Orc6 is reported to play during replicative helicase loading in S . cerevisiae ( Takara and Bell , 2011; Fernandez-Cid et al . , 2013; Frigola et al . , 2013 ) . However , in the course of analyzing our sequence alignments of Orc6 , we noticed that significant homology actually exists between metazoan , S . pombe , and S . cerevisiae Orc6 ( Figure 4—figure supplement 1 ) . This homology includes both the TFIIB repeat domains and the conserved helix in the Orc6 C-terminal domain that harbors the MGS mutation , indicating that all of the functional domains present in metazoan Orc6 are conserved in budding yeast Orc6 ( Figure 4—figure supplement 1 ) . However , in contrast to S . pombe and metazoan Orc6 , budding yeast Orc6—as well as related Saccharomycotina ( Candida albicans ) and Pezizomycotina ( Aspergillus fumigatus and Neurospora crassa ) —contain an insertion of ∼100–200 amino acids between the two TFIIB-like helical domains , an addition that may have complicated former analyses ( Figure 4—figure supplement 1 ) . This insertion contains phosphorylation sites for cyclin-dependent kinases , as well as an RXL motif reported to bind the cyclin Clb5 , and has been implicated in contributing to the prevention of DNA re-replication ( Nguyen et al . , 2001; Wilmes et al . , 2004; Chen and Bell , 2011 ) . Since the C-terminal domain of metazoan Orc6 associates with ORC by binding to Orc3 , we reasoned that the conservation of this region in budding yeast might extend to this interaction as well . Despite previous studies , the recruitment mechanism of S . cerevisiae Orc6 into ORC has been controversial . The C-terminal 62 amino acids of S . cerevisiae Orc6 have been shown to bind ORC1–5 , likely through interaction with Orc3 or Orc5 ( Chen et al . , 2007 ) . However , Orc6 has also been shown to bind Orc2 but not Orc3 or Orc5 ( Sun et al . , 2012 ) . Hence , we investigated whether S . cerevisiae Orc6 could bind to Orc2 or Orc3 , and whether a mutation of the conserved tyrosine , which corresponds to the MGS mutations in human Orc6 , might affect this interaction . In pull-down experiments from insect cell lysate co-expressing pairwise combinations of budding yeast Orc2 , Orc3 , and Orc6 , we observed a clear interaction between Orc2 and Orc3 , and between Orc3 and Orc6 ( Figure 9A ) . Orc2 , when expressed without Orc3 , was typically degraded , preventing an assessment of whether it efficiently bound Orc6 or not . We next tested whether a mutation of the conserved MGS tyrosine to serine ( Y418S ) could support the binding of S . cerevisiae Orc6 to Orc3 ( the MGS tyrosine in S . cerevisiae Orc6 highlighted by our alignment differs from that proposed previously [Tyr277] , due to a misalignment of the region in the original study [Bicknell et al . , 2011a] ) . Strikingly , the MGS-like mutation in S . cerevisiae Orc6 abolished the interaction with Orc3 ( Figure 9B ) . However , in contrast to what we observed with Drosophila and human proteins , the MGS-like mutation was not sufficient to prevent Orc6 binding to the Orc2–Orc3 subcomplex or to ORC1–5 ( Figure 9C , D ) . Because Orc6 has been previously reported to interact with Orc2 ( Sun et al . , 2012 ) , our result suggests that S . cerevisiae Orc6 is cooperatively recruited into ORC through interaction with both Orc2 and Orc3 . In agreement with this interpretation , Orc6 was still recruited into ORC lacking the Orc2 subunit , albeit at reduced levels , whereas Orc6 recruitment was completely lost in the absence of Orc3 ( omission of Orc3 further led to concomitant loss of Orc2 ) ( Figure 9E ) . Thus , unlike in metazoans , S . cerevisiae ORC appears to possess two Orc6 binding sites , one in Orc2 and one in Orc3 . Although the interaction with Orc3 is mediated by the C-terminal domain of Orc6 , it is not clear which domain is responsible for binding to Orc2 . However , since the Orc2–Orc6 interaction appears unique to S . cerevisiae ( compare Figures 3B , 8 and 9 ) , we propose this contact may be mediated by a yeast-specific domain inserted between the two TFIIB-like domains of Orc6 , which may in turn bind to the non-conserved N-terminal domain of yeast Orc2 that has been implicated previously in Orc6 interactions ( Sun et al . , 2012 ) . Consistent with our interpretation , an MGS-like mutation in yeast Orc6 supports yeast growth , whereas premature stop codons that result in C-terminally truncated Orc6 proteins do not ( Figure 9—figure supplement 1 ) . Taken together , our results reconcile previously conflicting data regarding Orc6 recruitment into ORC and highlight the importance of the conserved C-terminal segment of Orc6 in the formation of the hexameric ORC assembly through Orc6 binding to Orc3 . This interaction appears functionally relevant , as Orc6 C-terminal amino acid residues are required for yeast viablility ( Figure 9—figure supplement 1 ) ( Chen and Bell , 2011 ) , and yeast Cdt1 fused to the C-terminal domain of Orc6 alone is able to load MCM2–7 to replication origins in vitro ( Chen et al . , 2007 ) . 10 . 7554/eLife . 00882 . 020Figure 9 . The C-terminal domain in S . cerevisiae Orc6 interacts with Orc3 and an MGS-like mutation abrogates this interaction . ( A ) S . cerevisiae Orc6 binds Orc3 . His-tagged Orc2 , Orc3 or Orc6 were co-expressed with MBP–Orc2 , MBP–Orc3 or MBP only in High5 cells and cell extracts were subjected to pull-downs ( PD ) using amylose beads . Orc6 and Orc3 , as well as Orc2 and Orc3 , interact directly with each other . MBP–Orc2 was very unstable when expressed without Orc3 . The asterisk indicates a degradation product of either MBP–Orc3 or Orc2 . ( B ) An MGS-like mutation in the C-terminal domain of S . cerevisiae Orc3 abolishes binding of Orc6 to Orc3 . Wild-type ( WT ) Orc6 or Orc6 containing a mutation corresponding to the MGS mutation in human Orc6 ( Y418S ) were co-expressed with either MBP–Orc3 or MBP only in insect cells . Only wild-type S . cerevisiae Orc6 , but not the Y418S mutant , bound to MBP–Orc3 in amylose pull-downs . ( C ) The MGS-like mutation in S . cerevisiae Orc6 allows incorporation of Orc6 into the Orc2-Orc3-Orc6 subcomplex . Orc6-WT and Orc6-Y418S were co-expressed with untagged Orc2 and MBP–Orc3 in High5 cells . In amylose pull-downs , a ternary complex containing Orc2 , Orc3 , and Orc6 was recovered for both Orc6-WT and Orc6-Y418S . Since Orc6-Y418S does not interact with Orc3 alone ( panel B ) this result suggests that a second binding site for Orc6 exists in S . cerevisiae Orc2 which is not sensitive to the MGS-like mutation in the C-terminal domain of Orc6 . ( D ) The S . cerevisiae Orc3–Orc6 interaction is not essential for hexameric ORC formation . S . cerevisiae ORC subunits were co-expressed in High5 cells with Orc6-WT or Orc6-Y418S and cell extracts were subjected to pull-downs ( PD ) using amylose beads binding to MBP–Orc1 . All six subunits co-purify with MBP–Orc1 . Due to the similar migration of Orc6 and Orc4/Orc5 , bead-bound Orc6 was detected by Western blotting using anti-His antibody . No Orc6 was recovered in control pull-downs from cells expressing Orc6 and MBP . ( E ) At least two different binding sites recruit S . cerevisiae Orc6 into ORC . S . cerevisiae ORC subunits were co-expressed in insect cells to reconstitute ORC containing all six ORC subunits ( - ) , ORC lacking Orc2 ( no Orc2 ) and ORC lacking Orc3 ( no Orc3 ) . Amylose pull-downs targeting MBP–Orc1 result in reduced Orc6 incorporation into ORC in the absence of Orc2 , and no Orc6 is recovered when Orc3 is absent . Omission of Orc3 also resulted in loss of co-purification of Orc2 . As in ( D ) Orc6 was detected by Western blotting ( WB ) . Asterisks in ( D and E ) indicate a protein contaminant binding to amylose beads . In ( A–E ) Coomassie-stained SDS-PAGE gels are shown for pull-downs and Western blots ( WB ) of whole cell extract verify that all His-tagged subunits were expressed . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 02010 . 7554/eLife . 00882 . 021Figure 9—figure supplement 1 . The C-terminus of S . cerevisiae Orc6 is essential for yeast growth . ( A ) Schematic of Orc6 constructs tested for complementation . Orc6 lacking C-terminal amino acid residues was generated by introducing stop codons at the appropriate positions in the plasmid expressing Orc6 . The results of the complementation analysis are summarized . ( B ) Complementation analysis of S . cerevisiae Orc6 . Plasmids expressing no Orc6 ( empty vector ) , Orc6 wild-type ( WT ) or mutated Orc6 as indicated were transformed into strain ySC166 carrying a degron fusion of the chromosomal copy of Orc6 ( Chen et al . , 2007 ) . Transformants were selected for the plasmid in permissive growth conditions , and then replica-plated and grown in non-permissive conditions . The number of viable colonies is significantly reduced in cells expressing Orc6 lacking parts of the conserved C-terminal domain . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 021
5 out of the 6 ORC subunits are well conserved between yeast and metazoans , predicting that substantial structural homology should exist between the respective ORC complexes ( Duncker et al . , 2009 ) . However , this paradigm has been challenged by apparent differences in shape between previous ScORC and DmORC EM reconstructions; for example , ScORC forms a somewhat flattened crescent , whereas DmORC has a pronounced third dimension and adopts a more helical conformation ( Speck et al . , 2005; Clarey et al . , 2006 ) ( compare Figure 1—figure supplement 2A , Figure 10A ) . These differences were too significant to allow the subunit order determined for ScORC to be mapped directly onto DmORC ( Chen et al . , 2008; Sun et al . , 2012 ) , an analysis complicated by the low-resolution nature of initial ORC EM models ( Speck et al . , 2005; Clarey et al . , 2006 , 2008 ) . 10 . 7554/eLife . 00882 . 022Figure 10 . Comparison of metazoan and S . cerevisiae ORC architectures and model of how the MGS mutation in Orc6 impedes ORC formation . ( A ) Different views of S . cerevisiae ORC ( EMDB 1156 , Speck et al . , 2005 ) and the new Drosophila ORC structure ( this study ) are compared ( a more recent , higher-resolution EM model of budding yeast ORC in isolation ( Sun et al . , 2012 ) has been not deposited into the EM database and so could not be used here ) . The localization of ORC subunits as reported for S . cerevisiae ORC ( Chen et al . , 2008; Sun et al . , 2012 ) and Drosophila ORC ( this study ) are indicated . ( B ) Model for AAA+ subunit architecture and Orc6 recruitment in ScORC and DmORC . In contrast to Drosophila , yeast Orc6 also interacts with Orc2 ( Sun et al . , 2012 ) . The MGS mutation in the C-terminal domain of Orc6 impedes binding of Orc6 to the Orc3 insertion and recruitment of Orc6 into metazoan ORC , likely resulting in reduced origin licensing . Numbers indicate the respective ORC subunit , ‘Ins’ the Orc3 insertion , and ‘IIB’ the TFIIB domains of Orc6 . DOI: http://dx . doi . org/10 . 7554/eLife . 00882 . 022 The improved resolution EM structure of DmORC presented here , together with experimental mapping of the ORC subunits in the EM structure , provides the first organizational model for the architecture of metazoan ORC ( Figure 10B ) . Characteristics observed in the current structure , but not in previous ones , include a more featured AAA+-ring and likely subunit borders ( Figure 2B ) . Importantly , this new EM structure of ATPγS-bound DmORC is more similar to budding yeast ORC EM structure ( Speck et al . , 2005; Sun et al . , 2012 , 2013 ) , although the fly complex displays more surface features and is thicker when viewed from the lateral edge of the AAA+ core ( Figure 10A ) . Moreover , the AAA+ subunit order determined here for DmORC—Orc1-Orc4-Orc5-Orc2–Orc3—agrees with that published for ScORC ( Figure 10B ) . Although both the published budding yeast ORC structures and the current DmORC structure lack the resolution to unambiguously locate AAA+ and winged helix domains of the ORC subunits , they nonetheless demonstrate fundamental similarities in ORC structure from yeast to metazoans . One notable difference we do observe between yeast ORC and DmORC is that Drosophila Orc6 is flexibly tethered to ORC through Orc3 and is hence invisible in the EM 3D structure ( Figures 3A and 10B ) . By comparison , S . cerevisiae Orc6 has been reported to contribute more significantly to EM density of yeast ORC ( Chen et al . , 2008 ) and has been reported to bind to Orc2 and not Orc3 ( Sun et al . , 2012 ) . Our pull-down data suggest that the more stable conformation seen for budding yeast Orc6 may be due to the existence of binding sites for the subunit on both Orc2 and Orc3 within ORC1–5 , as opposed to just Orc3 in the Drosophila complex ( Figure 9 ) . In Drosophila , Orc6 is an essential protein , with reduced Orc6 levels resulting in reduced DNA replication in both cell culture and flies ( Chesnokov et al . , 2003; Balasov et al . , 2009 ) . Replicative functions have been attributed to the N-terminal TFIIB-like region of Orc6 , whereas the C-terminal domain is required for cytokinesis and interacts with the septin Pnut ( Chesnokov et al . , 2003; Balasov et al . , 2007 , 2009; Huijbregts et al . , 2009 ) . Our results here show that the C-terminus of Orc6 is also critical for DNA replication . Orc6 is a stable component of Drosophila ORC , and ORC lacking Orc6 cannot support DNA replication in vitro in ORC-depleted Drosophila extracts ( Chesnokov et al . , 2001 ) . Since the C-terminus of Orc6 mediates the association of Orc6 with ORC ( Figures 4–7 ) , we predict that truncations or mutations abrogating this interaction should impair DNA replication . Indeed , the expression of C-terminally truncated or mutated Orc6 transgenes in orc6-null flies results in reduced levels of Mcm4 associated with chromatin ( Figure 7C ) , consistent with reduced levels of replicative helicase loading . These observations may explain the previously described cell-cycle arrest in mitosis that coincides with broken and uncondensed chromosomes , as well as reduced BrdU incorporation in neuroblasts observed in these fly strains , suggesting DNA is underreplicated ( Balasov et al . , 2009 ) . Moreover , Orc6 constructs lacking part of the C-terminal domain are compromised for initiating DNA replication in Drosophila extracts ( Balasov et al . , 2007 ) . Our results provide a mechanistic explanation for these observations: since Orc6 contributes to recruitment of ORC to chromatin through the binding of its TFIIB-like domains to DNA ( Balasov et al . , 2007 ) , loss of the interaction between Orc6 and ORC1–5 likely reduces the levels of chromatin-associated ORC and impairs ORC-dependent pre-RC assembly . The more moderate replicative phenotype of Orc6 mutations that affect Orc6 recruitment into ORC , compared to Orc6 mutations that affect the DNA binding activity of Orc6 ( Balasov et al . 2007 , 2009 ) , suggests that Orc6 might make contacts with non-ORC pre-RC components that help recruit ORC to DNA during pre-RC assembly . Since a majority of ORC in fly ovary and embryo extracts is not bound to origins , we could not test if mutant Orc6 is recruited into the pre-RC with our assays . Orc6 is one of several pre-RC components found mutated in Meier-Gorlin syndrome ( MGS ) patients . In certain MGS patients , compound heterozygous mutations inactivate one ORC6 allele and introduce a missense mutation in the other allele that substitutes serine for tyrosine in a conserved C-terminal helix of the Orc6 protein . We show that tyrosine-to-serine mutation reduces the affinity of Orc6 for the core ORC1–5 complex in both Drosophila and humans ( Figures 6 and 8 ) , a finding that would appear to implicate this subunit in the formation of the pre-RC . Interestingly , earlier observations have appeared inconsistent with such a role of Orc6 in some metazoans; for example , in contrast to the Drosophila system , Orc6 is only loosely or transiently associated with ORC1–5 in humans as well as in other metazoans , while in vitro DNA replication experiments using Xenopus extracts have suggested that Orc6 may not be required for DNA replication in this context ( Chesnokov et al . , 1999; Dhar and Dutta , 2000; Chesnokov et al . , 2001; Dhar et al . , 2001; Gillespie et al . , 2001; Vashee et al . , 2001 , 2003; Giordano-Coltart et al . , 2005; Ranjan and Gossen , 2006; Siddiqui and Stillman , 2007 ) . Nevertheless , evidence is emerging that Orc6 is essential for DNA replication in these species as well . First , depletion of Orc6 in human cell lines causes licensing and DNA replication defects ( Prasanth et al . , 2002; Stiff et al . , 2013 ) . Second , human Orc6 is required in vitro for efficient DNA replication in Orc6-depleted Xenopus extracts ( Liu et al . , 2011 ) ; the initial discrepancies between the Drosophila system and other metazoan systems may reflect different extents of immunodepletion of Orc6 from Xenopus extracts due to the use of different antibodies . Third , tethering of human Orc6 to DNA in human cells creates a functional origin and recruits other factors of the licensing machinery ( Thomae et al . , 2011 ) . Finally , immunodepletion of Orc6 from HeLa cell extracts impairs pre-RC assembly on plasmid DNA ( Thomae et al . , 2011 ) . Our data lend credence to the idea that the MGS mutation identified in human Orc6 is hypomorphic , and that the diminished affinity of MGS mutant Orc6 for ORC1–5 compromises pre-RC formation . Consistent with this interpretation , we observe reduced but weak binding of a Drosophila Orc6 construct bearing the MGS mutation to either Orc3 or ORC1–5 in equilibrium binding experiments and pull-down assays ( Figure 6B , D ) . Nonetheless , additional interactions of Orc6 with pre-RC components other than ORC subunits during origin licensing could also contribute to Orc6 incorporation into the pre-RC , explaining why the MGS mutation is not lethal in humans . In addition to its canonical function during replication initiation , metazoan Orc6 is also essential for cytokinesis ( Prasanth et al . , 2002; Chesnokov et al . , 2003; Bernal and Venkitaraman , 2011 ) , helps to control centrosome copy number and , along with other pre-RC components , appears to be essential for formation of primary cilia ( Stiff et al . , 2013 ) . The C-terminal domain of Orc6 harboring the MGS mutation is important for cytokinetic function , both in Drosophila and humans ( Chesnokov et al . , 2003; Bernal and Venkitaraman , 2011 ) . At this point it is not known whether the MGS mutation specficially impairs Orc6’s role in cytokinesis , or whether the C-terminal domain of Orc6 is involved in regulating centrosome copy number and cilia formation in humans . In addition , it is not understood how the defects in origin licensing caused by other MGS mutations would feed back to ciliogenesis ( Stiff et al . , 2013 ) . Nevertheless , given the fact that many independent mutations in different human ORC subunits , where some have been shown to affect replication functions directly , and hypomorphic mutations in the ATR replication check point protein lead to primordial dwarfism , it is hard to escape the inference that loss of ORC replication licensing functions are indeed epistatic to the MGS phenotype . Interestingly , MGS patients do not appear to have an increased predisposition to cancer , suggesting chromosome stability is not compromised ( de Munnik et al . , 2012a , b ) . How these observations are linked , how they relate to the phenotypes observed in MGS patients , and how specific MGS mutations intercept these processes will be important tasks for future studies . S . cerevisiae Orc6 and metazoan Orc6 have been shown to perform distinct roles in initiating DNA replication: metazoan Orc6 binds DNA through its TFIIB-like domains and helps recruiting ORC to chromatin ( Chesnokov et al . , 2001; Balasov et al . , 2007; Liu et al . , 2011 ) , while S . cerevisiae Orc6 does not bind DNA but has been shown to be directly essential for the loading step of the MCM double hexamer onto DNA ( Lee and Bell , 1997; Balasov et al . , 2007; Chen et al . , 2007; Duncker et al . , 2009; Evrin et al . , 2009; Remus et al . , 2009; Gambus et al . , 2011; Liu et al . , 2011; Takara and Bell , 2011; Fernandez-Cid et al . , 2013; Frigola et al . , 2013 ) . Up to now , this difference has seemed consistent with the apparent lack of significant sequence homology between these proteins . However , we found that the C-terminal helix that is crucial for Orc6 association with metazoan ORC is actually highly conserved in S . cerevisiae Orc6 , both in sequence and in function ( Figure 4—figure supplement 1 , Figure 9 ) . Moreover , we found that the sequence homology between metazoan and S . cerevisiae Orc6 extends to the TFIIB-like folds identified in metazoan Orc6 ( Figure 4—figure supplement 1 ) . Interestingly , budding yeast Orc6 lacks multiple basic amino acids that are conserved in metazoan Orc6 , resulting in a lower isoelectric point for its TFIIB domains ( pI = 7 . 5 ) compared to the Drosophila protein ( pI = 9 . 0 ) . This difference in surface charge may explain the distinct DNA binding propensities of metazoan and S . cerevisiae Orc6 ( Lee and Bell , 1997; Balasov et al . , 2007; Chen et al . , 2007; Liu et al . , 2011 ) . Nonetheless , because the TFIIB domains are preserved across eukaryotic Orc6 orthologs , it is tempting to speculate that these folds also play an important conserved function during origin licensing . While recent reports indicate that Orc6 may not be essential for MCM2–7 recruitment to origins , it is essential for loading of MCM double hexamers onto origin DNA ( Fernandez-Cid et al . , 2013; Frigola et al . , 2013 ) . The loading function of Orc6 may involve binding of Cdt1 or another pre-RC component through direct protein interactions , as has been reported for S . cerevisiae Orc6 and Cdt1 ( Chen et al . , 2007 ) . Notably , of the two regions of Orc6 that were shown to independently interact with Cdt1 , each contain one of the TFIIB-like domains identified here in yeast Orc6 ( Chen et al . , 2007 ) , suggesting Cdt1 binding may be conserved in metazoan Orc6 as well . Along these lines , acquisition of the MGS mutation in human Orc6 would be predicted to reduce licensing capacity and MCM loading by impeding Orc6 recruitment into ORC and pre-replicative complex formation , respectively ( Figure 10B ) . Future efforts using reconstituted systems will be needed to test these concepts further .
Drosophila and human Orc6 proteins were expressed in insect cells . Individual genes were cloned into LIC converted baculovirus transfer vectors adding an N-terminal 6 × His-tag . N- and C-terminal truncations of Drosophila Orc6 ( amino acids 94–257 , 147–257 , 187–257 , 1–200 , 1–220 , 1–232 ) were generated by PCR amplification of specified regions and cloned as N-terminal 6 × His-fusions into baculovirus transfer vectors by ligation-independent cloning . Point mutations in Orc6 were obtained by site-directed mutagenesis . All clones were verified by DNA sequencing . Bacmids and baculoviruses were generated as described above . For purification of all human and Drosophila Orc6 proteins , 2 l of High5 cells were infected with baculoviruses expressing respective wild-type or mutant Orc6 proteins . After 48 hr , cells were harvested and lysate was prepared as described in detail above for Drosophila ORC . Briefly , cells were resuspended in 50 ml lysis buffer ( 50 mM Tris HCl pH 7 . 8 , 50 mM imidazole pH 7 . 8 , 600 mM KCl , 10% glycerol , 1 mM β-mercaptoethanol , 200 µM PMSF , 1 µg/ml leupeptin ) and lysed by sonication . Lysate was clarified by centrifugation and a 0 . 78 M ( NH4 ) 2SO4 precipitation , followed by another centrifugation step . Subsequently , Orc6 was bound to a 5 ml HisTrap HP column ( GE Healthcare Biosciences ) and washed with 100 ml lysis buffer . Salt was reduced to 300 mM KCl by doing an additional wash with 50 ml buffer ( 50 mM Tris HCl pH 7 . 8 , 50 mM imidazole pH 7 . 8 , 300 mM KCl , 10% glycerol , 1 mM β-mercaptoethanol ) . Orc6 was eluted by increasing the imidazole concentration to 250 mM over 30 ml . Fractions containing Orc6 were pooled , concentrated and further purified by gel filtration chromatography on a HiPrep 16/60 Sephacryl S-200 HR column ( GE Healthcare Biosciences ) . To probe interactions of ORC subunits in vivo when co-expressed in insect cells , 50 ml of High5 cells were co-infected with baculovirus expressing respective subunits . Bait subunits were tagged N-terminally with MBP and prey subunits tested for interaction with the bait were tagged with a 6 × His tag at the N-terminus . This allowed us to ensure expression of the prey protein in whole cell extract by Western blotting . 2 days post infection , High5 cells were harvested by centrifugation at 1 , 000 × g and were resuspended in 1 . 4 ml lysis buffer ( 50 mM Tris HCl pH 7 . 8 , 300 mM KCl , 10% glycerol , 1 mM DTT , 200 µM PMSF , 1 µg/ml leupeptin ) . Cells were lysed by gentle sonication for 10 s and subsequently centrifuged at ∼16 , 000 × g for 30 min ( NH4 ) 2SO4 was added to a final concentration of 0 . 78 M to the lysate and samples were incubated on ice for 30 min . After an additional centrifugation at ∼16 , 000 × g for 30 min , the clarified lysate was added to 50 µl of Amylose beads ( New England Biolabs ) to bind MBP-tagged bait subunit . After 30 min incubation , beads were washed three times with 1 ml of lysis buffer . Bound proteins were eluted either by addition of 40 µl of 20 mM maltose in lysis buffer or by 40 µl of SDS loading dye . Eluted proteins were analyzed by SDS-PAGE and subsequent staining with Coomassie Brilliant Blue . To confirm that 6 × His-tagged prey proteins are expressed , whole cell extracts were analyzed by Western blotting . 100 µl of High5 cells were harvested and resuspended in 100 µl of SDS-loading buffer . Samples were boiled and 5 µl were loaded on an SDS-PAGE gel . After electrophoresis , proteins were transferred to Immobilon-PSQ membrane ( EMD Millipore ) using a Trans-Blot semi-dry electrophoretic transfer cell ( Biorad ) . Membranes were probed for His-tagged proteins using a monoclonal anti-His antibody ( GE Healthcare Biosciences ) at a dilution of 1:10 , 000 in PBS-T containing 5% ( wt/vol ) nonfat milk . HRP-conjugated goat anti-mouse IgG ( Pierce Thermo Fisher Scientific , Rockford , IL ) was diluted 1:20 , 000 in PBS-T and used as a secondary antibody . Heterozygous flies orc635/Cy , GFP-orc6-wt; orc635/Cy , GFP-orc6-W228A/K229A; orc635/Cy , GFP-orc6-D224A/Y225A contain endogenous orc6 deletion and GFP–Orc6 fusion transposons in which designated amino acids were mutated to alanine . 10–12 freshly dissected ovaries were crushed with a glass homogenizer in 100 µl of high salt IP buffer ( 25 mM HEPES pH 7 . 6 , 12 . 5 mM MgCl2 , 100 mM KCl , 0 . 1 mM EDTA , 450 mM NaCl , 0 . 01% Triton X100 ) and extracted for 1 hr at 4°C with continuous rotation . Extracts were pre-cleared by centrifugation at 15 , 000 × g for 15 min . Supernatant was diluted three times with regular IP buffer ( 25 mM HEPES 7 . 6 , 12 . 5 mM MgCl2 , 100 mM KCl , 0 . 1 mM EDTA , 0 . 01% Triton X100 ) . Protein A-Sepharose ( BioVision , Milpitas , CA ) and rabbit polyclonal Orc2 antibodies were incubated with supernatant for 3 hr , washed three times with IP buffer ( 20–30 min ) and diluted in 10 µl of IP buffer . Samples were boiled in loading buffer , separated in 10% SDS-polyacrylamide gels and transferred to Immobilon-P membrane ( EMD Millipore ) . ORC subunits were detected by Western blotting with anti-GFP monoclonal antibody ( Clontech Laboratories , Mountain View , CA ) as well as anti-Orc3 or anti-Orc5 antibodies . In addition , GFP fluorescence of samples was directly measured immediately after washing . 100 µl were transferred into a 384-well plate and GFP fluorescence was detected on Multi-Mode Microplate Reader Synergy 2 ( BioTek , Winooski , VT ) with excitation filter 485/20 , emission filter 528/20 and sensitivity 110 . Data were normalized to orc635/Cy strain fluorescence emission intensity . Homozygous ( orc635/orc635 , GFP-orc6-wt; orc635/orc635 , GFP-orc6-W228A/K229A; orc635/orc635 , GFP-orc6-D224A/Y225A ) or heterozygous ( orc635/Cy , GFP-orc6-wt; orc635/Cy , GFP-orc6-W228A/K229A; orc635/Cy , GFP-orc6-D224A/Y225A ) third instar larvae were selected based on fluorescence of the Cy-YFP balancer . Brains were dissected and imaginal discs removed . Soluble fraction of MCM2–7 was extracted with 450 mM NaCl , 0 . 5% NP-40 in PBS buffer for 1 hr . Insoluble proteins were pelleted , the pellet washed three times , and pellets from 8–10 brains per lane were loaded onto an SDS-PAGE gel and Mcm4 detected by Western blotting with Mcm4 rabbit polyclonal antibodies . Orc6 was expressed in yeast under the control of its own promoter using plasmid pSPB66 ( a kind gift from SP Bell ) . Point mutations and premature stop codons were introduced by site-directed mutagenesis and verified by DNA sequencing . The yeast Orc6 degron strain ySC166 ( Chen et al . , 2007 ) was transformed with pSPB66 or mutant derivatives and transformants were selected for the plasmid in permissive growth conditions ( Cu2+ , dextrose , 25°C ) . Subsequently , transformants were replica-plated and growth assessed in non-permissve conditions ( no Cu2+ , galactose , 37°C ) . Multiple protein sequence alignments were performed with MAFFT ( Katoh et al . , 2005; Katoh and Toh , 2008 ) . Accession numbers used are as follows . Orc3 alignment with other eukaryotic ORC subunits and archaeal Orc1/Cdc6: Sulfolobus solfataricus Orc1–1 ( PDB code 2qby , chain A ) , Sulfolobus solfataricus Orc1–2 ( AAK41068 ) , Sulfolobus solfataricus Orc1–3 ( PDB code 2qby , chain B ) , Aeropyrum pernix Orc1 ( PDB code 2v1u , chain A ) , A . pernix Orc2 ( PDB code 1w5s ) , Pyrobaculum aerophilum Cdc6 ( PDB code 1fnn; AAL62992 ) , Pyrococcus furiousus Orc1 ( AAL80141 ) , Archeoglobus fulgidus Cdc6 ( AAB90989 ) , Drosophila melanogaster Orc1 ( AAF59236 ) , D . melanogaster Orc2 ( AAF55006 ) , D . melanogaster Orc4 ( AAF47276 ) , D . melanogaster Orc5 ( AAC46956 ) , S . cerevisiae Orc3 ( AAB38249 ) , C . albicans Orc3 ( EAK93535 ) , S . pombe Orc3 ( AAF05949 ) , N . crassa Orc3 ( EAA36220 ) , Arabidopsis thaliana Orc3 ( AAT37463 ) , Oryza sativa Orc3 ( BAC56110 ) , Homo sapiens Orc3 ( AAT38109 ) , Mus musculus Orc3 ( NP_056639 ) , X . tropicalis Orc3 ( AAI35907 ) , Danio rerio Orc3 ( AAH45352 ) , D . melanogaster Orc3 ( AAF58411 ) , Aedes aegypti Orc3 ( EAT44270 ) . Alignment of Orc6 protein sequences: S . cerevisiae ( AAA21822 ) , C . albicans ( EAK99180 ) , S . pombe ( NP_596222 ) , A . fumigatus ( EDP49624 ) , N . crassa ( CAD37038 ) , A . thaliana ( AEE30748 ) , O . sativa ( EAZ40750 ) , maize ( ACG41692 ) , soybean ( ACU17985 ) , H . sapiens ( AAD32666 ) , M . musculus ( AAD32667 ) , Rattus norvegicus ( AAI01872 ) , X . tropicalis ( AAI35542 ) , D . rerio ( AAH56528 ) , Branchiostoma floridae ( EEN60699 ) , Saccoglossus kowalevskii ( XP_002740148 ) , D . melanogaster ( AAF58890 ) , A . aegypti ( EAT37789 ) , Bombyx mori ( ADD10142 ) , Trichoplax adhaerens ( EDV21277 ) , Phytophthora infestans ( EEY55501 ) , Dictyostelium discoideum ( EAL65038 ) , Polysphondylium pallidum ( EFA84700 ) . | Cell division is essential for organisms to be able to grow , to repair tissues and to proliferate . However , cells can only divide once they have successfully replicated their DNA . Many different molecules are involved in these two processes , including a large multi-protein assembly called the origin recognition complex that helps to start the process of DNA replication . This complex contains six proteins but relatively little is known about its structure . It is also unclear how much origin recognition complexes ( ORCs ) differ between species . Now , Bleichert et al . have found a way to stabilize a specific conformation of Drosophila ORC , and have gone on to determine its structure at a higher resolution than was previously possible . This approach revealed that the arrangement of protein subunits in Drosophila ORC is similar to that found in yeast ORC . Most of the ORC subunits have similar amino acid sequences in both species . However , the Orc6 subunit was regarded a notable exception for a long time , with the yeast and Drosophila versions of this subunit having different sequences of amino acids . Bleichert et al . show that the Orc6 subunits actually have important similarities , both in sequence and in function . In particular , the C-terminus of the Orc6 protein contains similar amino acids in both yeast and Drosophila . Moreover , it performs the same role—binding to another subunit—in both yeast and Drosophila . As well as being important for cell division , human ORC has been implicated in Meier-Gorlin syndrome , a type of dwarfism . Mutations in three of the six ORC subunits , including Orc6 , have been found in people with Meier-Gorlin syndrome . The mutations in Orc6 that are associated with this syndrome are in the C-terminus , which suggests that some symptoms of the syndrome may be caused by DNA replication not being initiated correctly . Consistent with this idea , Bleichert et al . show that the introduction of the Meier-Gorlin syndrome mutation into Orc6 prevents this subunit from binding to the rest of ORC , and similar mutations do not support DNA replication in in vivo experiments . These results should increase our understanding of the function of Orc6 and its role in Meier-Gorlin syndrome , and also provide new insights into the changes in ORC architecture that have occurred during evolution . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2013 | A Meier-Gorlin syndrome mutation in a conserved C-terminal helix of Orc6 impedes origin recognition complex formation |
In birds and higher mammals , auditory experience during development is critical to discriminate sound patterns in adulthood . However , the neural and molecular nature of this acquired ability remains elusive . In fruit flies , acoustic perception has been thought to be innate . Here we report , surprisingly , that auditory experience of a species-specific courtship song in developing Drosophila shapes adult song perception and resultant sexual behavior . Preferences in the song-response behaviors of both males and females were tuned by social acoustic exposure during development . We examined the molecular and cellular determinants of this social acoustic learning and found that GABA signaling acting on the GABAA receptor Rdl in the pC1 neurons , the integration node for courtship stimuli , regulated auditory tuning and sexual behavior . These findings demonstrate that maturation of auditory perception in flies is unexpectedly plastic and is acquired socially , providing a model to investigate how song learning regulates mating preference in insects .
Vocal learning in infants or juvenile birds relies heavily on the early experience of the adult conspecific sounds . In humans , early language input is necessary to form the ability of phonetic distinction and pattern detection in the phase of auditory learning ( Doupe and Kuhl , 1999; Kuhl , 2004 ) . Because of the strong parallels between speech acquisition of humans and song learning of songbirds , and the difficulties to investigate the neural mechanisms of human early auditory memory at cellular resolution , songbirds have been used as a predominant model in studying memory formation during vocal learning . In juvenile songbirds , a small subset of neurons in the higher-order auditory cortex responded selectively to a song experienced in the early exposure , and thus were thought to be the neuronal substrate for song memory formation ( Yanagihara and Yazaki-Sugiyama , 2016 ) . However , it remains unclear how the neurons that represent the sound memory are incorporated into the higher-order integration center to direct the sensorimotor output . The auditory system of Drosophila melanogaster has attracted increasing attention in recent years for the huge progress in understanding its underlying neural mechanisms ( Clemens et al . , 2015; Kamikouchi et al . , 2009; Zhou et al . , 2015 ) . The courtship song produced by wing vibration of males during the courtship ritual has been studied most among the communication sounds in flies ( Laturney and Billeter , 2014 ) . The courtship song , thought to be the primary cue affecting the female’s choice of the courting male ( Crossley et al . , 1995; Villella and Hall , 2008 ) , comprises two components: trains of pulses called pulse song and sequences of humming called sine song ( von Schilcher , 1976a ) . Although the function of the sine song is not well understood , sound playback experiments have demonstrated that the pulse song promotes copulation in paired flies ( Kyriacou and Hall , 1982; Ritchie et al . , 1999 ) . Receptivity of females is improved by playback of an artificial pulse song , which reduces female rejection responses and shortens the time to copulation ( Bennet Clark and Ewing , 1969; von Schilcher , 1976a; von Schilcher , 1976b ) . An artificial pulse song also increases sexual behavior in males , even without the presence of females , stimulating ‘chaining behavior’ , in which males chase each other and form male-male chains ( Crossley et al . , 1995; Yoon et al . , 2013 ) . This chaining behavior presumably arises from the increase of sexual arousal , which induces a male to join the courtship strived by other nearby males ( Eberl et al . , 1997 ) . Intriguingly , the quality of the pulse song affects sexual arousal . The temporal gap between the pulses in the pulse song , namely the inter-pulse interval ( IPI ) , differs among sibling Drosophila species ( Cobb et al . , 1989; Ewing and Bennet-Clark , 1968; Ewing and Manning , 1967 ) and is thought to be the crucial parameters for sexual arousal and species recognition . Indeed , D . melanogaster males prefer the pulse song with a certain range of IPIs including 35 ms , the mean IPI of this species ( Yoon et al . , 2013; Zhou et al . , 2015 ) . This bias towards the conspecific pulse song raises a question of how IPI selectivity is formed , making the fruit fly a simple model to investigate the mechanism underlying sound perception . Fruit flies detect sound with antennal ears and , specifically , with mechanosensory neurons of Johnston’s organ ( JO ) ( Kamikouchi et al . , 2009 ) . Regarding the two key features of Drosophila pulse song , intra-pulse frequency ( IPF ) and IPI , the antennal ear is mechanically tuned to detect the conspecific IPF , and the brain is hypothesized to process the conspecific IPI ( Riabinina et al . , 2011 ) . Recently an auditory pathway to perceive the pulse song that underlies the mating decision was delineated in Drosophila males . This pathway includes mechanosensory neurons in JO ( JO neurons ) , aPN1 neurons ( also known as AMMC-B1 neurons ) , vPN1 neurons , and pC1 neurons ( Kamikouchi et al . , 2009; Vaughan et al . , 2014; Zhou et al . , 2015 ) . In males , the pC1 cluster includes the courtship command-like P1 neurons . Multi-stage transformations by neurons in this auditory pathway refine the perception of IPIs until the response of the pC1 neurons matches the behavioral response to songs with different IPIs . These studies illustrate how the tuning towards the conspecific song with 35 ms IPI can be achieved , and raise the question of how this IPI preference emerges . Although it is traditionally believed that the courtship behavior of flies is innate ( Auer and Benton , 2016; Baker et al . , 2001; Hall , 1994 ) , the programmed courtship machinery is susceptible to variables in development such as sleep deprivation ( Kayser et al . , 2014 ) , social isolation ( Kim et al . , 1998; Pan and Baker , 2014 ) and juvenile social experience ( McRobert and Tompkins , 1988 ) . Pioneering studies on zebra finches ( Chen et al . , 2017; Cousillas et al . , 2006; Woolley et al . , 2010; Yanagihara and Yazaki-Sugiyama , 2016 ) and bats ( Razak et al . , 2008 ) suggest that auditory selectivity in these animals developed in an experience-dependent manner . Accordingly , we hypothesized that in young flies , IPI preference might also be refined by the experience of songs from nearby males , which might modulate the partner selection in sexual behaviors . In this study , we examined whether Drosophila IPI selectivity was tuned by the auditory experience . Based on the sexual behaviors of males and females upon song playback , we established a new behavioral paradigm in which the flies were exposed to specific sound patterns for long periods before their IPI preference was evaluated . Surprisingly , we found that the experience of conspecific song , but not heterospecific song , tuned IPI perception in both males and female flies . Furthermore , we found that this experience-dependent IPI tuning relied on GABA synthesis , and that the ionotropic GABAA receptor of pC1 neurons gated IPI tuning in females . Our discovery establishes a new and simple system to study how the experience-dependent auditory plasticity is incorporated into higher-order integration center to modulate sexual behaviors at the molecular and cellular levels .
In Drosophila melanogaster , IPIs ranging from 35 ms to 75 ms induce the sexual behavior of males vigorously ( Yoon et al . , 2013 ) . Since the mean IPI of the courtship song in D . melanogaster is about 35 ms ( Cowling and Burnet , 1981 ) , it seems noteworthy that 75 ms IPI , which is out of the melanogaster IPI range ( Arthur et al . , 2013 ) and likely comes from another Drosophila species ( for example , an evolutionarily far species Drosophila rosinae in fasciola subgroup ) ( Costa and Sene , 2002 ) , induces sexual behavior as strongly as 35 ms IPI . We noticed that male flies that showed similar levels of response to both 35 ms and 75 ms IPI songs had been wing-clipped soon after eclosion and thus lacked experiences of wing-emitted sound ( Yoon et al . , 2013 ) . Because Drosophilids gather in groups in feeding sites ( Powell , 1997 ) , we reasoned that flies probably had experiences of the courtship songs of other males in social interactions , and tested how the auditory experience affected the IPI selectivity . To evaluate how the experience of wing-emitted sound from other males affects later acoustic preference , we measured the chaining behavior of males that were reared for five to six days in the following three conditions: ( 1 ) grouped flies without wings , ( 2 ) grouped flies with intact wings , and ( 3 ) single-reared flies with intact wings . The wings of males in the latter two groups were clipped only one day before the chaining test . For the chaining test , we used two types of artificial pulse songs: 35 ms IPI and 75 ms IPI songs to represent conspecific and heterospecific songs , respectively . Consistent with our previous report ( Yoon et al . , 2013 ) , flies grouped without wings responded strongly to both conspecific and heterospecific songs ( Figure 1A ) . In contrast , flies grouped with wings preferred conspecific over heterospecific song ( Figure 1B ) . This selective response was not observed in single-reared flies with wings ( Figure 1C ) . Together , these results indicate that the presence of other males with wings is required to shape the IPI preference in males . To investigate whether the prior sound experience modifies the IPI selectivity , we established a training procedure containing a training session and a subsequent test session ( Figure 2A and B ) . In the training session , we exposed wing-clipped single males to conspecific or heterospecific artificial song for 6 days after eclosion , which served as ‘auditory experience’ to flies . Naïve flies were also prepared in the same manner as experienced flies except for the exposure to the training sound . In the test session , we monitored their behavioral performance using chaining test . Conspecific song induced a strong chaining behavior of males in both naïve and experienced groups , irrespective of the training sound ( Figure 2C ) . In contrast , heterospecific song induced a strong chaining behavior in naïve but not in experienced flies when flies were trained with conspecific song ( Figure 2D , red line ) . Flies trained with heterospecific song retained their response to the heterospecific song ( Figure 2D , orange line ) . These results indicate that male flies selectively diminish the response to heterospecific song only after having experienced conspecific song . Females decide whether to mate with courting males ( Dickson , 2008 ) . To test whether the mating decision of females could also be tuned by a prior auditory experience , we probed song effects on copulation behavior ( Figure 3A ) . First , we paired naïve females with naïve wing-clipped males to confirm the IPI selectivity in promoting copulation as reported ( Bennet Clark and Ewing , 1969 ) . Compared with the test condition without sound playback , either conspecific ( 35 ms ) or heterospecific ( 75 ms ) song playback promoted copulation significantly ( Figure 3B ) . Both songs promoted copulation equally , showing that naïve females had no selectivity between these two songs . In contrast , playback of songs with shorter ( 15 ms ) or longer ( 105 ms ) IPIs did not promote copulation . These results are consistent with previous findings that only songs with certain IPIs in a specific range promoted copulation ( Bennet Clark and Ewing , 1969 ) . Then we tested whether previous sound experience affects female copulation behavior . We trained the females with conspecific or heterospecific song , in the same way as for males ( Figure 2B ) , and then tested the female receptivity to a mute male with song playback ( Figure 4 ) . To examine the song training effect on females , naïve or trained females were paired with naïve wild-type males that were wing-clipped for copulation test . With playback of conspecific song , females accepted mating with mute males regardless of the song experience during the training session ( Figure 4A ) . In contrast , with heterospecific song playback , the copulation rate dramatically decreased in females trained with the conspecific song ( Figure 4B , red line ) . Training with heterospecific song did not affect the receptivity to the heterospecific song ( Figure 4B , orange line ) . Both the significant increase of the copulation latency ( Figure 4C ) and the highest inhibition index of the copulation rate ( Figure 4D ) supported the conclusion that training of the conspecific song reduced female acceptance during the heterospecific song test . Taken together , previous experience of the conspecific song renders females more selective about the song when deciding to accept mating . Apparently , prior experience of the conspecific song fine-tunes the selectivity of the sound-evoked behavioral responses of both males and females , while prior experience of the heterospecific song does not . We next sought to identify the mechanism of this experience-dependent tuning of auditory behavior . In mammals , auditory experience governs the maturation of GABAergic inhibition that tunes the perception of sound in the auditory cortex ( Dorrn et al . , 2010 ) . Thus we asked whether GABA signaling was involved in the auditory plasticity that we found , by testing the receptivity of female flies with reduced GABA synthesis . We knocked down Glutamic acid decarboxylase 1 ( Gad1 ) , a gene encoding the major GABA synthesis enzyme , in putative GABAergic neurons ( Gad1-GAL4 > UAS-Gad1 RNAi; see Materials and methods for fly strains ) in females , and trained them with conspecific or heterospecific song . The copulation tests with conspecific song playback revealed that both Gad1 knockdown and control ( Gad1-GAL4 > RNAi background w1118 ) females in experienced groups responded to conspecific song as strongly as naïve females , irrespective of training experience ( Figure 5A ) . In contrast , when we used heterospecific song in the tests , Gad1 knockdown females showed two phenotypes different from the control group ( Figure 5B ) . The first phenotype came after training of conspecific song ( Figure 5B , red lines ) ; while control females reduced receptivity like wild-type females ( Figure 5B , right ) , receptivity of Gad1 knockdown females stayed at the same level as in naïve females ( Figure 5B , left ) . This result suggests the necessity of GABA in this experience-dependent IPI tuning . The second phenotype appeared after heterospecific song training ( Figure 5B , orange lines ) ; Gad1 knockdown flies decreased their copulation rate dramatically when compared with naïve flies ( Figure 5B , left ) , whereas control flies ( Figure 5B , right ) and wild-type flies ( Figure 4B ) did not . These results demonstrate that although the response to the conspecific song in females was neither interrupted by Gad1 knockdown nor by training ( Figure 5A ) , the response to heterospecific song was vulnerable to Gad1 knockdown and training ( Figure 5B ) . Training with both conspecific song and heterospecific song might have modified properties of the neural circuit for the processing of heterospecific song . GABA synthesis is necessary to show the plasticity induced by conspecific song training , and to defend against the modulation induced by heterospecific song training as well . Together , these results prove that GABA synthesis is necessary for the IPI tuning induced by conspecific song training , which is reminiscent of the involvement of GABA in auditory plasticity exhibited in mammals and songbirds ( Dorrn et al . , 2010; Kotak et al . , 2008; Yanagihara and Yazaki-Sugiyama , 2016 ) . P1 neurons , a male-specific subset of pC1 neurons , are the mating command-like neurons that receive multimodal input from olfactory , gustatory , and auditory systems ( Auer and Benton , 2016 ) . Multimodal sensory information is transmitted to P1 neurons through excitatory and inhibitory pathways to achieve a stringent control of courtship decision-making in males ( Clowney et al . , 2015; Koganezawa et al . , 2016 ) . In these pathways , GABA transmits inhibitory signals to P1 neurons via GABAA-type Rdl receptors ( Kallman et al . , 2015; Koganezawa et al . , 2016 ) . Similarly , female pC1 neurons , the counterpart of male pC1 neurons ( Koganezawa et al . , 2016 ) , regulate female receptivity by evaluating sexual signals from males including the courtship song and the male-specific pheromone cVA ( Zhou et al . , 2014 ) . Under the hypothesis that GABA signaling via Rdl receptors might also regulate female pC1 neurons , we asked whether pC1 neurons in females were the target neurons of GABA that mediates the experience-dependent IPI tuning . We knocked down the expression of Rdl by driving Rdl RNAi specifically in female pC1 neurons , defined by the intersection of an enhancer trap line NP2631 and dsxFLP ( Koganezawa et al . , 2016 ) . Consistent with the aforementioned results , in the conspecific song test both Rdl knockdown and control females in experienced groups responded similarly as naïve females did , irrespective of training experiences ( Figure 6A ) . In the heterospecific song test , however , Rdl knockdown females , but not control and wild-type ones , kept the receptivity to the heterospecific song even after training with the conspecific song ( Figure 6B ) . Accordingly , knockdown of Rdl in pC1 neurons abolishes the experience-dependent tuning of the IPI , indicating that GABA mediates this IPI tuning via GABAA receptors in pC1 neurons of females . Interestingly , training with heterospecific song induced no changes in both Rdl knockdown and control groups ( Figure 6A and B ) . This result contrasts with that in Gad1 knockdown flies , in which the experience of heterospecific song reduced the female receptivity upon exposure to heterospecific song ( Figure 5B ) . Rdl receptors in female pC1 neurons are thus unlikely to be the direct target of GABA signaling to defend against the modulation induced by the training of heterospecific song .
Temporal pattern of sound is a crucial feature in the communication signals of many animals , such as in bird songs , frog calls , cricket chirps , and human speech ( Pollack , 2001 ) . Particularly in lower-vertebrates and insects , understanding the simple patterns of sounds used in communication , such as the specific pulse rate , is important in deciphering the meanings of these signals ( Alexander , 1962; Bass and McKibben , 2003; Schöneich et al . , 2015 ) . Fruit flies use the pulse songs with a species-specific IPI during courtship ( Ewing and Bennet-Clark , 1968 ) . In this study , we found that the flies’ initial wide-ranging IPI preference was refined by early auditory experience . Since the IPI distribution in the recorded natural courtship song is particularly enriched at around 35 ms ( Arthur et al . , 2013 ) , young adult flies are highly likely to be exposed to this conspecific IPI emitted by other males . This experience might tune the IPI preference and predispose partner selection in sexual behavior later in life . Indeed , our results prove that social interaction during early adulthood tunes the IPI preference towards the conspecific IPI ( 35 ms ) ( Figure 1 ) . This beautiful coordination between innate preference and experience-dependent refinement allows enough flexibility in mating , and reduces the risk of crossbreeding between species , which contributes to species isolation . We find that only the experience of conspecific song tunes the auditory preference , while the experience of heterospecific song does not . This asymmetric learning of conspecific and heterospecific songs suggests that naïve flies can already distinguish conspecific song from heterospecific song , since only the former is capable of modifying their later preference behavior . We previously reported that male D . melanogaster showed equal behavioral preference towards IPIs between 35 ms and 75 ms ( Yoon et al . , 2013 ) , which were used as conspecific and heterospecific songs in the present study . However , another report showed that male D . melanogaster preferentially responded to 35 ms over all other IPIs ( Zhou et al . , 2015 ) . This discrepancy can now be explained by the experimental difference between these two studies , whether the male flies kept in a group have the experience of carrying wings ( Zhou et al . , 2015 ) or not ( Yoon et al . , 2013 ) . As for how long the necessary experience is , and whether a critical period exists , further study is needed to answer these questions . Whether nature or nurture plays dominant roles in the formation of animal behavior has been debated for a long time , yet the courtship behavior of D . melanogaster , including its underlying sensory perception , has long been recognized to be innate . Numerous empirical evidences have supported the capability of single-reared flies to perform all the courtship steps spontaneously and completely ( Auer and Benton , 2016; Baker et al . , 2001; Hall , 1994 ) . However , our results reveal that the specific sound experience is necessary to refine the auditory preference in sexual behavior , which for the first time suggests a mechanism of learning in the song discrimination of flies . In fact , animals in many species learn their mating preferences . One notable example is sexual imprinting , the process whereby mating preferences are affected by learning the species-specific characteristics at a very young age ( Irwin and Price , 1999 ) . As observed in birds ( Ten Cate , 1999 ) , fishes ( Kozak et al . , 2011 ) , and sheep and goats ( Owens et al . , 1999 ) , an early period of social interaction with parents or siblings helps the learner discriminate sex and species by learned phenotypic traits , and affects mating preference in the future ( Verzijden et al . , 2012 ) . Here we provide evidence that fruit flies refine the IPI preference by sexual imprinting , which would reinforce reproductive isolation together with innate auditory perception . This sexual imprinting of courtship song is apparently different from the lessons learned from the successful courtship experience ( Saleem et al . , 2014 ) or unsuccessful courtship attempts ( Griffith and Ejima , 2009 ) , by which male flies become more competitive over other males , or learn to avoid either mated or heterospecific females . Previous behavioral studies also indicated that social experience in juvenile stage affected adult courtship behaviors of insects . In crickets , juvenile experience of acoustic sexual signals influenced the development of three traits in adult: reproductive tactics , reproductive investment , and body condition ( Bailey et al . , 2010 ) . In fruit flies , young males courted by mature males with intact wings mated significantly faster than those that had been stored alone , suggesting auditory experience in immature stage might affect later courtship ( McRobert and Tompkins , 1988 ) . Consistent with these observations and going deeper , our study directly demonstrated , with the underlying mechanisms , that auditory experience during the immature stage shaped perception of courtship song , and directed the sexual behavior at the adult stage . Our findings greatly expand the understanding of the experience-dependent auditory plasticity in insects , whose mechanism is consistent with that of mammals and finches . In vertebrates , maturation of excitation-inhibition balance that governs sound perception requires acoustic experience . In rats , developmental sensory experience balances the excitation and inhibition in the primary auditory cortex ( A1 ) ( Dorrn et al . , 2010 ) , whose stereotyped sequential occurrence sharpens spike timing ( Wehr and Zador , 2003 ) . Hearing loss hinders the maturation of GABAergic transmission mediated by GABAA receptors in the auditory cortex of gerbils ( Kotak et al . , 2008 ) . In zebra finch , experience-dependent recruitment of GABAergic inhibition in the auditory cortex is necessary to form the memory template of the tutor song ( Yanagihara and Yazaki-Sugiyama , 2016 ) . In flies , our results also suggest that song experience recruits GABAergic inhibition on the auditory pathway , and the coordination of excitation and inhibition controls auditory responses and behavioral output ( Figure 7 ) . Interestingly , the phenotypes of Gad1 knockdown in GABAergic neurons and Rdl knockdown in pC1 neurons were different when females were tested with heterospecific song ( Figures 5 and 6 ) . This finding suggests that there are at least two distinct GABAergic pathways to control the experience-dependent auditory plasticity . How these GABAergic pathways are organized cooperatively to shape the IPI preference awaits further analysis . Interestingly , the combination of excitation and inhibition that modulates the mating decision in flies is not restricted to the auditory system , but is also conserved in olfactory and gustatory systems ( Auer and Benton , 2016; Clowney et al . , 2015; Kallman et al . , 2015 ) . The difference is that the sexual circuitry in the chemosensory modalities is thought to be hard-wired ( Auer and Benton , 2016; Hall , 1994; Pan and Baker , 2014 ) , while the inhibition we find in the auditory system matures with experience . Intriguingly , all these inhibitions found in olfactory , gustatory , and auditory pathways function directly on the pC1 neurons , strengthening the role of pC1 neurons as a crucial neural circuit node for multimodal integration ( Auer and Benton , 2016; Clowney et al . , 2015; Kallman et al . , 2015 ) . The discovery that only the training of conspecific song refines the IPI preference of wild-type flies is reminiscent of vocal learning in zebra finches , which preferentially learn the courtship song of their own species ( Brenowitz and Woolley , 2004; Doupe and Kuhl , 1999 ) . The courtship song preferences in female zebra finches are shaped by the developmental auditory experience ( Chen et al . , 2017 ) , which shares great similarity with that in fruit flies . The IPI in the courtship song of flies resembles the temporal gap between syllables in the finch song , which serves as a ‘barcode’ for song identity ( Araki et al . , 2016 ) . The auditory study of these two model organisms therefore might complement and enlighten each other in exploring the mechanism of experience-dependent plasticity in conspecific sound perception , potentially contributing to the understanding of language acquisition in humans . Unlike zebra finch , fruit flies rarely see their parents . Our results demonstrate that learning from adolescent peers is sufficient to modulate the perception of IPIs ( Figure 1 ) . In the natural environment , young flies possibly learn from young flies as well as mature flies . Taken together , our findings open a new research field to use the fruit fly , with its abundant molecular-genetic tools and simple neural circuits , to study the experience-dependent auditory information processing and sensorimotor output , which are challenging to examine at the molecular and cellular levels in zebra finches and primates including humans .
D . melanogaster was raised on standard yeast-based media at 25°C and in 40% to 60% relative humidity on a 12 hr light/dark cycle . Canton-S ( Hotta-lab strain , a gift from K . Ito ) was used as a wild-type strain . For knockdown experiments , the following transgenic flies were used: w; Gad1-GAL4 ( Ng et al . , 2002 ) ( a gift from K . Ito ) , UAS-Gad1 RNAi ( GD line; RRID: FlyBase_FBst0459538 ) and its control line w1118 ( VDRC ID: 60000 ) ( Vienna Drosophila Resource Center ) , UAS-Rdl RNAi ( VALIUM20; RRID: BDSC_52903 ) and its control line TRiP RNAi ( RRID: BDSC_36304 ) ( Bloomington Drosophila Stock Center ) , and tubP>GAL80>; NP2631-GAL4/CyO; dsxFLP/TM2 ( Koganezawa et al . , 2016 ) ( a gift from D . Yamamoto ) . Genotypes of flies used for each experiment are listed in Supplementary file 1 . Flies that were 6 to 7 day after eclosion were used for behavioral tests . The wings of males were clipped on the day of eclosion , unless otherwise noted . The neurons labeled by Gad1-GAL4 show essentially consistent distributions with those identified by in situ hybridization against Gad1 mRNA ( Okada et al . , 2009 ) . Silencing these Gad1-GAL4 positive neurons in the adult stage did not affect fly survival ( Muthukumar et al . , 2014 ) . The Gad1 RNAi used in this study was reported to knock down the Gad1 mRNA level to approximately 60% of wild type ( Jeong et al . , 2016 ) . In our study , no obvious behavioral defects were observed in Gad1 knockdown flies , and male Gad1 knockdown flies still responded normally to conspecific courtship song when tested at 7 days after eclosion ( Figure 5—figure supplement 1 ) . The efficacy of UAS-Rdl RNAi has been demonstrated ( Franco et al . , 2017; Koganezawa et al . , 2016 ) . Virgin males were collected within 10 hr after eclosion , and then housed in three different conditions: ( 1 ) grouped without wings , ( 2 ) grouped with intact wings , and ( 3 ) single-reared with intact wings . Flies housed in the first condition ( grouped without wings ) were prepared as described previously ( Yoon et al . , 2013 ) . In brief , their wings were clipped with forceps during brief anesthesia on ice soon after eclosion and the males were kept in a male-only group of 6 to 8 . Flies housed in the second ( grouped with intact wings ) and third ( single with intact wings ) conditions were kept with intact wings for 5 to 6 days , either in a group of 6 to 8 male flies or singly . Only one day before the test , the wings of flies housed in the second and third conditions were also clipped . The chaining behavior of all the males housed in three conditions was tested 6 to 7 days after eclosion . A protocol for the training session is described in more detail at Bio-protocol ( Li et al . , 2018 ) . Training session started on the day of eclosion . Adult virgin males and females were collected within 8 hr after eclosion under anesthesia on ice , and the wings of males were clipped . Each fly , whether a male or a female , was introduced gently to a training capsule and placed in front of a loudspeaker ( FF225WK , FOSTEX , Foster Electric Company , Tokyo , Japan ) . As experienced group , flies were continuously exposed to one particular training song for 6 days of training ( Figure 2A ) . Training song was an artificial pulse song comprised of the repetition of 1 s pulse burst and a subsequent 2 s pause , in which the pulses in the pulse burst had an IPI of 35 ms ( ‘conspecific song’ ) or an IPI of 75 ms ( ‘heterospecific song’ ) ( Yoon et al . , 2013 ) . Intrapulse frequency ( IPF ) of both IPI songs was set to be 167 Hz . As naïve group , flies were placed in front of the loudspeaker for 6 days after eclosion but not given any sound exposure . During the training session , each fly was accommodated singly in a training capsule . A training capsule was made of a glass tube cut out from a Pasteur pipette , two pipette tips , mesh and mending tape ( Figure 2B ) . Pipette tips , whose volumes are 1 ml , were cut to make the larger ends about 20 mm long . Two of these 20 mm pieces were hooked to a glass tube at its both ends . The size of a glass tube was about 27 mm long , with the internal diameter of 5 . 2 mm and the external diameter of 6 . 5 mm . Both exits of the glass tube were sealed with a piece of mesh stocking ( made of nylon and polyurethane ) , which allowed free passage of air but not the fly . A thin layer of fly food , standard Drosophila yeast-based medium , was paved at the bottom of the glass tube . The food in each capsule was renewed every 36 hr . Training capsules were placed within latticework of a container , named a ‘sound nest’ ( Figure 2B ) . One of the mesh-ends of each training capsule faced the loudspeaker , so that sound could be delivered to each chamber with minimal disturbance . The distance between loudspeaker and the near end of the training capsules was 24 mm . All the setups for the training were placed into a soundproof box ( W450 mm × L450 mm × H450 mm ) . Sound playback was controlled by the Windows Media Player on a tablet PC ( Windows 8 . 1 , Diginnos DG-D08IWB , Dospara , Tokyo , Japan ) , and delivered by a loudspeaker with a digital power amplifier ( Lepai LP-2020A + NFJ Edition , Bukang Electrics , Jieyang , China ) . The mean baseline-to peak amplitude of sound particle velocity was 8 . 6 mm/s when measured at the near end of the training capsules , and 6 . 6 mm/s at the far end of the training capsule . The sound particle velocity was identical for all training sounds . After the 6 day training , male flies were collected into a group of seven without anesthesia , and transferred to a vial containing fly food . Female flies were still kept in the training capsules singly without sound playback until the copulation test . After one-night rest without any sound playback , all flies ( 7 days after eclosion ) were subjected to the behavioral tests in the next morning ( ZT 0–3 ) ( Figure 2A ) . A protocol for the test session is described in more detail at Bio-protocol ( Li et al . , 2018 ) . Statistical analysis was performed with R ( version 3 . 0 . 3 ) . Mann-Whitney U test ( two-tailed ) was used to compare two groups of samples in the chaining behavior . Kaplan-Meier curves were generated using R and Log rank test was performed to compare females’ accumulative copulation rate between two groups in the copulation tests . The Kruskal–Wallis test ( two-tailed ) followed by Scheffe’s test was used to compare the copulation latency . The detailed statistical results are shown in Supplementary file 2 . The boxplot was drawn with ggplot2 package of R . Boxplots display the median of each group with the 25th and 75th percentiles and whiskers denote 1 . 5x the inter-quartile range . | Many mammals and birds have a critical period in youth when hearing the vocal cues of their parents helps them to learn the specific features of their communication sounds . Scientists have been studying the brains of humans , birds and other animals to find out what is happening in their brains when the animals hear these sounds . However , the brains of these species are too complex to fully understand how early vocal influences shape the brain networks that control behavior . Therefore , scientists often use ‘simpler’ organisms , such as insects , to study these processes . For example , fruit flies use a series of courtship behaviors – including mating calls – to attract their potential mates . To produce a courtship song , males vibrate their wings , which consists of short pulsed songs and sequences of humming . The time interval between the pulses is specific to a species . Until now it was thought that these mating calls are innate behaviors that cannot be learned or modified . To test this , Li et al . clipped the wings of male fruit flies so they could not produce their own song . First , they placed the females with the males and played one species-specific courtship song , and one from a different species . Both songs resulted in successful copulation and did not affect the female’s choice . To find out if a previous experience of a courtship song can influence the behavior of the fruit flies , Li et al . raised one group hearing their species-specific song and the other with a song from a different species . The results showed that females growing-up with their species-specific song , rejected males when a song of another species was played . However , the females accustomed to the other species’ song did not change their song preference and receptivity towards males . The same was also true for males: male fruit flies raised with their species-specific song later ignored another species’ song , which usually increased their mating drive . Li et al . further identified a specific region in the brain of the fruit flies known to be important for courtship , and a key molecule that regulated this behavior . These findings suggest that far from being innate , the mating preference in fruit flies can be learned and influenced by social experience . A next step will be to find out if fruit flies also have critical period for learning vocal cues and if so , how it is regulated at the molecular and neural levels . A better understanding of how fruit flies learn and discriminate sounds may bridge knowledge gaps in research using humans and other mammals . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Auditory experience controls the maturation of song discrimination and sexual response in Drosophila |
Expression levels of CX3CR1 ( C-X3-C motif chemokine receptor 1 ) on immune cells have significant importance in maintaining tissue homeostasis under physiological and pathological conditions . The factors implicated in the regulation of CX3CR1 and its specific ligand CX3CL1 ( fractalkine ) expression remain largely unknown . Recent studies provide evidence that host’s misfolded proteins occurring in the forms of polymers or amyloid fibrils can regulate CX3CR1 expression . Herein , a novel example demonstrates that polymers of human ZZ alpha-1 antitrypsin ( Z-AAT ) protein , resulting from its conformational misfolding due to the Z ( Glu342Lys ) mutation in SERPINA1 gene , strongly lower CX3CR1 mRNA expression in human peripheral blood mononuclear cells ( PBMCs ) . This parallels with increase of intracellular levels of CX3CR1 and Z-AAT proteins . Presented data indicate the involvement of the CX3CR1 pathway in the Z-AAT-related disorders and further support the role of misfolded proteins in CX3CR1 regulation .
Interactions between the chemokine receptors and chemokines , but also other proteins , peptides , lipids , and microbial products , play a critical role in the recruitment of inflammatory cells into injured/diseased tissues ( Bachelerie et al . , 2014 ) . Many human diseases involve altered surface expression of chemokine receptors , which can lead to a defective cell migration and inappropriate immune response . Most of the human peripheral blood mononuclear cells ( PBMCs ) express CX3CR1 ( Bazan et al . , 1997 ) , also known as the G-protein-coupled receptor 13 ( GPR13 ) or fractalkine receptor , a mediator of leukocyte migration and adhesion . In the central nervous system , CX3CR1 is largely expressed by microglial cells ( brain macrophages ) ( Ransohoff , 2009 ) , which are involved in neurodegenerative diseases like Alzheimer’s disease . The major role of CX3CR1-expressing cells is to recognize and enter tissue following CX3CL1 ( fractalkine or also called neurotactin ) gradient , and to crawl or ‘patrol’ in the lumen of blood vessels ( Auffray et al . , 2007 ) . Since CX3CR1/CX3CL1 axis is also involved in the synthesis of anti-inflammatory cytokines and has a significant role in cytoskeletal rearrangement , migration , apoptosis , and proliferation , its dysregulation is associated with the development of cardiovascular diseases , kidney ischemia–reperfusion injury , cancer , chronic obstructive pulmonary disease ( COPD ) , neurodegenerative disorders , and others ( Harrison et al . , 1998; Ning et al . , 2004; Rius et al . , 2013 ) . Some studies indicate that CX3CR1 deficiency contributes to the severity of infectious diseases ( Bonduelle et al . , 2012 ) and promotes lung pathology in respiratory syncytial virus-infected mice ( Das et al . , 2017 ) . Animals with deletion of CX3CR1 show impaired phagocytosis ( Thome et al . , 2015 ) , which is vital to prevent unwanted inflammation . It is clear that CX3CR1-expressing cells have tissue-specific roles in different pathophysiological conditions . Nevertheless , a comprehensive knowledge on the regulation of CX3CR1 expression is still missing . Current findings suggest that divergent proteins with a common propensity to form extracellular oligomers interact with chemokine receptors and affect their expression levels . For example , Alzheimer’s peptide , Aβ , interacts with CX3CR1 and significantly reduces its expression in cultured microglial cells and in Alzheimer’s brain ( Cho et al . , 2011 ) . Similarly , highly aggregated extracellular Tau protein binds to CX3CR1 , promotes its internalization , and reduces expression in microglial cells ( Bolós et al . , 2017 ) . In concordance , polymers of human Z alpha-1 antitrypsin ( Z-AAT ) , resulting from protein misfolding due to the Z ( Glu342Lys ) mutation in SERPINA1 gene , lower CX3CR1 mRNA expression in human PBMCs , which parallels with increased intracellular CX3CR1 and Z-AAT protein levels .
Inherited alpha-1 antitrypsin deficiency ( AATD ) is a rare genetic condition caused by SERPINA1 gene mutations . Homozygous Z AATD mutation is the most clinically relevant among Caucasians ( prevalence is about 1:2000-1:5000 ) that is characterized by low plasma levels of AAT protein ( 10–15% compared to the wild type , MM AAT , 1 . 3–2 g/l ) and the presence of intracellular and circulating Z-AAT polymers ( Tan et al . , 2014 ) . The liver is the major producer of AAT , therefore the accumulation of Z-AAT polymers in hepatocytes is a marker for diagnosing AATD ( Janciauskiene et al . , 2011 ) . The intracellular Z-AAT polymers have also been identified in other AAT-expressing cells like monocytes and macrophages ( Belchamber et al . , 2020 ) . The accumulation of polymers is harmful for AAT-producing cells , whereas the circulating Z-AAT polymers are not able to execute the tasks of AAT protein , a major inhibitor of serine proteases having a strong immunomodulatory potential . Based on the facts that: ( i ) circulating Z-AAT polymers contribute to the risk of developing pathologies ( Parmar et al . , 2002; Strnad et al . , 2020 ) , ( ii ) pathogenic oligomeric proteins affect CX3CR1 expression ( Bolós et al . , 2017 ) , and ( iii ) CX3CR1/CX3CL1 axis plays a significant role in immunity ( Imai and Yasuda , 2016 ) , we aimed to investigate CX3CR1 expression in PBMCs of ZZ AATD individuals . For this , in collaboration with German Alpha1 Patient Association and Aachen University , was prepared RNA from freshly isolated PBMCs of 41 clinically stable ZZ AATD volunteers independently of their clinical diagnosis or treatment with intravenous AAT , a specific augmentation therapy ( Janciauskiene and Welte , 2016 ) . For comparison , PBMCs isolated from healthy volunteers having normal plasma AAT levels were used . Additionally , a limited amount of RNA sample was available from PBMCs isolated from a cohort of 12 ZZ AATD emphysema patients at Leiden University Medical Center , The Netherlands ( Figure 1—figure supplement 1 ) . Independent of individual’s age , clinical diagnosis ( healthy , lung or liver disease ) , or augmentation therapy , the CX3CR1 mRNA expression turned to be much lower in ZZ AATD PBMCs than in PBMCs from non-AATD controls ( median [range]: 4 . 1 [2 . 7–5 . 5] vs . 18 . 5 [13–26 . 6] , p<0 . 001 ) ( Figure 1A ) . A previous study has shown that CX3CR1−/− mice have significantly higher plasma levels of CX3CL1 than wild-type mice ( Cardona et al . , 2008 ) . A diminished expression of CX3CR1 might be related to increased levels of soluble CX3CL1 , an exclusive ligand for CX3CR1 ( Bachelerie et al . , 2014 ) . However , the concentration of plasma CX3CL1 was low , did not differ between ZZ AATD and non-AATD individuals ( Figure 1B ) , and did not correlate with CX3CR1 mRNA in PBMCs . The expression and release of CX3CL1 is generally low in the absence of inflammatory insults ( Umehara et al . , 2004 ) showing that at the time point of blood donation all volunteers were under stable clinical condition . Although CX3CR1 is preferentially expressed on monocytes , other cells also express this receptor ( Landsman et al . , 2009 ) . Previous reports indicated that exogenous IL-15 is a negative regulator of CX3CR1 expression in human CD56+ NK cells ( Barlic et al . , 2003; Sechler et al . , 2004 ) . However , plasma levels of IL-15 were lower in ZZ AATD than in non-AATD ( pg/ml , median [range]: 6 . 6 [5 . 9–6 . 9] , n = 23 vs . 7 . 63 [6 . 63–8 . 1] , n = 21 , p=0 . 001 ) , excluding a possible link between IL-15 and CX3CR1 mRNA levels . Because ZZ AATD individuals , differently from non-AATD , have about 90% lower blood concentration of Z-AAT protein , which may influence cellular microenvironment ( Ramos et al . , 2010 ) , a relationship between CX3CR1 and Z-AAT plasma levels cannot be excluded . However , no correlation was found between CX3CR1 mRNA in PBMCs and plasma levels of Z-AAT measured by nephelometry ( data not shown ) . Next , plasma Z-AAT polymers were measured , as the biomarkers of all carriers of the Z allele ( Tan et al . , 2014; Janciauskiene et al . , 2002 ) . As anticipated , only minor amounts of polymers were detected in plasma of non-AATD individuals while plasma of ZZ AATD contained high amounts of polymers ( µg/ml , mean [SD]: 4 . 1 [6] , n = 18 vs . 1399 . 8 [750] , n = 20 , respectively ) . Since most of the ZZ AATD individuals received intravenous augmentation therapy with plasma purified AAT protein , ZZ AATD individuals were segregated into subgroups who receive or not receive therapy . There were no significant differences in Z-AAT polymer levels between the subgroups: ( µg/ml , median [range]: non-augmented 1506 . 6 [854–1781] , n = 17 vs . augmented 1348 . 5 [779 . 5–1529] , n = 23 , respectively ) . A previous study used a sandwich ELISA based on 2C1 antibody and found that circulating Z-AAT polymers range between 8 . 2 and 230 . 2 μg/ml in ZZ AATD ( Tan et al . , 2014 ) whereas much higher circulating levels of Z-AAT polymers were detected by using the single monoclonal antibody ( LG96 ) -based ELISA . These discrepancies can be due to the differences between antibody specificities . For example , 2C1 showed high affinity for polymers formed by heating M- or Z-AAT at 60°C ( Miranda et al . , 2010 ) while LG96 antibody recognizes naturally occurring/native Z-AAT polymers without requiring sample heating . To answer , why some individuals have higher plasma levels of Z-AAT polymers than monomers ( measured by nephelometry ) is of great importance for the further studies . Most interestingly , in ZZ AATD individuals , was found a trend toward an inverse relationship between CX3CR1 mRNA in PBMCS and plasma Z-AAT polymers ( r2 = −0 . 31 , n = 38 , p=0 . 055 ) ( Figure 1C ) . This latter finding prompted more extensive investigation whether Z-AAT polymers affect CX3CR1 expression when added to healthy donor PBMCs for 18 hr , ex vivo . Lipopolysaccharide ( LPS , from Escherichia coli , 1 µg/ml ) was included as a known reducer of CX3CR1 expression ( Pachot et al . , 2008; Sica et al . , 1997 ) . Indeed , polymeric Z-AAT in a concentration-dependent manner lowered CX3CR1 mRNA ( Figure 2—figure supplement 2 ) whereas repeated experiments using Z-AAT at a constant concentration of 0 . 5 mg/ml reduced CX3CR1 mRNA more than twice as compared to non-treated controls ( Figure 2A ) . Accordingly , LPS and polymer containing Z-AAT preparation significantly decreased surface expression of CX3CR1 , specifically in CD14+ monocytes and NK cells ( Figure 3 ) . By contrast , cellular levels of CX3CR1 protein increased in PBMCs treated with Z-AAT polymers or LPS ( used as a positive control ) as compared to non-treated controls ( Figure 2B ) . The CX3CR1 protein was present in detergent resistant lipid raft fraction of PBMCs treated with Z-AAT ( Figure 2—figure supplement 3A ) . Total cell lysates and lipid raft fractions from Z-AAT-treated PBMCs , in contrast to those prepared from M-AAT-treated or non-treated PBMCs , contained high amounts of AAT polymers ( Figure 2C , Figure 2—figure supplement 3B ) . The laser scanning confocal microscopy of double-labeled specimens showed a co-localization of Z-AAT polymers with CX3CR1 protein ( Figure 2D ) . Furthermore , 3D reconstruction of cross sections visualized larger Z-AAT aggregates surrounded by cellular extensions in a cap-like formation , suggesting that cells may react differently depending on the size of Z-AAT polymers ( Figure 2E ) . It cannot be excluded that Z-AAT polymers , similar like polymers of Tau protein , interact with CX3CR1 and get internalized ( Chidambaram et al . , 2020 ) . This may determine the fate of CX3CR1 mRNA expression , that is , sequestered intracellularly and not returning to the cell surface , CX3CR1 protein may induce signaling pathways lowering CX3CR1 expression . To achieve a definitive answer how Z-AAT or other types of protein polymers regulate CX3CR1 levels , detailed mechanistic studies are required . In general , along with transcriptional regulation , chemokine receptors trafficking is of great importance to understand ( Kershaw et al . , 2009 ) . Under the same experimental conditions , monomeric M-AAT had no effect on CX3CR1 mRNA expression relative to housekeeping gene HPRT1 ( mean [SD]: 24 . 9 [2 . 9] controls , n = 5 vs . 23 . 7 [1 . 3] , n = 5 , NS ) , and protein levels ( Figure 2B ) . Likewise , monomeric Z-AAT protein does not affect CX3CR1 mRNA and protein levels , and heat-induced polymers of M-AAT showed no effect on CX3CR1 expression as well ( Figure 2—figure supplement 4 ) . Probably , specific conformational properties and/or molecular sizes of Z-AAT polymers are required for their interaction with CX3CR1 . For example , cell surfaces express CX3CL1 as a constitutive oligomer ( three to seven molecules ) , which is essential for efficient interaction with CX3CR1 ( Hermand et al . , 2008; Ostuni et al . , 2020 ) . Numerous chemokines tend to self-associate that determines their activity ( Proudfoot et al . , 2003 ) , and therefore certain Z-AAT polymers may resemble chemokine structures competing for the same receptor ( s ) . In some experimental models , Z-AAT polymers expressed strong chemotactic properties ( Parmar et al . , 2002; Lomas and Mahadeva , 2002 ) . When chemokine receptors are engaged in chemotaxis , they can be removed from the cell surface by the ligand–receptor internalization ( Springer , 1994 ) , which might explain a decrease of CX3CR1 in ZZ PBMCs . Interestingly , the soluble form of CX3CL1 , even when used at a high concentration of 500 ng/ml , does not antagonize Z-AAT polymer effects on CX3CR1 mRNA and protein levels , and by itself showed no effect on CX3CR1 mRNA or protein levels ( Figure 2—figure supplement 5 ) , although some studies reported that CX3CL1 reduces CX3CR1 expression ( Pachot et al . , 2008; White et al . , 2014 ) . In solution CX3CL1 remains monomeric , even at high concentrations ( Hermand et al . , 2008; Mizoue et al . , 1999 ) whereas , as mentioned above , membrane-bound CX3CL1 occurs as oligomer . These two forms of the CX3CL1 perform differential roles ( Winter et al . , 2020 ) , and therefore it cannot be excluded that oligomeric , but not a soluble , form of CX3CL1 would compete with Z-AAT for CX3CR1 interaction in vivo . The CX3CR1 helps to define the major subsets of human monocytes because classical monocytes express much lower levels of CX3CR1 than non-classical monocytes ( Ziegler-Heitbrock et al . , 2010 ) . After in vitro challenge with LPS for longer periods ( like for 18 hr ) , human monocytes are known to increase in the mRNA and membrane expression of CD14 , a receptor for LPS ( Landmann et al . , 1996 ) . The enhancement of CD14 expression after treatment of PBMCs with Z-AAT strikingly resembled LPS ( Figure 4 and Figure 4—figure supplement 1 ) . This raised a suspicion that Z-AAT preparations might contain endotoxin . According to the limulus amebocyte lysate test , endotoxin levels of Z-AAT preparations were below detection limit ( 0 . 01 EU/ml ) . Moreover , LPS significantly induced expression of TNFα , IL-6 , and IL-1β while polymer containing Z-AAT preparations had no effect ( Figure 4—figure supplement 2 ) . Besides , LPS but not Z-AAT significantly increased release of cytokines ( IL-1β , pg/ml , median [range]: LPS 1342 . 9 [1008–1834] vs . Z-AAT 3 . 2 [2 . 5–5 . 9] vs . controls 2 . 5 [2 . 1–3 . 7] , n = 4 independent experiments; TNFα , ng/ml , mean [SD]: LPS 19 . 5 [2 . 5] vs . Z-AAT vs . controls , undetectable , n = 4 and IL-6 , ng/ml , median [range]: LPS 15903 . 5 [14 , 626–17 , 262] vs . Z-AAT 5 . 4 [2 . 9–6 . 1] vs . control [1 . 7 ( 1 . 0–2 . 3 ) , n = 4] ) . Therefore , the effect of Z-AAT preparations on CD14 is valid and unrelated to a potential LPS contamination . Although both Z-AAT polymers and LPS induce CD14 expression and similarly affect CX3CR1 expression and protein levels , data imply that Z-AAT polymers and LPS do not share the same signaling mechanisms . As a side note , it has been reported that CD14++ monocytes have the lowest expression of CX3CR1 ( Appleby et al . , 2013 ) . Low and high surface CX3CR1 levels are suggested to delineate two functional subsets of murine blood monocytes: ‘inflammatory’ and ‘resident monocytes’ ( Geissmann et al . , 2003 ) . This dichotomy appears conserved in humans as CD14+ CD16− , and CD14low CD16+ monocytes resemble ‘inflammatory’ and ‘resident’ monocytes . Previous study demonstrated that peripheral blood monocytes of clinically healthy young adults ( 30 years of age ) with ZZ AATD have significantly higher mRNA and surface expression of CD14 as compared to age matched MM subjects ( Sandström et al . , 2008 ) . Authors thought that the higher CD14 expression reflects early pathological processes whereas according to the current findings this phenomenon seems to relate with the circulating Z-AAT polymers . During steady state , non-classical monocytes expressing CX3CR1 patrol healthy tissues through crawling on the resting endothelium but these monocytes are required for a rapid tissue invasion at the site of infection or inflammation ( Auffray et al . , 2009; Cros et al . , 2010 ) . Previous work evidenced that the non-classical subset of monocytes , characterized by high expression of CX3CR1 , is almost absent in ZZ AATD emphysema patients ( Stolk et al . , 2019 ) . Moreover , plasma levels of AAT polymers were found to correlate with the levels of endothelium-related markers like sE-selectin and sICAM-1 ( Aldonyte et al . , 2004 ) . Beyond , in a small cohort of ZZ AATD emphysema patients , was found a strong inverse association between lung function , based on percentage ( % ) predicted transfer factor for carbon monoxide ( TLCO%pred ) and forced expiratory volume in 1 s ( FEV1%pred ) , and plasma levels of Z-AAT polymers: ( TLCO%pred [r2 = −0 . 75 , n = 9 , p=0 . 02] and FEV1%pred [r2 = −0 . 82 , n = 9 , p=0 . 006] ) . Thus , higher levels of Z-AAT polymers and lower numbers of CX3CR1-positive cells may favor the development of lung injury and disease . A decrease in the expression of CX3CR1 on human monocytes has been shown in patients with atopic dermatitis ( Echigo et al . , 2004 ) and septic shock ( Pachot et al . , 2008 ) . To date , many functional aspects of the CX3CR1–CX3CL1 axis have been suggested , including the adhesion of immune cells to vascular endothelial cells , chemotaxis , the crawling of the monocytes that patrol on vascular endothelial cells , the retention of monocytes of the inflamed endothelium to recruit inflammatory cells , and the survival of the macrophage . Considering the above , these different aspects of interactions between PBMCs and Z-AAT or other polymers occurring due to genetic or post-translational protein modifications require further investigations in dedicated clinical and experimental studies .
The study cohort consists of 41 clinically stable ZZ AATD volunteers collected in collaboration with German Alpha1 Patient Association and Aachen University independently on their clinical diagnosis or treatment with intravenous AAT and 21 non-AATD healthy controls . The institutional review board of Aachen University ( EK 173/15 ) provided ethical approval for individuals recruited in Germany . For Z-AAT polymer determination , we added 12 ZZ AATD emphysema patients recruited at Leiden University Medical Center . In addition , ZZ AATD emphysema patients ( four males and five females ) were enrolled with mean ( SD ) : age 51 ( 6 . 6 ) years , forced expiratory volume in 1 s percent predicted ( FEV1%pred , 66 . 3; Pachot et al . , 2008 ) and transfer factor of the lung for carbon monoxide percent predicted ( TLCO%pred , 64; Sica et al . , 1997 ) . The plasma levels of Z-AAT polymers in these cases were median ( range ) 714 . 2 ( [412 . 9–2270 . 4] µg/ml ) . Leiden University Medical Center provided ethical approval ( project P00 . 083 and P01 . 101 ) for the additional study groups . For all individuals , detailed medical records data were anonymized . All participants issued a written informed consent according to the ethical guidelines of the Helsinki Declaration ( Hong Kong Amendment ) as well as Good Clinical Practice ( European guidelines ) . Total PBMCs were isolated from freshly obtained peripheral blood ( within 6 hr ) using Lymphosep ( C-C-Pro , Oberdorla , Germany ) discontinuous gradient centrifugation according to the manufacturer’s instructions as described previously ( Frenzel et al . , 2014 ) . Thereafter , cells were lysed with RLT buffer for RNA analysis or suspended in RPMI-1640 medium ( Gibco , Thermo Fisher Scientific , Waltham , MA ) and plated into non-adherent 12-well plates ( Greiner Bio-One , Kremsmünster , Austria ) for the further analyses . Isolation of total RNA , synthesis of cDNA and mRNA analysis using Taqman gene expression assays ( Thermo Fisher Scientific , Waltham , MA , Table 1 ) were performed as described previously ( Frenzel et al . , 2014 ) . Real-time PCR was carried out in duplicates . RNA quality was checked on agarose gels . The AAT polymer ELISA using the monoclonal antibody LG96 ( deposited under access number DSM ACC3092 at German Collection of Microorganisms and Cell Cultures ) was developed by Candor Biosciences . Normal M-AAT was used for a negative control . Recovery ratio , signal-to-noise ratio , calibration curve , sample stability under different storage conditions were tested and all tests passed . A cross-reactivity with M-AAT was not reported in any of the tests . Nunc MaxiSorp flat-bottom 96-well plates ( Thermo Fisher , Waltham , MA ) were coated overnight at 2–8°C with monoclonal antibody LG96 , at 1 µg/ml in coating buffer pH 7 . 4 ( Candor Biosciences , Wangen , Germany ) . After a 2 hr blocking step , the plasma samples were applied in the previously determined dilutions made in LowCross-Buffer ( Candor Biosciences ) , which also served as a blank . Incubation was performed for 2 hr at room temperature ( RT ) . For detection , the captured antigen was incubated with antibody ( LG96 ) -horseradish peroxidase ( HRP ) conjugate ( 1:2000 ) for 2 hr . The conjugate was prepared in advance with the HRP Conjugation Kit Lightning-Link ( Abcam , Cambridge , UK ) according to the manufacturer’s instructions . For signal development SeramunBlau fast2 microwell peroxidase substrate ( Seramun , Heidesee , Germany ) was used . The incubation was performed at RT for 12 min in the dark and the reaction was stopped with 2 M H2SO4 . Plates were analyzed at 450 nm by microplate reader ( Dynex , Chantilly , VA ) equipped with Dynex Revelation 4 . 21 software . Measurements were carried out in triplicates . Plasma M- and Z-AAT was isolated by affinity chromatography using the AAT-specific Alpha-1 Antitrypsin Select matrix ( GE Healthcare Life Sciences , Cytiva , Sheffield , UK ) according to the manufacturer’s recommendations . For Z-AAT preparation plasma from volunteers not receiving AAT augmentation therapy was pooled . To change the buffer in the M- and Z-AAT protein pools to Hank’s balanced salt solution ( HBSS , Merck , Darmstadt , Germany ) , Vivaspin centrifugal concentrators with 10 , 000 MWCO ( Vivaproducts , Littleton , MA ) were used . Plasma purified human AAT ( 99% purity , Respreeza , Zemaira , CSL Behring , Marburg , Germany ) was changed to HBSS by the same method . Protein concentrations were determined using Pierce BCA Protein Assay Kit ( Thermo Fisher , Waltham , MA ) . The quality of the M- and Z-AAT preparations was confirmed on Coomassie gels ( 10% SDS-PAGE , Figure 2—figure supplement 1A ) and by analyzing endotoxin levels with Pierce Chromogenic Endotoxin Quant Kit according to the manufacturer’s guidelines ( Thermo Fisher , Waltham , MA ) using TECAN Infinite M200 PRO ( Männedorf , Switzerland ) . In both , M- and Z-AAT preparations , endotoxin levels were below the detection limit ( assay sensitivity: 0 . 01–0 . 1 EU/ml ) . Z-AAT was isolated by affinity chromatography using AAT-specific Alpha-1 Antitrypsin Select matrix as described above . After the isolation Z-AAT , protein was diluted with sterile 0 . 9% NaCl ( Fresenius Kabi , Bad Homburg , Germany ) , and Vivaspin-20 , 100 kDa centrifugal column units ( Sartorius , Göttingen , Germany ) were used to separate Z-AAT monomers from polymers . Protein concentrations were determined using the Pierce BCA Protein Assay Kit ( Thermo Fisher Scientific , Carlsbad , CA ) according to manufacturer’s instructions . The Z-AAT protein monomers were confirmed by using 7 . 5% SDS-PAGE without sample heating and without β-mercaptoethanol ( Figure 2—figure supplement 1B ) . PBMCs ( 5 × 106 cells/ml ) were incubated for 18 hr at 37°C , 5% CO2 either alone , or with Z- or M-AAT proteins , or LPS ( 1 µg/ml , E . coli O55:B5 , Sigma-Aldrich , Merck , St . Louis , MO ) . In some experiments , a recombinant CX3CL1 protein ( R&D Systems , Bio-Techne , Minneapolis , MN ) was used . Protein was reconstituted at a concentration of 25 µg/ml in sterile PBS containing 0 . 1% BSA ( Sigma-Aldrich ) and added to PBMCs at various concentrations up to 500 ng/ml either alone or together with Z-AAT ( 0 . 5 mg/ml ) for 18 hr . Afterward , cells were used for RNA isolation , flow cytometry or Western blot analysis . For Western blot , PBMCs were lysed in RIPA buffer ( Sigma-Aldrich ) , supplemented with protease inhibitor cocktail ( Sigma-Aldrich ) . For some Western blot experiments , we extracted detergent resistant lipid raft associated proteins from insoluble cell fractions using UltraRIPA kit according to the supplier’s instructions ( BioDynamics Laboratory , Tokyo , Japan ) . Equal amounts of lysed proteins were separated by 7 . 5% or 10% SDS-polyacrylamide gels ( under reducing conditions for CX3CR1 and non-reducing for total AAT or AAT polymer analysis ) prior to transfer onto polyvinylidene difluoride membranes ( Merck-Millipore , Burlington , MA ) . Membranes were blocked for 1 hr with 5% low fat milk ( Carl Roth , Karlsruhe , Germany ) followed by overnight incubation at 4°C with specific primary antibodies: polyclonal rabbit anti-human AAT ( 1:800 ) ( DAKO A/S , Glostrup , Denmark ) , mouse monoclonal anti-AAT polymer antibody ( clone 2C1 , 1:500 , Hycult Biotech , Uden , The Netherlands ) , rabbit polyclonal anti-CX3CR1 ( 1:500 , Abcam , Cambridge , UK ) , or HRP-conjugated monoclonal anti-β-actin antibody ( 1:20 , 000 , Sigma-Aldrich , Merck , St . Louis , MO ) for a loading control . The immune complexes were visualized with anti-rabbit or anti-mouse HPR-conjugated secondary antibodies ( DAKO A/S ) and enhanced by Clarity Western ECL Substrate ( BioRad , Hercules , CA ) . Images were acquired by using Chemidoc Touch imaging system ( BioRad ) under optimal exposure conditions and processed using Image Lab v5 . 2 . 1 software ( BioRad ) . For quantification , the signal intensity of the CX3CR1 protein band in each lane was divided by the corresponding β-actin band intensity ( normalization factor or loading control ) . Afterward , the normalized signal of each lane was divided by the normalized target signal observed in the control sample to get the abundance of the CX3CR1 protein as a fold change relative to the control . Plasma samples from 22 ZZ AATD and 21 non-AATD controls were analyzed for CX3CL1/Fractalkine using Duoset kit ( R&D Systems , Minneapolis , MN , assay sensitivity 0 . 072 ng/ml , detection range 0 . 2–10 ng/ml ) . Cell-free culture supernatants were analyzed directly or stored at −80°C . ELISA Duoset kits for TNF-α ( assay detection range 15 . 6–1000 pg/ml ) , IL-1β/IL-1F2 ( assay detection range 3 . 91–250 pg/ml ) , and IL-6 ( assay detection range 9 . 38–600 pg/ml ) were purchased from R&D Systems ( Minneapolis , MN ) and were used according to the manufacturer’s instructions . Plates were measured on Infinite M200 microplate reader ( Tecan , Männedorf , Switzerland ) . Measurements were carried out in duplicates . PBMCs ( 2 × 106 cells per condition ) were incubated with LPS ( 1 µg/ml ) , M-AAT ( 1 mg/ml ) , or Z-AAT ( 0 . 5 mg/ml ) for 18 hr . Staining was performed with phycoerythrin ( PE ) -conjugated mouse monoclonal anti-CX3CR1 antibody ( clone 2A9-1 Invitrogen , Thermo Fisher Scientific , Carlsbad , CA ) , fluorescein ( FITC ) -conjugated mouse monoclonal anti-CD14 antibody ( clone TuK4 , Life Technologies , Thermo Fisher Scientific , Carlsbad , CA ) , allophycocyanin ( APC ) -conjugated mouse monoclonal anti-CD16 antibody ( clone 3G8 , Immunotools , Friesoythe , Germany ) , or BV-480-conjugated anti-CD56 mouse monoclonal antibody ( Clone NCAM16 . 2 , BD Biosciences , San Jose , CA ) alone or in combinations . Dead cells were excluded by a staining with 7-amino-actinomycin D . Samples were measured on a BD FACSAria Fusion machine and analyzed with FlowJo v10 ( Becton , Dickinson and Company , Franklin Lakes , NJ ) . The gating strategy is shown in Figure 3—figure supplement 1 . Human total PBMCs ( 2 × 106 ) were plated onto glass coverslips and incubated alone or with Z-AAT polymers ( 0 . 5 mg/ml ) in RPMI medium for 18 hr at 37°C and 5% CO2 . Cells were then washed with PBS , fixed with 3% paraformaldehyde in PBS for 20 min , and continued with or without permeabilization with 0 . 5% Triton X-100 in PBS for 5 min at RT . For immunolabeling , cells were co-incubated with primary antibodies against human CX3CR1 ( rabbit polyclonal IgG [1:500] , Abcam , Cambridge , UK ) and anti-AAT polymer antibody , LG96 ( 1:5000 , mouse monoclonal ) for 1 hr , at RT . After washing , the cells were incubated with corresponding secondary antibodies ( 1:1000 ) conjugated to AlexaFluor-488 ( goat anti-rabbit ) or AlexaFlour-594 ( goat anti-mouse ) both from Thermo Fisher Scientific , Rockford , IL . After final wash , the cells were mounted on microscope slides using ProLong Gold Antifade Mountant with DAPI ( Thermo Fisher Scientific , Carlsbad , CA ) . Images were acquired using confocal laser microscope FluoView 1000 ( Olympus , Shinjuku , Japan ) equipped with a 60× oil immersion objective and differential interference contrast in sequential mode . Confocal z-stacks were collected with a 0 . 25 µm increment . Data were analyzed and visualized by using Sigma Plot 14 . 0 . One-tailed Student’s t-test was applied to compare two sample means on one variable . When more than two groups were involved in the comparison , one-way ANOVA was used . Data were presented as mean ( SD ) . If normality test failed , the nonparametric Kruskal-Wallis one-way analysis or Mann-Whitney rank sum test was performed , and data were presented as median ( range ) . For correlation analysis , the Pearson’s linear correlation method was used to measure the correlation for a given pair . A p-value of less than 0 . 05 was considered significant . | Proteins can lose their structure and form polymers because of mutations or changes in their immediate environment which can lead to cell damage and disease . Interestingly , polymers formed by a variety of proteins can reduce the levels of CX3C chemokine receptor 1 ( CX3CR1 for short ) that controls the behaviour of immune cells and is implicated in a range of illnesses . Inherited ZZ alpha-1 antitrypsin deficiency is a rare genetic condition that highly increases the risk of liver and lung diseases . This disorder is characterised by mutant alpha-1 antitrypsin proteins ( AAT for short ) reacting together to form polymers; yet it remains unclear how the polymers affect different cells or organs , and lead to diseases . To investigate this question , Tumpara et al . examined whether polymers of mutant AAT influence the level of the CX3CR1 protein in specific classes of immune cells . Experiments revealed that in people with AAT deficiency , certain blood immune cells express lower levels of CX3CR1 . Regardless of age , clinical diagnosis , or treatment regimen , all individuals with ZZ alpha-1 antitrypsin deficiency had AAT polymers circulating in their blood: the higher the levels of polymers measured , the lower the expression of CX3CR1 recorded in the specific immune cells . When Tumpara et al . added polymers of mutant AAT to the immune cells of healthy donors , the expression of CX3CR1 dropped in a manner dependent on the polymer concentration . According to microscopy data , AAT polymers occurred inside cells alongside the CX3CR1 protein , suggesting that the two molecular actors interact . In the future , new drugs that remove these polymers , either from inside cells or as they circulate in the body , could help patients suffering from conditions associated with this abnormal protein aggregation . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"medicine",
"short",
"report",
"cell",
"biology"
] | 2021 | Polymerization of misfolded Z alpha-1 antitrypsin protein lowers CX3CR1 expression in human PBMCs |
Plants rely on transcriptional dynamics to respond to multiple climatic fluctuations and contexts in nature . We analyzed the genome-wide gene expression patterns of rice ( Oryza sativa ) growing in rainfed and irrigated fields during two distinct tropical seasons and determined simple linear models that relate transcriptomic variation to climatic fluctuations . These models combine multiple environmental parameters to account for patterns of expression in the field of co-expressed gene clusters . We examined the similarities of our environmental models between tropical and temperate field conditions , using previously published data . We found that field type and macroclimate had broad impacts on transcriptional responses to environmental fluctuations , especially for genes involved in photosynthesis and development . Nevertheless , variation in solar radiation and temperature at the timescale of hours had reproducible effects across environmental contexts . These results provide a basis for broad-based predictive modeling of plant gene expression in the field .
Plants have evolved responses to complex environmental fluctuations that take place at time scales that vary from seconds to years and shape plant developmental and physiological responses . Variations in environmental signals , including temperature , water levels , solar radiation , biotic interactions and resource availability , are often unpredictable and need to be integrated and transduced to changes in gene expression , which may then be associated with physiological and/or morphological adaptations ( Ahuja et al . , 2010; Weston et al . , 2008 ) . Predicting the adaptive responses occurring in natural environments is a key challenge in plant biology . This is an undertaking that will not be achieved without understanding how genes and functional genetic networks are regulated in response to fluctuating stimuli out in nature ( Richards et al . , 2009 ) . Studies of plant transcriptional responses to environmental perturbations are nearly exclusively undertaken in controlled , static laboratory conditions that are divergent from what is seen in the natural world . While these laboratory experiments have enriched our knowledge of the molecular pathways involved in abiotic stimulus responses , it is clear that organismal phenotypes and the genetic architecture of various traits differ between controlled laboratory and field conditions ( Malmberg et al . , 2005; Mishra et al . , 2012; Weinig et al . , 2002 ) . This so-called “laboratory/field ( lab-field ) gap” is often referred to when trying to explain why the improvement of crops for resistance to abiotic stresses , for example , has not met the expectations arising from advances in genomics technologies ( Cabello et al . , 2014 ) . Given that environmental stresses are one of the main constraints on crop performance , understanding the mechanisms that plants rely on to cope with challenging conditions in the field will be an important asset for crop improvement . In addition to the dynamic nature of field conditions , the co-occurrence of multiple dynamic signals is another major cause of discrepancies in plant responses/phenotypes between laboratory and field conditions , as laboratory studies generally investigate the short-term effects of single environmental perturbations . The study of plant response to multiple concurrent stimuli , however , can provide insights into both ecological adaptation and , in the context of crop species , crop performance . Concurrent stresses have a major impact: the combination of heat and drought stresses , for example , has been found to be more detrimental to crop yield than the addition of either stress alone ( Mittler , 2006 ) . The interplay of multiple dynamic factors also affect plants at the physiological level , particularly in the case of photosynthesis: mesophyll conductance depends on both CO2 concentration and irradiance ( Kaiser , et al . , 2015 ) , while drought can undermine the otherwise positive effect of high temperature on photosynthetic efficiency ( Pfannschmidt and Yang , 2012 ) . Some of the effects of combinations of multiple environmental perturbations on gene expression cannot be predicted from the individual treatments , as has been shown in tobacco ( Rizhsky et al . , 2002a ) , Arabidopsis thaliana ( Rasmussen et al . , 2013 ) and sorghum ( Johnson et al . , 2014 ) . Because these studies were conducted under controlled conditions , little is known about how the levels and patterns of gene expression are modulated by multiple dynamic abiotic factors in the field ( Izawa , 2015 ) . Another feature of plant responses to environmental signals in nature is that they depend on a variety of ecological contexts ( e . g . seasonality , macroclimate ) . Indeed , a reason invoked by plant molecular biologists to avoid experiments in natural environments despite the lab–field gap is the perceived low reproducibility of results that would be generated under unpredictable fluctuating conditions ( Izawa , 2015 ) and the difficulty to detect transcriptional signals of interest ( Travers et al . , 2007 ) . These concerns arise mainly from the known sensitivity of gene expression to the overall environmental background , which in nature cannot be controlled . Thus , the extent to which the climatic or ecological context affects molecular environmental responses of interest in plants remains to be determined . The transcriptional effect of the environmental context is all the more critical considering that , for crop species , different types of agricultural settings or climates will impact dynamic gene expression responses that can translate into key phenotypes . Understanding both the effect of the environmental context on transcriptomic responses and the integration of multiple stimuli is essential to the development of predictive models for gene expression that can be generalized to wide ranges of agronomical settings and anticipated climatic conditions . A few studies have investigated the dynamic relationships between gene expression and fluctuating environmental conditions in the field in A . thaliana ( Richards et al . , 2012 ) , Oryza sativa ( Nagano et al . , 2012 ) and Andropogon gerardii ( Travers et al . , 2010 ) but they provide limited insight on the integration of multiple abiotic stimuli and the effect of environmental contexts on transcriptional responses . The study we report here is the first to focus on identifying the concurrent effects of multiple environmental factors on gene expression under natural climatic fluctuations in a crop species . Moreover , we examine the macro-environmental context of genome-wide gene expression by measuring transcriptome variation in contrasting seasons and field types . We analyzed the global gene expression patterns in O . sativa over a period of 1 month in two fields typical of the main modes of rice cultivation . We conducted these experimentsover two seasons , dry and wet , in three rice landraces . We used model selection methods to relate the main variations in global gene expression to variation in several environmental/developmental parameters with simple linear equations . The genetic background of the plants had limited effect on the environmental response of the global transcriptome . We show that additive effects of several environmental factors drive the field expression of multiple co-expressed gene clusters . We also found that the field context reshaped a large part of transcriptional patterns , while the effect of season was more limited . This sensitivity to the environmental context was particularly important for groups of co-expressed genes involved in photosynthesis and development . Finally , we used previously-published field expression data from Nagano et al . ( 2012 ) to show that only a small part of the relationships between weather fluctuations and gene expression identified under a tropical climate could be detected in temperate conditions .
Our experiment was designed to specifically assess the effect on rice global gene expression of climatic fluctuations , different types of field environments and the genetic background . We conducted two phases of field cultivation - one during the dry season ( January–February ) and one during the wet season ( July–August ) at the experimental rice station of the International Rice Research Institute ( IRRI ) , Los Baños , Laguna , in the island of Luzon in the Philippines in 2013 ( Figure 1A ) . The plants were cultivated in two adjacent fields typical of two different systems of rice cultivation . One field was cultivated following irrigated lowland practices: it was flooded with shallow water under constant irrigation and seedlings were transplanted to the field after being raised for 3 weeks in seedbeds ( referred to as irrigated field ) . The second field was managed according to upland cultivation practices: it was rainfed , not irrigated and had been directly seeded ( referred to as rainfed field ) . Each field was divided into two subfields constituting the biological replicates . We grew three different landraces of rice: ( i ) Azucena , an upland adapted landrace; ( ii ) Pandan Wangi , traditionally used for lowland cultivation only in the dry season; and ( iii ) Palawan , another upland adapted landrace that is used only during wet season cultivation . 10 . 7554/eLife . 08411 . 003Figure 1 . Modeling environmental and developmental effects on rice gene expression in the field . ( A ) Experimental design: fifteen sampling timepoints for each of the 16 season/field/replicate/genotype series amounts to 240 samples , representative of 30 different sets of climatic conditions , represented here by the 30 red lines on the graphs for temperature and precipitation ( accumulated mm per 5 min ) along the sampling period of each growing season . ( B ) Processing raw RNA sequencing data into main transcriptomic variation for different subsets of our data . ( C ) Modeling potential differences between fields and seasons for climatic/developmental response within a cluster mean . An example for cluster 15 of the two-season analysis , where the model selected is common to both fields in the wet season but field specific in the dry season . Az , Azucena landrace; Pa , Palawan landrace; PW , Pandan Wangi landrace . DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 003 To avoid the major shift in gene expression patterns induced by the transition to the flowering stage ( Sato et al . , 2011 ) , which would confound our detection of environmental effects , we sampled rice leaf tissue during 1 month of vegetative growth ( 15 sampling timepoints , 2 days apart ) . Sampling was carried out 4 hr after sunrise to minimize circadian-driven transcriptional variation . Each sample included six young leaves , each from a different plant , to minimize variation in individual plant microenvironment . We measured global gene expression using RNA sequencing ( Figure 1B ) . We excluded from our analysis genes for which we detected sequencing reads for less than 20 samples out of the 60 samples in each genotype per season subset of the data . The expression data for the 22 , 144 remaining genes was log-transformed and the biological replicates were averaged . Our goal is to relate gene expression variation over time to variation in climatic conditions and plant developmental stage , and assess how these relationships are affected by season , field type and genetic background . We focused on trends in gene expression variation common to a high number of genes: after removing 1251 genes with a low coefficient of variation and 2962 genes with a low mean expression , we grouped the remaining genes into co-expressed gene clusters ( Figure 1B ) . The number of clusters chosen was the highest that satisfied the constraint that no more than 5% of all the genes in the analysis belonged to “non-representative” small clusters , defined as containing less than 1% of all the genes in the analysis . We used the mean expression profile of all genes in each cluster as a representation of the variation in expression within that cluster . We used a model selection approach to explain gene expression patterns by environmental and developmental variation . This approach relies on selecting a linear combination of environmental/developmental ( ED ) input parameters that both minimizes model mean squared error ( MSE ) , quantifying the difference between the model and the expression data , and limits model complexity ( i . e . , avoiding over-fitting ) . A preliminary analysis showed that allowing for more than three parameters per equation over-fit the model more often than improving it , so we limited the number of parameters per linear equation to three . A typical ED equation had the following form: cluster mean = αED1 + βED2 + γED3 , where ED1 , ED2 and ED3 are ED parameters , and α , β and γ are linear regression coefficients . The ED parameters used in these models were measurements of current conditions at the time of sampling , recent changes in temperature , humidity , wind speed and solar radiation , temperature fluctuations , and short-term and long-term averages for all climatic conditions ( Table 1 ) . We included parameters for non-linear effect of short-term solar radiation on gene expression , because this type of effect has been observed on photosynthesis rate ( Li et al . , 2009 ) . There were two parameters for field soil moisture , at 30 and 15 cm below ground ( measured with tensiometers in the rainfed field and estimated to be constant at soil saturation value in the irrigated field ) , and a binary parameter for the field ( irrigated or rainfed ) . A parameter was designed to represent developmental stage , using fixed values for the transplanting stage , end of tillering production and the heading time . Our set of ED parameters included those that correlated with each other , so we averaged nearly similar parameters ( r > 0 . 98 ) and added to our model selection approach the constraint that two parameters with a Pearson correlation coefficient over 0 . 85 could not be selected in the same equation . 10 . 7554/eLife . 08411 . 004Table 1 . Climatic parameters for the environmental/developmental models and their abbreviations . Parameters calculated from the weather data are mostly ordinary averages ( linear: L ) for a large range of time windows before sampling but also include exponentially transformed averages ( non-linear: NL ) modeling stronger effects at either low ( NL- ) or high ( NL+ ) values . For the most dynamic climatic parameters , we calculated differences ( D ) between the measurement at the sampling timepoint and the measurement short amounts of time before sampling . To estimate fluctuations , we calculated averages of the residual term ( R ) from the seasonal decomposition of daily variation into cyclic and trend components . The abbreviation for each climatic factor and parameter type and time-window is given in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 004Average/change for the lastTemperature ( tp ) Relative humidity ( hu ) Solar radiation ( so ) Wind speed ( wd ) Atmospheric pressure ( ps ) Rainfall ( ra ) Sampling time15 min ( 15 min ) LLL NL− NL+LLLShort-term averages1 hr ( 1 hr ) LLL NL− NL+LLL4 hr ( 4 hr ) LLL NL− NL+LLL24 hr ( 24 hr ) LLLLLLLong-term averages3 d ( 3 d ) LLLLLL6 d ( 6 d ) LLLLLL10 d ( 10 d ) LLLLLL15 d ( 15 d ) LLLLLLRecent change20 min ( δ20 min ) DDDD1 hr ( δ1 hr ) DDDD2 hr ( δ2 hr ) DDDDFluctuations1 hr ( ε1 hr ) R4 hr ( ε4 hr ) R24 hr ( ε24 hr ) R We designed our approach to take potential differences in climatic response between fields , genotypes and seasons into account . To assess whether disparities in transcriptional patterns could be explained using distinct ED equations , we considered different ED models for each cluster . The simplest model is a single equation for the whole cluster mean . A more complex model would combine two different equations , for example , in the case of a field-specific model: cluster mean = αiED1 + βiED2 + γiED3 in the irrigated field and cluster mean = αrED4 + βrED5 + γrED6 in the rainfed field . If we were to take into account all possible differences between fields , genotypes and seasons , we would get models with as many as eight equations , which would be difficult to interpret . We therefore limited the maximum number of equations to four by applying the method to one season at a time , testing different equations between fields and genotypes; or considering only one genotype and examining the possibility of different equations between fields and seasons ( this latter case is represented in Figure 1C ) . We chose between the models comprising one to four equations using the Bayesian Information Criterion ( BIC ) , a statistical tool to limit over-fitting that includes a complexity penalty , calculated on the global model for all subsets . We first used our model selection approach on the dry season data alone , as this was the season with the greatest phenotypic differences between the rainfed and irrigated fields . We wanted to determine the extent of genotype and field differences in gene expression variation during that season and how well the ED models can explain these differences . For a given cluster , we quantified the differences in variation of gene expression between the two genotypes by calculating the correlation between the two genotype-specific subsets of the cluster mean . The same method was used to evaluate differences in expression patterns between fields . Only 19 out of the 56 clusters ( 4663 genes ) had a correlation coefficient between genotypes below 0 . 8 ( Figure 2A ) . Clusters with a low genotype correlation had a high model MSE , which showed that the ED models did not adequately explain these genotype differences . We found much more extensive differences in gene expression patterns between field environments ( Figure 2B ) , with 48 clusters ( 15 , 103 genes ) that had a correlation coefficient below 0 . 8 . While all clusters with high correlation between field environments ( r > 0 . 8 ) had ED models that fitted the cluster mean well ( model MSE < 0 . 12 ) , low MSE models were also selected for several clusters with strong dissimilarities between field environments . In some cases , ED models could thus explain strong field differences . We did not investigate the genotype effect further and instead used the genotypes as biological replicates that were averaged for the analysis of both seasons , incorporating all the expression data concatenated into 60 data points ( Figure 1A ) . 10 . 7554/eLife . 08411 . 005Figure 2 . In the dry season analysis , expression patterns of the two genotypes are highly correlated for most clusters while for the few other clusters , differences between the genotypes are poorly explained by the models ( high model MSEs for clusters with low correlation ) . ( A ) Model MSE vs . correlation ( Pearson coefficient ) between the genotypes for the 56 dry season cluster means . ( B ) Model MSE vs . correlation between the fields for the 56 dry season cluster means . MSE , mean squared error DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 005 In this two-season analysis , in addition to identifying which ED parameters the expression of each gene cluster can be related to , we are assessing whether the field environment and the season affect the identified transcriptional response . As an example , the simple season-specific ED model selected for cluster 9 is presented in Figure 3A ( Supplementary file 1A–C ) . The model consists of a single linear equation for both fields and both seasons . It combines a negative term for soil moisture at 15 cm depth , a positive term for 1-hr average of solar radiation ( exponentially transformed ) and a positive term for the change in temperature during the last 2 hr . In this model , the soil moisture input parameter ( Figure 3B , Supplementary file 1B ) allows for modeling gene expression differences between field environments while keeping a common model for both field environments , in particular the higher expression in the rainfed field environment in the dry season . The selection of a common model for the dry and wet seasons shows that the gene expression response to climatic factors of this cluster was largely independent of the season . 10 . 7554/eLife . 08411 . 006Figure 3 . Cluster 9 of the two-season analysis: its environmental/developmental model and some of the genes it contains that have a potential function in environmental response . ( A ) Gene expression for the cluster mean ( grey ) and spread ( calculated as 10% and 90% quantile of all genes in the cluster for each data point; grey area ) , and cluster model ( red ) . ( B ) Scaled climatic parameters in the model equation for each season . sm 15: soil moisture at 15 cm; so 1h NL-: one hour average of solar radiation transformed to increase the effect of low values; tp δ2h: change in temperature from 2 hr ago . ( C ) Gene expression pattern ( scaled ) of six genes in cluster 9 with a potential function in environmental response . DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 006 We used genotype correlation within a cluster as a measure of replicability of ED effects: the higher the genotype correlation , the more gene expression response was driven by factors common to both genotypes ( i . e . , climatic and developmental factors ) . While the median genotype correlation of all genes in the analysis was 0 . 55 , the median of all gene cluster means was 0 . 90 , showing that averaging expression profiles over many genes remarkably reduces sources of non-replicability . We focused on 27 gene expression clusters with a genotype correlation greater than 0 . 9 ( Figure 4 , Supplementary file 1C ) . They encompass 11 , 371 ( 63% ) of the 17 , 931 genes in the analysis , as they include most of the largest gene clusters . These clusters are also the ones with the best ED models ( i . e . , with a small error and low complexity , such as the one depicted in Figure 3 , Supplementary file 1C ) . For these clusters , we observed extensive field effect on the climatic/developmental response , with only 6 out the 27 clusters ( 2308 genes ) showing high correlation ( r > 0 . 8 ) between fields in both seasons . As expected from the much lower precipitation levels during the dry season , which accentuated the difference in water availability between fields ( Figure 1A ) , the dry season climatic conditions generated wider between-field differences in gene expression compared with the wet season ( Figure 4 , Supplementary file 1C ) . The season effect was slightly less prevalent , as 11 out the 27 clusters ( 4967 genes ) were modeled with the same equation for both seasons ( Supplementary file 1C ) , indicative of season-independent responses . 10 . 7554/eLife . 08411 . 007Figure 4 . Classification of the 53 clusters from the two-season analysis based on field correlation in the dry and wet seasons . Each dot represents a cluster with the size of the dot proportional to the number of genes in the cluster . The 27-gene clusters with igh genotype correlation ( r > 0 . 9 , Pearson coefficient ) are represented with a filled circle and divided into four groups depending on whether their correlation between fields in each season is below or above 0 . 8 ( dashed lines ) , referred to as high or low correlation . Group 1 clusters have a high correlation between fields in both seasons; group 2 clusters have a high correlation in the wet season but not in the dry season; group 3 clusters have a high field correlation in the dry season but not in the wet season; group 4 clusters have a low field correlation in both seasons . DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 007 We analyzed gene clusters according to their field response to understand how distinct modes of cultivation affect gene expression under the same climate . The expression of a gene cluster can be affected by the field environment in two ways: ( i ) distinct responses to climatic/developmental factors and/or ( ii ) a shift in expression level , representing different ways in which the effect of the field environment can be integrated with the climatic response and developmental program . Enrichment in specific functions or pathways within the different types of gene clusters can be indicative of the role of certain processes in the adaptation to distinct field environments . We divided the 27-gene clusters into four groups based on whether they showed different expression responses to climatic/developmental factors in each of the two fields ( correlation between the expression patterns of the two fields below 0 . 8 ) in one or two seasons ( Figure 4 ) . We also calculated the difference in mean expression between the fields for each season . To investigate the molecular processes that were most affected by climatic variation and field environments , we conducted a gene ontology ( GO ) term enrichment analysis for each cluster ( Supplementary file 1E ) . Correlations between ED parameters make it difficult to ascertain the causal factor of gene expression change from the parameter selected in a model . This is why , for the interpretation of ED models , we grouped those of the parameters that showed high correlations to each other ( Figure 5A ) . The parameters ( sometimes included in a group of highly correlated parameters ) selected for the 27-gene clusters most representative of ED responses are shown in Figure 5B . 10 . 7554/eLife . 08411 . 008Figure 5 . Summary of the ED models selected for the 27-gene clusters with high genotype correlation in the two-season analysis . ( A ) Justification for the grouping of model parameters: heat-map of the correlation between the ED parameters selected at least once in the models . For some parameters that have strong negative correlations with other parameters , we used negative values to better see groups of correlated parameters independently of the sign of the correlation . ( B ) The selection of a parameter into a model is represented by a colored box , orange for a positive term and purple for a negative term . Field correlation is the correlation between irrigated and rainfed field profiles for each season . Field difference is the average expression level in the rainfed field minus the average expression level in the irrigated field . ED , environmental/developmental DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 008 The functional characterization of genes involved in the response to abiotic stresses is mostly conducted with single environmental perturbations under controlled conditions . Few genes have been functionally characterized for a role in the response to combined environmental stresses . We have found multiple clusters with patterns of expression showing additive effects of several environmental signals . Because changes in the expression of a gene in reaction to perturbations in an environmental factor are usually interpreted as this gene having a function in the physiological response to that factor , our results suggest the genes in these clusters are involved in the response to simultaneous environmental changes . We assessed this further by looking at whether genes shown to have a role in the response to a specific abiotic stimulus also responded transcriptionally to other environmental signals . We compiled lists of genes that had already been demonstrated to have a role in the response to light , water availability or temperature in rice as well as rice homologs of such genes in other species ( mainly the model system A . thaliana ) . We looked at whether the expression patterns of these genes were consistent with their putative function . We focused our attention on the candidate genes that showed a replicable environmental response in our data ( genotype correlation above 0 . 8 ) and belonged to one of the 22 gene clusters whose main ED model terms are environmental parameters ( Supplementary file 1C ) , which we refer to as genes with a strong environmental regulation . Here we examine the correspondence of gene expression and climatic variation between our Philippine experiment and a previously published rice field transcriptomic study conducted in different ( temperate ) climatic conditions ( Nagano et al . , 2012 ) . From the latter experiment conducted in Japan , we selected 52 daytime timepoints that spanned nearly 7 weeks of vegetative growth in irrigated fields for the Japonica cultivar Nipponbare . We centered ( subtracted the mean ) the data by time of day to eliminate circadian clock effects . We used the same climatic measurements as those found in our experiment , which were recorded every minute . We refer to this data as the partial Nagano dataset ( PND ) . To determine to what extent our results would hold true under a different climate , we performed a new analysis of our data that excluded measurements from the rainfed field ( using the irrigated field half of the two seasons analysis dataset ) , as the experiment in Japan was exclusively conducted in an irrigated field ( Figure 1B ) . We used these results to test whether ED parameters selected in the models for our data could also explain expression variance of the same clusters of genes in the PND . The models selected for the 60 gene cluster means calculated from our data ( Supplementary file 1G ) could be represented by a single equation for the irrigated field in both seasons ( season-independent ) or of distinct equations for the irrigated field in each season ( season-specific ) . We reasoned that climatic response patterns that were not consistent between the two seasons of our experiment ( season-specific models ) were unlikely to be reproduced in the PND . Consequently , we only tested the transferability of models for those of our clusters with a season-independent model , when the model explained more than half of the variance of the cluster mean . For each of these 36 clusters , we calculated the mean expression profile in the PND of the genes in the cluster . We determined whether one or several parameters in the model of the cluster mean over our irrigated field data could model the cluster mean over the PND with the same coefficient signs , by comparing all possible models with combinations of these parameters with the BIC . Non-null models were selected for 24 of the 36 PND cluster means ( Figure 7 ) . They explained between 13 and 86% of the expression variance over the 52 data points of the PND . The parameters that could explain expression variation in our experiment and the PND were predominantly developmental stage and parameters for short-term solar radiation and temperature , especially in the models that explained a large part of the expression variance . Only one long-term climatic parameter was conserved in several PND cluster mean models: the 15-day average of wind speed , which was selected in four models , including one that explained 52% of the expression variance . These results show that a large part of climatic effects on gene expression vary with the type of climate; in particular , long-term effects . 10 . 7554/eLife . 08411 . 010Figure 7 . Transferability of two seasons irrigated field models to an independent dataset of rice gene expression under temperate climate . Models were selected for 60 clusters using our expression data in the irrigated field during the wet and dry seasons . The parameters from these models that could be transferred to explain the mean expression of the cluster mean in the PND are shown with a green box , the ones that were not transferred are indicated with a black box . The variance of the PND cluster mean explained by the partially transferred model is represented with decreasing shades of blue . PND , partial Nagano dataset . DOI: http://dx . doi . org/10 . 7554/eLife . 08411 . 010 As we found that several clusters enriched for genes involved in photosynthesis and development were among the most sensitive to the field context , we investigated whether this was also true for the seasonal and climatic context . We performed a GO term enrichment analysis on the clusters from the irrigated field analysis ( Supplementary file 1I ) . We identified two clusters highly enriched for genes related to photosynthesis ( p<10–27 and 10–26 , respectively ) , which were both modeled with season-dependent models . This result indicates that the transcriptional response of groups of co-expressed genes involved in photosynthesis is affected by the seasonal context . Two clusters ( 13 and 25 ) were enriched for genes associated with the developmental process and cell cycle . Cluster 13 was modeled with a season-dependent model while cluster 25 could be modeled with the same equation for both seasons of our experiment but no parameter of this equation could be transferred to model expression in the PND . This shows that the transcriptional regulation of some genes involved in development is not only sensitive to the field context but also to the season and climate type context . To compare our method with the method from Nagano et al . ( 2012 ) with respect to the detection of climatic effects , we conducted another analysis of the PND using our model selection approach , determining clusters and models independently of the results obtained on our data ( Supplementary file 1H ) . About a half of the models determined with the method by Nagano et al . ( 2012 ) do not have an environmental term and those that do , include the effect of a single climatic factor per gene , with a term that can be non-linear with diurnal changes in sensitivity . This makes their models intrinsically incompatible for a systematic direct comparison with ours . We thus chose to focus on two sets of genes identified by each method as having a clear environmental response and no developmental effect , and then evaluate the results of the alternative method on these genes . First , we selected from the results of our analysis genes highly correlated with their respective cluster means ( r > 0 . 8 ) and belonging to three clusters with a low error model ( MSE < 0 . 25 ) that included neither a developmental stage term nor a climatic parameter strongly correlated with the developmental stage . Among these 390 genes , 156 ( 40% ) had a Nagano et al . ( 2012 ) model that included an environmental term for the timepoints we considered ( between 8:00 and 16:00 ) . Second , we chose genes with Nagano et al . ( 2012 ) models , including no developmental or circadian terms and with a linear environmental term for either temperature or solar radiation ( which have the most reproducible effects ) that explained more than 50% of the expression variance . When looking at the results of our analysis for these 214 genes , we found that 155 ( 72% ) of them were in a cluster with at least one environmental term and a model error lower than 0 . 35 ( R2 > 0 . 7 ) . Our method was thus more efficient in matching the Nagano et al . models with a clear environmental term than the Nagano et al . ( 2012 ) method was in the reciprocal comparison . This indicates that our approach has a higher sensitivity to the effect of climatic factors on gene expression , in addition to its ability to detect the co-occurring effects of multiple environmental parameters .
We used a model selection approach to identify relationships between major variations in global gene expression and environmental conditions/developmental stage . Focusing our analysis on groups of genes showing consistent variation across two different genotypes in each of the two seasons , we determined that most of these representative expression patterns could be explained through the combined effects of several environmental parameters , related to distinct climatic/soil-related factors and/or on different time-scales . Co-occuring abiotic stresses trigger complex responses that cannot be predicted from the effect of single stresses ( Prasch and Sonnewald , 2015 ) , especially at the level of transcriptional regulation . Our dry season experiment , which included a drought period , presented ideal conditions to understand how limited water availability is integrated with other climatic signals at the transcriptional level in the field in comparison to controlled conditions . One common finding of studies on the combined effects of drought and heat is that a high number of genes are differentially expressed only by the combination of both stresses ( Johnson et al . , 2014; Rizhsky et al . , 2004 , 2002b ) . However , these results might be due to the fact that the differential expression induced by individual stresses is under the threshold of significance , while the addition of both small effects passes the differential expression threshold . Part of the assumed specificity could therefore be an artifact of the analysis . Although we cannot make a direct comparison with our analysis , as we do not have conditions testing the effect of drought alone , we can contrast gene expression responses between irrigated and rainfed fields during the dry season to assess how drought impacts the response to other abiotic stimuli and test whether there is indeed a specific response to drought combined with short-term climatic signals . Excluding clusters from groups 3 and 4 , the expression of which is impacted by non-drought related field effects in the wet season , the different effects induced by drought in the rainfed field for clusters modeled with terms for short-term averages of temperature/solar radiation in the irrigated field during the dry season were ( i ) show the same response to short-term temperature/solar radiation with no or little shift in expression ( clusters 18 , 21 and 24 ) ; ( ii ) show an additive effect of drought over the short-term climatic response ( clusters 4 , 9 , 12 and 30 ) ; or ( iii ) not display the short-term climatic response under drought ( clusters 1 , 8 , 10 and 11 ) . Only cluster 44 , containing 204 genes , responded to short-term temperature/solar radiation variations in the dry season , specifically in the rainfed field . Our results therefore show that the main behaviors under combined drought and short-term abiotic signals are an addition of both responses and response to either one of the stimuli ( often integrated with a long-term climatic response ) , while responses specific to the combination of both stresses are rare . Genes that have been previously characterized for their role in a specific type of abiotic stimulus – light , temperature or water availability – and with a replicable expression pattern in our experiment generally displayed a transcriptional response to that stimulus . In many cases , these genes also seemed to respond to other abiotic stimuli , demonstrating the extensive connections between environmental pathways occurring in complex natural conditions . Some genes did not , however , have a pattern of expression consistent with what was measured in previous laboratory studies . For example , OsCDPK7 and OsMIOX have been shown to be induced by drought ( Duan et al . , 2012; Wan et al . , 2007 ) . Nevertheless , in our data , there were no differences in mean expression between the irrigated and rainfed fields in the dry season for OsCDPK7 , while OsMIOX was less expressed in the rainfed field than in the irrigated field during both seasons . These discrepancies may be explained by the fact that abiotic stresses are generally analyzed individually . Some genes induced by drought , when water availability is the only varying environmental parameter , might be mostly responsive to temperature during a combination of drought and heat stress . Another possible explanation is the difference in time-scale of our field study compared with most laboratory-based experiments . The effect of an abiotic stress is often analyzed within hours of the stress treatment; in our field experiment , we examined climatic variables that changed constantly and gradually over a month . The time-scale of the study is especially critical in the case of drought stress , because in natural conditions , it usually occurs over the course of days or weeks . More generally , covering longer time frames than laboratory studies allowed us to observe environmental effects interact with plant developmental processes in an integrated way ( Allahverdiyeva et al . , 2015 ) , as well as examine the effects of long-term seasonal climatic effects . Our results thus show that measuring gene expression under natural conditions and for long periods of time can help better assess the effect of multiple abiotic stimuli on gene expression . In this study , we assessed the impact of different seasons , field environments and types of climate on weather-driven patterns of gene expression . Most of the seasonal differences were limited to the rainfed field and were related to the prolonged period , with very little precipitation occurring during the dry season . When analyzing only the irrigated field data , we found that 44 out 60 clusters ( representing 72% of the genes in all clusters ) were modeled with a common equation for both seasons ( Supplementary file 1G ) . A fair level of reproducibility was also observed between different years of irrigated field culture for rice ( Nagano et al . , 2012 ) . In contrast , the type of field in which rice was cultivated had a major effect on the response to climatic conditions of many genes . A large part of the differences in climatic response between field environments that we detected were especially pronounced in the dry season and linked to the water status of the rainfed field . Some expression differences , however , seemed independent of this drought effect , as they were detectable in the wet season . Finally , we used independent expression data to assess the effect of a different type of climate on transcriptional responses in irrigated conditions . We found that only a small subset of the gene expression patterns observed under the tropical climate of our experiment could be generalized to the temperate conditions of the Nagano et al . ( 2012 ) study . This result shows that we can expect to reproduce findings about gene expression despite seasonal variations , but we should be careful about generalizing results to other climates or soil conditions than the ones their study was set in . When investigating the reproducibility of transcriptional responses , we found that gene expression patterns driven by short-term averages of solar radiation and temperature , and to a lesser extent long-term wind speed , were the most consistent responses across seasons , fields and climates . Short-term averages of solar radiation and temperature belong to the group of parameters that was used to model most of the 27 gene co-expression clusters , with the highest correlation between genotypes in our two-season analysis . In particular , the exponentially transformed short-term averages ( 15 min , 1 or 4 hr ) of solar radiation ( NL− ) that model stronger effects of variations in the lowest range of irradiance were selected to model at least one entire season or field subset for 12 of the 27 gene clusters of our analysis and could systematically be transferred from the models of our irrigated field data to explain the expression variation of the same clusters of genes observed under the Japanese temperate climate ( Supplementary file 1G ) . In the conditions of the Nagano et al . ( 2012 ) experiment , short-term solar radiation variation was weakly correlated with variation in temperature and precipitation , showing that this effect is likely to be specific to irradiance level . The reason why variation in the lowest range of solar radiation values seem to have a stronger effect on gene expression ( compared with higher irradiance levels ) for some clusters needs to be investigated . One hypothesis is that responses to low light can rely mostly on changes in gene expression because they do not require as rapid a response as high light conditions ( Pfannschmidt and Yang , 2012 ) . In contrast , we found that the effect of longer-term variation ( 10- , 15-day averages ) in temperature and solar radiation were much less reproducible across seasons , fields and climates than short-term effects . One explanation is that , as these responses may be integrated over long periods of time , there is greater opportunity for them to be modulated by broad climatic and developmental factors . The only long-term environmental factor that could be transferred to model gene expression under the Japanese climate for several clusters was the 15-day average of wind speed . In our two-season analysis , it was selected to model across both seasons and fields cluster 32 , which is enriched for thigmotropism genes . While there have already been studies aiming at identifying genes responsive to mechanical cues ( Lee et al . , 2005 ) , they were limited to hour-scale laboratory experiments . Our results show that , in nature , the effect of wind on gene expression might be more prevalent on a week-scale , which is consistent with the long-term accommodation to repetitive wind loads observed in poplar ( Martin et al . , 2010 ) . Further study of the genes in cluster 32 should contribute to the understanding of gene regulation in response to wind necessary to unravel the cellular processes involved in the time integration of mechanosensing ( Moulia et al . , 2011 ) . The analysis for GO annotation enrichment of the 27 clusters with the most reproducible expression patterns revealed functionally related groups of genes that are co-expressed in response to environmental and developmental signals . These results point to the biological processes most critical for plant physiological response to dynamic environments . The GO terms for biological processes with the most significant enrichment were nucleic acid metabolic process ( and its parent term nucleobase-containing compound metabolic process ) and photosynthesis . Nucleic acid metabolic process related genes were highly enriched in clusters 4 and 6 ( p<10–18 and 10–22 , respectively ) , which were also both enriched for developmental process and cell cycle associated genes . The strong enrichment in nucleic acid metabolic process related genes highlights the importance of a tight transcriptional regulation of this molecular process , probably linked to the control of cell division , in the adaptation of the developmental program to environmental fluctuations ( Rymen and Sugimoto , 2012 ) . Both clusters had a negative correlation between field expression profiles in the dry season , while the field correlation was above 0 . 75 in the wet season , during which both cluster means were modeled with a positive term for the developmental stage parameter . This indicates an extensive impact of drought on the expression pattern of the genes in these clusters . As cluster 6 was modeled with several short-term environmental parameters , our results show that the transcriptional regulation of cell division-related aspects of growth integrates with the developmental program several time-scales of environmental stimuli . Clusters 1 , 5 and 13 were greatly enriched for photosynthesis related genes ( p<10–17 , 10–9 and 10–19 , respectively ) . These gene clusters were all down-regulated in response to drought , in accordance with previous results in A . thaliana ( Chaves et al . , 2009 ) . Clusters 1 , 5 and 13 were also modeled with short-term climatic parameters . These effects were detected mostly for the irrigated conditions , with positive terms for solar radiation in the case of clusters 1 and 5 . In contrast , cluster 13 was modeled with a negative term for short-term temperature in both fields . These differences indicate that different components of the photosynthetic machinery might need to respond to different environmental stimuli . Although it has been thoroughly demonstrated that a large part of environmental effects on photosynthesis are regulated at the physiological and biochemical levels ( Kaiser et al . , 2015 ) , the fine regulation of transcript abundance for these photosynthesis genes shown here confirms that the control of gene expression is also an important component of the modulation of photosynthetic activity ( Pfannschmidt and Yang , 2012 ) . Focusing on our irrigated field data and the data from an independent field experiment ( Nagano et al . , 2012 ) , we found that the expression patterns of groups of co-expressed genes involved in photosynthesis were affected by the seasonal context ( independently of drought ) and the environmental response of groups of genes associated with development was dependent on the climatic context . This suggests that the transcriptional regulation of both photosynthesis and development relies on complex mechanisms , resulting in the integration of numerous layers of environmental cues , short-term climatic fluctuations as well as steadier aspects of the plant surroundings . Understanding the dynamics of gene expression is a major challenge in biology . This is particularly difficult in the context of complex environments found in nature , as it requires unraveling the concurrent effects of multiple , fluctuating environmental signals on transcriptional patterns . Nevertheless , the ability to establish the environmental response of whole transcriptomes can have wide applications , including controlling engineered gene circuits ( Uhlendorf et al . , 2012 ) and predicting gene patterns in untested conditions ( Bonneau et al . , 2007; Danziger et al . , 2014; Nagano et al . , 2012 ) . Our results suggest that while field environments can result in complex responses , one can nevertheless identify co-expressed gene clusters the mean expression of which can be accurately modeled with climatic , field , seasonal and developmental factors . Further work can integrate such models with other approaches , including gene network inference ( Bonneau et al . , 2007 ) , genotype-by-environment interactions ( Marais et al . , 2013 ) and phenotypic modeling ( Aikawa et al . , 2010; Satake et al . , 2013 ) , to provide a more comprehensive picture of plant responses in natural environments . This , in turn , can be used in the design of improved crops ( Hammer et al . , 2006; Mochida et al . , 2015 ) and the prediction of the ecological effects of climate change ( Stafford et al . , 2013 ) .
We used rice landraces for which seeds were available at the IRRI in Los Baños , Philippines , and that we knew are traditionally used for either lowland or upland cultivation . Azucena and Palawan are Filipino upland landraces while Pandan Wangi is a lowland Indonesian landrace . Azucena and Padan Wangi were grown in the dry season , while in the wet season we grew Azucena and Palawan . The rice plants were grown at IRRI in a 25 × 90 m field divided into equal size rainfed and irrigated sub-fields . The dry season and wet season experiments took place during the months of January–February 2013 and July–August 2013 , respectively . The two seasons differed for the amount and frequency of precipitation as well as for temperature , which is generally lower during the dry season . During our experiment , the mean daytime temperature was 26 . 8°C for the dry season and 27 . 6°C for the wet season . The total precipitation during the experiment period was 84 . 5 mm in the dry season and 427 mm in the wet season . Precipitation of more than 1 mm was observed for 6 out of the 29 days of the dry season experiment and 22 out of the 29 days of the wet season experiment ( Supplementary file 2 ) . In the irrigated field , one 21 day-old seedling was transplanted per hill with a spacing of 20 × 30 cm . For the rainfed field , direct seeding was performed with four seeds per hole , with a spacing of 20 × 30 cm; seedlings were thinned out to one per hill after 2 weeks . The composition of the fertilizer used was 120:30:30 nitrogen/phosphorus/potassium ( NPK ) ; it was applied as recommended . Carbofuran insecticide/nematicide was applied at 1 and 15 days after sowing ( DAS ) and herbicide was applied at 1 DAS in the upland ecosystem . Manual weeding and general plant protection were performed as needed . The first sampling took place 16 days after transplanting seedlings in the irrigated field in the dry season and 23 days after transplanting in the wet season . Each sample consisted of six young leaves ( of approximately the same size throughout the experiment ) from six individual plants . Each leaf was immediately frozen in liquid nitrogen upon collection . We tried to reduce as much as possible the effect of circadian variation on gene expression , first between sampling time-points , by always starting the collection exactly 4 hours after sunrise . Second , to avoid a shift in expression within a sampling time-point due to the delay between the first and last collected samples , we ensured fast sampling by marking beforehand each plant and leaf to be collected . Collection took , on average , 13 min ( between 11 and 15 min , except for the first time-point of the dry season , which took 18 min ) . Averages of temperature , relative humidity , wind speed , solar radiation , precipitation and atmospheric pressure at the field site were recorded every 15 min using Wireless Vantage Pro2 ISS with 24-hr fan aspirated radiation shield from Davis Instruments ( Hayward , CA ) . Measurements started 15 days before the first sampling day . After the ninth sampling of the dry season and during all of the wet season , from 1 hr before sampling to the end of sampling , we recorded climatic averages every minute . There was a weather station in the irrigated and rainfed fields , one of which also included a solar radiation sensor . No significant difference in measurements was detected between the fields , so the data of the station with the solar radiation sensor was used . We used 60 cm long 2710ARL tensiometers from Soilmoisture Equipment Corp . ( Goleta , CA ) placed 30 and 15 cm deep in the soil at four different locations , two per replicate of the rainfed field . Plant height and tiller number were measured every 6 days for the same plants all along each season until the end of sampling ( four plants per replicate in the dry season , twelve plants per replicate in the wet season ) . Frozen leaf tissue was ground manually with a mortar and pestle cooled in liquid nitrogen . Total RNA was extracted from about 200–400 µl of ground tissue using the RNeasy Plant Mini Kit ( Qiagen , Venlo , The Netherlands ) , following the manufacturer’s protocol and eluting the RNA in 40 µl of water . RNA was treated with Baseline-ZERO DNase ( Epicentre , Madison , Wisconsin ) according to the manufacturer’s instructions then cleaned up with the Qiagen RNeasy Mini Kit and eluted in 32 µl of water . We assessed RNA quality using nanodrop and electrophoresis on an agarose gel . Total RNA , 4 µg , were depleted of ribosomal RNA using Epicentre Ribo-Zero Magnetic Kit for plant leaf tissue . We purified the depleted RNA with the Agencourt RNAClean XP kit ( Beckman Coulter , Brea , CA ) . We constructed RNA libraries using the Epicentre ScriptSeq v2 RNA-Seq Library Preparation Kit , purifying the complementary DNA and libraries with the Agencourt AMPure XP System . We added barcode index using the Epicentre ScriptSeq Index PCR Primers . We quantified the libraries by Qubit ( Life Technologies , Norwalk , Connecticut ) , with the DNA HS kit . Libraries quality and average fragment size was determined using the 2100 Bioanalyzer ( Agilent , Santa Clara , CA ) with high sensitivity DNA reagents and DNA chip . We quantified the libraries on the LightCycler 480 ( Roche , Nutley , NJ ) using the KAPA ( Wilmongton , MA ) Library Quantification Kit . Libraries were sequenced using HiSeq 2000 ( Illumina , San Diego , CA ) 51 bp paired-end sequencing , with either 6 or 8 libraries per lane . Each sample provided a mean of 58 million sequencing reads . The reads were aligned to the O . sativa Nipponbare – release 7 of the MSU Rice Genome Annotation Project reference genome ( Kawahara et al . , 2013 ) – 373 , 245 , 519 base pairs of non-overlapping rice genome sequence from the 12 rice chromosomes . Also included are the sequences for chloroplast ( 134 , 525 bp ) , mitochondrion ( 490 , 520 bp ) , Syngenta pseudomolecule ( 592 , 136 bp ) , and the unanchored BAC pseudomolecule ( 633 , 585 bp ) . The annotation contains 56 , 143 genes ( loci ) , of which 6457 had additional alternative splicing isoforms resulting in a total of 66 , 495 transcripts . We used Tophat ( Kim et al . , 2013; Trapnell et al . , 2009 ) version 2 . 0 . 6 to align the reads , discarding low-quality alignments ( quality score below 1 ) . To count the number of reads that uniquely mapped to genes we used HTSeq ( Anders et al . , 2015 ) version 0 . 6 . 1 . We compensated for variable sequencing depth between samples using the median-of-ratios method of DESeq2 ( Love et al . , 2014 ) version 1 . 6 . 3 , and further performed a variance stabilizing transformation provided by the same package . We used the normalized count data for downstream analysis . This data is available in the Gene Expression Omnibus database ( http://www . ncbi . nlm . nih . gov/geo/ ) under the accession number GSE73609 . We conducted a multi dimensional scaling of the normalized expression data where the samples clustered by genotype and field and to a lesser extent , season . We used these results to detect potential sample mislabeling and identified one sample switch . We also found that the sample for the first replicate of the Pandan Wangi rainfed field in the dry season , sixth time-point , did not cluster with any of the genotype/field groups so we removed it from the analysis and replaced it with a duplication of the second replicate . We considered each subset of the data that consisted of one genotype and one season and excluded from the normalized expression dataset genes for which we detected no read for more than 40 samples in any of these subsets , which reduced the dataset to 22 , 144 expressed genes . We transformed the obtained value with the following function: log2 ( x + 1 ) , as to keep positive values of expression and averaged the biological replicates . We calculated the coefficient of variation of the log-transformed expression over the 240 data points for each expressed gene . We identified 1251 genes with a coefficient of variation below 0 . 01 , which were considered as having a stable expression in our experiment and were removed from the analysis . A preliminary cluster analysis showed that genes with a very low expression level had a weak correlation with the center of their clusters , thus 2962 genes with a mean lower than 1 were not included in the clustering step . To remove the absolute differences in expression level between genotypes and seasons , which we did not intend to model , each genotype/season subset was centered . For each of the analyses , the expression data for each gene was scaled over the whole profile . Our clustering method was the Partitioning Around Medoids algorithm from the “cluster” package version 1 . 15 . 2 ( Maechler et al . , 2015 ) in R ( R development Core Team , 2011 ) using 1 − r , with r the Pearson correlation coefficient , as the distance between expression profiles . To choose the number of clusters ( k ) we first used several clustering indices but they gave inconsistent results and usually indicated a k lower than 20 . To avoid running the risk of under-clustering , which would have resulted in averaging genes responding to very different ED factors and therefore losing transcriptional signal , we chose to constrain k only on the fact that most clusters should represent a major trend in transcriptional variation . We therefore chose k as the highest number of clusters for which no more than 5% of all the genes in the analysis belonged to “non-representative” small clusters , defined as containing less than 1% of all the genes in the analysis . The gene co-expression clusters were ordered according to the number of genes they contained so that cluster 1 was the biggest cluster . Two sets of gene annotations were obtained from Gene Ontology ( http://geneontology . org/page/about ) from the November 2014 release . For the first one , the annotation database was queried via MSU locus identifiers; for the second one , the database was queried via Uniprot Ids , obtained via a mapping from MSU Id to OMA Id , and then OMA Id to Uniprot Id ( mapping files available in the current release of OMA ) . A third set of annotations was obtained directly from OMA ( http://omabrowser . org/oma/about/ ) . All three annotation sets were then combined non-redundantly in order to produce the final annotation file for rice genes . The enrichment analysis was conducted using the GOstats package in R ( Falcon and Gentleman , 2007 ) . To represent each cluster , we calculated the average scaled expression of all the genes in the cluster . This “cluster mean” was used as the output data in our modeling approach . Climatic , field and developmental parameters constituted the input data . We calculated averages of temperature , relative humidity , wind speed , solar radiation , precipitation and pressure for 15 min 1 , 4 , and 24 hr; and 3 , 6 , 10 and 15 days before sampling . Short-term changes in temperature , humidity , wind speed and solar radiation were determined by calculating the difference between the value of each measurement at the sampling time and 20 min ( δ20 min ) , 1 hr ( δ1 hr ) and 2 hr ( δ2 hr ) earlier , using 5 , 10 , and 30 min averages , respectively . We evaluated temperature fluctuations by decomposing daily variation in temperature with the seasonal decomposition by loess ( stl ) function in R and calculating 1 hr ( ε1 hr ) , 4 hr ( ε4 hr ) and 24 hr ( ε24 hr ) before sampling averages of the remainder of the decomposition . The value used for the soil moisture parameters was a mean of measurements from four tensiometers , two in each replicate of the rainfed field . Exponentially transformed values of solar radiation used the following equations: NL + ( x ) = exp ( ( x-400 ) /200 ) NL- ( x ) = exp ( ( 400-x ) /200 ) We designed a parameter that would give an estimate of the developmental stage , rather than age , of the plant to be able to compare appropriately the plants in the rainfed and irrigated fields as they followed different developmental itineraries relative to their age in days . This parameter used three stages as fixed points: transplanting stage ( given a value of 0 ) , corresponding to the actual transplanting event for the irrigated field and determined for the rainfed plants as the stage where they were the same height as just transplanted irrigated plants , end of tillering production ( 40 ) and heading stage ( 100 ) . Intermediary time-points were calculated linearly between these fixed points . Input parameters were centered per genotype and season and scaled over the whole dataset to match the expression data . When input parameters were more highly correlated than the highest genotype correlation of all clusters ( r = 0 . 98 ) , the correlated parameters were averaged together , as we would not have enough precision in the expression data to discriminate between them . The same method was used on each cluster of a given set to select a model . It is described here for the analysis of one genotype in two seasons Source code 1; the analysis of two genotypes in one season is identical and the analysis of the irrigated field with averaged genotype follows the same principle . We used microarray data from two studies ( Nagano et al . , 2012; Sato et al . , 2011 ) available on the GEO website ( accession numbers: GSE36040 and GSE21494 ) , that was already normalized and log-transformed . We used only data from sampling points collected at 8:00 , 10:00 , 12:00 , 14:00 and 16:00 as these times were far enough from sunrise ( around 4:30 ) and sunset ( around 19:00 ) . We only used data from samples collected before flowering , from 15 to 62 days after transplanting . Sunrise time varied little enough during that period ( from 4:23 to 4:34 ) that it could be estimated that a series of samples collected at the same time of day would have negligible variation in circadian clock . There were eight samples collected at 8:00 , 12 samples at 10:00 , eight samples for single replicates and 24 samples for triplicates of eight sampling points at 12:00 , eight samples at 14:00 and eight samples at 16:00 . We excluded from the analysis genes with an expression value below −7 in more than 17 samples , keeping 19 , 837 genes , 16 , 659 of which overlap with the ones in our analysis . We averaged the biological triplicates , resulting in a total of 52 data points . For every gene , we centered individually each of the separate profiles corresponding to a given time of day , thus eliminating potential circadian clock driven differences between these times of sampling . This data constituted the PND . We tested the transferability of the models determined with the irrigated field data of our experiment , limiting our evaluation to models that were season independent and explained more than half of the variance of the cluster mean . Using the same gene distribution as our irrigated field clusters , we calculated cluster means for the PND . We calculated climatic parameters in the same way as we did for our experiment . The developmental stage parameter was the number of days after transplanting . Input parameters were centered per time of day to match the expression data . We applied the second step of our model selection method to these profiles to select models common to all five time-of-day profiles , using as a subset of parameters the ones in the model determined for our irrigated field data instead of the hdi pre-selection . We only kept a parameter in the new model if it was fitted with a coefficient of the same sign as in the original model . For the independent analysis of the PND , we used the same method as for our data , which produced 60 gene clusters . The model selection method was a simplified version of the one used for our data . We added to the set of input parameters short-term averages of 8 and 12 hr to account for a possible longer effect of same day conditions , as some samples were collected later in the day than in our experiment . We only selected models common to all five time-of-day profiles , using a unique set of parameters selected from the cluster mean in its entirety and choosing from all possible linear regressions with no more than three parameters using the BIC . We used the Field Transcriptome Database in Oryza sativa ( http://fitdb . dna . affrc . go . jp/ ) to identify the genes showing a clear environmental response detected by the Nagano et al . models , choosing the ones that had models where both the R2 of the overall model and the R2d of the environmental parameter were over 0 . 5 , with no developmental or circadian terms and an environmental term for either temperature or solar radiation with no gate or a sinusoidal gate and a dose-dependent response . | Plants need to be able to sense and respond to changes in temperature , light levels and other aspects of their environment . One way in which plants can rapidly respond to these changes is to modify how genes involved in growth and other processes are expressed . Therefore , understanding how this happens may help us to improve the ability of crops to grow when exposed to drought or other extreme environmental conditions . Most previous studies into the effect of the environment on plant gene expression have been carried out under controlled conditions in a laboratory . These findings cannot reflect the full range of gene expression patterns that occur in the natural environment , where multiple factors ( e . g . sunlight , water , nutrients ) may vary at the same time . Therefore , it is important to also analyze the effect of fluctuations in multiple environmental factors in more complex field experiments . Plessis et al . developed mathematical models to analyze the gene expression patterns of rice plants grown in the tropical environment of the Philippines using two different farming practices . One field of rice was flooded and constantly supplied with fresh water ( referred to as the irrigated field ) , while the other field was dry and only received water from rainfall ( the rainfed field ) . The experiments show that temperature and levels of sunlight ( including UV radiation ) have a strong impact on gene expression in the rice plants . Short-term variations in temperature and sunlight levels also have the most consistent effect across the different fields and seasons tested . However , for many genes , the plants grown in the irrigated field responded to the changes in environmental conditions in a different way to the plants grown in the rainfed field . Further analysis identified groups of genes whose expression combined responses to several environmental factors at the same time . For example , certain genes that responded to increases in sunlight in the absence of drought responded to both sunlight levels and the shortage of water when a drought occurred . The next step is to test more types of environments and climates to be able to predict gene expression responses under future climatic conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"plant",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2015 | Multiple abiotic stimuli are integrated in the regulation of rice gene expression under field conditions |
Reconstructing the lineage of cells is central to understanding how the wide diversity of cell types develops . Here , we provide the neurosensory lineage reconstruction of a complex sensory organ , the inner ear , by imaging zebrafish embryos in vivo over an extended timespan , combining cell tracing and cell fate marker expression over time . We deliver the first dynamic map of early neuronal and sensory progenitor pools in the whole otic vesicle . It highlights the remodeling of the neuronal progenitor domain upon neuroblast delamination , and reveals that the order and place of neuroblasts’ delamination from the otic epithelium prefigure their position within the SAG . Sensory and non-sensory domains harbor different proliferative activity contributing distinctly to the overall growth of the structure . Therefore , the otic vesicle case exemplifies a generic morphogenetic process where spatial and temporal cues regulate cell fate and functional organization of the rudiment of the definitive organ .
A major challenge in developmental biology is to explain how spatiotemporally controlled cell specification and differentiation occur alongside morphogenesis in the construction of functional organs . The inner ear is an attractive model to take on this challenge since it is accessible to manipulation , and it undergoes dynamic self-organization . It contains a manageable number of distinct cell types , which develop rapidly on an organized schedule to generate the functional units of the mature organ – the sensory patches . The key cell types of the inner ear , the supporting cells and the hair cells of the sensory patches , and the sensory neurons that innervate them , originate early during embryonic development from progenitors located in the otic vesicle , a 3D-structure arising from the otic placode adjacent to the developing hindbrain ( Durruthy-Durruthy et al . , 2014; Raft et al . , 2007; Sapède et al . , 2012; Satoh and Fekete , 2005 ) , and they are easy to score by morphology , position and specific markers ( Haddon and Lewis , 1996; Raft and Groves , 2015; Whitfield , 2015; Wu and Kelley , 2012 ) . All these features have established the inner ear as a model widely used for the study of patterning and cell specification ( Atkinson et al . , 2015; Cai et al . , 2015; Fritzsch et al . , 2006a , 2006b; Whitfield and Hammond , 2007; Wu and Kelley , 2012 ) . Despite a good understanding about the molecular hierarchies , the knowledge of how individual progenitors behave throughout patterning , proliferation , and morphogenesis remains elusive . Neuronal vs . sensory specification is achieved through well-defined bHLH transcription factors: atoh1 for hair cell formation ( Millimaki et al . , 2007; Bermingham et al . , 1999 ) , neurog1 for sensory neuron determination ( Andermann et al . , 2002; Ma et al . , 1998 ) , and neuroD for sensory neuron differentiation and survival ( Jahan et al . , 2010; Kim et al . , 2001 ) . Signals arising in the surrounding tissues regionalize the otic vesicle along axes ( Maier and Whitfield , 2014; Radosevic et al . , 2011; Riccomagno et al . , 2002 , 2005; Sapède and Pujades , 2010 ) , and this multiple step process implies a gradual restriction of cell fates over time ( Whitfield and Hammond , 2007; Wu and Kelley , 2012 ) . However , the phenotypes of targeted mutants for these signaling pathways are not always easy to reconcile ( Raft and Groves , 2015 ) , due to the limited comprehension of how developmental gene regulatory networks are integrated . For this , cellular data are needed as it can address how patterns can be achieved while the cells proliferate and the tissue undergoes morphogenesis , which may affect cell positioning and exposure to signals , and therefore cell specification . Recent developments in 4D-microscopy imaging and cell tracking tools permit now simultaneous measurements at high spatial-temporal coverage and resolution , and therefore the assessment of cell lineages and cell behaviors including displacements and proliferations ( Amat et al . , 2014; Blanpain and Simons , 2013; Faure et al . , 2016; Keller , 2013; Li et al . , 2015; Olivier et al . , 2010; Truong et al . , 2011 ) . Thus , it is time to progress in filling the void between gene regulatory networks and tissue architecture . With this purpose , we reconstructed the otic neurosensory lineage and investigated their single cell behavior by using in vivo imaging technologies paired with image processing tools ( Figure 1 , Figure 1—figure supplement 1; Faure et al . , 2016 ) . Our dynamic analyses revealed some surprising results such as the impact of neuroblast delamination and otic vesicle morphogenesis on the size and shape of this progenitor domain , and further that place and order of neuroblast delamination foreshadow their position within the statoacoustic ganglion ( SAG ) . The comparative map of neuronal and sensory progenitors in the whole otic vesicle allows understanding how their distribution changes over time , being largely segregated with a small region of putative overlap . These findings provide the cellular data helping to understand how gene regulatory networks may work during development , tissue degeneration and regeneration . 10 . 7554/eLife . 22268 . 003Figure 1 . Expansion of the neuroblast delamination domain and formation of the SAG rudiment . ( a ) Overview of the imaging and image processing strategy: inner ears of zebrafish embryos stained for cell membrane , nucleus and cell fate markers were imaged between 14-42 hpf . Image datasets were processed by nucleus center detection , cell tracking and cell shape segmentation . Data were validated and curated ( Figure 1—figure supplement 1 ) . ( b–d ) Time-lapse stills showing the posterior expansion of the neuroblast delamination domain over time; 3D-rendering of segmented epithelial neuroblasts ( green ) in context of the otic structure ( plasma membranes in magenta ) at indicated times; insets display only the segmented delamination domain with the otic vesicle contour in white . ID Dataset: 140210aX; see Figure 1—figure supplement 2d for additional analyses . ( e–g ) Time-lapse stills showing a segmented delaminating neuroblast ( red; Video 2 ) ; ( e’–g’ ) magnifications of framed regions in ( e–g ) . ID Dataset: 140426aX . ( h–i ) Still images from Video 1 displaying: otic tissue architecture ( h ) , and cellular distribution ( i ) upon SAG formation . Reconstructed cell centers are color-coded according to cell position/identity ( see legend ) . ID Dataset: 140423aX . SAG/ALLg , statoacoustic/anterior lateral line ganglia . AM/PM , anterior/posterior maculae . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 00310 . 7554/eLife . 22268 . 004Figure 1—figure supplement 1 . 3D+time image analysis pipeline . Information about plasma membranes , nuclei and cell fates was collected upon imaging the inner ears of zebrafish embryos for several hours ( 14-42 hpf; Table 1 ) under a Zeiss Lightsheet Z . 1 microscope ( 3D+t SPIM imaging ) . The acquired data were preprocessed to generate the high-resolution datasets to be launched in BioEmergences platform ( Faure et al . , 2016; Olivier et al . , 2010 ) for cell center detection and automatic tracking . Data were validated , curated and analyzed using an ad-hoc strategy based on Mov-IT , a custom-made graphical interface ( Faure et al . , 2016 ) , which offers the tools for segmentation and tracking of cells to accurately reconstruct their positions , movements and divisions . The high-resolution datasets and reconstructed lineages were used for qualitative and quantitative studies of the indicated biological processes ( Table 2 ) . The cohort of embryos used in the study can be found in Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 00410 . 7554/eLife . 22268 . 005Figure 1—figure supplement 2 . Posterior expansion of the otic neuroblast delamination domain . Tg[neuroD:GFP] embryos were injected with lyn-TdTomato mRNA at 1cell-stage and imaged from 14 . 5 hpf onwards . Embryos express GFP ( green ) in neuronal progenitors and differentiating neuroblasts , and TdTomato in all cell membranes ( magenta ) . In the case of the inner ear , GFP is expressed in epithelial neuroblasts just prior to delamination and in the SAG neuroblasts . ( a–c ) Still image views of the ventral otic vesicle at the indicated time points showing the quick expansion of the delamination domain in the otic epithelium within 2 hr from anterolateral to posteromedial regions; note that at this stage the rudiment of the adjacent ALLg is already visible . ( a’–c’ ) Transverse views are digital reconstructions along the lines indicated in ( a–c ) and illustrate that the onset of neuroblasts’ delamination progresses from lateral to medial domains ( see arrowheads ) . ALLg , anterior lateral line ganglion; SAG , statoacoustic ganglion; nt , neural tube . The otic vesicle contour is depicted in white . ID Dataset: 140210aX . ( d ) Plot depicting the posterior expansion of the neuroblast delamination domain as assessed by neuroD-GFP expression . The position of posterior-most neuroD-GFP expressing cells in the otic epithelium , and the anterior and posterior edge of the otic vesicle ( dorsal view ) were assessed over time ( see scheme ) . The plot displays the position of posterior-most GFP epithelial cells as a percent of the AP otic vesicle length ( ID datasets: 140306aX , 140125aX , 140210aX ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 005
In order to study the spatiotemporal dynamics of neuroblast delamination we in vivo imaged Tg[neuroD:GFP] embryos , which express GFP in neuronal progenitors just prior to and after delamination from the otic epithelium . We observed the first delamination events at 18 hpf in the most anterolateral region of the otic floor ( Figure 1—figure supplement 2a–a’ ) . This domain expanded towards middle and posterior regions with cells delaminating very close to the neural tube ( Figure 1—figure supplement 2b–c´ ) , as had been described by analysis of serial transverse sections from 22 hpf onwards ( Haddon and Lewis , 1996 ) . Segmentation of the epithelial neuroD-domain allowed us to illustrate the quick expansion of this territory from anterolateral to posteromedial ( Figure 1b–d ) , and when comparing three different embryos this emerges as a common trend ( Figure 1—figure supplement 2d ) . After the delamination domain is established many more cells delaminate from this territory , accumulating in the SAG just below the epithelium in close intimacy with the anterior lateral line ganglion ( ALLg; Figure 1i , Video 1 ) . 10 . 7554/eLife . 22268 . 006Video 1 . Early organization of neuroblasts within the SAG . Tg[cldnb:lynGFP]Tg[Brn3c:GFP] embryos injected with H2B-mCherry mRNA were imaged , and reconstructed cell centers were color-coded according to their location/identity ( see legend ) . The projection view video ( large panel ) simultaneously displays the topological organization of the cell group selection and tissue architecture as a projection of the GFP channel ( plasma membranes in grey ) in x , y , z-axes . The distinct visualization modes displayed on the right hand side allow for a detailed 3D-visualization of data during the analyses . Orthogonal views are used to validate cell tracking , the oblique slice view allows orienting the orthoplane along the embryonic axes , and the rendering view permits to display validated cell centers in the context of the whole image volume . ID Dataset: 140423aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 006 Neuroblast delamination implies that cells exit the otic epithelium basally and therefore undergo cell shape changes . Segmentation of individual cells -delineation of the cell contours by computational tools- when transiting from the epithelium to the SAG showed that the neuroblast cell body moves basally in less than one hour after undergoing apical constriction ( see red cell in Video 2; Figure 1e–g’ ) . Neighboring neuroblasts often delaminate consecutively to accumulate in the SAG , which quickly becomes an adjacent compact mass extending beneath the ventral floor of the otic vesicle along the anteroposterior axis by 22 hpf ( Figure 1h–i; Video 1 ) . To understand the origin of SAG-neuroblasts and how delamination coordinates in space and time we backtracked these neuroblasts to their progenitor state and followed their dynamics . 10 . 7554/eLife . 22268 . 007Video 2 . Segmentation of delaminating neuroblasts . Tg[cldnb:lynGFP]Tg[Isl3:GFP] embryos were injected with H2B-mCherry mRNA at 1cell-stage , and single delaminating neuroblasts ( n = 5 ) were automatically segmented . A representative segmentation ( red colored cell ) is shown . Transverse and lateral views ( top and bottom rows ) with their respective high magnifications on the right hand side . Note that the neuroblast changes shape and exits the otic epithelium basally within one hour . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 007 As early as 20 hpf the SAG rudiment already innervates the different maculae , and by 48 hpf the SAG is composed of two segregated neuronal populations that although displaying the same molecular signature innervate different sensory patches ( Sapède and Pujades , 2010; Zecca et al . , 2015 ) . We were eager to explore the spatiotemporal features controlling these neuronal populations and to gather information on population dynamics and lineage relationships . We first investigated whether neuroblasts were primed to different SAG-populations by backtracking SAG-neuroblasts to their progenitor state , and assessing individual cell position and tissue movements over time systematically . For this , we used the Tg[cldnb:lynGFP]Tg[Isl3:GFP] embryo injected with H2B-mCherry ( ID Dataset: 140426aX , Tables 1–2 ) . 10 . 7554/eLife . 22268 . 008Table 1 . Cohort of embryos and datasets used for the study . Datasets used in this study with corresponding information about transgenic embryos and mRNA injections . Temporal frequency of image acquisition ( timestep imaging ) and corresponding imaging sequences are depicted . Each dataset corresponds to one imaged embryonic inner ear for the corresponding time period , except for dataset 140402aX in which both ears were imaged . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 008ID dataset Transgenic embryo mRNA injection Timestep imaging Imaging sequences 140210aXTg[neuroD:GFP]lyn-TdTomato5 min 14 . 5–31 . 5 hpf140125aXTg[neuroD:GFP]lyn-TdTomato5 min 16–37 . 5 hpf140306aXTg[neuroD:GFP]lyn-TdTomato H2B-cerulean5 min 12–32 . 9 hpf140426aXTg[cldnb:lynGFP]Tg[Isl3:GFP]H2B-mCherry4 min 18–36 . 2 hpf140430aXTg[cldnb:lynGFP]Tg[Brn3c:GFP]H2B-mCherry4 min 17–37 . 2 hpf140423aXTg[cldnb:lynGFP]Tg[Brn3c:GFP]H2B-mCherry4 min 16–26 . 5 hpf140507aXTg[Brn3c:GFP]H2B-mCherry5 min 24–43 hpf140326aXTg[Brn3c:GFP]lyn-TdTomato10 min 25–45 hpf140519aXTg[Brn3c:GFP]H2B-mCherry MO-neurog15 min 24–42 hpf140513aXTg[Brn3c:GFP]H2B-mCherry MO-neurog15 min 24–32 . 9 hpf140402aXTg[Brn3c:GFP] Tg[Isl3:GFP]lyn-TdTomato H2B-cerulean10 min 25–34 hpf10 . 7554/eLife . 22268 . 009Table 2 . Datasets used for the study of the different biological questions . Datasets used for each addressed biological question , and Figures in which the corresponding analyzed data are displayed . Each dataset corresponds to one imaged embryonic inner ear , except for 140402aX in which both ears were imaged . All datasets correspond to control samples , except for 140519aX and 140513aX that correspond to MO-neurog1 embryos ( see Table 1 ) . Note that we have performed deep and detailed analyses in few datasets , and supported the observations and conclusions with partial analyses of other datasets mainly included in supplementary figures . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 009Biological insightID datasetFiguresExpansion delamination domain140210aX 140125aX 140306aXFigure 1b–d Figure 1—figure supplement 2 Figure 1—figure supplement 2d Figure 1—figure supplement 2dSegmentation delamination domain140210aXFigure 1b–dSegmentation delaminating neuroblast140426aXFigure 1e–g' Video 2Otic vesicle architecture140423aXFigure 1h–iMovIT tools140423aXVideo 1Delamination dynamics140426aXFigure 2aNeuroblasts lineage140426aX 140423aXFigure 2b–g' Videos 3–5 Figure 3—figure supplement 1 Figure 2—figure supplement 1Clonal behavior of neuroblasts140426aXFigure 3a–c Figure 3—figure supplement 1cNeuronal progenitor map140426aXFigure 3d Video 8Spatiotemporal hair cell differentiation140507aX 140326aX 140402aX 140519aXFigure 4a–c , e Video 6 Figure 4—figure supplement 1a Figure 4—figure supplement 1c Figure 4—figure supplement 1b Figure 4f–h , j Video 6 Figure 4—figure supplement 1aHair cell progenitor map140507aX 140326aX 140519aXFigure 4d Video 7 Video 8 Figure 4—figure supplement 1c Figure 4i Video 7Spatiotemporal cell proliferation140507aX 140519aXFigure 5a–f Video 9 Figure 5g–lLocal cell density/NN-distances140507aX 140519aX 140430aX 140326aXFigure 5m Video 10 Figure 5—figure supplement 1a , b Figure 5n Video 10 Figure 5—figure supplement 1b Figure 5—figure supplement 1b Figure 5—figure supplement 1bOtic vesicle volume140426aX 140507aX 140430aX 140513aX 140519aXFigure 5—figure supplement 1e Figure 5—figure supplement 1c , e Figure 5—figure supplement 1e Figure 5—figure supplement 1e Figure 5—figure supplement 1d , e Analysis of the dynamics of this process revealed that once the first delamination events are observed , delamination occurs massively . The lineage tree shows that within six hours a big bulk of neuroblasts delaminates ( see white , yellow and orange lines in Figure 2a corresponding to single delaminated neuroblasts ) , and many neuroblasts can divide after exiting ( see lines branching in Figure 2a ) as previously described ( Vemaraju et al . , 2012 ) . This lineage analysis allowed ascribing delamination time ( see hpf in x-axis in Figure 2a ) and delamination position to each neuroblast . Then , we color-coded the reconstructed cell centers according to these two criteria and displayed them ( i ) early when in the otic vesicle ( Figure 2b , d , f ) , and ( ii ) late when within the SAG ( c-c’ , e-e’ , g-g’ ) . The first observation was that neuroblasts delaminated randomly in space and time ( see intermingled colors of the reconstructed cell centers in Figure 2b ) , but their organization within the SAG relies on the delamination order as delaminated cells aggregate laterally to the preexisting ganglion ( Video 3 ) , thereby generating a mediolateral ( ML ) growth pattern with early-delaminated cells ( white ) located more medially than late-delaminated ones ( red ) ( Figure 2c–c´; see Figure 2—figure supplement 1a–b for additional analysis ) . While the inner ear grows and undergoes morphogenesis , the SAG becomes squeezed in the space between the ventromedial wall of the ear and the neural tube , such that its organization is gradually converted from ML to dorsoventral ( DV ) , with early-delaminated sensory neurons located in close contact with the neural tube ( Video 3; Figure 2c ) . Thus , the time of neuroblast delamination foreshadows the ML gradient of neuronal differentiation in the SAG . 10 . 7554/eLife . 22268 . 010Figure 2 . The organization of cells within the SAG relies on specific temporal and spatial cues . ( a ) Flat representation of the neuroblast lineage tree with branches indicating cell divisions . The x-axis displays the time of embryonic development in hours post-fertilization ( hpf ) . Neuroblast lineages are displayed from the moment of delamination onwards and ordered and color-coded according to delamination timing ( intervals: 18–20 hpf white , 20–22 hpf yellow , 22–24 hpf orange , 24–30 hpf red ) . Some cells were not tracked until the end of the sequence , and are depicted as interrupted lines . The extensive cell loss during the early stages of delamination ( 18–22 hpf ) was verified in a second embryo; in both cases , about 25% ( 23 . 2% and 26 . 8% ) of the otic epithelial cells at 18 hpf exit by delamination in the consecutive four hours . ( c–c’ , e–e’ , g–g’ ) Neuroblasts within the SAG ( n = 144 of roughly total n = 250 ) were backtracked to their progenitor state in the epithelium ( n = 98; b , d , f; Videos 3 and 5 ) . Cell lineages were color-coded for: time of delamination ( b–c’; same intervals as in ( a ) ) , position in the epithelium along the AP ( d–e’ ) , or ML ( f–g’ ) axes . Note that ML organization of neuroblasts within the SAG ( c–c’ ) relies on their delamination order , and that the blue/white/red epithelial pattern ( d–e’; neuroblasts AP position ) but not the green/white/red one ( f–g’; neuroblasts ML position ) is maintained in the SAG over this time period ( 18–30 hpf ) . Reconstructed cell centers were displayed as colored-dots together with the corresponding raw images ( plasma membranes in grey level ) . ( b , d , f ) dorsal views; ( c , e , g ) ventral views; ( c’ , e’ , g’ ) lateral views . Anterior is always to the left . For this analysis , Tg[cldnb:lynGFP] Tg[Isl3:GFP] line was injected with H2B-mCherry mRNA at 1cell-stage ( Tables 1–2 ) . ID Dataset: 140426aX; see Figure 2—figure supplement 1 for additional analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01010 . 7554/eLife . 22268 . 011Figure 2—figure supplement 1 . Time of delamination and position of epithelial neuroblasts prefigure their location within the SAG . Tg[cldnb:lynGFP] Tg[Brn3c:GFP] embryo was injected with H2B-mCherry at 1cell-stage , imaged and analyzed from 16 hpf ( Table 1 ) . Reconstructed cell centers from neuronal progenitors were color-coded for time of delamination ( a ) , or position along the AP axis in the otic epithelium ( c ) , and followed from 18 hpf to 24 hpf . Note that: ( i ) among the delaminated neuroblasts from 18 hpf to 24 hpf , those delaminating earlier ( white cell centers in a-b ) are more medially located in the SAG than those delaminating later ( purple cell centers in a-b ) ; and ( ii ) the relative position of neuronal precursors along the AP axis in the otic epithelium is conserved within the SAG ( see cyan anterior cells vs . red posterior cells in c-d’ ) . ID dataset: 140423aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01110 . 7554/eLife . 22268 . 012Video 3 . The order of delamination foreshadows the mediolateral gradient of neuroblasts differentiation within the SAG . A cohort of 144 delaminated neuroblasts was analyzed for time of delamination . Reconstructed cell centers were color-coded according to four delamination intervals: 18–20 hpf white , 20–22 hpf yellow , 22–24 hpf orange , 24–30 hpf red . Note that neuroblasts exit randomly from the delamination domain . Those delaminating earlier are located more medially in the SAG than the later delaminating ones prefiguring the gradient of differentiation . Reconstructed cell centers were displayed as colored dots together with the corresponding volume rendering images ( plasma membranes in grey level ) with higher intensity on the left hand side . Tg[cldnb:lynGFP] Tg[Isl3:GFP] embryo was injected with H2B-mCherry mRNA at 1cell-stage . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 012 Then , we analyzed the epithelial origin of two functionally distinct SAG populations: anterior , mainly involved in vestibular , and posterior responsible for vestibular and acoustic functions ( Haddon and Lewis , 1996 ) . We found that the relative position of progenitors foresees their final location in the SAG: neuroblasts from the anterior portion of the SAG delaminated from the anterolateral floor of the otic vesicle , while the ones of the posterior SAG derived from the posteromedial otic epithelium ( Video 4 ) . Next , we addressed whether the epithelial coordinates prefigure the position of neuroblasts in the SAG primordium . We observed that the anteroposterior ( AP ) coordinates of neuroblasts delamination ( see differently colored epithelial cell center populations in Figure 2d ) generally defined their relative AP position within the SAG ( Figure 2e–e’; Video 5; see Figure 2—figure supplement 1c–d’ for additional analysis ) . These data show that progenitors delaminating within similar spatial regions along the AP axis maintain similar relative positions later on , and indeed the analysis of their trajectories indicates that cells , while in the otic epithelium , maintain their neighbor relationships . No such correlation was observed regarding the ML axis , because cells that were originally separated in the epithelium were found close in the SAG ( Figure 2f–g’; Video 5 ) , suggesting that the AP and time cues prevail . These findings establish a link between the place and order of neuroblast delamination and their neuronal ( functional ) identity within the SAG , and suggest the existence of a spatial and temporal regulation in the otic epithelium of the fate and exit of SAG neuroblasts . 10 . 7554/eLife . 22268 . 013Video 4 . The anterior and posterior SAG neuronal populations derive from the anterior and posterior otic epithelium , respectively . Neuroblasts belonging to distinct SAG neuronal populations , anterior or posterior ( cyan/red ) , were backtracked to their otic epithelial origin prior to delamination . Note that cells of the anterior and posterior SAG delaminate from the anterior and posterior otic epithelium , respectively , and they maintain this relative position . The video is played backwards . These neuroblasts are the same set of cells as used in Video 3 , but analyzed for AP position in the SAG instead of delamination time . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01310 . 7554/eLife . 22268 . 014Video 5 . Position of epithelial neuroblasts along the anteroposterior axis prefigures their location within the SAG . Reconstructed cell centers from neuronal progenitors were color-coded for position along the anteroposterior ( AP ) ( left ) or mediolateral ( ML ) ( right ) axes in the otic epithelium and followed from 18 hpf to 30 hpf . Note that the relative position of neuroblasts along the AP but not the ML axis is maintained from the otic epithelium to the SAG . These neuroblasts are the same set of cells as used in Video 3 , but analyzed for epithelial position along the AP/ML axes instead of delamination time . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 014 We were interested in understanding whether cell proliferation had any role in conferring neuroblasts categories . To investigate this question we used the previous dataset ( ID Dataset: 140426aX , Tables 1–2 ) to undertake the neuroblast clonal analysis ( Figure 3a ) , which showed that SAG-neuroblasts originate from a neuronal progenitor pool with different division behaviors: one third of the analyzed neuroblasts divided before exiting the epithelium ( n = 42 red cells in Figure 3a , Figure 3—figure supplement 1c ) , one third divided after delamination ( n = 40 orange cells in Figure 3a; Figure 3—figure supplement 1c ) , and a similar fraction of neuroblasts did not divide within this time interval ( n = 34 blue cells in Figure 3—figure supplement 1c ) . Thus , cell position within the otic epithelium is not relevant for these cell division behaviors because no specific spatial distribution can be observed ( Figure 3—figure supplement 1c ) . In addition , cell division and delamination are independent events , in contrast to what was proposed for epithelial invagination of the Drosophila tracheal placode ( Kondo and Hayashi , 2013 ) . 10 . 7554/eLife . 22268 . 015Figure 3 . Clonal analysis of neuroblasts . ( a ) Lineages of neuroblasts ( epithelial: colored; delaminated: white ) ordered by time of division and grouped according to the division behavior -dividing before ( red , top ) or after ( orange , bottom ) delamination . Each line corresponds to a single neuroblast . Discontinued lines represent cells that were not tracked further . The x-axis displays the corresponding time of embryonic development ( hpf ) . ( b ) Box plot illustrating the temporal delay in delamination between sister cells ( Figure 3—figure supplement 1d ) . ( c ) Illustration of neuroblast division behavior colored as in ( a ) . ( d ) Dynamic map of neuronal progenitors ( orange circles ) and their epithelial neighboring cells ( grey circles ) in the context of the whole otic vesicle ( grey dots ) over time; see Video 8 for the 24 hpf animation . The color intensity of cell centers depicts the position of cells along the dorsoventral axis of the otic vesicle . The map was built after following the lineages from 18 hpf to 26 hpf of all encircled cells . Note how neuroblast delamination impacts on the size and position of the progenitor domain ( orange cell centers ) over time . Tether cells are depicted as black circles . For this analysis , Tg[cldnb:lynGFP] Tg[Isl3:GFP] line was injected with H2B-mCherry mRNA at 1cell-stage ( Tables 1–2 ) . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01510 . 7554/eLife . 22268 . 016Figure 3—figure supplement 1 . Spatial distribution of epithelial neuroblasts according to division behavior or delamination time . Tg[cldnb:lynGFP] Tg[Isl3:GFP] embryo was injected with H2B-mCherry at 1cell-stage , imaged and analyzed from 18 hpf to 36 . 2 hpf ( Table 1 ) . Images of ( a ) nuclei as Maximal Intensity Projection of few orthoplanes of the ventral otic vesicle , and ( b ) plasma membranes as 3D-rendering are displayed . ( c ) Spatial distribution of neuroblasts , whose reconstructed cell centers were color-coded according to their division behavior -before or after delamination- ( Figure 3 ) . Out of 116 tracked epithelial neuroblasts , 42 divide before delamination ( red ) , 40 do so after delamination ( orange ) , and 34 do not divide within this time window ( blue ) . Note that there is no preferential spatial distribution of cells for these features . ( d ) Reconstructed cell centers of neuroblasts ( n = 131 ) were color-coded according to time of delamination: 18–20 hpf white , 20–22 hpf yellow , 22–24 hpf orange , 24–30 hpf red . Neuroblasts giving rise to two sister cells falling into distinct delamination intervals are shown as bicolored cell centers ( n = 11 ) . ID Dataset: 140426aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 016 Neuroblasts dividing before delamination give rise to two daughter cells that will delaminate ( Figure 3a–c ) . In most cases , sister cell delaminations happened within a two hours interval , however in a few cases the time of delamination between two sister cells can be as long as seven hours ( Figure 3b ) . Neuroblasts giving rise to two daughter cells that delaminated within a delay interval of more than two hours were evenly distributed in the otic epithelium ( n = 11/116 bicolored cell centers , Figure 3—figure supplement 1d ) . Overall , these analyses suggest that ( i ) epithelial neuroblasts’ divisions give rise to cells that will delaminate ( Figure 3c ) , and ( ii ) cell division is not necessary as a trigger for delamination . Finally , we used the same dataset ( ID Dataset: 140426aX; Tables 1–2 ) to understand how the neuronal progenitor domain changes upon cell delamination and morphogenesis of the inner ear over time ( orange circles in Figure 3d; n = 131 at 18 hpf ) . We followed as well the behavior of the surrounding otic epithelial cells as a repair for the edge of the progenitor domain ( see grey circles in Figure 3d , n = 64 at 26 hpf ) . Due to the massive cell delamination in a relatively short time ( more than 150 delamination events in the period 18-26 hpf ) , the neuronal progenitor domain undergoes dramatic size and shape changes ( Figure 3d , compare the territory containing orange cell circles at 18 hpf and 26 hpf ) , as described by Haddon and Lewis ( 1996 ) . Upon delamination a large amount of cells exits from the otic epithelium: about 25% of total otic cells at 18 hpf are lost by delamination between 18 hpf and 22 hpf . The exit of neuroblasts and the growth of the vesicle are part of the morphogenetic processes that progressively restricted the progenitor field to a more lateral region of the ventral floor of the vesicle ( orange cell centers in Figure 3d , 26 hpf ) . Hair cells are key cell types for building a functional sensory patch , and they are specified in the otic vesicle in the same way in all vertebrates . However , one difference is that in the fish the first hair cells differentiate while neuronal precursors are still delaminating , and that hair cells continue to be produced throughout life ( Haddon and Lewis , 1996 ) . This raises the question of how neuronal and hair cell production is coordinated from progenitor domains located very close within the otic epithelium , and this led us to compare the progenitor maps for hair cells and neurons . Up to date , the formation of the anterior ( AM ) and posterior ( PM ) maculae ( the first sensory patches to arise ) has been followed up by gene or transgene expression mainly in 2D , and therefore their proper allocation within the 3D-otocyst was difficult to assess . In order to understand how these sensory patches arose we followed in vivo the incorporation of newly differentiated hair cells . The first hair cells to arise -the so called tether cells ( Riley et al . , 1997 ) - constituted the posterior pole of each of the maculae ( white-dotted cells in Figure 4a–c; Figure 4—figure supplement 1b , c ) and prefigured the position of these sensory patches within the otic vesicle ( Video 6 ) . Next , these two sensory patches gradually increased their size incorporating differentiated hair cells at their anterior pole with a specific pattern ( see color-coded cell centers in Figure 4b–c; Video 6; ID Dataset: 140507aX; and see Figure 4—figure supplement 1b , c for additional samples exhibiting the same pattern ) . At the same time , growth and morphogenesis of the vesicle led to a structure in which the AM remained anterior and ventral , and the PM positioned in the posterior ventral edge of the medial wall of the vesicle . Once hair cells differentiate they become postmitotic ( Video 6 ) . Thus , the formation of the sensory patches depends on a pool of progenitors providing postmitotic differentiated hair cells . AM and PM develop asynchronously , with the incorporation of new hair cells being delayed in the PM ( Figure 4—figure supplement 1; Sapède and Pujades , 2010 ) . 10 . 7554/eLife . 22268 . 017Figure 4 . Spatiotemporal pattern of hair cell differentiation and map of sensory progenitors . Differentiated hair cells were tracked during 18 hr in control and MO-neurog1 , and reconstructed cell centers were color-coded according to the differentiation time displayed in the legend ( Video 6 ) . ( a–b , f–g ) Spatiotemporal pattern of hair cell differentiation of the anterior/posterior maculae ( AM/PM ) ; reconstructed colored cell centers overlaid with the corresponding raw images ( hair cell fate in grey level ) from Tg[Brn3c:GFP] embryos; ( c , h ) reconstructed colored cell centers in lateral view . Note how the temporal but not the spatial development is altered in the MO-neurog1 PM ( see Figure 4—figure supplement 1 ) . ( d , i ) Map of hair cell progenitors in the whole otic vesicle ( Videos 7–8 ) ; the maps were generated by backtracking the differentiated PM hair cells ( e , j ) . ID Datasets: 140507aX for control , 140519aX for MO-neurog1; see Figure 4—figure supplement 1 for additional analysesDOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01710 . 7554/eLife . 22268 . 018Figure 4—figure supplement 1 . Temporal pattern of hair cell differentiation in AM and PM . Tg[Brn3c:GFP] embryos injected with H2B-mCherry mRNA at 1cell-stage ( with/without MO-neurog1 ) were imaged from 24 hpf to 42 hpf . ( a ) Graphs showing the increase of differentiated hair cells in the anterior ( AM ) and posterior ( PM ) maculae over time; the final number of differentiated hair cells at 42 hpf is indicated . Each line corresponds to a differentiated hair cell plotted from and color-coded for time of differentiation ( see legend ) . Tether cells are depicted as white lines . Note the temporal differences in the development of the PM between control and MO-neurog1 embryos , with no major changes in the spatial pattern ( Figure 4 ) . Asterisk depicts an apoptotic hair cell . ID Datasets: 140507aX for control , 140519aX for MO-neurog1 . ( b ) Spatiotemporal pattern of hair cell differentiation of the AM and PM of both ears of Tg[Brn3c:GFP]Tg[Isl3:GFP] embryo injected with lyn-TdTomato and H2B-mCherry mRNA at 1cell-stage ( imaged from 25 hpf to 36 hpf ) . Differentiated hair cells were tracked and reconstructed colored cell centers overlaid with the corresponding raw images ( hair cell fate/SAG in grey level as 3D-rendering ) . Graphs showing the increase of differentiated hair cells in the AM and PM over time; the final number of differentiated hair cells at 36 hpf is indicated . ID Dataset: 140402aX . ( c ) Spatiotemporal pattern of hair cell differentiation of AM and PM . Tg[Brn3c:GFP] embryo injected with lyn-TdTomato mRNA at 1cell-stage and imaged from 25 hpf to 45 hpf . Graphs displaying the increase of differentiated hair cells in AM/PM . Hair cell progenitors and differentiated hair cells were tracked ( except for the one encircled in blue ) , and reconstructed colored cell centers overlaid with the corresponding raw images ( MIP of lynTomato and GFP signal of a few slices in grey level ) . ID Dataset: 130326aX . Legend in panel ( a ) applies to all plots and diagrams of the figure . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 01810 . 7554/eLife . 22268 . 019Video 6 . Spatiotemporal pattern of hair cell generation of sensory maculae in control and MO-neurog1 embryos . Tg[Brn3c:GFP] embryos were injected with H2B-mCherry mRNA at 1cell-stage ( in the presence or absence of MO-neurog1 ) , and differentiated hair cells of the anterior ( AM ) and posterior ( PM ) maculae were tracked during 18 hr ( Table 1 ) . The centers of hair cells were reconstructed and color-coded according to their time of differentiation as given by the onset of GFP expression . Top row displays reconstructed color-coded cell centers together with imaging data ( orthoslice views of the maculae with raw images of hair cells in grey level ) , middle row shows imaging data alone ( GFP signal as volume rendering ) , and bottom row displays only the reconstructed color-coded cell centers . Volume rendering and reconstructed cell centers panels rotate from dorsal to lateral view to illustrate the 3D-organization of hair cells within the maculae . ID Datasets: 140507aX for control , 140519aX for MO-neurog1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 019 Previous findings indicated that a specific pool of neuronal progenitors switches its fate to hair cells of the PM upon abrogation of neurog1 function ( Sapède et al . , 2012 ) . However , they were unable to clearly assess the position of this progenitor pool and how it behaved . To do so , we first undertook the dynamical analysis of maculae generation upon neurog1-downregulation by injection of translation-blocking morpholino ( ID Dataset: 140519aX , Tables 1–2 ) , which fully recapitulates the phenotype of neurog1hi1059 mutants ( Sapède et al . , 2012 ) . In the absence of neurog1 , sensory neurons do not form and supernumerary hair cells are produced in the PM ( Figure 4f–h; Video 6 ) . Tracking the differentiated hair cells allowed us to compare the dynamics of sensory development at the single-cell level: the overall spatial pattern of hair cell generation in the AM/PM was very similar between control and morphant embryos ( compare Figure 4b–c with g–h ) ; however the kinetics of hair cell production in the PM differed due to a boost of differentiation in MO-neurog1 from 34 hpf onwards ( Figure 4—figure supplement 1a ) . All supernumerary hair cells in the MO-neurog1 originate from newly differentiated cells , and not from differentiated hair cells re-entering the cell cycle ( Video 6 ) . To unveil the organization of the PM hair cell progenitor pool we generated the map of hair cell progenitors in the whole otic vesicle over time ( ID Datasets: 140507aX , 140326aX ) . For this , differentiated hair cells at 42 hpf were backtracked to their progenitor state , and information about differentiation time ( color-code ) and progenitor distribution was combined ( Figure 4d–e; Figure 4—figure supplement 1c ) . 3D-reconstructions revealed that progenitors for hair cells of the PM were distributed over the ventromedial domain of the otic vesicle at early stages of embryonic development ( see color-coded cell centers in Figure 4d and Figure 4—figure supplement 1c ) , and that upon neurog1 inhibition the hair cell progenitor pool expanded along the anteroposterior dimension allocating more medially ( ID Dataset: 140519aX; Figure 4i , Video 7 ) . Furthermore , the position of hair cell progenitors foreshadows the organization of differentiated cells within the sensory patch , demonstrating that otic epithelial cells did not rearrange during early stages of hair cell differentiation . However , the relative positions of the maculae changed upon morphogenesis , resulting in the growth of the PM towards the anterior region while growth of the AM was mainly anterior and towards dorsal ( Figure 4c , h ) . Finally , our analysis allowed us to compare the progenitor maps for hair cells and neurons at the onset of hair cell differentiation ( 24 hpf ) : they were largely segregated with a small region of putative overlap ( Video 8 ) , consistent with the existence of a pool of dual progenitors ( Sapède et al . , 2012 ) . 10 . 7554/eLife . 22268 . 020Video 7 . Position of the posterior macula hair cell progenitor pools in control and MO-neurog1 embryos . Dynamic display of the posterior macula hair cell progenitors in control and MO-neurog1 embryos at 24 hpf . Progenitor pools were determined from the backtracking of differentiated hair cells in the Tg[Brn3c:GFP] line injected with H2B-mCherry at 1cell-stage . Hair cell progenitors are color-coded for time of differentiation and plotted in the context of the whole otic vesicle ( grey dots depict reconstructed cell centers of the otic epithelium ) . Tether cells are shown as black circles . The animation displays a rotation of otic vesicles around the anteroposterior axis . ID Datasets: 140507aX for control , 140519aX for MO-neurog1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 02010 . 7554/eLife . 22268 . 021Video 8 . Maps of neuroblasts and PM hair cell progenitors in the whole otic vesicle . The neuronal progenitors ( orange; Figure 3d ) and the posterior macula hair cell progenitors ( blue; Figure 4d ) are plotted in the context of the whole otic vesicle ( grey dots ) at 24 hpf . Tether cells are shown as black circles . The animation displays a rotation of otic vesicles around the anteroposterior axis . Note that the two progenitor domains are adjacent , and neuroblasts are located more ventrally while sensory progenitors are more medially . ID Datasets: 140426aX for neuronal progenitors , 140507aX for PM hair cell progenitors . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 021 As overall shape and size of the otic vesicle remain relatively constant during neuroblast delamination , we assumed that there was some homeostatic process involving cell proliferation elsewhere and we further validated this hypothesis . To understand the proliferative behavior of the distinct otic territories , we reconstructed during 16 hr the lineages of 51–64 neighboring cells located either in the non-sensory ( dorsolateral region of the otic vesicle; Figure 5a–c , Video 9 ) or in the sensory domain ( ventromedial epithelial region; Figure 5d–f , Video 9 ) . For this we used the Tg[Brn3c:GFP] H2B-mCherry injected embryo ( ID Dataset: 140507aX ) . Cell behavior was assessed according to cell division ( dividing/non-dividing ) or cell differentiation ( progenitor/differentiated ) status . Cell position was monitored over time to identify spatial organization related to these features ( Video 9 ) . Cells within the non-sensory domain actively proliferated ( Figure 5a–c; Video 9 ) ; in contrast , half of the analyzed cells in the ventral region did not divide and some differentiated into hair cells ( Figure 5d–f; Video 9 ) , suggesting that once hair cell progenitors are committed they divide less . Interestingly , even though the non-sensory domain displayed higher proliferative activity , it was less compacted ( note neighboring reconstructed cell centers are more spaced in Figure 5a–b than in d–e ) , supporting the idea that it contributes to the overall growth of the vesicle during this time window . To quantify this , we calculated the nearest neighbor distance ( NN-distance ) for all cells in these domains over time and plotted the reconstructed cell centers color-coded for their NN-distance in the whole vesicle ( see that green cell centers are more compacted than blue ones in Figure 5—figure supplement 1a; see Figure 5—figure supplement 1b for analyses of additional specimens ) . 10 . 7554/eLife . 22268 . 022Figure 5 . Heterogeneous cell behavior in the non-sensory and sensory domains . Neighboring cells in the non-sensory and sensory domains of control ( a–f ) and MO-neurog1 ( g–l ) were tracked and reconstructed cell centers were color-coded according to cell proliferation/differentiation status ( see legend in ( c ) ; Video 9 ) ; they were plotted on the top of the corresponding raw images ( a–b , d–e , g–h , j–k; nuclei in grey level ) , or in graphs over time ( c , f , i , l ) displaying the total number of cells in each domain and their status in the course of the video . Note the differences in the graphs between non-sensory and sensory domains , but not between control and MO-neurog1 embryos . ( m–n ) Estimated local cell densities at 24 hpf are represented by color-coded cell centers across the whole otic epithelium ( Video 10 ) . Tg[Brn3c:GFP] embryos injected with H2B-mCherry and with/without MO-neurog1 at 1cell-stage were used for full lineage reconstruction ( Tables 1–2 ) . Anterior is always to the left . ID Datasets: 140507aX for control , 140519aX for MO-neurog1; see Figure 5—figure supplement 1 for additional analyses . ( o ) Graphic depicting the total number of cells in the otic vesicles for wild type ( control , n = 3 ) , neurog1hi1059/ hi1059 mutant in the Tg[Isl3:GFP] background ( n = 3 ) , and MO-neurog1 embryos ( n = 2 ) at 24 hpf . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 02210 . 7554/eLife . 22268 . 023Figure 5—figure supplement 1 . Tissue architecture in sensory and non-sensory domains . Tg[Brn3c:GFP] injected with H2B-mCherry mRNA at 1cell-stage were imaged and analyzed . ( a ) Plots at three time points showing the distribution of tracked neighboring otic cells within the non-sensory ( blue ) or sensory ( green ) domains in the context of the whole otic vesicle ( grey dots ) . Color intensities represent the nearest neighbor distance ( NN-distance , see legend ) ; graph shows the median and quartiles of the NN-distance for each domain over time . Note that cells within the non-sensory territory are more spaced than cells in the sensory domain . ID Dataset: 140507aX . ( b ) Graph showing the differences in average NN-distances between the ventromedial and dorsolateral domains ( top ) , or ventral and dorsal territories ( bottom ) of several ears . Note that cells within the dorsolateral/dorsal territories are more spaced than cells in the ventromedial/ventral domains . Cell selections in the context of the whole ear for 140426aX are shown in the left-hand side . ID Datasets: 140507aX , 140519aX , 140430aX , 140426aX . ( c–e ) Comparison of the volumes of otic vesicles of control and MO-neurog1 embryos at 24 hpf . ( c–d ) Lateral views of otic vesicle volumes depicting the epithelial surfaces: basal/outer ( grey mesh ) and apical/inner ( green mesh ) ; insets in ( c–d ) display the corresponding dorsal views . ID Datasets: 140507aX for control , 140519aX for MO-neurog1 . ( e ) Graph showing the average volume of otic vesicles for control ( n = 3 ) and MO-neurog1 ( n = 2 ) embryos in cubic µm . Note the increase of the average volume in MO-neurog1 . Tg[Brn3c:GFP] embryos were injected with H2B-mCherry and with/without MO-neurog1 at 1cell-stage . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 02310 . 7554/eLife . 22268 . 024Video 9 . Cell proliferative activity in the non-sensory and sensory domains . Tg[Brn3c:GFP] embryos injected with H2B-mCherry at 1cell-stage were used for full lineage reconstruction ( from 24 hpf to 38 hpf ) of 51 and 64 neighboring cells located in the non-sensory and sensory domains of the otic vesicle , respectively . Transverse view on the left is to better illustrate the position of the cell population domains along the axes . Reconstructed cell centers color-coded according to cell division/differentiation status and plotted in context of the raw images ( nuclei channel as volume rendering in grey levels ) . ID Dataset: 140507aX . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 024 Similar cell behaviors and cell organization were observed in the MO-neurog1 with no increase in proliferative activity in the sensory domain ( Figure 5g–l ) , revealing that upon neurog1 abrogation hair cell progenitors do not divide more actively , and therefore supernumerary hair cells derive either from expansion of the progenitor pool , or from precocious differentiation of progenitors . Since in MO-neurog1 cells do not undergo delamination , we determined how these non-delaminated cells located within the epithelium . For this , we estimated the distribution of local cell densities over the entire otic epithelium ( Figure 5m–n; Video 10 ) . While upon neurog1 abrogation the number of cells was greatly increased ( Figure 5o; control = 384 ± 17 cells n = 3 vs . neurog1hi5109/hi5109 = 479 ± 14 cells n = 3 or MO-neurog1 = 506 ± 14 cells n = 2 ) , the global pattern of cell densities did not change: cells in either condition are most densely packed in the anterolateral and ventral region and most spaced in the dorsal domain ( see red and blue cells in Figure 5m–n; Video 10 ) . This suggests that the impact of non-delaminated cells in the compaction of the otic epithelium is low . Thus , if in the MO-neurog1 cell density is not altered but cell numbers are higher in the vesicle , the volume of the otic structure should be larger than the control one . Indeed , when assessing the volumes by 3D-point-cloud segmentation the otic vesicles of MO-neurog1 were larger than the control ones ( Figure 5—figure supplement 1c–e ) . Altogether these results show that the otic epithelium is a heterogeneous tissue , where proliferative activity and cell compaction differ between sensory and non-sensory domains during the generation of the neurosensory cellular elements . 10 . 7554/eLife . 22268 . 025Video 10 . Local cell densities across the whole epithelium in control and MO-neurog1 . All cells of the otic epithelium in a control and a MO-neurog1 embryo are plotted and color-coded according to their estimated local cell density at 24 hpf . Tether cells are depicted in black for better orientation . The animation displays a rotation of otic vesicles around the anteroposterior axis . Note that the anteroventral territory displays higher cell density in both cases . ID Datasets: 140507aX for control , 140519aX for MO-neurog1 . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 025
We provide information about cellular/population dynamics and lineage relationships of neurosensory elements in the inner ear from the reconstruction of their lineage trees from 4D in vivo data . These results enable us to: ( i ) understand the proportions of the system , ( ii ) reveal the impact of morphogenesis in the spatiotemporal distribution of neurosensory cell progenitor pools , and ( iii ) provide the cellular data to complement the well described gene regulatory networks involved in neurosensory specification . Our strategy has been to analyze in depth and detail few selected embryos , and support our findings with lower power analyses of other specimens . This analysis of progenitor populations in their native environment over extended timespans provides the opportunity to understand in vivo cell behaviors , both at the single cell level and at the cell population level . However , due to the technically demanding nature of the experiments and analyses the sample size could not be very high . All biological systems have intrinsic noise , and therefore variability: absolute cell numbers may differ between embryos , the onset of developmental processes might occur at slightly different times , or even distinct transgenic backgrounds may behave different . Therefore , to overcome these possible biases we analyzed additional specimens with lower power analysis and demonstrate that our approach can yield reproducible results in terms of cell type proportions and cellular behaviors . To assure the robust outcome of a functional inner ear , with sensory patches containing the precise number of mechanosensory hair cells properly innervated by sensory neurons arranged in a precise topology ( Video 11 ) , the developmental strategy used by distinct progenitor populations differs: neuronal specification is concomitant with proliferation ( before/after delamination ) , while hair cell specification and differentiation lead to postmitotic cells indicating that the final number of sensory cells relies on the control of the progenitor pool . Interestingly , the position of hair cell progenitors foreshadows the organization of differentiated cells within the sensory patch , demonstrating that although the otic vesicle undergoes morphogenesis and cells do extensively proliferate , progenitors do not rearrange within the epithelium during early stages of hair cell differentiation . 10 . 7554/eLife . 22268 . 026Video 11 . Innervated sensory patches in the embryonic inner ear . Animation of a Tg[Brn3:GFP]Tg[Isl3:GFP] embryo displaying the sensory patches with differentiated cellular neurosensory elements in green at 48 hpf . Differentiated hair cells of the maculae ( AM/PM ) and cristae ( ac/lc/pc ) are innervated by sensory neurons of the SAG , which shows the typical segregation into anterior and posterior portion alongside with the segregated projections to the hindbrain ( Sapède and Pujades , 2010 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 22268 . 026 The inner ear comprises two distinct functional modalities –vestibular and acoustic- carried out by different SAG populations . However , their epithelial origin and how SAG neuroblasts acquire their specific identity are still open questions . Here we show the importance of place and time of neuroblast delamination in their allocation within the SAG , and shed light on how distinct information may converge in the progenitor cells . Delamination place confers position along the AP axis of the SAG , and therefore function , most probably as the result of the integration of patterning signals involved in the emergence of the different domains ( Maier and Whitfield , 2014; Radosevic et al . , 2011 ) . Additionally , delamination time prefigures the relative position of early-delaminated neuroblasts vs . late-delaminated ones and therefore , the gradient of neuronal differentiation within the SAG . Consequently , an epithelial neuroblast will need to integrate spatial ( its position in the epithelium ) and temporal ( its time of delamination ) information for its final allocation within the SAG . However , to achieve the fully functional organ , complexity needs to increase later on possibly by some new delamination events , division of SAG neuroblasts and changes in morphology . Given that auditory neurons , which innervate the PM , accumulate later than the vestibular neurons ( Sapède and Pujades , 2010 ) , it is possible that the majority of neuroblasts becoming auditory neurons may delaminate from posterior domains , since no specific increase in cell proliferation has been observed at later stages in the posterior part of the SAG ( Vemaraju et al . , 2012 ) . However , if this is the mechanism and how it is controlled remain to be explored . The knowledge of how the gradient of neuronal differentiation within the SAG is built up helps the comprehension of its differential gene expression pattern according to the neuronal differentiation state ( Zecca et al . , 2015 ) , and the organized distribution of differentiated neurons ( Vemaraju et al . , 2012 ) . Thus , our results provide ( i ) a model to discuss how the selection of different progenitors and the determination of their relative population size could be regulated , which would not be possible to derive by gene expression only , and ( ii ) a framework to challenge the system –both in silico and in vivo- upon modification of key factors for cell fate decisions and patterning . Interesting to note is that during a relatively short period of time , 18-30 hpf , a big proportion of epithelial neuroblasts delaminate . This drives a dramatic remodeling of the neurosensory progenitor domain that may lead to changes in the exposure of progenitor domains to sources of signals , such as Shh from the floor plate and the ventral neural tube . This loss of epithelial cells is probably compensated for by the active proliferation in the dorsal non-sensory territory in order to keep the homeostasis of the organ . Finally , we provide the dynamic map of neurosensory progenitors based on in vivo cell lineage studies supplying a global and temporal perspective to previous otic neurosensory lineage analyses , which were mainly focused on the spatial dynamics of gene expression ( Durruthy-Durruthy et al . , 2014 ) . This permits targeting and challenging progenitor pools specifically . Therefore , our findings establish a ground to further explore intrinsic vs . extrinsic models for cell fate determination , and will contribute to the mechanistic understanding of the developmental gene regulatory networks .
Zebrafish embryos were obtained by mating of adult fish using standard methods . All fish strains were maintained individually as inbred lines . All protocols used have been approved by the Institutional Animal Care and Use Ethic Committee ( PRBB–IACUEC ) , and implemented according to national and European regulations . All experiments were carried out in accordance with the principles of the 3Rs . Wild type zebrafish strain was AB/Tu ( RRID:ZIRC_ZL1/RRID:ZIRC_ZL57 ) . Tg[neuroD:GFP] expresses GFP in neuronal progenitors ( Obholzer et al . , 2008 ) and differentiating neuroblasts ( Zecca et al . , 2015 ) , and Tg[Isl3:GFP] ( also called Isl2b ) expresses GFP in afferent sensory neurons of cranial ganglia ( Pittman et al . , 2008 ) . Tg[cldnb:lynGFP] labels the plasma membranes of the otic and anterior lateral line structures ( Haas and Gilmour , 2006 ) and allows to visualize the kinocilium of differentiated hair cells . Tg[Brn3c:GFP] expresses GFP in differentiated hair cells of the ear and lateral line system , coinciding with the onset of differentiation ( Xiao et al . , 2005 ) . Embryos homozygous for the neurog1hi1059 mutation ( Golling et al . , 2002 ) in the Tg[Isl3:GFP] background were obtained by incross of heterozygous carriers ( Sapède et al . , 2012 ) ; the presence of the neurog1hi1059 allele was identified by PCR genotyping fin-clips or embryo tails genomic DNA . Tg[neurog1hi1059/+ Isl3:GFP] fish were crossed and embryos fixed overnight at 4°C in 4% Paraformaldehyde ( PFA ) and washed with 0 . 1% Tween-20 in PBS ( PBST ) at room temperature ( RT ) . To label the plasma membranes , embryos were incubated for 12 hr at RT with Phalloidin-Alexa658 ( Invitrogen , Carlsbad , CA ) diluted 1:20 in PBST containing 1 . 5% Triton X-100 . They were washed with PBST and incubated with DAPI to stain nuclei . Embryos were then used for the analysis of the total number of cells within the otic vesicle ( Figure 5o ) . Embryos were injected at 1cell-stage with translation-blocking morpholino oligomers ( MOs ) obtained from GeneTools , Inc . : neurog1-MO 5’ -ACG ATC TCC ATT GTT GAT AAC CTG G-3’ ( Cornell and Eisen , 2002 ) . 5 ng of MO-neurog1 was injected as previously described ( Sapède et al . , 2012 ) . Morphants display the same phenotype as neurog1hi1059/hi1059 mutants ( Sapède et al . , 2012 ) . For in vivo imaging morphant embryos were used for practical reasons: it is not possible to phenotype the mutants as early as the stage we start the imaging ( 24 hpf ) , and since neurog1hi5109/hi5109 need to be obtained by incross of neurog1hi5109/+ fish because the line cannot be maintained in homozygosis , only 25% of embryos are neurog1hi5109/hi5109 with the inconveniences that this poses for life imaging . For mRNA expression , capped H2B-mcherry , H2B-cerulean ( Olivier et al . , 2010 ) or lyn-TdTomato mRNAs were synthetized with mMessage mMachine kit ( Ambion ) . Embryos were injected at 1cell-stage and let to develop until the desired stages . Stained fixed samples were mounted in 1% Low Melting Point ( LMP ) -agarose on glass-bottom Petri dishes ( Mattek ) and imaged on a Leica SP5 inverted confocal microscope using a 20x objective ( NA 0 . 7 ) . Embryos were anesthetized in 0 . 17 mg/ml tricaine in system water and mounted in 0 . 75% LMP-agarose in glass capillaries size 2 ( volume 20 µl , BRAND GMBH ) . Time-lapse imaging was performed at 26 . 5°C ( to avoid melting of LMP-agarose ) on a Zeiss Lightsheet Z . 1 microscope using a 20x or a 40x objective and the developmental stage was corrected accordingly ( at 26 . 5°C development is delayed about 0 . 7 fold ) . Nuclei , plasma membranes and cell fate were recorded simultaneously for the entire system ( Figure 1a , Figure 1—figure supplement 1 ) . The cohort of embryos and datasets used in this study are depicted in Tables 1 and 2 . Each dataset corresponds to the imaging of a distinct embryo inner ear , except for 140402aX in which both ears were imaged . Note that we have performed deep and detailed analyses in few datasets , and supported the observations and conclusions with partial analyses of additional datasets mainly included in supplementary figures . | Our ears , eyes and other sensory organs collect information about the world around us . In the inner ear – which is responsible for balance and hearing – specialized cells known as hair cells detect sounds and body position . This information is passed on to other cells called sensory neurons , which relay the information to the brain . Both of these cell types originate from a pool of cells known as “progenitors” located in a structure called the otic vesicle within the embryo , where the progenitor cells follow different sets of instructions to make hair cells or sensory neurons . Although we have identified many of the genes that are important for setting these instructions , it is not known how the progenitor cells behave in the inner ear or how they follow these guidelines . Dyballa et al . used high-resolution imaging to reconstruct the histories of individual hair cells and sensory neurons in the inner ear of zebrafish embryos . The experiments combined techniques that allowed individual cells to be tracked over time and showed what types of cell they developed into . Dyballa et al . used these data to develop a map of the progenitor cells in the whole otic vesicle . The findings help to fill the void in our understanding of the link between gene activity and tissue architecture in the inner ear . The next challenge is to use these findings as a basis to further explore models for how sensory neurons and hair cells develop , and to understand how to overcome the inner ear degenerati as we age . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2017 | Distribution of neurosensory progenitor pools during inner ear morphogenesis unveiled by cell lineage reconstruction |
Cell size uniformity in healthy tissues suggests that control mechanisms might coordinate cell growth and division . We derived a method to assay whether cellular growth rates depend on cell size , by monitoring how variance in size changes as cells grow . Our data revealed that , twice during the cell cycle , growth rates are selectively increased in small cells and reduced in large cells , ensuring cell size uniformity . This regulation was also observed directly by monitoring nuclear growth in live cells . We also detected cell-size-dependent adjustments of G1 length , which further reduce variability . Combining our assays with chemical/genetic perturbations confirmed that cells employ two strategies , adjusting both cell cycle length and growth rate , to maintain the appropriate size . Additionally , although Rb signaling is not required for these regulatory behaviors , perturbing Cdk4 activity still influences cell size , suggesting that the Cdk4 pathway may play a role in designating the cell’s target size .
Uniformity of cell size is a consistent feature of healthy tissues . While different cell types can differ greatly in size , cells within a given tissue tend to be strikingly similar ( Ginzberg et al . , 2015 ) . In fact , in some tissues , loss of cell size uniformity is a diagnostic marker of malignancy ( Greenough , 1925 ) . Such observations raise the intriguing question of whether there are dedicated mechanisms that restrict cell size to a specific range . Is size uniformity the product of cellular processes that monitor cell size and correct deviations from a target size ( Ginzberg et al . , 2015; Lloyd , 2013 ) ? Studies of yeast have long postulated the existence of cell-autonomous size control mechanisms . Evidence of cell size checkpoints in the cell cycle of both budding yeast and fission yeast was first reported in 1977 ( Johnston et al . , 1977; Fantes and Nurse , 1977 ) . In both species , it was found that the length of the cell cycle is selectively extended in small cells , giving them more time to grow before their next division . Two classes of models have been proposed to explain the biophysical basis of size sensing in yeast ( Wood and Nurse , 2015 ) . Geometric models propose that cells measure dimensions such as cell length or surface area by relying on highly specific intracellular localizations of cell-cycle regulators such as pom1 or cdr2 ( Martin and Berthelot-Grosjean , 2009; Bhatia et al . , 2014; Pan et al . , 2014 ) . Alternatively , titration-based models suggest that the concentration of a cell cycle activator overcomes the concentration of a cell cycle inhibitor in a growth-dependent manner ( Turner et al . , 2012; Schmoller et al . , 2015 ) . Several studies have suggested that the cell cycle of animal cells , like that of yeast , also includes cell size checkpoints ( Killander and Zetterberg , 1965a; Dolznig et al . , 2004; Gao and Raff , 1997 ) . In 1965 , Zetterberg and Killander reported that proliferating fibroblasts that have recently entered S-phase are more uniform in size than sister cells that have just exited mitosis ( Killander and Zetterberg , 1965a ) . Based on these and other observations , they concluded that cells that are smaller than a critical size are maintained in G1 for longer periods of growth ( Killander and Zetterberg , 1965b ) . Subsequent studies have also suggested that size-sensing mechanisms selectively promote S-phase entry in large but not small cells ( Dolznig et al . , 2004; Gao and Raff , 1997 ) . The observation that cell cycle progression is cell-size-dependent raises the question: at which point in the cell cycle does size sensing occur ? A defining event in the eukaryotic cell cycle is the ‘restriction point’ , the time early in G1 when cells commit to undergoing another cell division cycle if growth factors are present ( or transition to a quiescent G0 state in the absence of growth factors ) ( Pardee , 1974 ) . In budding yeast , some studies suggest that the restriction point ( START ) and the cell size checkpoint are one and the same ( Cross , 1995 ) . In animal cells , reports of size-dependent S-phase entry have suggested that the observed cell size checkpoint may operate later in G1 , after the restriction point , leaving this an open question ( Foster et al . , 2010; Zetterberg and Larsson , 1991 ) . Despite the studies described above , research addressing the question of whether animal cells , like yeast , autonomously sense and regulate their own size remains inconclusive ( Lloyd , 2013 ) . While the work of Zetterberg and Killander focused on the heterogeneous behavior of cells sharing a common extracellular environment , much of the research on animal cell growth focused on extracellular factors that elicit large changes in cell size . These studies , which highlighted pathways controlling cell growth and cell cycle progression that could be differentially activated , led many to conclude that animal cell growth and cell cycle progression are independent processes ( Conlon et al . , 2001; Echave et al . , 2007 ) . Since the 1960s , two alternate models have been proposed to explain cell size homeostasis , the adder model and the sizer model ( Conlon and Raff , 2003 ) . According to the adder model , size homeostasis is not the result of size-sensing mechanisms . Instead , size homeostasis is the outcome of a balance between a constant amount of mass that cells accumulate each cell cycle and the reduction in cell mass that accompanies cell division . At the core of the adder model is the assumption that small and large cells accumulate the same amount of mass over the course of the cell cycle . Since large cells lose a greater amount of mass upon division ( e . g . half of a large cell is more than half of a small cell ) , size variation is constrained . In contrast to the adder model , the sizer model assumes that size homeostasis is the product of size-sensing mechanisms that selectively restrict the growth of large cells or promote the growth of small cells . As the studies mentioned above illustrate , the extent to which the sizer model and adder model describe size homeostasis of animal cells remains unresolved ( Lloyd , 2013 ) . Furthermore , almost all literature on cell size homeostasis , whether supporting the sizer model or the adder model , is confined to the context of cells that are actively proliferating . In the adder model , size homeostasis is a consequence of cell divisions . In discussions of the sizer model , research has almost exclusively investigated the existence of cell size checkpoints , that is processes that inhibit cell cycle progression for cells that are smaller than a target size ( Wood and Nurse , 2015 ) . This focus on cycling cells is puzzling since most cells in an animal body are terminally differentiated , do not undergo cell cycles , yet still display size uniformity . In this light , it is compelling to consider the possibility that at least some mechanisms of size homeostasis are not dependent on cell size influencing cell cycle progression , but instead involve an effect of cell size on the rate of cell growth . One reason questions about cell size control are difficult to answer is that , while it is easy to envision experimental assays of cell size , assays of size sensing are more challenging to conceptualize . In this study , we describe new experimental approaches to assay size sensing by monitoring cell size variance . To resolve the ambiguity inherent in our previous approach to this question ( Kafri et al . , 2013 ) , we also develop methods to separately assay the influence of cell size on cell cycle progression and on growth rate . With these approaches , we determine that animal cells monitor their own size and correct deviations in cell size . Our results show that , like yeast , animal cells that are smaller than their target size spend longer periods of growth in G1 . Surprisingly , however , we found that in addition to a G1-length extension in small cells , animal cells also employ a conceptually different strategy of size correction . During two distinct points in the cell cycle , anabolic growth rates are transiently adjusted so that small cells grow faster and large cells grow slower . These periods of growth rate adjustment function to lower cell size variability , promoting size uniformity . While cellular growth rates have previously been observed to vary with cell cycle stage ( Kafri et al . , 2013; Goranov et al . , 2009; Son et al . , 2012 ) , to our knowledge such a reciprocal coordination of growth rate with cell size has not been previously observed in any organism . Throughout nature , negative feedback circuits maintain traits and quantities within their appropriate range . Properties such as body temperature and body plan proportions are specified by processes that sense and correct deviations from a target value . The results presented here suggest that the size homeostasis of animal cells is actively maintained by cell-autonomous negative feedback mechanisms that sense and correct aberrations in cell size by adjusting both cell cycle length and growth rate .
Previous literature has defined cell size as cell volume ( Cadart et al . , 2017; Tzur et al . , 2009 ) or cell mass ( Conlon and Raff , 1999 ) . In this study , we define cell size by a cell’s total macromolecular protein mass , as this metric most closely reflects the sum of anabolic processes typically associated with cell growth ( Mitchison , 2003 ) and with activity in growth-promoting pathways such as mTORC1 ( Laplante and Sabatini , 2012 ) . In contrast , cell volume is a more labile phenotype , sensitive to ion channel regulation and fluctuations in extracellular osmolarity . In S . cerevisiae , cell size is thought to be regulated by a cell size checkpoint in G1 ( Johnston et al . , 1977; Amodeo and Skotheim , 2016 ) . To test whether a similar size checkpoint functions in animal cells , we made single-cell measurements of cell size , cell cycle stage , and cell age ( i . e . time elapsed since last division ) , in order to investigate the relationship between these properties . We used time-lapse microscopy to image live HeLa and Rpe1 cells for a period of 1–3 days . At the end of the imaging session , cells were immediately fixed and stained with AlexaFluor 647-Succinimidyl Ester ( SE-A647 ) , a quantitative protein stain that we previously established as a an accurate measure of cell mass ( Kafri et al . , 2013 ) . This experiment provides three measured properties for each cell . Firstly , movies recorded by time-lapse microscopy reveal the amount of time elapsed between each cell’s ‘birth’ via mitosis and its fixation , which we refer to as the cell’s ‘age’ at the time of fixation . Secondly , staining with SE-A647 reveals each cell’s size at the time of fixation . SE-A647 staining is sensitive enough to detect the change in cell mass that occurs in less than three hours of growth ( less than 15% of the cell cycle ) ( Figure 1—figure supplement 1 ) . Thirdly , using florescent cell cycle indicators , we determine each cell’s cell cycle stage at the time of fixation . To quantify ‘cell age’ from the recorded time-lapse movies , we customized computational methods to identify cell boundaries , track cell motion , and monitor thousands of individual cells throughout the course of their cell cycle . To determine cell cycle stage , we used cell lines stably expressing mAG-hGem , a fluorescent reporter of APC activation and G1 exit ( Sakaue-Sawano et al . , 2008 ) . Thus , our experimental design independently quantifies age , size , and cell cycle stage for each cell . This experimental approach allowed us to monitor how the mean cell size changes , as a function of both age and cell cycle stage . A priori , we considered two alternative outcomes , which are depicted in Figure 1A and B . If S phase entry is not regulated by cell size checkpoints ( Figure 1A ) , a cell’s size should exclusively reflect the amount of time that it has been growing ( i . e . its age ) , but not whether it has passed the G1/S transition . In contrast , if the transition from G1 to S phase is selectively accessible only to cells that have exceeded a critical size threshold ( Figure 1B ) , we would expect S-phase cells to be larger on average than cells in G1 , even when comparing G1 cells and S-phase cells of identical ages . The results of this experiment are shown in Figure 1C and D . Figure 1C shows the average size of HeLa cells as a function cell age . The transient slowing of cell growth about 10 hr after birth ( the average age of S-phase entry ) is consistent with a behavior we and others have previously identified ( Kafri et al . , 2013; Gut et al . , 2015 ) . As expected , the average size of the oldest cells is approximately double that of the youngest . Additionally , comparing the size distribution of the youngest cells ( first 1 . 5 hr after birth ) to that of mitotics ( identified by their rounded shape using phase-contrast microscopy ) similarly shows a doubling in cell size over the course of the cell cycle , as expected ( Figure 1—figure supplement 2 ) . The size distribution of mitotics is identical to that of the oldest cells tracked , confirming that the data in Figure 1C represents growth over the course of the entire cell cycle . In Figure 1D , cell size vs . age is plotted separately for G1 and post-G1 subpopulations . Consistent with the existence of a cell size checkpoint ( Figure 1B ) , Figure 1D shows that S-phase cells are significantly larger than G1 cells of the same age , even though both have been growing for the same amount of time since birth . This suggests that cells exit G1 in a size-dependent manner . Furthermore , the mean size of G1 cells plateaus as cells begins to enter S-phase , consistent with a mechanism where cells enter S-phase upon reaching a particular size threshold . The size difference between identically aged G1 and S-phase cells is statistically significant ( student’s t-test p<3 . 92e-04 ) and was observed in three additional independent replicate experiments ( p<0 . 009 , p<5 . 80e-08 , p<9 . 07e-15 ) . If cells enter S-phase only after attaining a critical size , cells that are born small are expected to have longer periods of growth in G1 . To test this prediction , we assayed the correlation of cell size at birth with a direct measurement of G1-length in single cells . Monitoring the size of live cells is not possible with the SE-A647 staining technique . Instead , we imaged cells with time lapse microscopy and monitored the size of their nuclei ( estimated as the area covered by the nucleus in a widefield image ) , as a proxy for cell size ( Figure 2—figure supplement 1 ) . In yeast , the nucleus is known to grow continuously throughout all stages of cell cycle and is correlated with cell size ( Jorgensen et al . , 2007 ) . To test whether this is also the case in our experimental system , we measured the correlation between nucleus size and cell size ( i . e . SE-A647 intensity ) in an unperturbed population of HeLa cells ( Figure 2A ) . As expected , this correlation is significant ( Pearson’s r = 0 . 68 ) . Figure 2B and C show that both cell size and nucleus size continually increase throughout the cell cycle . Figure 2C also highlights that our measurements ( both SE-A647 and nucleus size ) are sufficiently accurate to resolve the small size increases that accumulate in less than 3 hr of growth ( <15% of doubling time ) . Additionally , the correlation of nuclear size with cell size is strong enough that a plot showing mean nuclear size vs . cell age reproduces the features of the mean cell size vs . cell age curve , including the transient slowing of cell growth around S-phase entry ( Figure 2—figure supplement 2 ) . Last , we confirmed that growth in nucleus size is independent of DNA synthesis , by treating cells with aphidicolin to prevent DNA replication . Aphidicolin-arrested cells continued to grow in both cell size and nucleus size ( Figure 2—figure supplement 3 ) , maintaining a correlation between the two . When we monitored nuclear growth in single cells expressing nuclear-localized cell cycle markers , we observed a significant and reproducible negative correlation between the size of the nucleus at birth and G1 duration ( Figure 2D ) . Hela cells that had smaller nuclei at birth spent longer amounts of time in G1 . This result , along with the results presented in Figure 1 , indicates that cells that are smaller at birth compensate with lengthened periods of growth in G1 and is consistent with the model of a cell size threshold at G1 exit . This finding was also independently confirmed in Rpe1 cells , in a separate study in our labs ( Liu et al . , 2018 ) . While the correlation of nucleus area and G1 length is statistically significant ( p<6 . 7×10−8 , Figure 2D ) and reproducible ( N > 4 ) , it is not a strong correlation ( Pearson’s r = −0 . 41 ) . One reason for this may be that our measurement of nucleus size is not a perfect correlate of cell size . Also , as we show later in this study , G1 length is not the only means by which cells correct their size . As is elaborated in the discussion , correction of fluctuations in cell size by mechanisms other than cell cycle checkpoints would tend to loosen the demand on the G1-length extension mechanism and weaken the correlation between cell size and G1 length . To further test whether information about cell size is communicated to the cell cycle machinery , we examined the effect of slowing down cellular growth rates on the length of G1 . If cells leave G1 only when a particular size has been reached , slowing down their growth rate would be expected to prolong G1 , as cells would require more time to reach the size threshold . Growth rate can be slowed down with drugs such as the mTORC1 inhibitor rapamycin . Culturing Rpe1 cells in 70 nM rapamycin for several days caused a 60% decrease in the average growth rate . As predicted , we observed an 80% increase in the average G1 duration of these cells . Because of the lengthened cell cycle , the profound effect of rapamycin on growth rate caused only a 20% reduction in cell size ( Figure 3A–E ) . Previous studies have observed that rapamycin influences both growth rate and G1 length ( Hashemolhosseini et al . , 1998; Wiederrecht et al . , 1995; Brown et al . , 1994 ) and sometimes ascribed those effects to two distinct functions of the mTOR pathway ( Fingar et al . , 2004 ) . Our current results suggest that the influence of rapamycin on G1 length may be an indirect consequence of its influence on growth rate . According to this interpretation , the inhibition of cell growth by rapamycin results in a gradual reduction of cell size which , in turn , triggers compensatory increases in G1 length . This interpretation is consistent with the evidence of cell-size-dependent regulation of G1 length that we observe , even in the absence of rapamycin ( Figure 1D and Figure 2D ) . Additional support for this interpretation ( as well as analogous results in HeLa and other cell lines ) will be provided in later sections of this study that examine the timing of rapamycin’s effects on cellular growth rate , cell size , and cell cycle progression . In proliferating cells , cell size is the product of growth duration ( cell cycle length ) and growth rate , s=τ×v . Therefore , cells could potentially correct aberrations in size not only by modulating the amount of time ( τ ) that they grow , by adjusting G1 duration , but also by modulating how fast they grow ( v ) . To investigate this possibility , we examined the relationship between cell size and growth rate . If a cell can sense its own size and adjust its growth rate accordingly – just as a thermostat coordinates heat production with room temperature via short bursts of heat – we might expect these corrections to be transient and subtle . Experimental detection of transient changes in growth rates of single cells has proven to be challenging . To circumvent this difficulty , we derived an indirect inference method to assay whether growth rates of individual cells are dependent on their size . Using the coupled measurements of cell size and cell age described earlier ( Figure 1C ) , we calculated the variance in cell size as a function of cell age ( i . e . time since division ) . In the absence of any size-dependent growth rate regulation , the variance in cell size is expected to increase as cells grow . The reason for this expectation is that individual cells in a population will have some variability in their growth rates , with some cells growing slightly faster than others . The consequence of this cell-to-cell variability is that , in any given time interval , fast-growing cells will accumulate more mass than slow-growing cells , thereby increasing disparities in cell size ( Figure 4A ) . In fact , in a synchronized population of growing cells , variance in size can only decrease if small cells grow faster than large cells ( Figure 4B ) . To see that this is always true , consider a time interval during which cells grow from S1 to S2 . Cell size variance at any given time t2 is related to the cell size variance at any previous time interval , t1 , by: ( 1 ) Var ( S2 ) =Var ( S1 ) +Var ( ΔS ) +2Cov ( S1 , ΔS ) where ΔS is the change in size during the interval , that is the growth rate , and Cov ( S1 , ΔS ) is the covariance between initial cell size and growth rate . Therefore , the change in size variance follows: ( 2 ) ΔVar ( S ) =Var ( S2 ) −Var ( S1 ) =Var ( ΔS ) +2Cov ( S1 , ΔS ) where ∆S is the change in cell size accruing over the time period ∆t , that is the growth rate . Since variance is always positive ( by definition ) , cell size variance can decrease with time ( ΔVar ( S ) <0 ) if , and only if , the correlation of cell size and growth rate is negative , that is Cov ( S1 , ΔS ) <0 . This mathematical relationship can be exploited as a method of data analysis . It can be used to detect periods of size-dependent growth rate regulation by analyzing measurements of cell size , without the need to directly measure growth rates . Figure 4 shows that periods of decreasing cell size variance do , in fact , occur during the cell cycle of unperturbed HeLa and Rpe1 cells ( Figure 4C–F ) . During these periods , cells grow ( mean size increases ) , and yet become more similar in size . This indicates that cellular growth rates are regulated to correct aberrant cell sizes . To quantify the strength of this regulation , we normalized the change in size variance , ∆Var ( S ) =VarS2-Var ( S1 ) , by the amount of growth that has occurred ( i . e . the change in mean size , S- ) to define the coefficient of growth rate variation , designated Gcv . ( 3 ) Gcv={ΔVar ( S ) S−2−S−1ΔVar ( S ) >0−|ΔVar ( S ) |S−2−S−1ΔVar ( S ) <0 The Gcv-value analysis is a new method to interrogate size control in growing cells . As long as the mean cell size is increasing over time , Gcv can be interpreted as follows . If a cell’s growth rate is independent of its size , Gcv directly equals the coefficient of variation ( CV ) of cellular growth rates , Gcv=Var ( ΔS ) S−2−S−1=σgrμgr . This can be shown by substituting the relationship in Equation 2 into the formula for Gcv ( Equation 3 ) , noting that if growth rates are size-independent , ΔVar ( S ) >0 and CovS1 , ΔS=0 . Negative values of Gcv ( arising from dips in variance ) imply that growth rate is actually not independent of size , and that cell size and subsequent growth rate are negatively correlated . We can also note that if Gcv is much higher than is plausible for the CV of growth rates ( which will equal 1 if growth is a Poisson process ) , it is likely that growth is positively correlated with size , as in the case of exponential growth . Plotting Gcv versus cell age consistently reveals two distinct periods during which cell size and growth rate are negatively correlated ( i . e . Gcv is negative ) , in both HeLa and Rpe1 cells ( Figure 4G , H ) . While the exact timing of the drops in Gcv varied between experiments ( and a third , weaker drop is occasionally observed in HeLa cells ) , the observation of at least two periods during which Gcv is negative is highly reproducible . This result is striking because it suggests communication between cell size and cellular growth rate that is transiently established twice during the cell cycle , perhaps due to cell-cycle-dependent signalling linking size and growth rates . The analysis presented in Figure 4 does not discriminate G1 cells from S phase cells and , consequently , the observed drops in variance cannot be explained by the cell-size checkpoint gating G1 ( Figure 4—figure supplement 1 ) . Therefore , Figure 4 clearly reveals two distinct times in the cell cycle , where the growth rate of individual cells is selectively repressed in large cells or accelerated in small cells – increasing uniformity in cell size . The only alternative possibility is that cells are removed from the distribution , by either dividing or dying in a size-dependent manner . This is very unlikely to be the case . The dips below zero occur earlier than cells start dividing . ( This is true for at least the first dip in Rpe1 and both dips in HeLa , where the mean cell cycle length is 21 hrs , with a standard deviation of 1 . 6 hrs . A sample cell cycle length distribution is shown in Figure 4—figure supplement 2 ) . Furthermore , there was a very low rate of apoptosis in our experiments ( <1% of cells imaged from birth died before dividing ) , so the negative Gcv is not explained by cell death . To test the conclusions of the Gcv-value analysis , we asked whether the correlation of growth rate and cell size could be observed directly in live cells . Using time-lapse microscopy , we monitored nuclear growth in hundreds of live cells over several days . Comparing growth trajectories collected from the largest and smallest cells in the population provided a dramatic demonstration that growth rates are , indeed , reciprocally coordinated with cell size ( Figure 4I ) . Because growth rates became slower in large cells and faster in small cells , individual cells that were initially quite different in size gradually become more uniform in size , explaining the decreases in cell size variance shown in Figure 4E–H . Furthermore , as predicted by the Gcv-value analysis , the negative correlation of cell size and subsequent growth rate was not continuously present , but was transiently established twice during the cell cycle ( Figure 4J and Figure 4—figure supplement 3 ) . The Gcv-value analysis and live-cell tracking all indicate that cellular growth rates are regulated in a manner that reduces cell size variability . Taken together , our results suggest that cells employ two separate strategies to correct deviations from their appropriate size: ( 1 ) small cells spend more time in G1 and ( 2 ) small cells grow faster than large cells ( Figure 5 ) . A prediction of this dual-mechanism model is that perturbations that lengthen the cell cycle would be counteracted by a compensatory decrease in growth rate , allowing cells to accumulate the same amount of mass despite the longer periods of growth . Conversely , perturbations that reduce growth rate would be counteracted by a compensatory lengthening of G1 . As an initial test of this prediction , we treated Rpe1 cells with two well-characterized pharmacological inhibitors . To experimentally lengthen the cell cycle , we treated cells with SNS-032 , a potent inhibitor of Cdk2 . To experimentally decrease growth rates , we used rapamycin , an inhibitor of mTORC1 . We optimized working concentrations of both drugs to influence rates of cell cycle progression or growth without causing cell cycle arrest . We then applied the drugs to unsynchronized populations of cells and monitored cell count and average cell size over the course of three days in the presence of the drugs ( Figure 6A ) . We fixed samples at various timepoints after the drugs were applied and measured cell size ( SE-A647 intensity ) and cell count , obtaining a time series of cell size measurements and a corresponding time series of cell count measurements . Cell count was also independently monitored in live samples in each condition , using differential phase-contrast microscopy . Figure 6A ( top panel ) shows the influence of the growth rate inhibitor ( rapamycin ) and the cell cycle inhibitor ( SNS-032 ) on Rpe1 cell size . While SNS-032 treatment caused an increase in average cell size , rapamycin treatment caused a decrease in average cell size . This result is expected since SNS-032 causes cells to grow for longer periods of time before division and rapamycin causes cells to grow slower . For both drugs , the change in average cell size proceeded only during the initial phase of drug treatment , followed by stabilization of a new cell size that no longer changed with time . Note that the cell size data shown in Figure 6A represent population averages . In control populations , while single cells grow and divide , the average cell size remains constant and does not change with time . In drug-treated populations , we observed a gradual change in average cell size , followed by a plateau ( Figure 6A ) . To quantify this trend , we separately considered the early and late stages of the drug treatments . We define the early stage of each treatment as the time interval between drug addition and cell size stabilization ( approximately one day ) , and we define the late stage as the time interval between cell size stabilization and the experiment’s end . For the early stage of drug treatment , we used linear regression to characterize cell size as a function of time . For the late stage of drug treatment , average cell size was quantified from the mean ( linear regression with a slope of 0 ) . Figure 6A ( bottom panel ) shows the influence of SNS-032 and rapamycin on rates of cell division . As expected , the Cdk2 inhibitor SNS-032 caused an immediate decrease in proliferation rate . Notably , while rapamycin also slowed proliferation rate , this influence only occurred during the late stage of drug treatment . In contrast , rapamycin treatment induced an immediate effect on cell size . From the measurements of average cell size and cell count over time , two parameters were derived for each sample: average cell cycle length ( τ ) and average cellular growth rate ( v ) . To quantify cell cycle length , τ , measurements of cell count over time were fitted to exponential curves , Nt=N0eαt where Nt is cell count at time t and α=ln ( 2 ) τ . In all cases , fits were constructed with two independent exponential regressions , one estimating the average cell cycle length during the early stage of drug treatment ( Figure 6B , early time interval ) and the second estimating cell cycle length during the late stage of drug treatment ( Figure 6B , late time interval ) . This resulted in two estimates of cell cycle length for each drug treatment ( τearly , τlate ) as shown in Figure 6B–F . To estimate average growth rate , we relied on the trends of cell count over time and average cell size over time ( Figure 6A ) . Specifically , we calculated the rate of increase of the total population’s bulk mass and divided that by the cell count . To understand this calculation , consider a simple case where the average cell size does not change over time , as is the case in the control populations that are not treated with drugs . In such populations , even though average cell size does not change with time , bulk biomass ( Mt=cell count × cell size ) does increase , because of the expansion in cell number . To estimate the amount of mass that a single cell accumulates over a short time interval ( growth rate ) , we take the bulk amount of mass ( Mt ) that is accumulated by the total population during that interval and divide it by the number of cells in the population . In the general case , we calculate average growth rate with: ( 4 ) v=1NtdMtdtwhere Mt is given by Mt=Nt×St- and Nt and St- represent cell count and average cell size at time t . As with the calculation of cell cycle lengths , growth rate values were separately calculated for the early and late stages of drug treatment ( Figure 6B , early time interval and late time interval ) , resulting in two separate growth rate values per drug treatment ( vearly , vlate ) . Figure 6B–F show that during the early stage of drug treatment , SNS-032 primarily influences the length of cell cycle , τ , while rapamycin primarily influences the rate of cell growth , v . However , later in the drug treatment and after a change in cell size is observed , a coordination of growth rate and cell cycle length becomes apparent . Specifically , while SNS-032 caused an immediate lengthening of cell cycle , its influence on growth rate was observed only after prolonged drug treatment ( Figure 6B , E and F ) . This suggests that the influence of cdk2 inhibitor SNS-032 on growth rate is indirect and is mediated by a property that accumulates over time , presumable cell size . Conversely , rapamycin caused an immediate reduction in growth rate , while its influence on cell cycle length occurred much later ( Figure 6B–D ) . This delayed effect of rapamycin on cell cycle length is directly apparent from the abrupt change in the slope of the proliferation curves of rapamycin treated cells ( Figure 6A , bottom panel , rapamycin treatments ) . The sharp contrast between the temporal order of events that follow rapamycin and SNS-032 treatment is illustrated in Figure 6C–F , where growth rate is plotted as a function of cell cycle length for each time interval . The curves overlaid on Figure 6C–F delineate paired values of growth rate and cell cycle length that correspond to a fixed product , τ×v=constant . Data points that fall close to a given curve represent conditions that , while different in growth rate and cell cycle length , will yield cells of similar size . Comparing the measurements shown in Figure 6C–F to these iso-size curves shows that the compensatory adjustments of cell cycle length or growth rate seen during the late stage of rapamycin and SNS-032 treatments , respectively , counteract the initial effects of the drugs and return the product τ×v to its homeostatic value . This explains the relatively small influence of these drugs on cell size ( depending on drug dose , size decreases by 20–30% in rapamycin , and increases by 25–50% in SNS-032 ) , compared to their effects on growth rate and cell cycle length . A quantitative prediction of the dual-mechanism model shown in Figure 5 is that growth rate , v , and cell cycle length , τ , are coordinated to maintain cell size at its fixed target size , τ×v=target size . This means that growth rate and cell cycle duration are related by: ( 5 ) vi=CTS τiwhere vi and τi represent the growth rate and cell cycle length in experimental condition i , while CTS is a constant that defines the cell’s target size . To test the generality of this prediction , we repeated the experiment described in Figure 6 using a variety of cdk inhibitors to slow cell cycle progression , as well as several inhibitors of cell growth . Each drug was used at several different concentrations where cells maintain viability and proliferation . We used a simplified version of the procedure described above for rapamycin and SNS-032 ( without separating the time series into early and late stages ) to calculate the average growth rate and cell cycle length of Rpe1cells during a three-day incubation in each drug . Figure 7A and B show that most chemical perturbations of growth rate and cell cycle length had surprisingly small influences on Rpe1 cell size . Whiles drugs produced up to 4-fold changes in both growth rate and cell cycle length , the largest changes in cell size were close to 1-fold ( i . e . 50% change in cell size ) , with most treatments changing size by less than 30% . Consistent with the dual-mechanism model , treatment with all but one ( palbociclib , Figure 8A ) of the tested compounds resulted in coordinated changes of both cell cycle length and growth rate , such that cell size remained relatively unchanged . Also , consistent with the results in Figure 6 , growth rate inhibitors caused a small decrease in cell size ( Figure 7A , rightmost panel ) while cell cycle inhibitors caused a small increase in cell size ( Figure 7B , rightmost panel ) . Lastly , separate analysis of early and late stages of the drug treatments confirmed that inhibitors of cell cycle regulators cause immediate lengthening of the cell cycle followed by compensatory decreases in growth rate , while cycloheximide , torin-2 , and rapamycin induce immediate decreases in growth rate followed by compensatory increases in cell cycle length ( Figure 7—figure supplement 1 ) . Of the tested cdk inhibitors , only the cdk4/6 inhibitor palbociclib increased cell cycle length without triggering a compensatory decrease in growth rate , causing an unusual increase in cell size ( Figure 8A , B ) . Knockdown of cdk4 and cdk6 by siRNA yielded a similar increase in cell size ( Figure 8—figure supplement 1 ) . This result shows that , while the coordination of growth rate and cell cycle length is robust , it is dependent on mechanisms that can be perturbed or reprogrammed . In an attempt to further disrupt the processes that maintain cell size uniformity , we treated Rpe1 cells with the HSP90 inhibitor radicicol . HSP90 is known to suppress phenotypic variability ( Hsieh et al . , 2013 ) and has also been found to regulate both cell cycle progression ( Mollapour et al . , 2010; Bandura et al . , 2013 ) and mTORC1 mediated growth in response to cellular stress ( Conn and Qian , 2011 ) . Remarkably , culturing cells in radicicol not only decoupled growth rate and cell cycle length ( Figure 8A , B ) , but also increased cell size variability by 26% ( ±3 . 2% ) relative to untreated cells ( see Materials and methods ) . In contrast , cdk4/6 inhibition with palbociclib had no significant influence on cell size heterogeneity . This finding further illustrates that cell size uniformity is a regulated phenotype that can be perturbed by disrupting the coordination of growth and cell cycle progression . Figure 9A shows that the reciprocal influence of the various drug treatments on growth rate and cell cycle length in Rpe1 cells match the trend predicted by Equation 5 . Changes in growth rate , v , and cell length , τ , are not only reciprocally related , but also quantitatively coordinated such that the product , v×τ , is relatively constant . To test the generality of the compensation mechanism , these measurements were repeated , in four additional cell lines: U2OS ( Figure 9B ) , SAOS2 ( Figure 9C ) , HeLa ( Figure 9D ) and 16HBE ( Figure 9E ) . Because cell size is sensitive to perturbation of Cdk4 ( Figure 8A , B ) , we deliberately included cell lines that lack Rb signaling ( Rb-null SAOS2 , as well as HeLa and 16HBE which are expected to have disrupted Rb activity [Landry et al . , 2013; Ahuja et al . , 2005] ) alongside cell lines with intact Rb signaling ( Rpe1 and U2OS ) . Since the cell lines are differentially sensitive to the various drugs , drug concentrations were optimized separately for each cell line to find doses that cause a detectable change in cell cycle length and/or growth rate but don’t cause cell death or cell cycle arrest . The conditions shown in Figure 7 are represented by the gray circles in Figure 9A . For Figure 9B–E , the drug concentrations used , along with their effects on growth rate , cell cycle length , and cell size are shown in Figure 9—figure supplement 1–4 . As expected , palbociclib caused large increases in cell size in Rpe1 and U2OS cells , but not in the Rb-inactive cell lines . ( Note that , since drug doses were optimized to cause an effect , the palbociclib doses shown in Figure 9C and E are high ( 2–4 µM , Figure 9—figure supplements 1 , 2 ) , so the response of 16HBE cells may be due to an off-target effect . The Rb-inactive cell lines were completely insensitive to the palbociclib doses ( 60–500 nM ) used on Rpe1 and U2OS cells . ) Aside from the differences in palbociclib sensitivity , however , Figure 9A–E show very similar trends . The intact size compensation in Rb-inactive cell lines suggests that , while the cyclin D-Rb axis plays a role in specifying the target cell size , this pathway is not responsible for the coordination of growth rate with cell cycle length to maintain cell size uniformity . An interesting finding derived from Figures 7–9 is the difference in the way cell size is affected by perturbations of Cdk1/2 and Cdk4/6 . Inhibitors of Cdk1/2 and inhibitors of Cdk4/6 both cause an increase in the duration of cell cycle . However , this lengthening of cell cycle is compensated by decreased growth rates in the case of Cdk1/2 inhibitors , but not in the case of the Cdk4/6 inhibitor , palbociclib ( Figures 7–9 ) . In classical descriptions , Cdk2 and Cdk4 partner with cyclin E and cyclin D , respectively , to promote the G1/S transition . To explore whether cyclin E and cyclin D play divergent roles in the regulation of cell size , we generated two stable cell lines that overexpress either cyclin E1 or cyclin D1 upon induction with doxycycline . As a positive control , we also generated a cell line with doxycycline-inducible overexpression of p27 , which is expected to lengthen ( rather than shorten ) the cell cycle . Inducible expression was validated by western blotting ( Figure 10—figure supplement 1 ) . As expected , overexpression of both cyclin D and cyclin E resulted in higher proliferation rates ( i . e . shorter cell cycles ) , while overexpression of p27 resulted in slower proliferation rates ( longer cell cycles ) as compared to control ( Figure 10A , B , C ) . Curiously , despite the similar influence of cyclin E and cyclin D overexpression on cell cycle lengths ( Figure 10A , B ) , these genetic perturbations had markedly different influences on cell size ( Figure 10D , E ) . While overexpression of cyclin E shortened the cell cycle , the size distribution of cells overexpressing cyclin E was indistinguishable from control cells ( Figure 10A , D ) , due to a compensatory increase in growth rate ( Figure 9A ) . Similarly , while overexpression of p27 lengthened the duration of cell cycle , cells overexpressing p27 were only minimally different in size ( Figure 10C , F ) . In contrast , overexpression of cyclin D caused a shortening of cell cycle that was accompanied by dramatic reduction in cell size ( Figure 10B , E ) . The ability of cells to compensate for genetic perturbations of cyclin E but not for perturbations of cyclin D is consistent with our pharmacological perturbations and suggests a distinct role for cyclin D/Cdk4 in target size regulation . ( Note that this result differs from that obtained in drosophila , where cyclin D/Cdk4 overexpression did not affect proliferating cell size ( Datar et al . , 2000 ) . )
To date , the subject of cell size uniformity has remained largely unexplored . In this study , we use multiple experimental approaches ( Figures 1–4 ) to show that cells that have escaped their appropriate size range are corrected by two separate mechanisms: regulation of G1 length and regulation of cellular growth rate ( Figure 5 ) . To test this model , we show that chemical or genetic perturbations that alter cell cycle length consistently lead to reciprocal changes in growth rate , such that cell size remains constant ( Figure 6 Figure 7B ) . Similarly , perturbations that change growth rate lead to reciprocal changes in cell cycle length ( Figure 6 and Figure 7A ) . In both cases , the compensatory changes occur after prolonged drug treatment ( Figure 6 and Figure 7—figure supplement 1 ) , once the initial effects of the drugs have yielded a change in cell size , supporting the model illustrated in Figure 5 . To further test whether cell size is buffered from changes in cell cycle length , we overexpressed cyclin E or p27 to shorten or lengthen the duration of the cell cycle , respectively . In both cases , we found that the changing the cell cycle length resulted in a compensatory increase ( cyclin E ) or decrease ( p27 ) in growth rate to keep cell size constant ( Figure 10 ) . To directly visualize the negative correlation of cell size with G1 length , we used time lapse microscopy to show that cells that are born with a smaller nucleus spend longer periods of growth in G1 ( Figure 2D ) . While the correlation of cell cycle length and nucleus size is statistically significant and reproducible , there remains the question of why this correlation is small in magnitude . One possibility is suggested by the model in Figure 5 . Since deviations from target size are corrected not only by G1-length extension but also by adjustments to the rate of cell growth , the burden on each one of those separate mechanisms is weakened . In support of this idea , we repeatedly observe that when growth is inhibited with rapamycin the slope of nucleus size versus G1 length becomes about 50% steeper ( See Liu et . al . , Figure 3G versus 3I [Liu et al . , 2018] ) . While the possibility that the size of animal cells is regulated by cell cycle checkpoints has been previously debated , the possibility that feedback selectively suppresses growth rate in large cells is new and has only been sparsely explored ( Kafri et al . , 2013 ) . Accurately measuring the correlation of growth rate and cell size is challenging , due to the difficulty of measuring the growth rate of individual cells with high sensitivity . To circumvent this challenge , we developed the Gcv-value analysis to infer this correlation from measurements of cell size variance . With this approach , we not only provide the first definitive proof that individual animal cells autonomously monitor their own size and adjust their growth rate accordingly , but also provide new experimental assays for the study of growth rate regulation . The possibility that size homeostasis is maintained by growth rate regulation may have broad implications . In animals , most cells are terminally differentiated . These cells do not divide and yet , often undergo precise adjustments of cell size ( Nie et al . , 2013 ) . Also , it is often in such terminally differentiated tissues that size uniformity is most striking . This suggests the presence of size specification mechanisms that are not dependent on regular cell divisions . Furthermore , the size threshold model is fundamentally limited in its ability to correct the size of very large cells . If some cells grow fast enough to double even in a very short cell cycle , regulation of cell cycle length alone cannot constrain size variability in the population . A conclusion of this study is that the size of an animal cell is maintained at its target size value by homeostatic coordination of growth rate and growth duration . These results propose a conceptual distinction between a cell’s actual size and a cell’s target size , in the sense that a cell can temporarily be larger or smaller than its target size . Cells that are smaller than their target size grow faster and for longer periods of time . This finding raises the question of what mechanism dictates the cell’s target size . While this study , and Liu et al . , ( Liu et al . , 2018 ) show that cell size homeostasis is maintained by changes in both the rate and the duration of cell growth ( Figures 6 , 7 , 9 and 10 ) , we also show that cell size is sensitive to perturbations of Cdk4/cyclin D ( Figures 8 and 10 ) . Both pharmacological inhibition of Cdk4 and genetic overexpression of cyclin D cause significant changes in cell size value . These results are also consistent with the dramatic influence of Cln3 ( a cyclin D homologue ) on cell size in budding yeast ( Tyers et al . , 1992 ) . Altogether , these results suggest that cell size is strongly influenced by the Rb/cyclin D axis . In Figure 9 , we show the coordination of growth rate with cell cycle length in five different cell lines , including Rb-inactive cell lines as well as cell lines with intact Rb signaling . To our surprise , measurements of Rb positive and Rb negative cell lines produced similar results . This contrast between the strong influence of cyclin D/Cdk4 on cell size to the dispensability of Rb for size homeostasis may be telling . On one hand , Figure 9 shows that Rb is not required for maintenance of size homeostasis , nor for the coordination of growth rate with cell size length . On the other hand , the dramatic influence of cyclin D/Cdk4 on cell size ( Figure 8 , Figure 10 ) implies an important role for the Rb-cyclin D axis in size regulation . An intriguing possibility is that the function of the Rb-cyclin D axis is in dictating the specific cell size value ( i . e . ‘target size’ ) to be maintained rather than in ensuring cell size homeostasis at that target size . Consistent with the possibility that Cdk4 is involved in target size specification , palbociclib , a cdk4/6 inhibitor , causes a relatively uniform increase in cell size , suggesting a reprograming of target size . In contrast , the size increase caused by radicicol , an hsp90 inhibitor , is accompanied by an increase in the variability of cell size , suggesting a disruption of cell size homeostasis . While this evidence is not sufficient to determine whether cyclin D/Cdk4/Rb is involved in target size specification , it does warrant further explanation in a future study . The results presented here refute the premise that cell size is simply the quotient of independently regulated rates of cell growth and division ( Lloyd , 2013 ) . Instead , we show evidence of feedback between these two processes that robustly maintains cell size even in the face of strong perturbations of either growth or cell cycle progression . The presence of feedback that maintains a constant cell size , and minimizes variation in cell size , suggests the presence of a cell size sensor . With an understanding that this sensor influences growth and cell cycle progression in opposite ways , and with assays that can separately quantify each mode of regulation , we are in a position to uncover its molecular identity .
Cells were grown in Dulbecco’s Modification of Eagles Medium ( Cellgro , 10–017-CV ) , supplemented with 10% Fetal Bovine Serum ( FBS , Gibco , 26140 ) and 5% Penicillin/Streptomycin ( Cellgro 30–002 CI ) , incubated at 37°C with 5% CO2 . Measurements were made when cells were 50–75% confluent , to avoid the effects of sparse or dense culture on cell growth and proliferation . To distinguish S-phase cells from cells in G1 , we used cell lines stably expressing two nuclear-localized fluorescent cell cycle markers: ( 1 ) the geminin degron fused to Azami Green ( mAG-hGem ) and ( 2 ) the cdt1 degron fused to monomeric Kusabira Orange 2 ( mKO2-hCdt1 ) ( Sakaue-Sawano et al . , 2008 ) . Cells were seeded in 6-well , glass-bottom ( No . 1 . 5 ) plates one day prior to imaging . Cells were seeded at low densities such that they would not reach confluence during the course of the experiment , as described in Materials and methods-Cell culture . Leibovitz’s L-15 Medium ( ThermoFisher , 21083027 ) with 10% FBS was used during image acquisition , with a layer of mineral oil on top of the media to prevent evaporation . The microscope was fitted with an incubation chamber warmed to 37°C . Widefield fluorescence and phase contrast images were collected at 15 min intervals on a Nikon Ti motorized inverted microscope , with a Nikon Plan Fluor 10 × 0 . 3 NA objective lens and the Perfect Focus System for maintenance of focus over time . A Lumencor SOLA light engine was used for fluorescence illumination , and a Prior LED light source was used for transmitted light . A Prior Proscan III motorized stage was used to collect images at multiple positions in the plate during each interval . Images were acquired with a Hamamatsu ORCA-ER cooled-CCD camera controlled by MetaMorph software . After two days of time-lapse imaging , cells were fixed and stained as described below . Images of fixed cells were acquired on the same microscope , at the same stage positions , and an image-registration tool was designed in Matlab to match individual cells in the time-lapse movies to the corresponding cells in the fixed images . Cells were fixed in 4% paraformaldehyde ( Alfa Aesar , 30525-89-4 ) for 10 min , then permeabilized in chilled methanol for 5 min . Cells were stained with 0 . 04 ug/mL Alexa Fluor 647 carboxylic acid , succinimidyl ester ( SE-A647 , Invitrogen A-20006 ) , to label protein . DNA was stained with DAPI ( Sigma D8417 ) . To slow growth rate the following drug treatments were used: cycloheximide ( 1 , 0 . 6 , and 0 . 06 uM , Sigma C4859 ) , Torin-2 ( 10 , 5 , and 2 . 5 nM , Tocris 4248 ) , rapamycin ( 7 , 0 . 7 , and 0 . 07 uM , CalBiochem 553211 ) . To slow the cell cycle , the following drug treatments were used: BN82002 ( 25 , 12 . 5 , 6 . 2 , 3 . 1 , 1 . 6 , and 0 . 78 uM , Calbiochem 217691 ) , SNS-032 ( 39 and 9 . 8 nM , Selleckchem S1145 ) , PHA848125 ( 175 nM , Selleckchem S2751 ) , Cdk2 Inhibitor III ( 5 , 1 . 5 , 0 . 75 , and 0 . 38 uM , Calbiochem 238803 ) , Dinaciclib ( 10 , 5 , and 2 . 5 nM , Selleckchem S2768 ) , Palbociclib ( 500 , 250 , 125 , and 62 . 5 nM , Selleckchem S1116 ) . Radicicol ( Tocris , 1589 ) treatments were done at concentrations of 250 nM , 500 nM , and 1 uM . Each drug treatment was done in duplicate , alongside six control ( DMSO-treated ) samples , with several thousand cells in each sample . Cells were imaged using a Perkin Elmer Operetta high content microscope , controlled by Harmony software , with an incubated chamber kept at 37°C and 5% CO2 during live-cell imaging . A Xenon lamp was used for fluorescence illumination , and a 740 nm LED light source was used for transmitted light . To monitor proliferation of live cells , differential phase contrast images were collected every 6–12 hr , over the course of three days , using a 10 × 0 . 4 NA objective lens . A plot of the number of cells vs . time was fitted to an exponential growth model to calculate the average cell cycle time . Samples were fixed after 14 , 44 , and 68 hr in each drug . After fixation and staining with SE-A647 ( protein ) and DAPI ( DNA ) , fluorescence images were collected with a 20 × 0 . 75 NA objective lens . The bulk protein content ( total SE-A647 intensity of sample ) and number of cells were measured in each sample . From these measurements , we calculated the average growth rate and cell cycle length of cells in each condition , by fitting all data points ( from two replicates of each condition ) to an exponential growth model . Cell cycle length was also independently measured by monitoring the proliferation of live cells in each condition with differential phase-contrast microscopy , as described above . Entry clone vectors were obtained encoding open reading frames for CCND1 ( clone V57299 from Openfreezer ) , CCNE1 ( clone HsCD00045400 from DNA Resource Core , Harvard Medical School ) and CDKN1B ( clone HsCD00080627 from DNA Resource Core , Harvard Medical School ) . To alter cell cycle length by genetic manipulation , we generated stable Rpe1 cell lines with doxycycline-inducible overexpression of cyclin D1 , cyclin E1 , or p27 ( CDKN1B ) using the Retro-X Tet-One Inducible Expression System ( Clontech 634307 ) as per manufacturer’s instructions . At the start of each experiment cells were treated with doxycycline at concentrations of 1000 ng/mL ( for cyclin D1 and p27 overexpression ) , 100 ng/mL ( for cyclin E overexpression shown in Figure 8 ) , or 50 ng/mL ( for cyclin E overexpression shown in Figure 9 ) . Doxycycline concentrations were optimized for maximum effect on proliferation in each cell line . Proliferation was monitored by differential phase contrast imaging over the course of a 77 hr incubation in doxycycline , with periodic fixation and staining of samples to measure average growth rate and cell cycle length in each condition , as was done for the pharmacological perturbations described above . Each doxycycline induction was done in triplicate , alongside three untreated control samples of the same cell line . For comparison with drug-treated Rpe1 cells ( Figure 8A ) , growth rates and cell cycle lengths of each cell line were normalized so that the untreated controls of each cell line matched the average non-transduced Rpe1 control sample , since the engineered cell lines had slightly different basal proliferation rates . ON-TARGETplus SMARTpool siRNAs for the genes of interest ( Dharmacon L-003238–00 , L-003240–00 , L-003236–00 ) as well non-targeting negative control siRNAs were obtained from Dharmacon ( Lafayette , CO ) . DharmaFECT 1 Transfection Reagent ( Dharmacon T-2001 ) was used to transfect Rpe1 cells with each siRNA or combination of siRNAs , as indicated in Figure 8—figure supplement 1 . Cells were harvested at 48 and 72 hr post-transfection , for western blot verification of siRNA knockdown . Cells were also fixed and stained with SE-A647 at these time points , to measure cell size as described above . All image analysis ( cell segmentation , tracking , measurements of fluorescence intensity and nuclear size ) and data analysis was performed with custom-written tools in Matlab . The mean and variance of cell size as a function of age ( Figure 1C and Figure 4C–H ) were calculated by nonparametric regression , as described by Wasserman ( Wasserman , 2010 ) . To quantify the influence of drugs ( palbociclib and radicicol ) on the variation in cell size , we used the normalized median absolute deviation ( MAD ) , that is MAD ( x ) median ( x ) . We quantified the 2-d projected area of the nucleus ( i . e . the area of the image covered by the nucleus ) , which we found correlates well with cellular protein content ( Figure 2A–C and Figure 2—figure supplement 2 ) . As illustrated in Figure 2—figure supplement 1 , we tracked individual HeLa cells over time and monitored their cell cycle progression and nuclear growth . Note that , in these cells , the nucleus tends to elongate as it grows , rather than expanding as a sphere . ( This is significant because spherical expansion could yield a spurious correlation between size and growth of the 2-d projected area ) . We chose to monitor projected area so as not to make assumptions about nuclear shape in calculating volume , while bearing in mind the potential artifacts . We also trimmed the first 1 . 25 hrs of each trajectory , to avoid the effects of possible changes in nuclear shape as the cell flattens after mitosis . Trajectories ended at the onset of mitosis , when the nuclear envelope breaks down . | Animal cells come in many different sizes . In humans , for example , egg cells are thousands of times larger than sperm cells . Yet cells of any given type are often strikingly similar in size . The cells that line the surface of organs including the skin and kidneys are especially uniform; in fact a loss of size uniformity in certain tumors is a sign of malignancy . What kind of regulation could enable separate cells within a tissue to have the same size ? One possibility is that each type of cell is programmed with a specific target size , and that a cell can sense if it strays from its target and take steps to compensate . Animal cells sensing their own size was first reported in the 1960s , and now Ginzberg et al . confirm that human cells grown in the laboratory do indeed monitor their size and correct deviations from their target . It turns out that two separate and independent processes help to keep all the cells in the population roughly uniform in size . Firstly , proliferating human cells that are smaller than their target size spend longer growing before they divide . Secondly , at two time points between cell divisions , large cells adjust their growth rate such that they grow slower than small cells . To show these processes in action , Ginzberg et al . introduced mutations or chemicals that perturbed the length of time between cell divisions or the rate of a cell’s growth . As expected , most of these perturbations had only a modest influence on cell size , due to the cell’s compensatory strategies . Cells that had less time to grow compensated by more quickly making new protein molecules , meaning that they still had enough material to build two new cells by the time they had to divide . In contrast , if a cell’s division was artificially delayed , it reduced its growth rate to stop it from becoming too large . Similarly , cells grown in conditions that slow the production of proteins extended the time between their cell divisions to give them enough time to accumulate the material required for two new cells . In a recent related study , Liu , Ginzberg et al . identified some of the molecules that a human cell uses to sense its own size . Together these two studies now pave the road to answering a fundamental question in cell biology: what is the elusive cell size sensor ? Understanding how cells sense their size will open a window onto how quantitative information is programmed , sensed and communicated within living cells . These findings will shed also new light onto how cells specialize into cell types of different sizes , and what happens when cells lose the ability to sense or regulate their size in diseases like cancers . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"computational",
"and",
"systems",
"biology"
] | 2018 | Cell size sensing in animal cells coordinates anabolic growth rates and cell cycle progression to maintain cell size uniformity |
PlyC , a bacteriophage-encoded endolysin , lyses Streptococcus pyogenes ( Spy ) on contact . Here , we demonstrate that PlyC is a potent agent for controlling intracellular Spy that often underlies refractory infections . We show that the PlyC holoenzyme , mediated by its PlyCB subunit , crosses epithelial cell membranes and clears intracellular Spy in a dose-dependent manner . Quantitative studies using model membranes establish that PlyCB interacts strongly with phosphatidylserine ( PS ) , whereas its interaction with other lipids is weak , suggesting specificity for PS as its cellular receptor . Neutron reflection further substantiates that PlyC penetrates bilayers above a PS threshold concentration . Crystallography and docking studies identify key residues that mediate PlyCB–PS interactions , which are validated by site-directed mutagenesis . This is the first report that a native endolysin can traverse epithelial membranes , thus substantiating the potential of PlyC as an antimicrobial for Spy in the extracellular and intracellular milieu and as a scaffold for engineering other functionalities .
Streptococcus pyogenes ( Spy ) , a leading human pathogen , colonizes the skin and mucosal surface , and can lead to severe life-threatening conditions including bacteremia , necrotizing fasciitis , and streptococcal toxic shock syndrome ( Cunningham , 2000 ) , as well as a variety of superficial infections such as impetigo and pharyngitis . Streptococcal pharyngitis , one of the most common childhood illnesses in the United States , accounts for 15 million outpatient physician visits ( Hing et al . , 2008 ) and leads to an estimated economic burden of more than half a billion dollar per year ( Pfoh et al . , 2008 ) . Globally , Spy-associated diseases are correlated with high mortality and represent a worldwide health concern ( Carapetis et al . , 2005 ) . The ability of many pathogens , such as Spy , to invade host cells , survive at low levels in an intracellular niche , and seed reinfection poses a major challenge in developing effective treatment protocols to address this carrier state ( Rohde and Cleary , 2016 ) . Inside host cells , pathogens are protected from many antimicrobial agents and host immune effectors , and these evasion mechanisms directly contribute to long-term persistence ( Helaine et al . , 2014; Iovino et al . , 2014; Pentecost et al . , 2010 ) . Spy is well known for its ability to proliferate within host cells ( Barnett et al . , 2013 ) and escape autophagic degradation ( Sakurai et al . , 2010 ) . Notably , Spy can be recovered from clinical specimens of excised human tonsils ( Osterlund et al . , 1997 ) , even after antibiotic treatment . No effective approach has yet been identified that can specifically kill intracellular Spy , although successful strategies have been reported to control other intracellular bacteria , for example , by antibiotics that can accumulate within cells ( Ahren et al . , 2002; Mandell and Coleman , 2000 ) , encapsulating antibiotics into liposomes or nanoparticles for intracellular delivery ( Couvreur et al . , 1991 ) , or antimicrobial peptides ( Di Grazia et al . , 2014 ) . There is thus the need for an antimicrobial agent that targets both extracellular and intracellular Spy . Bacteriophage-encoded endolysins that act as peptidoglycan hydrolases represent a novel class of antimicrobials that kill bacteria by lysing the cell wall . Exogenously applied , these enzymes kill susceptible Gram-positive bacteria rapidly and with high specificity ( Loessner , 2005 ) . Numerous studies demonstrate the therapeutic potential of endolysins in various animal infection models of human disease , and no adverse side effects have yet been reported ( Daniel et al . , 2010; Pastagia et al . , 2013; Schuch et al . , 2014 ) . Therefore , endolysins have recently gained increasing attention as antimicrobial agents or ‘enzybiotics’ ( Fischetti , 2005 ) . However , despite a wealth of knowledge about their antibacterial potential , little is known about their capacity , either inherent or conferred by molecular engineering , to penetrate mammalian cells where they could target intracellular pathogens . One such Spy-specific endolysin , PlyC , has been characterized using biochemical and structural techniques and shown to be a 114 kDa multimeric holoenzyme . These studies demonstrated that PlyC comprises one PlyCA subunit , harboring two distinct catalytic domains that work synergistically to cleave two different bonds in the peptidoglycan , and eight identical PlyCB subunits that form a symmetrical ring for cell wall recognition ( McGowan et al . , 2012; Nelson et al . , 2006 ) . PlyC’s bactericidal activity against Spy in biofilms was shown in vitro ( Shen et al . , 2013 ) , as was its therapeutic potential in an in vivo model of upper respiratory Spy colonization ( Nelson et al . , 2001 ) . Here , we investigate the ability of PlyC to target and control intracellular Spy believed to be associated with streptococcal infections that are highly refractory to antibiotic treatment .
To mimic the pathogenic niche for Spy colonization and invasion , we established a co-culture model of human epithelial cells and Spy strain D471 to differentiate non-adherent , adherent , and intracellular streptococci . In experiments with human epithelial cell lines A549 ( Figure 1—figure supplement 1 ) or Detroit 562 ( data not shown ) , rates of Spy adherence ranged from 1 to 5% of the inoculum and rates of internalization ranged from 1 to 10% of the adherent streptococci , which are consistent with previous Spy/epithelial cell co-culture studies ( LaPenta et al . , 1994; Ryan et al . , 2001 ) as well as in vivo findings ( Kugelberg et al . , 2005 ) . Using this co-culture system , the bacteriolytic efficacy of three endolysins known to lyse Spy in vitro ( PlyC ( Nelson et al . , 2006 ) ; B30 ( Donovan et al . , 2006 ) ; and Ply700 ( Celia et al . , 2008 ) ) were evaluated for activity against intracellular Spy . Although we anticipated the enzymes would need to be engineered with cell penetrating peptides or transduction domains in order to acheive intracellular activity , treatment with 50 μg/ml WT PlyC reduced intracellular colonization ( colony forming units; CFUs ) by 95% ( p<0 . 001 ) within 1 hr , and lower concentrations displayed a dose response ( Figure 1a ) . In contrast , the B30 and Ply700 streptococcal phage endolysins failed to significantly decrease the intracellular Spy CFUs . We then assessed PlyC in a co-culture with primary human tonsillar epithelial cells grown from a tonsillectomy as a more clinically relevant model since these cells are known to be the major reservoir for recurrent Spy pharyngotonsillitis ( Osterlund et al . , 1997 ) . Roughly 90% of intracellular Spy were eliminated when treated with 50 μg/ml PlyC ( Figure 1b ) , similar to the effect in immortalized A549 epithelial cells , although the lower dose treatments did not demonstrate significant killing . At present , it is not known if the differences in efficacy are due to differences in the distribution of cellular receptors between the cell types or other phenotypic differences . Nonetheless , these data indicate that the native PlyC holoenzyme can be internalized by mammalian cells and that the endolysin retains bacteriolytic efficacy against Spy in the intracellular environment . 10 . 7554/eLife . 13152 . 003Figure 1 . PlyC eliminates intracellular Spy in a dose-dependent manner . ( a ) Spy/A549 co-cultures , treated with 10 µg/ml penicillin and 200 µg/ml gentamicin for 1 hr to remove extracellular bacteria , were incubated for another hour with antibiotic-free growth medium in dilutions of PlyC beginning at 50 μg/ml ( 440 nM ) , or with 50 μg/ml heat-denaturated PlyC ( 70°C for 30 min . ) , 50 μg/ml ( 1890 nM ) Ply700 , or 50 μg/ml ( 890 nM ) B30 . Viable intracellular Spy were enumerated as colonies on agar plates that had been incubated with serial dilutions of cell lysates . ( b ) A primary tonsillar epithelial cell co-culture was treated with 10 μg/ml penicillin and 200 μg/ml gentamicin for 1 hr after which the antibiotic was removed and the tonsil cells were incubated with 50 , 10 , or 1 μg/ml PlyC for 1h before lysis and enumeration of Spy . All experiments were performed in triplicate biological replicates , with mean and standard deviations displayed . Statistical analysis using Student's t-test is reported as *p<0 . 05; **p<0 . 005; ***p<0 . 001; ns: not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 00310 . 7554/eLife . 13152 . 004Figure 1—figure supplement 1 . Adherent and internalized colony counts of Spy recovered from co-culture assay . Determination of the adherent and internalized ( intracellular ) Spy CFUs in co-culture with A549 epithelial cells . A ~107 CFU innoculum dose was added to cultured A549 cells , incubated for 2 hours , and washed to remove non-adherent Spy . The remaining culture was treated with media or media supplemented with 10 μg/ml penicillin and 200 μg/ml gentamicin for 1 hr , followed by an additional wash step to remove antibiotics , lysis of A549 cells , and serial dilution plating to enumerate Spy CFUs . Media only-treated A549 cells represented adherent plus internalized Spy , whereas antibiotic-treated A549 cells represented internalized Spy . All experiments were performed in triplicate biological replates , with mean and standard deviations displayed . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 004 One explanation for the demonstrated intracellular lytic activity of PlyC is that this finding could be an artifact if the enzyme damages plasma membranes or is otherwise toxic toward eukaryotic cells . To test for this possibility , we used propidium iodide ( PI ) to visually distinguish between cells with and without membrane damage . Thirty-minute exposures to 100 μg/ml PlyC revealed no membrane damage compared to Triton X-100 treated controls ( Figure 2a ) . In addition , a trypan blue dye exclusion assay showed no difference between growth media control and 1 hr treatment with 100 μg/ml PlyC ( data not shown ) . Both assays suggest that exposure to or internalization of PlyC does not compromise the membrane integrity of epithelial cells . 10 . 7554/eLife . 13152 . 005Figure 2 . Membrane integrity is not compromised by PlyC , and internalization is mediated by the PlyCB subunit . ( a ) A549 epithelial cells were permeabilized with 0 . 02% Triton X-100 ( left panel ) , as shown by staining with propidium iodide ( red ) and DAPI ( nucleus , blue ) , whereas cells pre-incubated with 100 μg/ml PlyC for 1 hr showed no damage ( right ) . Scale bar is 5 μm . ( b ) Fluorescently labeled PlyC ( left panel ) and PlyCB ( middle panel ) , but not PlyCA ( right panel ) , are internalized in A549 cells upon incubation , as shown with protein ( 5 µg/ml ) conjugated with AlexaFluor555 ( green stain ) . Cells were incubated in serum-free F12K medium for 30 min at 37°C , fixed with 4% PFA in PBS and subsequently stained with DAPI ( blue ) . Arrows indicate internalized PlyC or PlyCB in vesicle-like structures with an average size of 0 . 5 μm . Scale bar is 10 μm . ( c ) Confocal microscopy of internalized PlyC colocalized with intracellular Spy . Nucleus ( blue , DAPI ) , Spy bacteria ( green , AlexFluor 488 conjugated wheat germ agglutinin ) , PlyC ( red , AlexFluor 555 conjugated ) , and actin filament ( magenta , AlexFluor 647 conjugated phalloidin ) are shown in the same focal plane . PlyC ( red ) is colocalized with Spy ( green ) in the merged image . The maximum intensity projection shows all Z-stacks simultaneously . Scale bar , 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 00510 . 7554/eLife . 13152 . 006Figure 2—figure supplement 1 . Purification of PlyC and its individual subunits . 12% SDS-PAGE of purified PlyCA ( 50 kDa subunit ) , PlyC holoenzyme ( 50 and 8 kDa subunits ) , and PlyCB ( 8 kDa subunit ) . Molecular mass markers as indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 006 Next , we determined whether internalization requires the PlyC holoenzyme or is specific to one of the PlyC subunits . PlyC , PlyCA , and PlyCB were purified ( Figure 2—figure supplement 1 ) and fluorescently labeled . At a concentration of 5 μg/ml , both the fluorescently PlyC holoenzyme and the PlyCB octamer , but not the PlyCA subunit , were internalized by A549 epithelial cells within 30 min ( Figure 2b ) . Both internalized PlyC and PlyCB were found to accumulate in vesicle-like structures with diameters of ≈0 . 5 μm . In confocal microscopy experiments , A549 epithelial cells were exposed to Spy and 30 min later to PlyC . Wild type ( WT ) PlyC , with full bacteriolytic activity , colocalized with intracellular Spy ( Figure 2c , merged image in the lower middle panel ) , where ruptured Spy cell walls are visible . No intact chain-like streptococci were observed through the entire Z-stack . Indeed , the maximum intensity projection ( i . e . merged Z-stack ) shows that all intracellular Spy were targeted and ruptured by internalized PlyC ( Figure 2c , lower right ) , and no extracellular Spy could be detected outside or between the host cell boundaries ( indicated by actin filament staining ) . Taken together , the data show that the subunit PlyCB mediates translocation of the PlyC holoenzyme across the plasma membrane , where it is subsequently trafficked and quantitatively lyses intracellular Spy . The PlyC crystal structure ( McGowan et al . , 2012 ) provided key insights into the origin of PlyCB–membrane interactions . At neutral pH , the PlyCB octamer is net positive ( pI = 8 . 5 ) , and its solvent accessible surface contains 48 cationic residues , two arginines and four lysines on each monomer ( Figure 3a ) . Cationic amino acid patches are known to frequently promote membrane adhesion and often result in membrane penetration of proteins , as described for cell-penetrating peptides ( CPPs ) ( Dietz , 2010 ) . To explore the roles of cationic residues in PlyC , we mutated the exposed Arg and Lys residues on PlyCB to Glu in order to independently determine the role of each mutation in Spy binding and lysis , as well as PlyC cellular uptake . Four of five holoenzyme constructs , PlyC ( PlyCBK70E , K71E ) , PlyC ( PlyCBK23E ) , PlyC ( PlyCBK59E ) , and PlyC ( PlyCBR66E ) , were soluble and tested for lytic activity in a turbidity reduction assay . A fifth mutant , PlyC ( PlyCBR29E ) was insoluble and could not be further studied . The holoenzymes incorporating PlyCBK23E and PlyCBK70E , K71E were fully active in external Spy degradation , but PlyC ( PlyCBK59E ) and PlyC ( PlyCBR66E ) only had ≈15% and 1% of WT PlyC activity , respectively ( Figure 3b ) . 10 . 7554/eLife . 13152 . 007Figure 3 . Cationic residues on the surface of PlyCB are important for Spy binding and epithelial cell internalization . ( a ) View of the membrane-binding interface of the PlyC holoenzyme ( PDB: 4F88 ) . The PlyCB octamer’s solvent-accessible surface is color-coded blue and red for areas of net positive and negative charge , respectively , with Arg and Lys side chains assigned as positive and Asp and Glu as negative . The PlyCA catalytic subunit , located behind the PlyCB octamer in this view , is shown in green . The rectangular region at top is enlarged at the right using the high-resolution model of PlyCB ( PDB: 4F87 ) that shows details of a cationic groove . Based on effects of mutations , this region appears important for interactions with both bacterial and mammalian cells . ( b ) Extracellular bacteriolytic activity of selected PlyC mutants on Spy , normalized to WT PlyC in a turbidity reduction assay . All experiments were performed in triplicate biological replicates , with mean and standard deviations displayed . ( c ) Microscopy of fluorescently labeled PlyCB and mutants on Spy ( left column ) and internalization into A549 epithelial cells ( right column ) , where arrows point to intracellular PlyCB . In distinction to the data in Figure 2c , Spy are not ruptured upon protein binding since the constructs lack the catalytic PlyCA domains . Scale bar is 5 μm in the left column and 10 μm in the right column . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 00710 . 7554/eLife . 13152 . 008Figure 3—figure supplement 1 . Fourier Transform Infrared Spectroscopy of WT PlyCB and binding groove mutants . Amide I and amide II vibrational bands of PlyCB , PlyCBK23E , PlyCBK59E , and PlyCBR66E . The amplitudes of the raw spectra were adjusted such that all spectra coincided at 1650 and at 1750 cm–1 . Therefore , the absorbance scale of 10 mAU ( 10–2 absorbance units ) applies strictly only to the WT PlyCB data set . For clarity , the spectra for PlyCBR66E , PlyCBK59E and PlyCBK23E were subsequently shifted by 10 , 20 and 30 mAU , respectively , and dashed lines are intended to guide the eye to band maxima . Within the amide I band , spectral contributions centered at 1660 cm–1 are usually assigned to α-helices and contributions centered at 1640 and 1680 cm–1 are attributed to β-sheets . Judged by this assignment , the secondary structure of the WT protein is largely conserved in PlyCBR66E and PlyCBK59E while the ratio of α and β structural motives may differ in PlyCBK23E . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 008 In its role in external Spy degradation , PlyCB is known to bind the streptococcal surface ( Nelson et al . , 2006 ) . Therefore , the same cationic to anionic mutations were made to PlyCB in the absence of the catalytic PlyCA and proper secondary structure folding was confirmed by infrared spectroscopy ( Figure 3—figure supplement 1 ) and analytical gel filtration ( data not shown ) . Of the four PlyCB constructs , only PlyCBK70E , K71E was insoluble and excluded from further study . The remaining PlyCB mutants were used to assess Spy binding and internalization by A549 epithelial cells . Fluorescence microscopy showed a lack of binding of the PlyCBR66E mutant to Spy while PlyCBK23E and PlyCBK59E , as well as a charge-conserving PlyCBR66K mutant bound comparably to WT ( Figure 3c , left ) . Moreover , the PlyCBK23E , PlyCBK59E , and PlyCBR66E mutants , but not PlyCBR66K , lost the ability to enter A549 epithelial cells ( Figure 3c , right ) . Thus , K23 of PlyCB is not involved in Spy binding but is involved in epithelial cell internalization , whereas K59 and , to a greater extent , R66 are involved in both Spy binding and internalization . In addition , the response of the charge-conserving PlyCBR66K mutant supports the hypothesis that electrostatic interaction involving R66 plays a significant role in PlyCB targeting both the Spy cell wall and the epithelial surface . To identify ligands on the plasma membrane that mediate PlyCB internalization , two approaches were taken . First , we investigated whether electrostatic interactions with glycosaminoglycans ( GAGs ) play a role in PlyCB internalization , as has been shown for cationic CPPs ( Tyagi et al . , 2001 ) and can be assessed by competing for the membrane-bound GAGs with soluble GAGs such as heparin or chondroitin sulfate-B ( CS-B ) . An excess of heparin or CS-B did not block PlyCB internalization into A549 epithelial cells ( Figure 4—figure supplement 1a–c ) . Furthermore , treatment of A549 cells with chondroitinase ABC or heparinase III prior to PlyCB incubation did not abrogate PlyCB internalization ( Figure 4—figure supplement 1d , e ) , ruling out a role for these GAGs as a PlyCB receptor . The second approach focused on anionic lipids , which are also implicated in the internalization of cationic CPPs ( Su et al . , 2010 ) . A phosphoinositide array ( PIP Strip ) showed that WT PlyCB interacts with phosphatidylinositol ( PI ) , phosphatidic acid ( PA ) , and phosphatidylserine ( PS ) but not with phosphatidylethanolamine ( PE ) , phosphatidylcholine ( PC ) , or any of the phosphatidylinositides ( Figure 4—figure supplement 2 ) , suggesting that the binding has elements of specificity . PlyCBK59E and PlyCBR66E that lack the ability to internalize in epithelial cells also lost the ability to bind PI and showed diminished binding to PA and PS . Surface plasmon resonance ( SPR ) was subsequently used to quantify PlyCB binding to sparsely-tethered bilayer membranes ( stBLMs ) ( McGillivray et al . , 2007 ) containing various anionic lipids in a background of 1 , 2-dioleoyl-sn-glycero-3-phosphocholine ( DOPC ) , where control experiments on pure DOPC showed no detectable protein binding . Typical SPR traces ( Figure 4a , b ) show that PlyCB binds phospholipids with largely different affinities and is presumably differentially organized at the surface of DOPC membranes that contain below and above 15 mol% 1 , 2-dioleoyl-sn-glycero-3-phospho-L-serine ( DOPS ) . At 10 mol% PS , adsorbed PlyCB washes out in a buffer rinse whereas it is largely retained upon rinsing in membranes with 20 mol% PS . Simultaneously with the SPR characterization , electrochemical impedance spectroscopy ( EIS ) ( McGillivray et al . , 2007 ) was used to monitor bilayer degradation upon protein binding . The Cole-Cole plot in Figure 4c , obtained with 30 mol% PS at the highest PlyCB concentration , illustrates that changes in electrical membrane characteristics were minimal . The EIS spectra were measured between 100 kHz and 1 Hz , but the spectra shown here cover only 100 kHz to 13 . 6 Hz in order to visualize the capacitive semi-circles well . Neither the membrane capacitance Cmem nor the membrane resistance Rdef ( Figure 4c , inset ) changed significantly ( for quantitative assessment , see Table 1 ) . 10 . 7554/eLife . 13152 . 009Figure 4 . PlyCB binds to PS-containing membranes . PlyCB in 150 mM NaCl at pH 7 . 4 was characterized on surface-ligated DOPC-based stBLMs with PS at various concentrations and its impact on membrane structure monitored with EIS . ( a , b ) Binding of PlyCB to stBLMs with 10 and 20 mol% DOPS , respectively . Each spike indicates an injection with an increased concentration of protein . A final buffer rinse removed the adsorbed protein quantitatively on 10 mol% PS , but only partially on 20 mol% PS , indicating that the association of the protein with the membrane surface differed depending on PS concentration . ( c ) Raw EIS spectra ( Cole-Cole plots ) of an stBLM ( 20 mol% DOPS ) as prepared ( dashed line ) and after incubation with 480 nM PlyCB ( solid line ) . Fitted parameters ( equivalent circuit model , see inset ) , corrected for electrode surface size , are given in Table 1 . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 00910 . 7554/eLife . 13152 . 010Figure 4—figure supplement 1 . Effects of soluble GAGs or GAG lyases on internalization of PlyC . ( a ) A549 epithelial cells were incubated with PlyCB-AlexaFluor555 ( red ) in serum-free medium F12K for 30 min at 37ºC , fixed by 4% PFA in PBS and stained with DAPI ( nucleus , blue ) . ( b ) Cells were pre-treated with 50 μg/ml of chondroitin sulfate-B for 30 min then incubated with PlyCB-AlexaFluor555 before being fixed and stained with DAPI . ( c ) Cells were pre-treated with 100 IU/ml of heparin for 30 min then incubated with PlyCB-AlexaFluor555 before being fixed and stained with DAPI . ( d ) Cells were pre-treated with 20 mIU/ml of chondroitinase ABC for 1 hr then incubated with PlyCB-AlexaFluor555 before being fixed and stained with DAPI . ( e ) Cells were pre-treated with 5 mIU/ml of Heparinase III for 1 hr then incubated with PlyCB-AlexaFluor555 before being fixed and stained with DAPI . Arrows indicate the internalized PlyCB . Scale bar is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 01010 . 7554/eLife . 13152 . 011Figure 4—figure supplement 2 . Assessment of PlyCB binding to phospholipids using a PIP Strip array . Layout of phospholipids used in the phospholipid arrays ( left panel ) and binding results of His-tagged WT PlyCB , PlyCBR66E , and PlyCBK59E . The blot was developed with the LumiSensor Chemiluminescent HRP Substrate Kit while using Mouse Anti-His mAb followed by Goat Anti-Mouse IgG [HRP] . Legend: 1 , lysophosphatidic acid ( LPA ) ; 2 , lysophosphocholine ( LPC ) ; 3 , phosphatidylinositol ( PI ) ; 4 , phosphatidylinositol-3-phosphate ( PI3P ) ; 5 , phosphatidylinositol-4-phosphate ( PI4P ) ; 6 , phosphatidylinositol-5-phosphate ( PI5P ) ; 7 , phosphatidylethanolamine ( PE ) ; 8 , phosphatidylcholine ( PC ) ; 9 , sphingosine-1-phosphate ( S1P ) ; 10 , phosphatidylinositol-3 , 4-bisphosphate ( PI ( 3 , 4 ) P2 ) ; 11 , phosphatidylinositol-3 , 5-bisphosphate ( PI ( 3 , 5 ) P2 ) ; 12 , phosphatidylinositol-4 , 5-bisphosphate ( PI ( 4 , 5 ) P2 ) ; 13 , phosphatidylinositol-3 , 4 , 5-trisphosphate ( PI ( 3 , 4 , 5 ) P3 ) ; 14 , phosphatidic acid ( PA ) ; 15 , phosphatidylserine ( PS ) , 16 , blank . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 01110 . 7554/eLife . 13152 . 012Table 1 . Best-fit parameters from fitting the EIS data in Figure 4c to an equivalent circuit model ( ECM ) . Specific capacitances and resistances were determined after normalization by the geometric electrode surface area ( a = 0 . 32 cm2 ) and multiplication by the roughness factor ( β = 1 . 4 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 012ECM parametersAs-prepared bilayer ( DOPC/DOPS = 7:3 ) Bilayer + PlyCB ( 480 nM ) Rsol ( Ω ) 129 . 0 ± 0 . 2106 . 0 ± 0 . 3Cstray ( nF ) 7 . 0 ± 0 . 113 . 0 ± 0 . 3Cmem ( μF/cm2 ) 0 . 710 ± 0 . 0020 . 739 ± 0 . 004Rdef ( kΩ cm2 ) 60 . 20 ± 0 . 0643 . 98 ± 0 . 12CPEdef ( μF/cm2 ) 8 . 17 ± 0 . 068 . 97 ± 0 . 12αdef0 . 925 ± 0 . 0030 . 870 ± 0 . 006 Raw SPR data , such as those in Figure 4a , b , were modeled in terms of Hill functions to determine protein affinities , Kd , to the membranes . Figure 5a displays results obtained with 30 mol% of various anionic lipids in DOPC-based stBLMs . PlyCB binds poorly to phosphatidylglycerol ( PG ) and PI and shows only low affinity to PA . In all these cases , rinsing the membranes with protein-free buffer reverts the SPR signal to baseline . In contrast , PlyCB binding to 30 mol% PS is extremely strong . Figure 5b shows PlyCB binding to stBLMs with PS concentrations between 10 and 30 mol% . At PS ≥20 mol% , the adsorption isotherms no longer followed a Hill function with N = 1 ( i . e . , a simple Langmuir isotherm ) but were well described by Hill functions with N > 1 ( Table 2 ) . On membranes that contained 30 mol% PS , both the affinity of the protein ( Kd = 40 ± 10 nM ) and its surface density were very high . Buffer rinses of stBLMs that contained PS at ≥15 mol% removed the bound protein only partially ( Figure 4b ) . While membrane binding of PlyCB is extremely strong at high PS concentration , its affinity decreases rapidly at lower PS content . At 15 mol% PS , the Kd increases to ≈0 . 5 μM , and at 10 mol% PS , PlyCB binding is reduced to the level of binding equal to 30 mol% PI or PG . This suggests that there is a threshold of PlyCB binding to PS-containing bilayers , where membranes at PS concentrations >15 mol% behave differently from membranes at ≤15 mol% . 10 . 7554/eLife . 13152 . 013Figure 5 . Roles of specific interaction of PlyCB with PS and electrostatic interaction in membrane binding . ( a ) Binding isotherms of WT PlyCB to membranes with different anionic phospholipids of the same concentration . The isotherm for DOPC:DOPS = 70:30 is only partially shown . Buffer rinses following protein incubations removed bound PlyCB from PG , PI , and PA-containing membranes as in the experiment shown Figure 4a . ( b ) WT PlyCB binding to binary membranes with DOPC and various concentrations of DOPS . ( c ) WT PlyCB binding to a ternary membrane composed of DOPC:DOPA:DOPS = 70:20:10 . ( d ) PlyCBR66E binding to a binary stBLM containing DOPC with 30 mol% DOPS . Data at cp > 10 μM were corrected for protein contributions to the optical index of the buffer ( see 'Materials and methods' ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 013 The observed threshold in PlyCB binding to PS-containing membranes suggests that both electrostatic attraction and specific binding contribute to PlyCB interacting with the membrane . If this is true , anionic lipids distinct from PS can provide the electrostatic component of the interaction . We measured the affinity of the protein to an stBLM composed of PC:PA:PS = 70:20:10 ( Figure 5c ) and observed that the affinity was higher ( Kd ≈ 1 . 2 μM ) than to membranes with 10 mol% of PS or 30 mol% PA as the only anionic lipid , but not as strong as to stBLMs containing ≥20 mol% PS . To confirm the PS specificity of PlyCB interaction with the membrane , we also measured the membrane affinity of PlyCBR66E ( Figure 5d ) . As expected , this mutant binds poorly to PS , comparable with WT PlyCB to PI-containing membranes ( compare Figure 5d and Figure 5a ) . These Kd values were converted into the free energy , ΔG = – kBT lnKd , between the membrane-bound and dissolved states of the protein ( details , see 'Materials and methods' ) . The major results of the series of experiments reported here are summarized in Figure 6 in terms of ΔΔG values , that is , changes in ΔG compared with that of DOPC with 30 mol% DOPS , the membrane we found most strongly attracted the protein and therefore had a strongly negative ΔG . 10 . 7554/eLife . 13152 . 014Figure 6 . PlyCB affinities for anionic lipid components in terms of free energy of membrane binding . Membranes with 30 mol% anionic lipids of various compositions are compared to WT PlyCB on 30 mol% PS . This graph visualizes that WT protein was bound to PG , PI , or PA much more weakly than to PS ( positive deviations of ΔΔG indicate lower affinities ) and that a low concentration of PS with a larger concentration of PA restored some of the protein binding . PlyCBR66E bound to 30 mol% PS with a similarly low affinity as that of WT PlyCB to 30 mol% PI . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 01410 . 7554/eLife . 13152 . 015Table 2 . Best-fit parameters from fitting Equation ( 1 ) to SPR data shown in Figure 5 . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 015ProteinstBLM compositionKd ( µM ) R∞ ( ng/cm2 ) NWT PlyCBDOPC:DOPG 70:30>> 50n . d . n . d . DOPC:soy PI 70:30> 50n . d . n . d . DOPC:DOPA 70:3026 . 9 ± 2 . 9170 ± 8 ( Langmuir fit ) DOPC:DOPS 90:10>> 50n . d . n . d . DOPC:DOPS 85:150 . 49 ± 0 . 02128 . 8 ± 2 . 71 . 0 ± 0 . 1DOPC:DOPS 80:200 . 04 ± 0 . 01251 . 7 ± 0 . 41 . 5 ± 0 . 1DOPC:DOPS 70:300 . 04 ± 0 . 01402 . 5 ± 0 . 52 . 2 ± 0 . 2DOPC:DOPA:DOPS 70:20:101 . 2 ± 0 . 1134 . 6 ± 1 . 2 ( Langmuir fit ) PlyCBR66EDOPC:DOPS 70:3062 . 2 ± 22 . 2331 ± 722 . 0 ± 0 . 05 Neutron reflection ( NR ) measurements prior to protein incubation confirmed that stBLMs of various anionic compositions all contained high-quality lipid membranes . Subsequently , neutron data were collected at multiple protein concentrations and , eventually , after a final rinse with protein-free buffer . In order to obtain high signal-to-noise data , PlyCB concentrations were chosen for which SPR showed high protein loading at the membrane surface . Each of the measurements was performed at multiple isotopic water ( H2O or D2O ) compositions of the buffer . This contrast-variation approach permits the determination of unique bilayer-protein complex structures , as shown earlier ( McGillivray et al . , 2009; Shenoy et al . 2012b ) , in terms of one-dimensional lipid/protein distributions as a function of the distance from the interface , z . These distributions are referred to as component volume occupancy ( CVO ) profiles ( Heinrich and Lösche , 2014 ) . PlyCB-membrane complexes on PC/PS stBLMs organize in different overall structures on 15 and 30 mol% DOPS , as shown in Figure 7 . At 15 mol% PS , PlyCB forms a diffuse layer of protein that extends about 200 Å out from the bilayer surface ( Figure 7a ) . Overall , the lateral protein density remains low at any z . Moreover , protein insertion into the membrane is negligible . The CVO profiles at 15 mol% PS in the bilayer suggest that PlyCB forms an open-mesh network on the membrane surface that has a thickness of ≈20 nm , and does not penetrate the bilayer . 10 . 7554/eLife . 13152 . 016Figure 7 . PlyCB forms distinct membrane–protein complexes at high and low PS content of the bilayer . CVO profiles ( Heinrich and Lösche , 2014 ) of membrane-bound PlyCB on DOPC stBLMs containing ( a ) 15 mol% and ( b , c ) 30 mol% DOPS . Neutron reflection experiments were performed on neat bilayers and on bilayers in contact with protein-containing solutions , and co-refined ( protein concentration in panels a and b: 5 μM and 600 nM , respectively ) . Panel ( c ) shows the same stBLM as in ( b ) after a buffer rinse to remove loosely associated protein from the membrane surface . Only the structures of the stBLMs in the presence of PlyCB protein are shown . The CVO profile in ( a ) shows accumulation of protein outside the bilayer , but no incorporation into the membrane . In distinction , the protein inserts deeply into the PS-rich bilayer , as indicated by the CVO profiles in panels b and c . The legend box in ( c ) applies to all panels . Confidence limits of the CVO profiles ( not shown for the lipid components but indicated by dashed lines for the protein distributions ) were determined by fitting the experimental data with a Monte Carlo Markov Chain . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 016 At high DOPS concentration , PlyCB associates with stBLMs in a different complex . Figure 7b shows the CVO profile of the protein on DOPC:DOPS 70:30 at cp = 600 nM . The protein is deeply inserted in the membrane and spans its entire thickness , headgroup to headgroup . Some protein is superficially adsorbed as evidenced by a protein CVO peak at a distance of ≈20 Å above the membrane surface . Flushing the membrane with protein-free buffer removed this adsorbed protein , as shown in Figure 7c . The CVO profile after the flush shows that the PlyCB protein retained within the membrane is almost symmetrically distributed across the tethered bilayer and occupies up to 10% of its volume . The protein CVOs shown in Figure 7a , b for stBLMs in contact with protein-containing solutions extend beyond the size of a single PlyCB octamer and could therefore only be modeled by continuous spline-based distributions . In contrast , the CVO profiles in Figure 7c , after the buffer rinse , indicate a compact protein conformation within the membrane . With the bilayer structure modeled as a continuous distribution of components ( Shekhar et al . , 2011 ) , we used the PlyCB crystal structure ( PDB: 4F87 ) ( McGowan et al . , 2012 ) and optimized its distance and orientation at the membrane , as described earlier ( Nanda et al . , 2010 ) . This rigid body modeling approach ( Figure 7c , black trace ) resembled the protein distribution determined in a spline fit ( Figure 7c , red trace ) , which suggests that the inserted protein spans the bilayer at a well-defined angle that leads to a match with the bilayer thickness . Data resampling ( Nanda et al . , 2010 ) determined the most likely model for the protein within the bilayer and established confidence limits for the model parameters . These are indicated as dashed lines in Figure 7 . The crystal structures of both the PlyC holoenzyme and the PlyCB octamer ( PDB: 4F88 and 4F87 , respectively; Figure 3a ) reveal a shallow , charged groove on the PlyCB membrane-binding surface surrounded by residues R29 , K59 , and R66 ( Figure 3a , inset ) . In combination with the residues identified to affect intracellular translocation , this suggests that the groove is a binding site for anionic lipids . To explore this hypothesis , a 1 . 7-Å X-ray crystal structure of PlyCBR66E was determined ( PDB: 4ZRZ ) . Superimposing this structure on that of a high-resolution structure of WT PlyCB ( PDB: 4F87 ) reveals that the mutated E66 side chain maintains the same location as R66 in the WT protein , and a hydrogen bond with E36 is preserved ( Figure 8a ) . However , R29 of PlyCBR66E adopts a new conformation and forms H bonds with E66 and E36 , effectively collapsing the groove . 10 . 7554/eLife . 13152 . 017Figure 8 . Interaction between PlyCB and PS requires structural integrity of a cationic binding groove . ( a ) Superimposed crystal structures of WT PlyCB ( PDB: 4F87 , cyan carbons ) and PlyCBR66E ( PDB: 4ZRZ , yellow , with green to emphasize the key residues R29 , E36 and the mutated E66 ) in the groove region . These are both high-resolution structures ( 1 . 4 Å and 1 . 7 Å , respectively ) of PlyCB alone . The orange sphere shows the location of a putative phosphate site observed in WT PlyCB in the presence of phosphate or phosphate-bearing ligands . This site holds a water molecule in the WT structure , but an empty site was observed in PlyCBR66E , apparently disrupted by the mutation . Mesh shows 2Fo-Fc electron density contoured at 1 . 6 σ for the three key residues in the mutant structure . The H-bond interaction between R66 and E36 is maintained in the mutant , suggesting that the observed side-chain conformations represent a stable configuration . ( b ) Local structure around the PS docking pose with lowest interface energy , as determined with Rosetta . The PS ( cyan carbons ) with partially removed fatty acid tails binds to the putative docking site adjacent to R29 , K59 , and R66 . The headgroup and two α carbons of the fatty acid chains are labeled in cyan . The PlyCB backbone is shown in green , with the side chains in the docking site ( sticks with green carbons ) together with their superimposed conformations from the holoenzyme crystal structure ( PDB: 4F88 , magenta carbons ) . Interestingly , the conformation of PlyCB with a double H-bond between R66 and E36 , which is the conformation seen in the PlyCB-alone high-resolution structure PDB: 4F87 ( Figure 3a , 6a ) , is reproduced in the PS docking . Proposed H bonds and salt bridges , in the calculated PS docking , as well as the distances between the corresponding atoms are also shown ( dashes and labels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 017 Cocrystallization of PlyCB with PS headgroup moieties and other candidate ligands , and soaking of these into PlyCB crystals , failed to show ordered ligands beyond the putative phosphate site of Figure 8a . As an alternative method , we used RosettaLigand ( Meiler and Baker , 2006 ) to model PS and PE docking to one subunit of the PlyCB octamer near the groove surrounded by R29 , K59 , and R66 . Because the burial of the lipid fatty tails cannot be adequately accounted for in this approach , these ligands were trimmed to three different lengths ( details , see 'Materials and methods' ) , with the longest construct retaining the α carbons of the fatty acid chains . In all , approximately 30 , 000 dockings were performed with each ligand . Possible complexes were selected using two criteria . The first candidate set was selected of models with the lowest interface energy scores ( Rosetta total energy of the complex minus the sum of the energies of the isolated protein and ligand ) and is composed of 152 PS and 125 PE candidates . Each of the candidates was visually inspected to exclude models in which ( 1 ) the phosphate group binds outside of the potential binding groove or ( 2 ) there would be unavoidable clashes if the fatty acid tails were complete , narrowing down the count of viable docking poses to 36 PS and 39 PE complexes . The best accepted PS model had an interface energy 3 . 74 kcal/mol lower than that of the best model for PE ( Figure 8b ) . In 11 PS candidates , including the three best models with different tail constructs , the seryl carboxyl formed a salt bridge with R29 and there was also a salt bridge between the phosphate group and R66 . In 18 PE candidates , including the three best models with different tail constructs , there was a salt bridge between the amine and E63 , as well as between the phosphate group and R66 . These electrostatic interactions dominated the energetics , and resulted in opposite orientations of PE and PS relative to the protein . A summary of the candidate set 1 is given in Table 3 . 10 . 7554/eLife . 13152 . 018Table 3 . Summary of the best set 1 candidate models by interface energy . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 018LigandSetBest model by interface energyInterface energy score ( kcal/mol ) Rosetta total energy ( kcal/mol ) H-bond and Salt bridges InteractionsDistances between corresponding atom pairs ( Å ) PS1A-6 . 64-1464 . 65Head - R292 . 90 , 2 . 77P - R662 . 67Tail - R663 . 45 , 2 . 931B-3 . 06-1461 . 49Head - R292 . 84 , 2 . 90 , 3 . 32P - R662 . 89 , 3 . 321C-3 . 15-1461 . 72Head - R292 . 87 , 3 . 43Head - E362 . 95P - R662 . 86PE1A-2 . 90-1465 . 81Head - E632 . 92P - R663 . 19Tail - R662 . 76 , 3 . 24Tail - R292 . 921B-2 . 92-1464 . 62Head - E632 . 95P - R662 . 88 , 3 . 431C-2 . 61-1466 . 02Head - E632 . 98P - R662 . 84 , 3 . 61Ligand in set 1A has the longest tail construct ( kept up to the α carbon ) . Ligand in set 1B only has head group and phosphate . Ligand in set 1C has one carbon after the ester linkage . Head is the ligand head group . P is the ligand phosphate moiety . Tail includes the ligand glycerol moiety and two fatty acid chains . Since energy criteria may not be reliable , the candidate set 2 was compiled by selecting the longest construct models that were excluded from candidate set 1 but contained at least one H bond or salt bridge between the docking site residues and the ligand headgroup or phosphate . This method generated 29 PS and 9 PE candidates . Considering the interfacial H bond energies and their bond lengths , 5 PS and 2 PE models were identified as good candidates that were missed in candidate set 1 ( Table 4 ) . 10 . 7554/eLife . 13152 . 019Table 4 . Good models in set 2 candidates . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 019LigandModel IDH-bond and Salt bridges InteractionsDistances between corresponding atom pairs ( Å ) PSgp7_D0145Head - R292 . 94P - R662 . 91Tail - R662 . 92gp7_D0430Head - R292 . 79P - R292 . 74P - R662 . 85 , 2 . 85Tail - K592 . 90gp8_D0131Head - R662 . 86 , 2 . 94P - 292 . 83gp1_F0153Head - R292 . 83 , 2 . 88P - R662 . 93gp9_F0363Head - R292 . 61P - R662 . 96 , 3 . 08Tail - R662 . 60PEgp1_F0391Head - E362 . 84P - R292 . 89gp9_F0228Head - E362 . 95P - R292 . 90Head is the ligand head group . P is the ligand phosphate moiety . Tail includes the ligand glycerol moiety and two fatty acid chains . Since this is the first report of the internalization of an endolysin into epithelial cells , we sought to further explore whether internalization occurs via an active or passive process . Addition of 20 μg/ml fluorescently labeled PlyCB readily produced punctate intracellular staining in A549 cells within 30 min at 37°C , but not at 4°C ( Figure 9a , b ) , indicating energy-dependent , active transport . We also observed that PlyCB colocalizes with cholera toxin subunit B ( CTxB ) , a lipid raft marker that binds GM1 gangliosides ( Latif et al . , 2003 ) , and both CTxB and PlyCB internalization were inhibited by chelation of cellular cholesterol after pre-treatment with filipin III ( Figure 9c , d ) . This suggests that PlyCB entry is mediated by a lipid-raft-dependent process such as caveolae-mediated endocytosis ( Lee et al . , 2008 ) . On the other hand , we ruled out other possible internalization pathways . Cytochalasin D , a macropinocytosis inhibitor that induces depolymerization of actin filaments ( Wakatsuki et al . , 2001 ) , did not affect PlyCB internalization , nor did PlyCB colocalize with transferrin , a ubiquitous marker of clathrin-dependent endocytosis ( data not shown ) . Furthermore , monodansylcadaverine , an inhibitor of clathrin-dependent endocytosis , did not prevent PlyCB internalization ( data not shown ) . 10 . 7554/eLife . 13152 . 020Figure 9 . Intracellular trafficking of PlyCB depends on caveolae-mediated endocytosis . ( a ) A549 cells were incubated with PlyCB-AlexaFluor488 ( green ) in serum-free medium F12K for 30 min at 37ºC , fixed by 4% PFA in PBS , and stained with DAPI ( nucleus , blue ) . Arrows indicate intracellular PlyCB vesicles . ( b ) PlyCB internalization was not observed when incubating PlyCB-AlexaFluor488 with cells at 4ºC . ( c ) Cells were incubated with PlyCB-AlexaFluor488 ( green ) and CTxB-AlexaFluor555 ( red ) in serum-free medium F12K for 30 min at 37ºC , fixed , and subsequently stained with DAPI ( blue ) . Arrows show co-localization ( yellow ) of PlyCB and CTxB . ( d ) Cells were pre-treated with Filipin III for 30 min , incubated with PlyCB-AlexaFluor 488 ( green ) and CTxB-AlexaFluor 555 ( red ) before being fixed and stained with DAPI . ( e ) A549 cells were transfected with CellLight early endosome-GFP ( green ) 24 hr prior to treatment of with PlyCB-AlexaFluor555 ( red ) and fixed after 60 min incubation with PlyCB . PlyCB-AlexaFluor555 co-localizes ( yellow ) with early endosome compartments as indicated by the bold arrows . The narrow arrows indicate non-endosomal vesicles containing PlyCB . ( f ) Cells were transfected with CellLight lysosome-GFP ( green ) 24 hr prior to treatment with PlyCB-AlexaFluor555 ( red ) and fixed after 60 min incubation with PlyCB . The bold arrows points to the co-localization ( yellow ) of PlyCB with lysosomal compartments . The narrow arrows indicate non-lysosomal vesicles containing PlyCB . Scale bars are 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 13152 . 020 To determine whether PlyCB utilizes the endocytic pathway for transport within epithelial cells , we analyzed the distribution of labeled protein to different subcellular localizations . Confocal Z-stack analysis ( Figure 9e , f ) showed internalized PlyCB partially colocalized with early endosomal and lysosomal compartments but was also found in vesicles not associated with these compartments . Taken together , these results and Figure 2c suggest that the internalized PlyC endolysin utilizes various intracellular transport mechanisms , some leading to lysosomal degradation but others to vesicle fusion with intracellular Spy . Further research will be needed to better understand the PlyC mechanism ( s ) of transport and intracellular interactions with Spy .
While extracellular Spy remains uniformly sensitive to penicillin , treatment failures are common because Spy can invade human host cells and adopt an intracellular niche ( LaPenta et al . , 1994; Marouni and Sela , 2004; Nobbs et al . , 2009 ) where the antibiotic is unable to penetrate . Internalization is implicated in recurring streptococcal infection because Spy can escape from endosomes and enter the cytosol in a streptolysin O-dependent manner , induce apoptosis of host cells , and subsequently repopulate the mucosal surface after antibiotic prophylaxis ends ( Ogawa et al . , 2011; O'Seaghdha and Wessels , 2013; Sharma et al . , 2016 ) . Consequently , there is growing interest in antimicrobial agents that can combat intracellular streptococci . Bacteriophage-encoded endolysins clearly show therapeutic potential against extracellular bacteria in vitro and in vivo , although to date , there have not been reports of wild type endolysins internalizing a eukaryotic cell . Accordingly , there have been extensive efforts to deliver these enzymes across the plasma membrane , the physical barrier that prevents internalized bacteria from being targeted by antimicrobials . CPPs have been suggested as fusion tags for endolysins to direct them against intracellular pathogens ( Borysowski and Gorski , 2010 ) . Toward this end , we did not expect WT PlyC to internalize cells , and therefore fused PlyC to TAT , one of the best characterized CPPs . In comparing the ability of TAT-labeled PlyC and WT PlyC ( used as a control ) to lyse intracellular Spy ( data not shown ) , we found that both lysed the pathogen inside of cells , inspiring the investigations reported here . PlyC is part of the lytic system of the C1 bacteriophage and , as such , has evolved to degrade the streptococcal peptidoglycan for release of progeny phage . Nonetheless , our results suggest a moonlighting role based on structural motifs within the PlyCB subunit whereby a cationic binding groove is a central determinant of interaction with plasma membranes and subsequent internalization . We propose that PlyC internalization is a three-step process that involves ( 1 ) membrane binding through electrostatic interaction and specific ligation with PS , ( 2 ) cellular entry by caveolae ( or lipid raft ) -dependent endocytosis , and ( 3 ) cytosolic transport where PlyC comes into contact with intracellular Spy . Interestingly , TAT-mediated protein transduction was reported to occur in a similar pathway ( Gump and Dowdy , 2007 ) , suggesting that PlyC internalization is comparable to cellular uptake of TAT . Whether eventual co-localization between PlyC and Spy involves endosomal escape or phagolysosomal fusion is currently unknown and requires further studies . Our SPR studies show that PS plays an eminent role in PlyCB membrane binding and suggest it is the cellular ligand that mediates protein binding to the plasma membrane . In semi-quantitative terms , this result is echoed in the Rosetta protein-ligand docking models . For PS and PE , only distinguished through the seryl carboxyl , the results for lipids with truncated fatty acid chains suggest that PS has a lower interface energy ( by >3 . 5 kcal/mol ) than that of PE . A part of this energetic difference comes from an interaction between the carbonyl group of the fatty acid tail and R66 , not observed for PE because of its consistently different orientation relative to the protein . For the shorter fragments ( ligand without fatty acid chains and glycerol backbones , see Materials and methods ) , the best models for PS and PE do not differ significantly in interface energy , at least partly because the tail carbonyl group is absent . Obvious caveats apply to these results: ( 1 ) in a real binding event , the phospholipid ligand is membrane-embedded which may restrict viable conformations in the bound state , despite the fact that the membrane is dynamic and flexible , and ( 2 ) trimming of the fatty acid tails may eliminate some contribution of the hydrophobic effect in the bound state , which may , however , be comparable for PS and PE . Despite these caveats , all low-energy candidate models demonstrate strong interactions between the ligand phosphate group and one of the two neighboring arginines – a type of electrostatic interaction that has been experimentally observed in other systems ( Woods and Ferre , 2005 ) – in support of the hypothesis that the lipid binding site is adjacent to R29 , K59 , and R66 . We note that the docking calculations are consistent with the experimental PlyCB-membrane binding data , even though the agreement could be fortuitous if various approximations in the docking analysis compensate . Indeed , the free-energy difference between docked PS and PE compares well with the difference in ΔG observed between PlyCB binding to PS-containing membranes and to membranes without this specific phospholipid ligand . PlyC binding to purely zwitterionic membranes is weak and has not been thoroughly studied , but the binding to non-specific anionic lipids or the binding of PlyCBR66E to PS are good proxies which both show ΔΔG values around 4 kcal/mol ( Figure 6 ) – a difference of the same order of magnitude as that observed in the docking result . We therefore suggest that the docking studies , while somewhat simplified , capture the essence of the specificity observed for PS lipid . PS is an important moiety of eukaryotic cellular membranes , preferentially found on the inner plasma membrane leaflet . Disruption of membrane asymmetry is known to follow apoptosis . On the other hand , recent studies ( Segawa et al . , 2011; van den Eijnde et al . , 2001 ) show that membrane inversion is not unique for cell death , and some viruses , such as the vesicular stomatitis virus , utilize PS as the cellular receptor ( Carneiro et al . , 2006 ) . It is therefore plausible that PS can also act as a receptor for PlyC , even more so since PS equilibration across the plasma membrane is more likely for cells that are stressed by pathogen invasion . In an analogous system , annexin A5 binds PS reversibly in a calcium-dependent manner and is often used as a marker for apoptosis . Upon binding to PS-containing membrane patches , annexin A5 forms a convex trimer which nanomechanically elicits membrane invagination and the formation of endocytotic vesicles that are actively transported by the cytoskeleton ( van Genderen et al . , 2008 ) . This translocation pathway is not restricted to apoptotic cells , because other proteins can be trapped in such annexin-A5-mediated portals of entry ( Kenis et al . , 2004 ) . Accordingly , PlyC might penetrate cells in a similar mechanism , as PlyCB possesses eight potential PS binding sites , each on the convex outer surface of the octamer . While further experimental evidence is required , it is tempting to speculate that parallels between the PS-driven membrane interaction of annexins and PlyC may produce membrane invaginations that lead to PlyC endocytosis . In conclusion , we identify a novel , effective and – as far as currently assessed – safe enzyme-based antimicrobial agent that specifically targets both extracellular and intracellular Spy . With the large pool of endolysins that bear distinct functional domains , our results not only show the potential of PlyC as a topical agent to combat refractory streptococcal infection on the skin and mucosal surface , but also provide a rationale to further investigate the intracellular potential of endolysins against other intracellular pathogens in a species-specific manner . Finally , the identification of PlyCB as the shuttle that catalyzes cellular entry of the catalytic PlyCA subunit opens the possibility that distinct payloads can be fused onto the PlyCB scaffold for delivery into mammalian cells .
Streptococcus pyogenes strain D471 , the primary Spy strain used in this study , is an M6 strain that has been characterized for its pathogenic mechanisms as well as its ability to be internalized by epithelial cells ( Ryan et al . , 2001 ) . Streptococci were routinely grown in liquid THY medium ( Todd-Hewitt broth , supplemented with 1% ( wt/vol ) yeast extract ) or on Columbia blood agar plates ( Thermo Scientific , Waltham , MA ) . For cloning and production of PlyC and its derivatives , a single colony of Escherichia coli , strain DH5α or strain BL21 ( DE3 ) , containing the construct was inoculated and grown in LB-Miller medium ( Luria broth containing 10 g/l NaCl ) with 100 μg/ml ampicillin . Overnight cultures of bacteria were pelleted and resuspended in fresh LB with 1/3 volume of 80% glycerol before being stored in a –80°C freezer . The constructs for pBAD24::plyC , pBAD24::plyCA , and pBAD24::plyCB were previously described ( Nelson et al . , 2006 ) . Mutagenesis studies utilized the QuikChange II Site-Directed Mutagenesis Kit ( Agilent Technologies , Santa Clara , CA ) and were conducted per manufacturer’s instruction . All constructs were verified by DNA sequencing before being transformed into the expression strain BL21 ( DE3 ) . The purifications of each construct were performed as previously described ( McGowan et al . , 2012 ) . Mutants were characterized for proper folding by analytical gel filtration as previously described ( Nelson et al . , 2006 ) and by Fourier Transform Infrared Spectroscopy ( FTIR ) with an MCT detector in a Hyperion microscope ( Nikon , Melville , NY ) and a Vertex 80 FTIR ( Bruker , Billerica , MA ) . Spy strain D471 was grown overnight at 37°C , washed in phosphate-buffered saline ( PBS ) , and resuspended to the desired concentration ( OD600 nm = 2 . 0 ) . One hundred microliter of Spy D471 containing suspension was mixed with 100 µl of purified PlyC or mutants ( 1 µg ) in a 96-well plate . The OD600 was measured on a SpectraMax 190 spectrophotometer ( Molecular Devices , Sunnyvale , CA ) every 15 s over a 20 min . period to monitor bacterial lysis . The activity of the PlyC mutants was normalized against wild-type ( WT ) PlyC endolysin , which represented 100% activity in the 20-min assay . Five microgram of purified PlyC , PlyCA , PlyCB or indicated mutants were reacted with the carboxylic acid , succinimidyl ester of AlexaFluor 555 or AlexaFluor 488 ( Invitrogen , Carlsbad , CA ) according to the manufacturer’s instructions . Unreacted dye was removed from labeled protein by application to a 5-ml HiTrap desalting column ( GE Healthcare , Marlborough , MA ) equilibrated with PBS . Cross-linked protein samples were subjected to analytical gel filtration on a Superose 12 column ( GE Healthcare ) calibrated with gel filtration standards ( Bio-Rad , Hercules , CA ) . Human alveolar epithelial A549 cells ( Human Lung Carcinoma cell line , CCL-185 ) were cultured in F-12K medium supplemented with 10% Fetal Bovine Serum ( FBS ) at 37°C , 5% CO2 , and 95% relative humidity . For primary epithelial cell cultures , the experimental protocol received Institutional Review Board approvals from both the Rockefeller University ( VAF-0621-1207 ) and the Weill Cornell Medical College ( nos . 0803009695 and 0806009857 ) . Individual patient consent for the use of tissue in research applications was obtained prior to the surgical procedure . Human primary tonsil epithelial cells were isolated and grown in Dulbecco's modified Eagles medium ( DMEM ) ( Thermo Scientific ) supplemented with 10% FBS and an antibiotic/antimycotic cocktail ( 100 µg/ml penicillin , 100 µg/ml streptomycin , 2 . 5 µg/ml amphotericin B ) . In order to evaluate the intracellular bacteriolytic efficacy of endolysins against internalized Spy , we established a coculture assay to quantify Spy adherence and invasion . Epithelial cells were grown to 80% confluent monolayers in 24-well tissue culture plates ( approximately 2×105 cells/well ) . An overnight culture of Spy strain D471 was washed in sterile PBS , resuspended in serum-free media and the concentration adjusted to ≈2×107 CFUs and incubated with epithelial cells at a multiplicity of infection ( MOI ) of 100 bacterial cells per one epithelial cell for 1 hr . Next , each well was washed 3× with PBS . One hundred microlter of a 0 . 25% trypsin/0 . 02% EDTA solution was added to each well to detach cells . Then , 400 μl of a 0 . 025% Triton X-100 solution in PBS was added to lyse the epithelial cells . Ruptured cells were visible within 5–10 min . Finally , the solution was serially diluted in PBS and plated on Columbia blood agar plates for enumeration of viable CFUs , which represent both adhered and internalized streptococci . For determination of internalized cell counts , 10 μg/ml penicillin and 200 μg/ml gentamicin were added to the co-culture for 1 hr prior to lysis in order to kill non-adherent and adherent , but not internalized bacteria ( as these antibiotics are not taken up by the epithelial cells ) . Then , internalized bacteria were serially diluted and plated on Columbia blood agar plates for enumeration after epithelial cell lysis . Thus , we differentiated non-adherent , adherent , and internalized Spy . To determine the efficacy of endolysins for eliminating internalized Spy , post-antibiotic treated co-cultures were washed 3× in PBS and incubated with endolysin for 1 hr before lysis and further enumeration of recovered Spy colonies . Epithelial cells were seeded onto 12 mm cover slips in 24-well tissue culture plates . When reaching 80% confluence , cells were washed twice with PBS prior to incubation with 20 μg/ml AlexaFluor labeled PlyC , PlyCB , or their mutants in serum-free medium for 30 min . The cells were again washed three times with PBS , fixed by 4% paraformaldehyde ( PFA ) , and mounted with ProLong Gold Antifade Reagent with 4' , 6-diamidino-2-phenylindole ( DAPI ) on glass slides for microscopic examination using an inverted scanning confocal microscope with Argon laser excitation ( Carl Zeiss LSM 710 , Germany ) . Images and Z-stacks were obtained with a Plan Apochromat 100×/1 . 4 objective lens and analyzed with the Zen 2010 digital imaging software ( Carl Zeiss ) . Alternatively , an Eclipse 80i epifluorescence microscope ( Nikon ) with X-Cite 120 illuminator ( EXFO , Canada ) , Retiga 2000R CCD camera , and NIS-Elements software ( Nikon ) was used for image acquisition and analysis . For the observation of intracellular distribution of internalized PlyCB , A549 epithelial cells were transfected with CellLight early endosome-GFP , CellLight lysosome-GFP , or CellLight actin-GFP ( Invitrogen ) 24 hr prior to counter staining with 20 μg/ml PlyCB conjugated with AlexaFluor ( Molecular Probes ) for 30 min . The cells were then washed with PBS twice , fixed with 4% PFA , and mounted with ProLong Gold Antifade Reagent with DAPI on a glass slide for confocal microscopy , as described above . To determine the internalization pathway , A549 epithelial cells were pre-incubated for 30 min with complete medium in the presence or absence ( control ) of specific inhibitors of endocytic pathways , including 50 μM monodansylcadaverine ( MDC , an inhibitor of clathrin-mediated endocytosis ) , 1 μg/ml filipin III ( an inhibitor of caveolae-mediated endocytosis ) , or 0 . 5 μM cytochalasin D ( CytD , an inhibitor of micropinocytosis ) , followed by two PBS washes and incubation with 20 μg/ml PlyCB conjugated with AlexaFluor555 for an additional 30 min . Along with the pharmacological inhibitors , the corresponding fluorescently labeled markers , including transferrin receptor ( a marker of clathrin-mediated endocytosis ) , cholera toxin subunit B ( CTxB , a marker of caveolae-mediated endocytosis ) , and phalloidin ( a stress fiber marker of micropinocytosis ) were used to confirm the effect of the inhibitors . All inhibitors and markers were from Life Technologies . For the membrane permeability assay , cells were either permeabilized with 0 . 02% Triton X-100 ( positive control ) or incubated with 100 μg/ml PlyC at 37°C for 30 min , washed with PBS twice , fixed with 4% PFA and mounted with ProLong Gold Antifade Reagent with DAPI on a glass slide . For the trypan blue assay , cells were seeded into 24-well tissue culture plates . When 80% confluent , cells were washed twice with PBS prior to incubation with various concentrations of PlyC at 37°C for 30 min . Next , cells were washed 3× in PBS and 100 μl of a 0 . 25% trypsin/0 . 02% EDTA solution was added to each well to detach cells from the bottom of wells . The trypsinized cells were then mixed with a 1:1 ( vol/vol ) trypan blue solution ( Thermo Scientific ) at 37°C for 30 min , manually counted in a Hausser hemocytometer ( Thermo Scientific ) and assessed for viability by measuring the proportion of non-viable cells stained by the dye versus unstained viable cells . A phosphoinositide array membrane ( PIP Strip , Invitrogen ) was used per manufacturer’s instructions to qualitatively assess the affinities of His-tagged PlyCB and selected mutants to phospholipids . To detect bound proteins that interact with specific phospholipids , mouse anti-His mAbs followed by goat anti-mouse IgG [HRP] were used and the blot was developed with the LumiSensor Chemiluminescent HRP Substrate Kit ( all from GenScript , Piscataway , NJ ) . 1 , 2-dioleoyl-sn-glycero-3-phosphocholine ( DOPC ) , 1 , 2-dioleoyl-sn-glycero-3-phospho-L-serine ( DOPS ) , 1 , 2-dioleoyl-sn-glycero-3-phospho- ( 1’-rac-glycerol ) ( DOPG ) , 1 , 2-dioleoyl-sn-glycero-3-phosphatidic acid ( DOPA ) and L-α-glycero-3-phosphatidylinositol ( soy PI ) were from Avanti Polar Lipids ( Alabaster , AL ) . The tether compound , Z 20- ( Z- octadec-9-enyloxy ) -3 , 6 , 9 , 12 , 15 , 18 , 22-heptaoxatetracont-31-ene-1-thiol ( HC18 ) , characterized as described elsewhere ( Budvytyte et al . , 2013 ) , was from Dr . D . Vanderah ( Institute for Bioscience and Biotechnology Research , Rockville , MD ) . Stock solutions of the lipids in chloroform were mixed to obtain targeted lipid compositions . The organic solvent was evaporated by placing the lipid mixtures under vacuum for 12 hr . Dried lipid films were hydrated in a high-salt aqueous buffer ( 1 M NaCl , 10 mM NaPO4 , pH 7 . 4 ) to obtain a lipid concentration of ≈5 mg/ml and sonicated until clear suspensions were obtained . These were extruded through a polycarbonate membrane ( Avanti ) with a pore size of 100 nm at least 21 times to obtain narrow distributions of unilamellar vesicles . Sparsely-tethered bilayer membranes ( stBLMs ) were formed as described earlier ( McGillivray et al . , 2007 ) . Briefly , phosphorous-doped , n-conducting silicon wafers ( El-Cat , Ridgefield Park , NJ ) for neutron reflection or microscopy glass slides for impedance spectroscopy and SPR measurements were used as substrates . Substrates were cleaned in Nochromix BX10 ( Godax Laboratories , Cabin John , MD ) solution with concentrated sulfuric acid , followed by extensive rinses with purified water ( EMD Millipore , Billerica , NY ) and pure ethanol ( Pharmco-Aaper , Shelbyville , KY ) , and dried in a nitrogen flow . They were loaded in a magnetron ( ATC Orion; AJA International , North Scituate , MA ) and coated with a ≈2 nm Cr adhesion layer and an Au layer with a thickness of ≈45 nm ( for EIS or SPR ) or ≈15 nm ( for NR ) . Self-assembled monolayers ( SAMs ) were prepared by overnight incubation of the Au-coated substrates in 0 . 2 mM ( total concentration ) ethanolic solution of HC18 and β-mercaptoethanol ( βME , Sigma-Aldrich , St . Louis , MO ) in a molar ratio of 30:70 . Upon removal from the incubation solution , the SAM-covered slides were immediately incubated with a vesicle suspension at high ionic strength for at least 2 hr , followed by a rinse with PBS of low ionic strength ( 50 mM NaCl ) at pH 7 . 4 that completed stBLM formation by rupture of adhering vesicles under osmotic shock . stBLMs on Si wafers were mounted in an electrochemical cell that implements a three-electrode configuration ( McGillivray et al . , 2007; Valincius et al . , 2006 ) : the gold-coated substrate , confined by Viton rings to a geometric area of ≈0 . 32 cm2 , served as the working electrode , a saturated silver-silver chloride ( Ag|AgCl|NaCl ( aq , sat ) ) microelectrode ( M-401F , Microelectrodes , Bedford , NH ) was used as a reference electrode , and a 0 . 25 mm diameter platinum wire ( 99 . 9% purity , Sigma-Aldrich ) acted as the auxiliary ( counter ) electrode . The typical roughness factor of the Au film , β ≈ 1 . 4 , was earlier determined ( Budvytyte et al . , 2013 ) . A Solartron 1287A potentiostat and 1260 frequency analyzer were used to measure impedance spectra of the stBLMs . An a . c . voltage with amplitude of 10 mV was applied across the sample while the measurements were carried out within a frequency range of 1 Hz to 100 kHz . Data acquisition was performed with the Zplot and Zview software packages ( Scribner Associates , Southern Pine , NC ) and the results analyzed by fitting the data to equivalent circuit models ( ECMs ) as described ( McGillivray et al . , 2007; Valincius et al . , 2006 ) . Briefly , the ECM contains elements that represent the solution resistance Rsol of the electrolytic buffer and the stray capacitance Cstray of the cabling and sample cell . The bilayer is represented by a capacitance Cmem parallel to a branch that represents membrane defects by their resistance Rdef and a constant-phase element ( CPE ) . The CPE has an impedance , Z = ( CPE ( iω ) α ) -1 , that is characterized by its constant-phase element coefficient , CPE ( a capacitance per unit area × tα–1 ) and its exponent α , which can vary between 0 and 1 . Thereby , a CPE becomes a conventional capacitance for α = 1 . While the membrane capacitance can also be represented by a CPE , we showed in earlier studies that the exponent α is close to one ( McGillivray et al . , 2007 ) . In distinction , αdef usually deviates significantly from one . A schematic of the ECM is shown as an inset in Figure 4c . A custom-built SPR instrument ( SPR Biosystems , Germantown , MD ) assembled in the Kretschmann configuration was used to quantify the affinity of PlyCB to stBLMs of various phospholipid compositions . SAM-covered Au-coated glass slides were index-matched with the prism , and stBLMs completed by vesicle fusion in situ . The SPR setup allows for simultaneous EIS measurements , and the quality of the lipid membranes was assessed before each binding experiment . stBLMs with a resistivity below 40 kΩ×cm2 were rejected . In the optical set-up , a fan of monochromatic light ( λ = 763 . 8 nm ) hits the sample at a range of incident angles , and a 2D CCD detector records the intensity of the reflected light . The position of the intensity minimum on the CCD defines the resonance angle , recorded as a function of time . The resonance angle corresponding to the neat bilayer is measured as a reference and its change recorded as increasing concentrations of PlyCB in PBS at 150 mM NaCl , pH 7 . 4 , are added to the stBLM . The temperature , typically set to 25ºC , is maintained within ± 0 . 01°C . SPR Aria ( SPR Biosystems ) was used for real-time data recording . The SPR response , R , is measured as a displacement of the reflection minimum on the CCD detector ( in pixels ) . The sensitivity of the instrument has been calibrated with a different protein – the PTEN tumor suppressor ( Shenoy et al . , 2012b ) – to be ΔΓ = 5 . 8 ng/cm2 per pixel of SPR response change . Time courses of R at each protein concentration , cp , were either recorded until equilibrated or fitted to an exponential function for determination of the equilibrium SPR response , Req . Quantification of low-protein affinities requires high protein concentrations in the buffer . If in excess of ~10 μM protein , this increases the optical index , n , of the medium in contact with the sensor surface , thus inflating the readout . Such measurements were corrected as follows . The refractive index increment is dn/dc ≈ 0 . 185 ml/g ( Barer and Joseph , 1954; Benesch et al . , 2000; Shenoy et al . 2012b ) . The sensitivity of the instrument is Δn = ( 6 . 4 ± 0 . 3 ) ×10–5/RU , where the response unit ( RU ) is the shift of the SPR reflection minimum on the detector in pixels ( Shenoy et al . 2012b ) . At a protein concentration , cp = 50 μM , one thereby expects a change in SPR response of 8 . 9 ± 0 . 4 RU due to protein in the buffer . To confirm this calculation , we measured the SPR response of a pure DOPC stBLM to dissolved PlyCB at concentrations up to 75 μM ( not shown ) . Results were consistent with the computed SPR signals under the assumption that the protein does not adsorb to the DOPC bilayer: dn/dc = 0 . 186 ± 0 . 011 ml/g . SPR at high cp ( > 10 μM ) were accordingly corrected . To quantify the protein affinity to the membranes in terms of its equilibrium binding constant Kd and surface density of bound protein at infinite concentration , R∞ , data were fitted to a Hill function with coefficient N , ( 1 ) Req = cpN R∞cpN + Kd where N = 1 corresponds to the Langmuir adsorption isotherm ( Schasfooort and Tudos , 2008 ) . Differences in binding Free Energies of membrane-associated proteins between two SPR measurements that yielded dissociation constants Kd ( 1 ) and Kd ( 2 ) , were determined as ( 2 ) ∆∆G=-kBT ln ( Kd ( 2 ) /Kd ( 1 ) ) NR measurements were performed at the NGD-Magik reflectometer at the NIST Center for Neutron Research . Reflectivities were typically collected for momentum transfers , qz , between 0 . 008 and 0 . 250 Å−1 . stBLMs were prepared on a SAM-coated silicon wafer and assembled in a flow cell . The flow-through sample cell design allows for in situ buffer exchange on the instrument . Protein-free and protein-loaded membranes were measured in at least two solvent isotopic contrasts that consisted of aqueous buffer prepared with D2O , H2O , or mixtures of the two . For each contrast , acceptable counting statistics are typically obtained after 6 hr . After the measurement of protein-loaded samples , the sample was gently rinsed and the stBLM measurement repeated to characterize protein unbinding . Analysis of NR data was performed using ga_refl ( Kienzle et al . , 2000-2010 ) . The structure of the lipid membrane within the stBLM was evaluated in a continuous distribution model ( Shekhar et al . , 2011 ) , and Catmull-Rom splines were used to parametrize the component volume occupancy ( CVO ) of the lipid components that form the bilayer , as well as the membrane-associated protein ( Heinrich and Lösche , 2014 ) . A Monte-Carlo resampling technique and Monte-Carlo Markov Chain algorithm ( Kirby et al . , 2012 ) were used to determine the confidence limits of the fitted parameter values ( Heinrich et al . , 2009 ) . Crystallization of PlyCBR66E , diffraction measurement , and structure refinement were as previously described for WT PlyCB ( McGowan et al . , 2012 ) . Briefly , crystals grew in sitting drops at room temperature from purified PlyCB at 14 mg/ml mixed with 43% methylpentanediol , 30 mM ammonium sulfate , 70 mM Na HEPES , pH 6 . 0 . X-rays were produced by a rotating anode generator ( MicroMax 007HF , Rigaku , The Woodlands , TX ) and diffraction data that yielded a resolution of 1 . 7 Å were measured by RAXIS IV++ image plates and processed using d*trek software ( Rigaku ) . The initial model of the mutant was made by omitting the amino acid side chains 29 , 36 , 59 , 63 , and 66 from the precedent structure ( PDB code: 4F87 ) . The omitted side chains were imaged using difference maps and added to the model during refinement . Refinement converged with Rf = 0 . 25 and rmsd bond lengths of 0 . 011 Å . The refined mutant structure shown in Figure 8a has been deposited in the PDB with accession code 4ZRZ . Initial atomic coordinates of the PlyCB octamer were taken from the PlyC holoenzyme X-ray crystal structure ( PDB code: 4F88 ) . To relieve preexisting strain according to the Rosetta’s energy function , an energy minimization was performed including all side chains with RosettaLigand ligand_rpkmin ( Davis and Baker , 2009; Meiler and Baker , 2006 ) . 216 PS structures were extracted from a fully hydrated DOPC:DOPS ( 3:1 ) bilayer MD model ( Shenoy et al . , 2012a ) . A corresponding set of PE structures was obtained by removing the seryl carboxyl group of PS . Because the burial of the lipid fatty tails cannot be adequately accounted for in this approach , the lipid tails were trimmed as follows: ( A ) fatty acid tails were kept up to the α carbon . ( B ) Only the phosphate and headgroup were kept . ( C ) The lipid headgroup was truncated after the sn-3 position of the glycerol . All 216 PS ( or PE ) conformations in the bilayer model were included in the ligand conformer library . A second conformer library containing 25 to 50 conformations , depending on the construct , was generated using Frog2 web interface ( Miteva etal . , 2010 ) . All docking trajectories were started from a position close to the PlyCB cationic groove containing R29 , K59 , and R66 , featuring the crystallographically observed putative phosphate site shown in Figure 8a . Ligand docking poses were sampled within a sphere of 5 Å radius centered at the initial site using RosettaLigand ( Davis and Baker , 2009; Meiler and Baker , 2006 ) , which uses a Monte Carlo search algorithm and supports modeling with flexible ligand and protein side chains . For each of the input ligands , 10 , 000–16 , 000 models were generated . To compile candidate set 1 , the models were ranked by their interface energies . To generate candidate set 2 , the models were selected if there existed at least one pair of non-hydrogen atoms between ligand and docking site residues ( Y26 , R29 , E26 , K59 , V62 , E63 , and E66 ) within 3 . 5 Å distance that were potentially capable of forming an H bond or salt bridge . | Streptococcus pyogenes is the bacterium that causes throat infections and other serious infections in humans . Antibiotics such as penicillin are used to treat active infections , but so-called “strep throat infections” often return after treatment . This is because S . pyogenes can enter the cells that line the throat and hide from the antibiotics , which cannot enter the throat cells . Endolysins are enzymes produced by viruses that attack bacteria , and these enzymes target and destroy the bacterial cell wall . A previous study revealed that an endolysin known as PlyC could destroy S . pyogenes bacteria on contact . PlyC and other endolysins have the potential to act as alternatives to common antibiotics , but before these enzymes can be developed as therapeutics , it is important to understand how they interact with human host cells . Like antibiotics , the PlyC endolysin was not expected to enter throat cells . However , Shen , Barros et al . have now discovered that not only can PlyC enter throat cells , it can essentially chase down and kill S . pyogenes that are hiding inside . Other similar enzymes could not act in this way , and further studies confirmed that PlyC could move around inside a throat cell without causing it damage . Shen , Barros et al . also determined that PlyC has a pocket on its surface that binds with a specific component of the throat cell membrane , a molecule called phosphatidylserine . This interaction – which is a bit like a lock and key – grants PlyC access into the cell . While it is clear that PlyC eventually kills S . pyogenes hiding inside throat cells , future experiments will aim to determine how PlyC moves around once inside an infected throat cell . Together , an understanding of how an endolysin enters cells and destroys hiding S . pyogenes will contribute to the development of endolysins with broader activity , which can be used as alternatives to common antibiotics . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"structural",
"biology",
"and",
"molecular",
"biophysics"
] | 2016 | A bacteriophage endolysin that eliminates intracellular streptococci |
The lateral entorhinal cortex ( LEC ) is thought to bind sensory events with the environment where they took place . To compare the relative influence of transient events and temporally stable environmental stimuli on the firing of LEC cells , we recorded neuron spiking patterns in the region during blocks of a trace eyeblink conditioning paradigm performed in two environments and with different conditioning stimuli . Firing rates of some neurons were phasically selective for conditioned stimuli in a way that depended on which room the rat was in; nearly all neurons were tonically selective for environments in a way that depended on which stimuli had been presented in those environments . As rats moved from one environment to another , tonic neuron ensemble activity exhibited prospective information about the conditioned stimulus associated with the environment . Thus , the LEC formed phasic and tonic codes for event-environment associations , thereby accurately differentiating multiple experiences with overlapping features .
The entorhinal cortex is thought to support rapid encoding of daily experiences by serving as an interface between the hippocampus and neocortical regions ( Eichenbaum , 2000; Squire , 1992 ) . In particular , lateral portions of the entorhinal cortex encode non-spatial information , such as objects ( Deshmukh et al . , 2012; Deshmukh and Knierim , 2011; Keene et al . , 2016; Tsao et al . , 2013 ) as well as visual and olfactory stimuli ( Igarashi et al . , 2014; Leitner et al . , 2016; Suzuki et al . , 1997; Xu and Wilson , 2012; Young et al . , 1997 ) . Recent evidence , however , demonstrates that the role of the LEC goes beyond a simple relay of non-spatial sensory information . For example , some cells in the LEC encode the location at which a specific object is placed ( Keene et al . , 2016 ) whereas other cells fire at the location of a previously encountered object or rewarding stimulus that is no longer present in the environment ( Deshmukh and Knierim , 2011; Tsao et al . , 2013 ) . These findings are consistent with the LEC’s role in recognizing objects in a specific context ( Hunsaker et al . , 2013; Van Cauter et al . , 2013; Wilson et al . , 2013 ) and suggest that the LEC may encode the content of events with the environmental context in which they take place . The selectivity of the LEC for the combination of sensory event and spatial/environmental context raises a question as to how strongly environmental context per se modulates cell firing in the LEC and how it compares to the modulation by more transient sensory events . We set out to address these points by recording the activity of LEC cells during six blocks of trace eyeblink conditioning , in which rats associated either auditory or visual conditioned stimulus with eyelid shock in two different conditioning environments . This paradigm was chosen because the acquisition and retrieval of memory are known to depend on the LEC ( Morrissey et al . , 2012; Ryou et al . , 2001; Tanninen et al . , 2013 , 2015 ) ; thus , inferences could be made about the necessity of any observed activity patterns for behavior . The use of the transient , discrete sensory stimuli , rather than objects , allowed for precisely controlling the onset and offset of sensory events . We found that the LEC cells encode the conjunction of a sensory stimulus and an environment phasically during the stimulus presentations , while also differentiating between environments tonically based on the history of stimulus presentations in those environments , revealing that stimulus-environment conjunctions are coded in the LEC at multiple time scales .
We recorded activity from cells in the LEC while seven rats underwent six blocks of trace eyeblink conditioning which required the subjects to form an association between a neutral conditioned stimulus ( CS ) and mild electric stimulation near the eyelid ( US ) over a 500 ms interval ( Figure 1A ) . The conditioned responses ( CRs ) were monitored by recording electromyogram from the eyelid . To examine the selectivity of cell firing for stimulus modality and conditioning environment , three of these blocks ( CS-US block ) included the pairings of either auditory or visual CS and the US in one of two visually distinct conditioning boxes ( Figure 1B ) . To examine the cells’ selectivity for stimulus relationship , in the remaining three blocks ( CS-alone block ) the CS was presented by itself in the box . The CS presentations were separated with pseudorandom inter-trial intervals ( ITI ) ranging from 20 to 40 s . Each rat daily received six trial blocks in a fixed temporal sequence ( an example pattern , Figure 1C; patterns used for each rat , Table 1 ) , enabling the rats to acquire the temporal context predictive of what would happen in the present trial block . Consistent with our previous findings ( Morrissey et al . , 2017; Takehara-Nishiuchi and McNaughton , 2008 ) , the rats gradually increased the expression of CRs in the CS-US paired trials but not CS-alone trials ( Figure 1D; Two-way repeated measures ANOVA , Session × Block , F75 , 450 = 2 . 99 , p<0 . 001 ) . The asymptotic level of CR expression during the three CS-US blocks was significantly different from that during the three CS-alone blocks ( follow-up one-way repeated measures ANOVA on CR% during the last session , F5 , 30 = 21 . 9 , p<0 . 001 , planned pairwise comparisons , ps < 0 . 05/6 ) , but it was comparable between all three CS-US blocks ( all ps > 0 . 7 ) . In addition , the increased frequency of eyeblink responses was not observed during the ITIs of any of six trial blocks ( Figure 1—figure supplement 1A; Session × Block interaction , F70 , 420 = 8 . 48 , p=0 . 717 ) , suggesting that the eyeblink responses were conditioned to the CS , but not to the conditioning environment ( see also , Morrissey et al . , 2017 ) . In the last session , the frequency of CR expression changed upon the transition from the CS-alone to the CS-US block within the first ten trials ( Figure 1E ) , suggesting that blocks of CS-alone trials did not simply extinguish associations acquired on previous days , rather they formed a distinct temporal context between earlier and later trials . On a trial-by-trial basis , all except one rat responded correctly on the first trial of two blocks in which the change in stimulus contingency was signaled by the change in the conditioning environment ( Figure 1F , the performance of Rat 2 , Figure 1G , the performance of four rats that underwent the trial blocks in the same temporal order ) . In contrast , when the stimulus contingency was changed in the same environment , the rats gradually adjusted the frequency of CR expression over ten CS-US paired trials , suggesting that the presence of the US served as a cue for the block transition . Thus , the rats formed three associative memories that shared a common , relational feature ( the CS-US association ) but differed in a discrete , physical feature ( the sensory modality of the CS ) or environmental context ( Box 1 or 2 ) . They also learned to use the change in conditioning environment and the presence of the US to rapidly infer whether the CS would be paired with the US in the current block . 10 . 7554/eLife . 28611 . 003Figure 1 . Rats acquired three associative memories with different stimulus and environmental features . ( A ) In trace eyeblink conditioning , animals associated a neutral conditioned stimulus ( CS ) with electrical shock near the eyelid ( US ) over a temporal gap ( 500 ms ) . With repeated pairings , rats developed anticipatory blinking responses before US onset ( CRs ) , which were monitored by recording electromyogram ( EMG ) from the upper eyelid ( a . u . , arbitrary unit ) . ( B ) Daily conditioning took place in a conditioning chamber which consisted of two visually distinct rooms ( Box 1 and 2 ) connected by a short walkway . ( C ) Rats first entered Box 1 and received an auditory stimulus ( ACS ) alone for 20 trials ( CS-alone block ) , followed by 50 ACS-US pairings ( CS-US block ) . They then moved to Box 2 , where they received 20 ACS alone trials followed by 50 ACS-US paired trials . Finally , they returned to Box 1 and received a visual stimulus ( VCS ) alone for 20 trials followed by pairings of the VCS and US for 50 trials . ( D ) The frequency of CR expression increased in the CS-US paired , but not in the CS-alone , trials ( mean CR% ± SEM; n = 7 rats; Sessions × Trial blocks interaction , F75 , 450 = 2 . 992 , p<0 . 001 ) . H1 and H2 show the rate of blinking when only the CS was presented before any conditioning . ( E ) In the last session , CR% ( in every 10 trials , mean ± SEM; n = 5 rats that received the CS-alone block before the CS-US paired blocks ) showed an abrupt transition upon the shift from the block of CS-alone trials ( light colors ) to the block of CS-US paired trials ( dark colors ) . ( F ) A trial-by-trial pattern of CR expression of Rat 2 during the last session . ( G ) The proportion of rats ( n = 4 rats underwent the trial blocks in the same temporal sequence ) that showed the CR in each trial during the last session . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 00310 . 7554/eLife . 28611 . 004Figure 1—figure supplement 1 . EMG activity during intervals between CS presentations . ( A ) The proportion of trials with significant eyelid EMG amplitude change ( pre-CRs ) remained low across sessions and did not differ between CS-alone and CS-US paired trials ( mean ± SEM; n = 7 rats; Sessions × Trial blocks interaction , F70 , 420 = 8 . 482 , p=0 . 717 ) . ( B ) As a measure of the general activity level during each trial blocks , eyelid EMG amplitude was averaged across the entire period of a block and converted to ranks . The ranked amplitude ( one line for each of seven rats ) was comparable among six trial blocks ( Kruskal-Wallis test , χ25 = 10 . 11 , p=0 . 072 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 00410 . 7554/eLife . 28611 . 005Table 1 . Temporal order of the six trial blocks used for each rat . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 005ID1ST2ND3RD4TH5TH6THRAT1ACS , Box 1ACS-US , Box 1ACS , Box 2ACS-US , Box 2VCS , Box 1VCS-US , Box 1RAT2ACS , Box 1ACS-US , Box 1ACS , Box 2ACS-US , Box 2VCS , Box 1VCS-US , Box 1RAT3ACS , Box 1ACS-US , Box 1ACS , Box 2ACS-US , Box 2VCS , Box 1VCS-US , Box 1RAT4VCS , Box 1VCS-US , Box 1ACS , Box 2ACS-US , Box 2ACS , Box 1ACS-US , Box 1RAT5ACS-US , Box 1ACS , Box 1ACS-US , Box 2ACS in Box 2VCS-US , Box 1VCS , Box 1RAT6ACS-US , Box 1ACS , Box 1ACS-US , Box 2ACS in Box 2VCS-US , Box 1VCS , Box 1RAT7ACS , Box 1ACS-US , Box 1ACS , Box 2ACS-US , Box 2VCS , Box 1VCS-US , Box 1 From the first day of conditioning , action potentials of cells in the LEC were extracellularly recorded with a chronically-implanted microdrive array containing twelve independently movable , four-channel electrodes ( tetrodes ) and two reference electrodes . The final locations of the tetrodes were evenly distributed along the anterior-posterior axis within the LEC ( Figure 2A , B ) . Firing rates of the majority of cells were less than 2 Hz ( Figure 2C ) , which is consistent with those reported in the previous studies ( Deshmukh et al . , 2012; Deshmukh and Knierim , 2011; Tsao et al . , 2013 ) . 10 . 7554/eLife . 28611 . 006Figure 2 . Recording locations and average firing rates of LEC units . ( A ) Schematic representations of the recording sites in the LEC along the anterior-posterior axis . The dots indicate final locations of tetrodes . The numbers show the distance from bregma ( mm ) ( B ) Example photomicrographs of the Nissl stained coronal sections with the tracks and final locations of the tetrodes ( circled ) . ( C ) Histogram of the averaged firing rate ( FR ) of the recorded LEC cells . In each cell , FRs during inter-trial intervals were separately calculated in six trial blocks , and the highest FR was used as the FR for a cell . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 006 Among 250 cells recorded across seven rats ( Table 2 ) , 113 cells ( 45 . 2% ) significantly changed their firing rate during the CS and subsequent intervals between CS offset and US onset relative to inter-trial intervals in at least one of six trial blocks ( a random permutation test , α = 0 . 05; Table 3 ) . Of these CS-responding cells , 8 . 8% showed selectivity for CS modality , meaning that firing rates differed between the auditory CS and visual CS trials in Box 1 , regardless of whether it was a CS-alone or CS-US paired trial ( a random permutation test , α = 0 . 05; Figure 3A , Cell 1 ) . In parallel , a separate set of CS-responding cells , 6 . 2% , was selective for the conditioning environment , exhibiting significantly different firing rates between the auditory CS presented in Box 1 and that in Box 2 , regardless of whether it was presented with the US or alone ( Cell 2 ) . More cells ( 12 . 4% ) differentially responded to the CS depending on both CS modality and conditioning environment ( Cell 3 ) . In parallel , 9 . 7% differentiated the firing response to the CS depending on whether it was presented alone or paired with the US at least in one of six trial blocks ( Relationship , Cell 4 ) . A considerable proportion of cells exhibited the selectivity for stimulus relationship together with the selectivity for stimulus modality ( 11 . 5% , Cell 5 ) , conditioning environment ( 7 . 1% , Cell 6 ) , or both ( 19 . 5% , Cell 7 ) . Overall , the majority of CS-responding cells were selective for more than one task variable ( Figure 3B ) . The proportion of cells in each selectivity category was comparable between superficial and deep layers of the LEC ( Table 3 ) . The magnitude of firing differentiation as measured by the distribution of a shuffle-corrected ‘Differential index’ ( below 0-no significant to 1-strongest differentiation ) , was similar between stimulus modality and conditioning environment ( Figure 3C , Kolmogorov-Smirnov test , p=0 . 330 ) while the magnitude of firing differentiation for stimulus relationship appeared to be weaker than those for modality ( p=0 . 103 ) and environment ( p=0 . 016 ) . These results suggest that a sizable proportion of LEC cells transiently signaled the stimulus-environment conjunction time-locked upon the CS presentation . 10 . 7554/eLife . 28611 . 007Figure 3 . Phasic firing patterns selective for stimulus-environment conjunction . ( A ) Examples of firing patterns of cells that were identified as selective for stimulus relationship ( CS alone vs . CS-US paired , Relationship ) , stimulus modality ( auditory vs . visual CS , Modality ) , or conditioning environment ( Box 1 vs . 2 , Environment ) and their combinations . Rasters and peristimulus time histograms ( PSTH ) represent activity patterns during the presentations of one of two CS ( auditory , ACS or visual , VCS ) in one of two conditioning environments ( Box 1 or 2 ) . Time 0 indicates CS onset . In each PSTH , the lines in light color show the firing pattern to the CS presented alone ( the first 20 trials ) while those in dark color show the pattern to the CS paired with the US ( the latter 50 trials ) . Light gray shadings indicate the CS , and black bars mask the artifact induced by the US . ( B ) Area-proportional three-venn diagram ( Micallef and Rodgers , 2014 ) showing a major overlap in selective cells for three task variables . The numbers show the percentage of cells in each category to total CS-responding cells . ( C ) Distributions of shuffle-corrected differential indices . A positive value means the significant selectivity ( random permutation tests , α = 0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 00710 . 7554/eLife . 28611 . 008Table 2 . Number of recorded units and sessions in each rat . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 008IDNumber of acquisition sessionsNumber of sessions with recorded unitsTotal number of unitsRAT 1171038RAT 2161261RAT 318511RAT 4161349RAT 52167RAT 6201355RAT 718112910 . 7554/eLife . 28611 . 009Table 3 . Number and percentage of cells with selective CS-evoked firing for three variables . Table summarizes the number of cells that significantly changed CS-evoked firing rates depending on CS-US relationship ( CS-alone trials vs . CS-US paired trials , R ) , CS modality ( auditory CS vs . visual CS , M ) , conditioning environment ( Box 1 vs . Box 2 , E ) , the combination of them , or none of them ( Non-S ) . The values in parentheses show the percentage of cells in each category to total CS-responding cells . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 009SiteNon-SRMER+MR+EM+ER+M+EOverall 11328 ( 24 . 8% ) 11 ( 9 . 7% ) 10 ( 8 . 8% ) 7 ( 6 . 2% ) 13 ( 11 . 5% ) 8 ( 7 . 1% ) 14 ( 12 . 4% ) 22 ( 19 . 5% ) Superficial layers 5616 ( 28 . 6% ) 5 ( 8 . 9% ) 5 ( 8 . 9% ) 3 ( 5 . 4% ) 7 ( 12 . 5% ) 4 ( 7 . 1% ) 7 ( 12 . 5% ) 9 ( 16 . 1% ) Deep layers 5712 ( 21 . 1% ) 6 ( 10 . 5% ) 5 ( 8 . 8% ) 4 ( 7 . 0% ) 6 ( 10 . 5% ) 4 ( 7 . 0% ) 7 ( 12 . 3% ) 13 ( 22 . 8% ) Most cells ( 137 , or 54 . 8% ) did not significantly change firing rates upon the onset of the CS relative to inter-trial intervals ( ITI ) . These cells maintained stable firing rates throughout the entire period of each trial block , but nearly all of these cells ( 97 . 9% ) changed ITI firing rates depending on which trial block a rat was in ( Figure 4A ) . The across-block difference in their firing rates was not simply due to the difference in perceivable features of the conditioning environment . Virtually no cells ( 0 . 7% ) were purely selective for conditioning environment: the cells generally did not exhibit significantly different firing rates in Box 1 blocks compared with Box 2 , independent of ACS-US pairings or ACS alone . Rather , a subset of non-responding cells , 9 . 2% , was selective for both environment and stimulus relationship , exhibiting differential firing rates for the ACS-US block compared with the ACS-alone block in only one of the conditioning boxes ( Cell 8 , relationship and environment ) . Furthermore , a separate set of cells ( 25 . 9% ) changed their ITI firing rates within the same conditioning environment depending on which of two CS was presented and whether the CS was paired with the US in a present trial block ( Cell 9 , relationship and modality ) . The largest proportion of cells ( 38 . 9% ) was classified as selective for the stimulus relationship , stimulus modality , and conditioning environment , resulting in unique firing rates for one of six trial blocks ( Cell 10 ) or different firing rates for each of six trial blocks ( Cell 11 ) . Importantly , their spike waveforms were comparable between six trial blocks ( bottom of each panel ) , suggesting that the substantial changes in ITI firing rates across the trial blocks were not due to instability of the recording ( e . g . , electrode drift ) . 10 . 7554/eLife . 28611 . 010Figure 4 . Tonic firing patterns selective for stimulus-environment conjunction . ( A ) Examples of firing patterns of cells during intervals between trials with rasters and peristimulus time histograms as in Figure 3A . The bottom panel shows each cell’s average waveforms recorded in one of four wires of a tetrode during six trial blocks . ( B ) Area-proportional three-venn diagram showing that the majority of cells were selective for more than one task variable . The numbers show the percentage of cells in each category to all cells . ( C ) Distribution of the shuffle-corrected differential indices as in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 010 The across-block difference in ITI firing rates was also observed in the CS-responding cells , and the proportion of cells for each category was comparable to that in the non-responding cells ( Table 4 ) . After CS-responding and non-responding cells were combined , the majority of these cells was classified as selective for more than two task variables ( Figure 4B ) . The proportion of cells in each selectivity category was comparable between superficial and deep layers of the LEC ( Table 4 ) . The magnitude of firing differentiation , as measured by the distribution of a shuffle-corrected ‘Differential index’ ( below 0-no significant to 1-strongest differentiation ) , was similar between stimulus modality and conditioning environment ( Figure 4C , Kolmogorov-Smirnov test , p=0 . 787 ) while the magnitude of firing differentiation for stimulus relationship was weaker than those for stimulus modality ( p=0 . 030 ) or conditioning environment ( p=0 . 002 ) . These findings suggest that regardless of firing rate changes during events , nearly all cells in the LEC signaled highly integrated information about sensory events taking place in a given environment during intervals between the events . 10 . 7554/eLife . 28611 . 011Table 4 . Number and percentage of cells with selective firing for task variables during intervals between trials . Table summarizes the number of cells that showed selective firing rates for three task variables during intervals between trials , as shown in Table 3 . The values in parentheses show the percentage of cells in each category to total non-responding , CS-responding , or all cells . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 011Response type/SiteNon-SRMER×MR×EM×ER×M×ENon-responding 1373 ( 2 . 2% ) 11 ( 8 . 0% ) 5 ( 3 . 6% ) 1 ( 0 . 7% ) 30 ( 21 . 9% ) 20 ( 14 . 6% ) 12 ( 8 . 8% ) 55 ( 40 . 1% ) CS-responding 1134 ( 3 . 5% ) 10 ( 8 . 8% ) 4 ( 3 . 5% ) 2 ( 1 . 8% ) 13 ( 11 . 5% ) 15 ( 13 . 3% ) 10 ( 8 . 8% ) 55 ( 48 . 7% ) Superficial layers 1105 ( 4 . 5% ) 11 ( 10 . 0% ) 5 ( 4 . 5% ) 1 ( 0 . 9% ) 19 ( 17 . 3% ) 15 ( 13 . 6% ) 11 ( 10 . 0% ) 43 ( 39 . 1% ) Deep layers 1402 ( 1 . 4% ) 10 ( 7 . 1% ) 4 ( 2 . 9% ) 2 ( 1 . 4% ) 24 ( 17 . 1% ) 20 ( 14 . 3% ) 11 ( 7 . 9% ) 67 ( 47 . 9% ) To examine the stability of selective firing patterns during the ITI period , firing fluctuation within a trial block was compared with that across six trial blocks . Firing rates of each cell were stable during the ITI period in a trial block , but drastically differed between the trial blocks ( Figure 5A left ) . The majority of cells ( 96 . 4% ) showed a smaller degree of firing fluctuation , measured by Kullback-Leibler divergence , within each trial block than across the trial blocks ( Figure 5A right , B ) . As a population , the degree of within-block fluctuation was significantly smaller than that of the between-block fluctuation ( Figure 5C , signed rank test , p<0 . 001 ) . 10 . 7554/eLife . 28611 . 012Figure 5 . LEC cells stably maintained the selectivity for stimulus-environment conjunction during the stimulus-free period . ( A ) Example of fluctuations of binned firing rates ( FR ) of a cell across the 9 s interval during each of six trial blocks . The degree of FR fluctuation was marginal across time bins in each trial block ( ‘Within’ , rows ) while FR greatly changed across six blocks ( ‘Between’ , columns ) as evident in the difference in Kullback-Leibler divergence from a uniform distribution ( KLD , right ) . ( B ) Distribution of the difference between the Between-KLD and Within-KLD . ( C ) Cumulative distribution of the Within- ( black ) and Between- ( red ) KLD . ( D ) Difference in the proportion of cells whose ITI firing rates were judged as selective for at least one of three task features depending on the bin size . The same criteria were applied to averaged firing rates in one 9 s bin ( 1 ) , four 2 . 25 s bins ( 1/4 ) , or eight 1 . 125 s bins ( 1/8 ) covering the 9 s ITI . ( E ) The proportion of selectivity categories ( see Table 4 ) of firing rates during a series of eight 1 . 125 s bins during the 9 s ITI . The 8th bin was the closest to CS onset . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 012 Another measure of firing stability was to examine the selectivity separately in a series of time bins during the ITI period . To test this , a methodological decision had to be made as to what temporal bin size was most appropriate for parsing the ITI . By re-running the test across multiple bin sizes , we found that the proportion of selective cells decreased as the bin size became shorter ( Figure 5D ) , most likely due to the reduced accuracy in estimating firing rates . Importantly , the proportion of selective cells and the type of selectivity was stable across the time bins regardless of their temporal proximity to the CS onset ( Figure 5E ) . These results suggest that firing rates of individual cells were largely homogeneous during the ITI period in a trial block . To examine whether individual cell firing was correlated with CR expression on a trial by trial basis , firing rates were compared between trials in which the rats showed CRs and those in which they did not . Across three blocks of CS-US pairings , 5 . 7–7 . 9% of cells’ firing rates during CS-US intervals were judged to be selective for CR expression ( 2 out of 35 cells in ACS-US in Box 1; 4 out of 61 cells in ACS-US in Box 2; 5 out of 63 cells in VCS-US in Box 1 ) . Virtually no firing rates during the ITI period were selective for CR expression ( 2 out of 35 cells in ACS-US in Box 1; 1 out of 61 cells in ACS-US in Box 2; 1 out of 63 cells in VCS-US in Box 1 ) . These results suggest that although the LEC is necessary for memory acquisition and expression of this associative learning paradigm ( Morrissey et al . , 2012; Tanninen et al . , 2015 ) , firing patterns of individual cells were not correlated with behavioral expression of the memory on a trial-by-trial basis . We next examined the selectivity of ensemble firing during trials and inter-trial intervals for sensory events and conditioning environment . Consistent with the observations from the single unit analysis ( Figures 3 and 4 ) , neuron ensembles appeared to form a unique firing pattern for each of six trial blocks , and the across-block difference was noticeable not only after , but also before the CS presentation ( Figure 6A ) . To quantify the degree to which the ensemble activity differentiated each of three task variables , the state vectors ( i . e . , the firing rates of 250 cells either during the trial or ITI period ) were compared within versus between trial blocks ( Figure 6—source data 1 ) . During trials , ensemble firing patterns differentiated between ACS and VCS trials ( Modality ) , between CS-alone trials and CS-US trials ( Relationship ) , and ACS trials in Box 1 and 2 ( Environment ) more strongly than odd- and even-numbered trials of the same block ( None , the upper limit; n = 20 randomly selected sets of 10 trials per trial block; one-way ANOVA , F4 , 195 = 238 . 975 , p<0 . 001; posthoc test Tukey HSD , all ps < 0 . 001 ) . The degree of differentiation was the strongest for stimulus modality ( vs . Environment , p<0 . 001; vs . Relationship , p<0 . 001 ) and was comparable to the differentiation of two trial blocks without any overlapping variables ( All , the lower limit , p=0 . 122 ) . Also , the differentiation of Relationship was weaker than the differentiation of Environment ( p<0 . 001 ) . During the ITI period , ensemble firing patterns differentiated two blocks with different stimulus modality more than those in different environments ( one-way ANOVA , F4 , 195 = 142 . 092 , p<0 . 001; posthoc Tukey HSD , p<0 . 001 ) , suggesting the stronger modulation by which CS had been presented than which conditioning box the rat was in . The differentiation for stimulus relationship was significantly weaker than that for stimulus modality ( p<0 . 001 ) and conditioning environment ( p<0 . 001 ) but was stronger than the chance-level differentiation ( p<0 . 001 ) . 10 . 7554/eLife . 28611 . 013Figure 6 . Ensemble activity showed a comparable level of selectivity for stimulus-environment conjunction during stimulus and non-stimulus periods . ( A ) Grayscale plots show the normalized firing rate during six trial blocks ( from left , ACS trials in Box 1 , ACS-US trials in Box 1 , ACS trials in Box 2 , ACS-US trials in Box 2 , VCS trials in Box 1 , and VCS-US trials in Box 1 ) . Cells were sorted based on the ACS-induced firing rate during the ACS-US block in Box 1 , from the largest increase ( cell #1 ) to the largest decrease ( cell #250 ) . Two white lines indicate the onset and offset of the CS while black bars mask US artifacts . ( B ) Matrices of the correlation coefficient ( r ) of ensemble firing rates ( State vector , SV ) between two of six trial blocks during CS-US pairings ( Trial ) and intervals between trials ( ITI ) . ( C ) During both task phases , the r for two blocks with different CS ( Modality; mean ± SEM , n = 20 runs with 10 subsampled trials ) was comparable to that for two blocks that differed in all task variables ( All ) . It was significantly lower than that in different conditioning boxes ( Environment ) and with different stimulus contingencies ( CS-alone blocks and CS-US blocks , Relationship ) . The r for Relationship was significantly higher than that for Environment but lower than that for odd- and even-numbered trials from the same block ( None ) . *p<0 . 001 , in posthoc Tukey HSD . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 01310 . 7554/eLife . 28611 . 014Figure 6—source data 1 . Ensemble activity showed a comparable level of selectivity for stimulus-environment conjunction during stimulus and non-stimulus periods . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 01410 . 7554/eLife . 28611 . 015Figure 6—figure supplement 1 . Ensemble firing patterns differentiated trial blocks more strongly depending on the CS modality and conditioning environment than CS-US relationship . ( A ) Support Vector Machine ( SVM ) classifier was applied to binned firing rates during six trial blocks . Representative examples of confusion matrices indicating proportions of trial types that were identified by the SVM classifier correctly ( warm colors along the diagonal ) or misidentified as a different trial type ( lighter colors off-diagonal ) . During CS-US pairings ( trial ) and inter-trial intervals ( ITI ) , nearly all misclassifications were due to the confusion between CS-alone trials and CS-US paired trials . ( B ) Classification accuracy ( mean ± SEM , 20 runs with 200 randomly subsampled cells ) for binary discrimination along stimulus relationship ( CS-alone vs . CS-US trials; Relationship ) , modality ( auditory vs . visual CS trials; Modality ) , and conditioning environment ( Box 1 vs . 2; Environment ) . In both task phases , the classification accuracy for Modality and Environment was higher than that for Relationship . *p<0 . 001 in posthoc Tukey HSD . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 015 Similar patterns were also observed when ensemble selectivity was quantified with a Support Vector Machine ( SVM ) classifier . During the trial and ITI periods , classification accuracy was well above chance level ( accuracy above chance; trial , 87 . 1 ± 0 . 3%; ITI , 82 . 7 ± 0 . 5% , n = 20 runs with 200 randomly sampled cells ) , suggesting the strong differentiations of ensemble firing patterns across six trial blocks ( Figure 6—figure supplement 1A ) . Subsequent binary discriminations of one of three task variables showed that in both the trial and ITI periods the classification accuracy for stimulus modality and conditioning environment was significantly higher than that for stimulus relationship ( Figure 6—figure supplement 1B; n = 20 runs with 200 randomly sampled cells , t-test with Bonferroni corrections , trial , both ps < 0 . 001 , ITI , both ps < 0 . 001 ) . The classification accuracy for the stimulus modality was higher than that for the conditioning environment during the ITI ( p=0 . 002 ) , but not the trial period ( p=0 . 693 ) . Collectively , the analyses with two independent measures of ensemble selectivity suggest that LEC formed ensemble codes that differentiated the six trial blocks not only during the trials but also during extended intervals between the trials . Ensemble firing during ITIs maintained high selectivity for the modality of the CS and the CS-US relationship even though the CS had been terminated tens of seconds before . It is therefore possible that the LEC prospectively signals which CS will be subsequently presented , by recovering the stimulus-environment associations learned through past experiences . Alternatively , the LEC may retrospectively hold information about which CS had been presented several seconds ago . To differentiate these possibilities , we examined the degree to which ITI ensemble firing changed upon the transition from one trial block to the next . This analysis was applied to 139 cells recorded from four rats which underwent six trial blocks in the same temporal sequence ( Table 1 ) . A state vector ( firing rates of 139 cells ) for an ITI ( template ) was compared against the vectors for the other ITIs using Pearson correlation coefficient ( Figure 7A ) . Ensemble firing in an ITI was similar to those in the other ITIs within the same trial block , though the similarity appeared to monotonically decrease across ITIs within each trial block . In contrast to the gradual changes , ensemble firing showed an abrupt transition after the shift from the ACS-US block in Box 1 to the ACS-alone block in Box 2 , ( Figure 7B , template = the 70th trial ) as well as the shift from the ACS-US block in Box 2 to the VCS alone block in Box 1 ( template = the 140th trial ) . These block shifts involving the move from one box to the other resulted in a greater change in the ensemble firing ( measured by a ‘Similarity Score’ , 1- no change , 0-uncorrelated ) than their spontaneous change within each block ( Figure 7C , n = 20 runs with randomly sampled 100 neurons; t-test with Bonferroni corrections , ps <0 . 001 ) . Notably , the block shift involving the change in CS modality and conditioning environment induced a greater change in ensemble firing than the block shift involving the environmental change alone ( p<0 . 001 ) . This difference was apparent during the last five recording sessions , but not during the first five recording sessions ( Figure 7—figure supplement 1 ) . Linear regression analyses showed that over nine blocks of five sessions , the degree of ensemble transition upon the block shift with the environmental and modality change was increased ( Figure 7D , R2 = 0 . 944 , p<0 . 001 ) while the degree of transition upon the block shift with the environmental change alone was decreased ( R2 = 0 . 508 , p=0 . 031 ) . 10 . 7554/eLife . 28611 . 016Figure 7 . LEC ensemble activity showed an abrupt transition upon the shift of trial block involving changes in the conditioning environment and the stimulus modality . ( A ) A matrix of state vector ( SV ) correlation in which the element rij represents the correlation coefficient between the SV computed during a 9 s interval ( ITI ) before the ith and jth CS presentation . The ITI number was continuously assigned to 210 ITIs according to the temporal order of six trial blocks as shown at the top of the matrix . The highlighted parts along the diagonal of the matrix ( black squares ) represent the r values between SVs within a trial block while the remaining parts represent those between different trial blocks . ( B ) The r for the last ITI in each trial block ( red line ) as a function of the temporal order of another ITI with which the r was computed . At the transition from the CS-alone block to the CS-US block in the same environment ( 20th , 90th , 160th ) , the r values with SVs in the same trial block ( ‘Within-block’ , light-gray shading ) was comparable to those with SVs in the following trial block ( ‘Between-blocks’ , dark-gray shading ) . In contrast , the r value abruptly dropped upon the block shift including the change in the environment ( 70th ) and CS modality ( 140th ) . ( C ) The Between-blocks r values ( mean ± SD , n = 20 repeats with 100 subsampled cells ) was significantly lower than the Within-block r values for two block shifts involving the environmental changes ( E , E+M; *p<0 . 001 posthoc Tukey HSD ) , but not in the three shifts involving the stimulus contingency change ( R ) . The Between-block r values were lower for the shift involving the changes in the CS modality and environment ( E+M ) than that involving the environmental change alone ( E; # , p<0 . 001 ) . ( D ) Over nine blocks of five sessions , the degree of ensemble transition ( Within-block r minus Between-block r ) increased for the block shift including the environmental and modality change ( E+M ) , but decreased for the block shift including the environmental change alone ( E ) . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 01610 . 7554/eLife . 28611 . 017Figure 7—figure supplement 1 . Changes in ensemble activity with experiences . A matrix of state vector ( SV ) correlation during the first five sessions ( Early ) and the last five sessions ( Late ) as Figure 7A . Ensemble firing patterns differentiated trial blocks more strongly depending on the CS modality than conditioning environment only after repeated exposures to the fixed temporal structure of trial blocks . DOI: http://dx . doi . org/10 . 7554/eLife . 28611 . 017 In contrast with the abrupt transition of ensemble activity triggered by the environmental change , ensemble similarity gradually decreased over several ITIs upon the shift from the CS-alone to CS-US trial blocks within the same environment ( Figure 7B , template = the 20th , 90th , and 160th trials ) . The change in ensemble firing during the first ITI was either smaller or comparable to the within-block change ( Figure 7C , ACS block in Box 1 , p=0 . 074; ACS block in Box 2 , p=0 . 239; VCS block in Box 1 , p<0 . 001 ) . Also , in two out of three block transitions , the degree of ensemble change did not significantly increase or decrease with experiences ( ACS block in Box 1 , R2 = 0 . 009 , p=0 . 805; ACS block in Box 2 , R2 = 0 . 882 , p<0 . 001; VCS block in Box 1 , R2 = 0 . 326 , p=0 . 108 ) . These results suggest that the transition of ensemble activity preceded the first presentation of the CS upon the block shifts involving the environmental change , but not the block shifts involving the contingency change within the same environment . And , in the former case , over two weeks of repeated exposures to the fixed temporal structure , the LEC ensembles gradually developed a greater sensitivity to the modality of the CS to be presented in the environment .
Accumulating evidence from behavioral and electrophysiological studies show that the LEC plays a critical role in associating events with the environments in which they take place ( Deshmukh and Knierim , 2011; Hunsaker et al . , 2013; Keene et al . , 2016; Tsao et al . , 2013; Van Cauter et al . , 2013; Wilson et al . , 2013 ) . We show here that cells in the LEC signaled this event-environment conjunction not only during events ( ‘phasic’ coding ) but also stably during extended periods between those events ( ‘tonic’ coding ) . The analyses of both single neurons and ensembles of neurons in the LEC revealed a highly selective code for conditioning epochs that were defined by both static , environmental features and associated transient stimuli . Selectivity was observed not only when stimuli were being presented but also , with comparable strength , during intervals between stimulus presentations . These results suggest that the LEC may signal not only what is happening but also what had happened previously in an environment . The identification of phasic and tonic codes was possible due to the types of stimuli that were used . Temporally precise , transient stimuli offered methodological advantages for discriminating between stimulus-evoked and stimulus-independent responses , thereby overcoming a limitation of many previous investigations which used physical objects to characterize neuron activity ( Deshmukh et al . , 2012; Deshmukh and Knierim , 2011; Keene et al . , 2016; Tsao et al . , 2013 ) . In the present data set , ~45% of cells responded to the CS , which falls within the range of selective cells for unimodal sensory stimuli ( 47–90% , Leitner et al . , 2016; Suzuki et al . , 1997; Xu and Wilson , 2012; Young et al . , 1997 ) but higher than that of cells selective for object location ( ~30% , Deshmukh and Knierim , 2011; Tsao et al . , 2013 ) . Some of these cells differentially responded to the CS depending on whether the CS was predictive of a US . Notably , these cells were also selective for the modality of the CS and environment in which the CS was presented ( Figure 3 ) . This conjunctive selectivity results in a clear separation of neural responses to the CS across six trial blocks ( Figure 6A ) , suggesting that the cells carry both behaviorally relevant and incidental information about the CS . In contrast to the strong selectivity for the stimulus and environmental features of the task , almost no cells in the LEC differentiated firing patterns depending on CR expression on a trial by trial basis . This finding suggests that the LEC may not be involved in the generation of CRs per se . Rather the LEC may play a role in differentiating the CS depending on the current situation . In parallel with stimulus-responsive cells , nearly all cells exhibited high selectivity for specific trial blocks by drastically discriminating between trial blocks during the non-stimulus periods ( Figure 4 ) . Ensembles were strongly selective for the trial blocks that used visual compared with auditory CS , yet only about half of these cells changed firing rates at the time of CS presentations and maintained stable firing patterns throughout the entire period of a trial block . Notably , virtually none of these cells were purely selective for the conditioning environment ( independently from the stimulus features ) , suggesting that they did not simply respond to perceivable differences between two conditioning boxes , such as patterns of the wall and lighting ( see also Lu et al . , 2013 ) . Moreover , given the absence of a significant difference in spontaneous eyeblink activity across six trial blocks ( Figure 1—figure supplement 1B ) , differential firing rates were not reflective of variations of activity level across blocks . Therefore , a plausible interpretation would be that these cells encode the conditioning environment with the history of sensory events that took place in that environment . The hypothesis that differential activity patterns between task epochs were driven by past events in those same contexts is consistent with prior observations of neural activity patterns in the LEC . In particular , prior experience with an object can later evoke selective firing patterns in the LEC even after that object has been removed: an ‘object-trace’ code ( Deshmukh and Knierim , 2011; Tsao et al . , 2013 ) . In these earlier studies , object-trace selectivity was observed in ~15% of LEC cells ( Tsao et al . , 2013 ) , while the present results showed that virtually all cells in the LEC differentiated sensory stimuli that had been experienced tens of seconds or longer in the past . This difference may be in part because our criteria for significance were based on the averaged firing rate across 9 s periods before CS presentations . This long time window serves as a kind of a ‘low-pass-filter’ that cancels out variations in spike timing across trials . In fact , the proportion of selective cells dropped down to ~50% when shorter bin sizes were used ( Figure 5D ) . Another critical difference between our study and previous studies is the number of times stimuli were presented . Over about two weeks , the rats of the present study repeatedly underwent the same temporal structure of the trial blocks , each of which included seventy presentations of the CS . The total number of repeated stimulus presentations , therefore , is far greater than the repetition of object presentations used in the previous studies ( Deshmukh and Knierim , 2011; Tsao et al . , 2013 ) . This task design could have facilitated the development of an internal model of the stimulus , which may be reflected in the high proportion of cells that maintain the selectivity during the stimulus-free periods . The strong selectivity for the combination of sensory and environmental features unique to each trial block was also observed at the ensemble level during both the stimulus and stimulus-free periods ( Figures 6 and 7 ) . Notably , changes in ensemble activity during stimulus-free periods were tightly coupled with the shift of trial blocks with the change in the conditioning environment . When rats moved from one conditioning box to the other , ensemble activity showed abrupt transition before the first presentation of the CS ( Figure 7A–C ) . Initially , the degree of ensemble transition was comparable regardless of whether the rats were to receive the different or same modality of the CS from/as the one in the previous block . Over two weeks , however , it became greater for the block shift with the different CS than the one with the same CS ( Figure 7C , D ) . This observation suggests that the LEC keeps track of what had happened in a given environment in the past and prospectively signals the information when the rat enters the environment next time . One may argue that LEC ensemble activity differentiated the trial blocks because of their temporal separation , rather than their difference in the stimulus and environmental features . In support of this view , some cells in the cortex change their responses to a sensory stimulus ( Fahy et al . , 1993; Riches et al . , 1991; Suzuki et al . , 1997 ) or environment ( Takehara-Nishiuchi et al . , 2013 ) over repetitive presentations of the stimulus , suggesting that they may keep track of the temporal order or repetition of experiences . In fact , within each trial block , the similarity of ensemble firing between stimulus-free periods decreased with time ( Figure 7B ) . This time-dependent change in the ensemble activity , however , was substantially smaller than the changes upon the shift of two trial blocks in the different environment ( Figure 7B ) , suggesting that time alone is not sufficient to account for the observed across-block changes in ensemble firing . Comparing the relative impact of time , environment , and past experiences on the LEC activity requires future studies which monitor the change of LEC ensemble activity while controlling the temporal order of experiences with and without changes in the environment or sensory events . To disentangle the coding properties of the LEC that resulted in trial block selectivity , our correlation analysis compared a relative difference in the selectivity for three features ( Figure 6 ) . In general , ensemble selectivity was higher for stimulus modality ( auditory vs . visual CS ) compared with conditioning environment ( Box 1 vs . 2 ) . This observation could be considered alongside previous observations that LEC ensembles show greater selectivity for object identity compared with object location ( Keene et al . , 2016 ) and suggests that the physical features of sensory events have a strong influence on the neuron activity in the LEC . In parallel the selectivity of LEC ensembles was weaker for stimulus relationship ( CS-US paired vs . CS alone ) than stimulus modality . In our paradigm , stimulus relationship was a critical factor for adaptive behavior while stimulus modality and conditioning environment were not . The stronger selectivity for behaviorally irrelevant over relevant features sharply contrasts with the opposite selectivity observed in neuron ensembles in the medial prefrontal cortex ( mPFC ) during the same behavioral paradigm ( Morrissey et al . , 2017 ) . These findings indicate a difference in information organization between the two regions , the LEC being more sensitive to incidental details of experiences and the mPFC being more sensitive to their behaviorally relevant features . The difference in feature selectivity may underlie the complementary roles that the hippocampal system and prefrontal cortex play for memory formation , consolidation , and assimilation ( Morrissey et al . , 2017; Preston and Eichenbaum , 2013 ) . Memories of daily experiences are thought to reside in a hippocampal-rhinal-neocortical network within which information becomes increasingly well integrated as it flows from the neocortical associative areas to the hippocampus ( Eichenbaum , 2000; Squire , 1992 ) . The presently observed selectivity of LEC cells suggests that the perceivable and relational information of experiences has already been highly integrated at the level of the LEC , even to a comparable level that was reported in the hippocampus ( Komorowski et al . , 2013; Leutgeb et al . , 2005; Moita et al . , 2003; Wood et al . , 1999 ) . The LEC , therefore , may not differ from the other regions regarding the level of information integration , but it may make a unique contribution to memory formation and retrieval by stably maintaining the highly integrated information throughout the experience . Such representations may serve as a scaffold in which efferent regions , such as the hippocampus , embed temporal dynamics of events within an experience , thereby facilitating accurately encoding the content of each experience . They also play a role in the retrieval of appropriate information based on the similarity between a present situation and situations encountered in the past . In summary , the present findings support an emerging view that all regions within the hippocampal system encode the highly integrated information of sensory and environmental features of an experience ( Keene et al . , 2016 ) . The LEC may be unique in that it signals the information not only during specific events but also stably throughout the experience . This coding property may support the LEC’s involvement in the accurate binding of physical , relational , and contextual features of everyday experiences ( Eichenbaum , 2000; Morrissey and Takehara-Nishiuchi , 2014; Squire , 1992 ) .
All surgical and experimental procedures were approved by the Animal Care and Use Committee at the University of Toronto . Seven male Long-Evans rats ( Charles River Laboratories , St . Constant , QC , Canada ) weighing 500–600 g at the time of surgery were used in the experiment . The rats were single-housed in Plexiglas cages in a colony room with ad libitum access to food and water . They were maintained on a reversed 12 hr light-dark circadian cycle , and all experiments took place during the dark part of the cycle . Prior to the experiment rats were handled 1–2 times per week . Microdrives and electrodes were constructed in-house following a modified procedure described in a previous report ( Kloosterman et al . , 2009 ) . A 3D-printed plastic base housed twelve individually movable tetrodes , two reference electrodes , and an electrode interface board ( EIB-54-Kopf , Neuralynx , Bozeman , MT , USA ) . Tetrodes were made by folding in four and twisting a 36 cm wire ( 12 µm polyimide coated nichrome wire , Sandvik , Stockholm , Sweden ) . Each tetrode was connected to the plastic base with a shuttle containing a custom-made screw which allowed a precise control of the tetrode movement . From the top , individual tetrode wires were connected to the interface board with gold pins . From the bottom , the electrode tips were cut , and gold plated to reduce impedance to 200–250 kΩ . At the base , the tetrodes were arranged in a bundle with a honeycomb pattern , with the total area of the bundle not exceeding 1 . 2 × 1 . 2 mm . During the surgery , the base of the drive with all tetrodes retracted was positioned at the surface of the brain in the right hemisphere ( 6 . 5 mm posterior , 5 . 3 mm lateral from Bregma at a 5° lateral angle ) secured to the skull with a self-adhesive resin cement . During the surgery , tetrodes were lowered either 3 mm ( rats 1–4 ) or 5–7 mm ( rats 5–7 ) . The reference electrodes were placed at the white matter tracts between the entorhinal cortex and the ventral hippocampus . Four stainless steel wires , two for delivering stimulation and two for recording electromyographic ( EMG ) activity , were implanted in the muscle of the left , upper eyelid and connected to the interface board . Two screws serving as a ground were placed in the skull above the cerebellum and the parietal cortex . Across 7–10 days following the surgery , the rats were daily connected to the recording system to monitor the signal while tetrodes were gradually lowered to the lateral entorhinal cortex ( LEC ) . The conditioning began at least seven days after the surgery . After each conditioning session , the tetrodes were lowered 30–65 µm to record different cells every day . Therefore , all cells were recorded only in one session . To allow for tetrodes to stabilize after each new placement and to avoid a potential electrode drift during the recording , the following recording session was conducted at least 20 hr after the electrode adjustment . Experimental apparatus consisted of two 20 × 20 × 70 cm boxes , connected with a 21 . 5 × 10 × 70 cm walkway ( Figure 1B ) . The rats could freely move between the boxes and walkway when a partition between them was lifted by the experimenter . The first box was dark with uniformly painted walls . The second box was lit and had walls painted in a black and white striped pattern . Daily recording sessions consisted of six blocks of classical trace eyeblink conditioning ( Figure 1C ) in which rats associated a neutral conditioned stimulus ( CS ) with an unconditioned stimulus ( US ) that was presented 500 ms after the offset of the CS . The CS was either 85 dB , 2 . 5 kHz pure tone ( ACS , Multipurpose Sound Generator ENV-230 , Med Associate Inc . , St . Albans , VT , USA ) or a white LED light pulsed at 50 Hz ( VCS ) . The US was a mild electric stimulus applied near the left upper eyelid ( a 100 Hz square pulse at 0 . 3–2 . 0 mA ) , generated by a stimulus isolator ( ISO-Flex , A . M . P . I . , Jerusalem , Israel ) . Seven rats received six blocks of this conditioning , each of which included one of two CS and took place in one of two visually distinct rooms ( Figure 1B ) in a fixed temporal sequence ( Figure 1C , Table 1 ) . For example , a rat was first placed in one of the rooms ( Box 1 ) and presented with the ACS alone for 20 times , followed by 50 pairings of the ACS and US . The rat was then gently forced to walk over to the other room ( Box 2 ) and received 20 presentations of ACS alone followed by 50 ACS-US pairings . Finally , the rat was forced to walk back to the original room and received 20 presentations of visual CS ( VCS ) alone followed by 50 pairings of VCS and US . During the first two days , the rats underwent the same temporal sequence of the blocks without the presentation of the US . Single-unit activity in the LEC and the electromyogram activity ( EMG ) in the eyelid were recorded by using a Cerebus neural signal processing system ( Blackrock Microsystems , Salt Lake City , UT , USA ) . A rat was connected to the signal processing system through an electrode interface board ( EIB-54-Kopf , Neuralynx , Bozeman , MT , USA ) and the Omnetics headstage adapter ( EIB-54K , Blackrock Microsystems ) . The signal was digitized at a headstage ( CerePlex M 64 , Blackrock Microsystems ) at 30 kHz and transmitted to the signal processing system . The threshold voltage for signal acquisition of action potentials was set at 50–63 μV . And the signal exceeding the threshold in one of the four channels of a tetrode was recorded for 1 ms . The signal from tetrodes was amplified and filtered above 250 Hz . EMG activity was continuously sampled at 10 kHz and filtered between 250–5000 Hz . Learning of the association between the CS and US was measured based on the frequency of adaptive conditioned responses ( CR ) which were defined as an increase in EMG amplitude prior to US onset ( Morrissey et al . , 2012 , 2017 ) . The instantaneous amplitude of the EMG signal was calculated as the absolute value of the Hilbert transform of the EMG signal . Then , two values were calculated in each trial: ( 1 ) the averaged amplitude of EMG signals during a 200 ms period before CS onset ( pre-CS ) and ( 2 ) the averaged amplitude during a 200 ms period before US onset or the corresponding interval in the CS-alone trials ( CR value ) . A threshold was defined as the average of all pre-CS values plus one standard deviation . If the pre-CS value was greater than 130% of the Threshold in a trial , the trial was discarded due to the hyperactivity of a rat immediately before the CS presentation . If the pre-CS value was below the Threshold and CR-value was greater than 110% of the Threshold , the trial was considered to have a CR . Finally , in the case when the pre-CS value was between the Threshold and 130% of the Threshold , the trial was considered to have a CR only if the CR value above the Threshold was five times greater than the pre-CS value above the Threshold . CR% was calculated as a ratio of the number of trials containing the CR to the total number of valid trials ( i . e . , total trials minus hyperactive trials ) . To examine whether eyeblink frequency that was not time-locked to the CS presentations increased with the conditioning , we applied the same analysis to EMG signals 480–280 ms before CS onset ( pre-CR% ) . CR% and pre-CR% were calculated separately in each of six trial blocks . It was compared across sessions and conditions with a two-way repeated measures ANOVA with sessions and trial blocks as within-subjects factors . To compare general activity level between the trial blocks , the amplitude of EMG trace was averaged across all inter-trial intervals of each trial block . Because the values were not distributed normally , they were converted to ranks and compared with a Kruskal-Wallis test . Putative units were isolated offline using an automated clustering software package ( KlustaKwik , K . D . , Harris , Rutgers , The State University of New Jersey , Newark , NJ , USA ) , followed by manual sorting ( MClust , D . A . Redish , University of Minnesota , Minneapolis , MN; Waveform Cutter , S . L . Cowen , University of Arizona , Tucson , AZ , USA ) . Subsequent data analyses used only units in which the amplitude and shape of spike waveform were consistent across the entire recording period and the distribution of the inter-spike intervals did not contain more than 1% of the spikes within a 2 ms refractory period . Individual cells were first categorized into two types , CS-responding and non-responding cells . A cell was categorized as a CS-responding if its firing rate during the trial period ( from CS onset to US onset , 600 ms ) was significantly different from its firing rate during the period before CS onset ( 1 s ) . The significance was tested by random permutation tests which examined whether the observed firing rate difference fell within the 5% upper tail of the distribution of chance firing difference estimated by randomly assigning each firing rate to either the trial or the pre-CS period ( 1000 re-assignments ) . Subsequently , the selectivity of firing rates for three task variables was quantified as a Differential Index , which compared mean firing rates between two conditions: Differential Index = ( Fr1 – Fr2 ) / ( Fr1 + Fr2 ) where Fr1 and Fr2 are averaged firing rates across trials in two conditions . To test the selectivity of CS-evoked firing rates , firing rates during the 600 ms period from CS onset to US onset were used . To examine the selectivity of firing rates during intervals between trials , firing rates during the 9 s period before CS onset ( ITI ) were used . Because the duration of intervals between the trials ranged from 20 to 40 s , the ITI period started at least 11 s after US offset . This 9 s period was chosen to avoid any contamination of lasting firing rate changes that continued for a few seconds after US offset . For the selectivity for stimulus relationship , Fr1 was the mean firing rate during CS-alone trials , and Fr2 was the mean firing rate during the corresponding CS-US paired trials with the same CS and in the same conditioning environment . For the selectivity for stimulus modality , Fr1 was the mean firing rate during trials with the auditory CS in Box 1 , and Fr2 was the mean firing rate during trials with the visual CS in Box 1 . For the selectivity for the conditioning environment , Fr1 was the mean firing rate during trials with the auditory CS in Box 1 , and Fr2 was the mean firing rate during trials with the auditory CS in Box 2 . Raw differential indices were converted to absolute values . Then , in each cell the differential index at chance was estimated as 5% upper tail of its corresponding distribution of differential indices with shuffled trial labels ( random permutation test , 1000 shuffles , α = 0 . 05 ) , and this value was subtracted from the value with the real trial label . A cell was judged as selective for a task variable if the shuffle-corrected differential index was greater than zero . During the 9 s ITI period , firing rates of each cell were binned into fifteen 600 ms bins and averaged across all trials in each of six trial blocks . The averaged , binned firing rates were divided by their sum , resulting in a distribution of firing rates across the bins . Kullback-Leibler divergence ( DKL ) between this distribution ( P ) and a uniform distribution ( Q ) was calculated as follows:DKL ( P||Q ) =∑iP ( i ) log2P ( i ) Q ( i ) As a measure for the within-block variation of firing rates , DKL was calculated across 15 bins in each of six trial blocks . As a measure for the across-block variation of firing rates , DKL was calculated across six trial blocks in each of 15 time bins . The median of the values was used as the within and across DKL for each cell . This analysis was applied to cells recorded during a block of CS-US trials in which a rat expressed the CR in minimum 20 trials ( CR trials ) and did not in minimum 20 trials ( non-CR trials ) . To match the number of two types of trials , 20 trials were randomly sampled from each trial type . The Differential index was calculated with FR1 as firing rates during the CR trials and FRs as firing rates during the non-CR trials . The significance of the selectivity was determined by random permutation tests as shown in ‘Single unit selectivity . ’ To compare the selectivity of population firing patterns during trials for three task variables , we first constructed firing rate vectors that consisted of binned firing rates of all cells ( state vectors ) in each trial ( CS + subsequent CS-US interval , 600 ms ) . In each block , ten state vectors were randomly selected and averaged , resulting in one state vector for each of six trial blocks . The firing rate of each cell in the state vectors was divided by its own maximum across the six blocks . We then calculated Pearson correlation coefficient ( r ) between the state vectors in two trial blocks with the ACS and VCS in the Box 1 ( Modality , ACS alone vs . VCS alone , ACS-US vs . VCS-US ) , two blocks with the ACS in the Box 1 and 2 ( Environment , ACS alone in Box 1 vs . ACS alone in Box 2 , ACS-US in Box 1 vs . ACS-US in Box 2 ) , or two blocks of ACS-alone trials and ACS-US paired trials in the Box 1 ( Relationship , ACS alone vs . ACS-US , VCS alone vs . VCS-US ) . As controls , we calculated r values of state vectors between odd- and even-numbered trials within the same trial block ( ACS-US paired in Box 1 , VCS-US paired in Box 1 , an estimate of the highest r value ) and pairs of trials without any overlapping task variable ( ACS alone in Box 2 vs . VCS-US paired in Box 1 , ACS-US paired in Box 2 vs . VCS alone in Box 1; an estimate of the lowest r value ) . We repeated these steps for twenty times , which generated 40 r values ( 2 pairs of trial blocks × 20 repeats ) for each comparison . The same analysis was applied to firing rates during 600 ms bins randomly selected from 9 s periods before the CS onset ( ITI period ) . The difference in the r values across these comparison types was quantified with a one-way ANOVA with posthoc comparison with Tukey HSD . To examine how the ensemble activity changed upon the transition from one trial block to the next , we selected 139 cells recorded from four rats which underwent the trial blocks in the same temporal order ( Rats 1 , 2 , 3 , 7 , see Table 1 ) . We first randomly selected 100 cells out of 139 cells and constructed a state vector for each inter-trial interval ( a 9 s period before CS onset ) . The firing rate of each cell in the state vectors was divided by its own maximum across all ITIs . Pearson correlation coefficient ( r ) was calculated for each state vector ( template ) with all other state vectors . This generated a 210 × 210 matrix of r values ( Figure 7A ) . The change in the state vectors after the block shift was quantified with the averaged r values of the last trial in a trial block with the first 19 trials in the following block ( Between-block Similarity Score ) . The change in the state vectors within a trial block was measured as the averaged r values of the last trial in a trial block and the 19 preceding trials ( 1st–19th CS-alone trials or the 31st to 49th CS-US trials ) in the same block ( Within-block Similarity Score ) . These steps were repeated for twenty times each of which used 100 randomly subsampled cells . The t-test with Bonferroni corrections was used to compare between the Between- and Within-block Similarity Scores in each comparison type or the Between-block Similarity Scores between the comparison types . To examine how the degree of ensemble transition changed over about two weeks of recording sessions , 139 cells were sorted based on the session on which they were recorded . We then repeated the above-mentioned analysis on a series of 49 cells taken from the sorted 139 cells with an increment of 10 cells . This roughly corresponded to examining ensemble firing in a block of five sessions with an increment of one session . The Between-block Similarity Score was subtracted from the Within-block Similarity Score . Linear regression analyses were used to test whether the value changed over the session blocks . As a secondary measure of ensemble selectivity , we used a machine learning algorithm , a Support Vector Machine ( SVM ) classifier ( Chang and Lin , 2011; Cortes and Vapnik , 1995 ) . The decoding accuracy of the classifier represented a degree to which ensemble firing patterns differentiated three task variables and their combination . The procedure was similar to that used in our previous study ( Morrissey et al . , 2017 ) . SVM classification was performed in MATLAB ( Mathworks , Natick , MA , USA ) with the algorithms from the open source LIBSVM library ( Chang and Lin , 2011 ) . The SVM classifier constructs a model based on a set of attributes from the training data and then predicts the target values based on the test data attributes . In the present study , the attributes were normalized firing rates of a neuron population in a trial from one of the six trial blocks . The target values were the identities of the sampled trials . Each run of SVM analysis used a set of 200 randomly subsampled cells ( total 20 runs ) . Population firing rate vectors were constructed by concatenating the responses of these cells during a 600-ms period from CS onset to US onset ( trial ) or a 600-ms period randomly selected from a 9 s period before the CS onset ( ITI ) in a trial from one of six trial types . In each cell , the firing rate in each trial was divided by the maximum firing rate of the cell among all trials . Note that the cells were recorded in separate sessions from seven rats , and thus we ignored any correlated activity between cells . The classifier constructed a model with the Gaussian radial basis function kernels . To maximize the classification accuracy , two SVM parameters , cost and gamma , were identified by performing a grid search over a range of values , using the firing rates from all available trials for a given target value . Then , a new population firing rate matrix was generated from firing rates during twenty trials of each of six trial blocks . These trials were randomly drawn , without replacement , from all trials in a given trial block . Then , in each cell , the firing rate in each trial was divided by the maximum firing rate of the cell among the 120 trials ( 20 trials × 6 blocks ) . Half of the trials ( 10 trials from each block ) were then used to train the SVM classifier with the parameters selected by the grid search . The remaining trials ( 10 trials from each block ) were then used to test the decoding accuracy after training . The process was repeated 20 times using a different sampling of 60 training and 60 test trials each time . The accuracy was defined as a proportion of correct predictions out of 1200 tests ( 60 test trials × 20 sets of randomly selected trials ) . To quantify the selectivity for one of three trial variables , the SVM classification was performed after collapsing six trial types into two trial types ( for the modality feature , trials with the auditory CS and visual CS; for the environmental feature , trials in Box 1 and Box 2; for the relationship feature , CS-alone and CS-US paired trials ) . Classification accuracy was compared between three task features with t-test with Bonferroni corrections . | The context in which an event occurs plays a large role in how the brain understands and responds to the event . While a key part of context is where we are , contexts can also change within the same space: different meetings are held at different times and with different people in the same room , and a grassy field can be a place of intense competition or a place to relax and gaze at clouds . However , we have little understanding of how the brain sets up and maintains a sense of context . A region of the brain called the lateral entorhinal cortex ( LEC ) responds to events as they happen , but may also maintain a record of past experiences , and helps us to learn new associations between events . To find out how LEC neurons might represent context , Pilkiw et al . measured the activity of individual LEC neurons in rats as they experienced different combinations of events and environments . In each trial , the rats were placed in one of two different rooms and exposed to one of two sensory cues ( sound or light ) six times , either alone or , to test learning , paired moments later with a mild stimulation to the eyelid . The gaps between the cues lasted from 20 to 40 seconds . As expected , some LEC neurons responded to the sensory cues , and varied their responses to cues depending on whether or not they were paired with eyelid stimulation . What was much more striking is that almost all cells in the LEC behaved very differently in different contexts , both in response to the cues and also during the long gaps between the cues . This suggests that the LEC provides the brain with information about the circumstances of an event , and may be the reason we expect different events under different circumstances – even if we are in the same place . We tend to underestimate how much we rely on context to remember events and to guide our behavior . Many disabling health conditions , including addiction , post-traumatic stress disorder and obsessive-compulsive disorder , are affected by context . For example , people who are trying to overcome drug addiction can often reduce their cravings by avoiding places and situations in which they previously used the drug in question . Understanding how the LEC represents context may therefore help us to develop treatments that target this brain region in order to alter harmful behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Phasic and tonic neuron ensemble codes for stimulus-environment conjunctions in the lateral entorhinal cortex |
Cardiorespiratory recovery from apneas requires dynamic responses of brainstem circuitry . One implicated component is the raphe system of Pet1-expressing ( largely serotonergic ) neurons , however their precise requirement neonatally for homeostasis is unclear , yet central toward understanding newborn cardiorespiratory control and dysfunction . Here we show that acute in vivo perturbation of Pet1-neuron activity , via triggering cell-autonomously the synthetic inhibitory receptor hM4Di , resulted in altered baseline cardiorespiratory properties and diminished apnea survival . Respiratory more than heart rate recovery was impaired , uncoupling their normal linear relationship . Disordered gasp recovery from the initial apnea distinguished mice that would go on to die during subsequent apneas . Further , the risk likelihood of apnea-related mortality associated with suppression of Pet1 neurons was higher for animals with baseline elevated ventilatory equivalents for oxygen . These findings establish that Pet1 neurons play an active role in neonatal cardiorespiratory homeostasis and provide mechanistic plausibility for the serotonergic abnormalities associated with SIDS .
Tissue oxygen levels are maintained within a narrow , life-sustaining range through the coordinated actions of oxygen flux through breathing , red blood cell loading-unloading , and circulation via heart pumping . This vital cardiorespiratory homeostasis is subserved by the interplay of neural circuits that must reliably function for the duration of an organism's ex utero life . During early neonatal life , mammals are especially vulnerable to homeostatic impairments given their relatively lower oxygen reserve due to smaller lung volume and faster decline in blood oxygen levels during hypoxic conditions owing to the steep desaturation kinetics of the still presiding fetal oxyhemoglobin ( Fewell , 2005 ) . Paradoxically , young mammals including human infants undergo more frequent interruptions in breathing , called apneas , that are typically coupled with heart rate slowing , bradycardias ( Daily et al . , 1969; Fewell et al . , 2005; Kelly et al . , 1985; Southall et al . , 1980 ) . At the same time , neonatal mammals have an especially robust , protective homeostatic response – referred to as autoresuscitation – which utilizes gasping after an apnea to increase blood oxygen levels to facilitate restoration of heart rate and eupneic breathing ( Gershan et al . , 1992; Guntheroth and Kawabori , 1975; Jacobi et al . , 1991; Saiki et al . , 2001 ) . Here we present progress in delineating aspects of the underlying neurobiology , querying the neonatal importance for Pet1-lineage neurons , largely serotonergic ( 5-hydroxytryptamine- , 5-HT-producing ) , in the apnea recovery response . Serotonergic transmission has been implicated in the autoresuscitation response in rodents ( Barrett et al . , 2016; Cummings et al . , 2011a; Erickson and Sposato , 2009; Sridhar et al . , 2003 ) and in cardiorespiratory modulation in humans and rodents , including in the sudden infant death syndrome ( SIDS ) ( Duncan et al . , 2010; Feldman et al . , 2003; Hodges and Richerson , 2010; Kinney and Thach , 2009; Peña and Ramirez , 2002; Ptak et al . , 2009; Ray et al . , 2011 ) . In mice , chronically disabling vesicular neurotransmission from 5-HT neurons ( ‘silencing’ them ) from mid-embryogenesis onward results in pups with diminished capacity to recover from induced asphyxic apnea ( Barrett et al . , 2016 ) . Impairment was observed in pups across postnatal ( P ) days 5–8 but no longer by P12 , suggestive of a neonatal period of heightened vulnerability to neurological dysfunction and cardiorespiratory stressors . Rodent pups with 80–90% reduction in medullary 5-HT content , resulting from perturbation chemically [5 , 7-dihydroxytryptamine treatment ( Cummings et al . , 2011b ) ] or genetically [germ line Pet-1 gene deletion ( Cummings et al . , 2011a; Erickson and Sposato , 2009; Erickson et al . , 2007 ) or tryptophan hydroxylase 2 ( Tph2 ) deletion ( Chen et al . , 2013 ) ] also showed impaired recovery from apneic challenges . These rodent data build mechanistic plausibility for the SIDS-associated findings of postmortem brainstem 5-HT neuron abnormalities ( Duncan et al . , 2010; Paterson et al . , 2006 ) , cardiorespiratory tracings showing prolonged and more frequent apneic and bradycardic events associated with progression leading to death ( Meny et al . , 1994; Poets et al . , 1999; Sridhar et al . , 2003 ) , and the epidemiological determination of a postneonatal critical period of heightened SIDS risk ( 2–4 months of age ) ( American Academy of Pediatrics Task Force on Sudden Infant Death Syndrome , 2005 ) . Yet evidence for an acute , real-time role for postneonatal serotonergic neurons in modulation of the autoresuscitation response remains lacking . Studies have largely involved chronic or extended 5-HT system manipulations spanning embryonic ( Barrett et al . , 2016; Chen et al . , 2013; Cummings et al . , 2011a; Erickson and Sposato , 2009 ) and/or postneonatal development ( Yang and Cummings , 2013 ) in which secondary , compensatory network changes can occur in addition to the primary , engineered serotonergic neuronal abnormality . Here we report progress in this area through studies in which we acutely induced Pet1-neuron perturbation in vivo at P8 and measured cardiorespiratory outcome and recovery across a chain of asphyxic-induced apneas . We used an inducible ( clozapine-N-oxide ( CNO ) -triggered ) neuronal inhibition strategy ( Ray et al . , 2011 ) involving the cognate , synthetic inhibitory G protein-coupled receptor hM4Di ( also referred to as Di ) ( Armbruster et al . , 2007 ) to disrupt at P8 the activity of a raphe neuron population defined by expression of a Pet1 BAC transgene . Pet1 gene expression serves largely as a serotonergic marker ( Fyodorov et al . , 1998 ) , with the Pet1 BAC as a driver offering genetic access to 5-HT neurons ( 5-HT+ , Pet1+ , tryptophan hydroxylase 2 ( Tph2+ ) cells ) , plus a small subset of raphe neurons that are negative for 5-HT while nonetheless positive for Pet1 expression ( 5-HT- , Tph2low , Pet1+ cells ) ( Barrett et al . , 2016; Okaty et al . , 2015; Pelosi et al . , 2014; Sos et al . , 2017 ) . Our results suggest that Pet1-neuron activity is required neonatally for maintaining baseline heart rate and ventilation and for normal survival rates in response to apneas . Indeed , CNO-Di-mediated perturbation of Pet1+ neurons at P8 renders pups significantly more likely to die after an apnea when compared to CNO-treated sibling controls . Further , we found that this acute manipulation of Pet1+ neurons primarily affected the respiratory components of apnea recovery while sparing much of the cardiac response – a cardiorespiratory uncoupling not predicted by earlier chronic perturbation studies ( Barrett et al . , 2016; Cummings et al . , 2011a; Cummings et al . , 2011b ) and which runs counter to the linear relationship between breathing and heart rate recovery present in control pups . Additionally , we found that a disordered gasp response to the first apnea characterized pups that succumbed to a subsequent apnea in the assay . As well , post hoc analyses identified specific respiratory features associating with autoresuscitation failure . These findings support a model in which Pet1-neuron activity is required neonatally for robust apnea recovery and may , by extension , inform strategies for pediatric autoresuscitation and SIDS prevention .
To study Pet1+ neurons neonatally , we pursued a chemogenetic Di strategy allowing for noninvasive , inducible , targeted neuronal perturbation suitable to the small size of P8 mouse pups and the physical and temporal constraints of our plethysmographic apnea-induction-recovery ( autoresuscitation ) assay . We applied the Flpe recombinase-encoding BAC transgenic driver Pet1-Flpe ( Jensen et al . , 2008 ) acting on the engineered Flp-responsive ROSA26 ( R26 ) allele designated Gt ( ROSA ) 26Sortm ( CAG-FSF-CHRM4* ( Di ) ) Dym ( denoted for ease in short-hand as RC-FDi ) ( Brust et al . , 2014; Ray et al . , 2011 ) to drive in Pet1 neurons expression of Di ( Armbruster et al . , 2007 ) , the Gi/o protein-coupled receptor with engineered selectivity for the injectable synthetic ligand CNO . Established previously , CNO-triggering of Di signaling in Pet1-lineage neurons using this transgenic approach results in hyperpolarization and diminished excitability in vivo and in vitro ( Brust et al . , 2014; Ray et al . , 2011; Teissier et al . , 2015 ) . Additionally , CNO-triggered Di signaling has been shown to inhibit synaptic transmission ( Stachniak et al . , 2014 ) . The Pet1-Flpe transgene drives Flpe expression from mid-embryogenesis onward in a majority of Pet1-expressing postmitotic neurons ( Jensen et al . , 2008 ) and reliably mediates recombination of RC-FDi ( Ray et al . , 2011 ) and other R26 engineered alleles ( Brust et al . , 2014; Jensen et al . , 2008 ) . Thus double transgenic Pet1-Flpe , RC-FDi pups allow for acute , inducible perturbation of Pet1 neurons for neonatal , whole-animal functional study before and following intraperitoneal ( i . p . ) CNO administration . Immunohistochemically-stained sections of the raphe from P8 double transgenic Pet1-Flpe , RC-FDi mouse pups , referred to as Pet1-Di pups , confirmed protein expression of the HA-tagged Di receptor in serotonergic ( Tph2+ ) neurons ( Figure 1A–C’’’ ) , consistent with previously published Pet1-Flpe driver specificity in neonatal pups ( Barrett et al . , 2016 ) . The raphe location , proportion , and intensity of HA-Di immunodetection signal was qualitatively similar across Pet1-Di pups from independent litters ( Figure 1—figure supplement 1 ) . In sibling control single transgenic RC-FDi pups , referred to as control-Di pups ( harboring the unrecombined RC-FDi allele and thus negative for Di transcription ) , no HA-Di immunosignal was detected ( Figure 1D–F ) . To assay cardiorespiratory function in P8 mouse pups under conditions of room air ( RA ) ( Figure 1G , open rectangle a ) and then apnea-inducing , asphyxia conditions ( Figure 1G , filled rectangles b-e ) , we used head-out plethysmography and ECG with continuous recording of breathing ( frequency f , breaths • min−1; and pressure changes associated with respiratory activity used to calculate tidal volume VT , ml • g−1 ) , heart rate ( HR ) , oxygen consumption ( V˙O2; ml • min−1 • g−1 ) , and body temperature while maintaining pup thermoneutrality ( TB at 36 ± 0 . 05°C ) through chamber temperature adjustment ( TA at 35–36 ± 0 . 05°C ) . From these measurements along with body mass ( g ) , values were determined for minute ventilation ( V˙E; ml • min−1 • g−1 ) and ventilatory equivalents for oxygen ( V˙E/V˙O2 ) . To assess if Pet1 neurons at P8 modulate RA cardiorespiratory parameters , measurements were collected prior and during neuron perturbation ( Figure 1G , open rectangle a versus filled rectangle b’ of Figure 1H , respectively ) . Initial baseline homeostatic characteristics showed no significant difference between Pet1-Di and control-Di pups , indicating that mere expression of Di in Pet1-lineage cells ( not yet triggered by CNO ) as well as harboring and expression of the Pet1-Flpe transgene were neutral in this assay ( Table 1 ) . Following CNO injection ( Figure 1H , filled rectangle b’ ) , double transgenic Pet1-Di pups ( referred to as Pet1-Di-CNO ) exhibited statistically significant decreases in V˙E and HR ( p=0 . 04 and p=0 . 02 respectively , Figure 2A’ and B’ ) not observed in control-Di pups ( referred to as control-Di-CNO , Figure 2A and B ) . Coefficient-of-variation calculations for breathing f and HR suggest comparable dispersion of the data obtained during CNO exposure as compared to baseline prior to CNO injection ( Figure 2—figure supplement 1 ) . Pet1-Di-CNO pups with the highest V˙E prior to CNO injection exhibited the largest V˙E drop upon CNO administration and those with the lowest V˙E prior to CNO exhibited modest , albeit not statistically significant , increases ( Figure 2A’ and and C’ ) . This suggests that the ventilatory neurocircuitry may engage Pet1 neurons to allow for greater deviation from a standard homeostatic set point , such that when Pet1 neurons are inhibited the ventilatory dynamic range narrows overall . An alternative , more complex and arguably less likely technical explanation posits that pups with highest baseline V˙E values are pups with highest Di expression levels ( within the distribution determined by R26/CAG expression variation ) such that CNO-triggering drives a greater cellular and circuit perturbation ultimately reflected in larger decreases in V˙E . However , HR findings do not lend support for this latter explanation , given that such a correlation was not present between the magnitude of HR changes upon CNO administration as compared to either pre-CNO HR , pre-CNO V˙E , or change in V˙E ( Figure 2D’ , Figure 2—figure supplement 2 ) . In contrast to these Pet1-Di-CNO-specific effects , no statistically significant effects were observed on ventilatory equivalents for oxygen ( V˙E/V˙O2 ) for either group ( control-Di-CNO and Pet1-Di-CNO ) ( Figure 2—figure supplement 3A , A’ , C , C’ ) . Both groups though showed a subtle decrease in oxygen consumption following CNO and return to the plethysmograph chamber ( Figure 2—supplement figure 3B , B’ ) . We next queried whether CNO-Di-mediated disruption of Pet1 neurons at P8 altered pup recovery from repeated , episodic , asphyxia-induced apneas ( experimental paradigm schematized in Figure 1G and H , modified from previous studies ( Barrett et al . , 2016; Cummings et al . , 2011a; Erickson and Sposato , 2009 ) . We found autoresuscitation to be less effective in Pet1-Di-CNO pups in comparison with control-Di-CNO pups , resulting in a significant increase in mortality ( Figure 3A , one-tailed Fisher Exact Test with Lancaster's Mid-P correction p=0 . 04 ) . The calculated odds ratio for pup death as an outcome of asphyxic apnea in the face of Pet1-Di-CNO versus control-Di-CNO is 6 . 5 , suggesting a substantially increased vulnerability to apneas when Pet1 neuron activity is acutely perturbed . HR and breathing responses during a successful versus failed autoresuscitation are shown in Figures 1H and 3B , respectively . In general , successful autoresuscitation is characterized by a brief primary apnea and rapid recovery of normal HR and eupneic breathing following the onset of gasping ( Figure 1H ) , whereas failed autoresuscitation is characterized by a prolonged primary apnea ( delayed gasping ) and an inability to recover eupneic breathing despite gasping , and ultimately failure to sustain HR at a recovered or near-recovered level ( Figure 3B ) . Seven of twenty-two Pet1-Di-CNO pups compared to one of fifteen control-Di-CNO pups failed during the course of the assay to restore eupneic respiratory rhythm resulting in death ( Figure 3A ) . To query components of the autoresuscitation response and their possible dependency upon normal Pet1-neuron activity , we examined the recovery from each asphyxia bout ( Figure 4 ) , determining the recovery latency time ( τ ) to achieve at least 3 s of breathing frequency f or HR at levels ≥ 63% of the eupneic f and HR observed as baseline immediately preceding the given apneic challenge ( Figure 4A , a’–d’; Figure 1G and H ) . Time to 63% recovery was chosen for analysis because of the sensitivity likely offered via sampling that part of the recovery response which shows the largest amount of system change ( system recovery ) , as predicted by the time constant ( τ ) of a first-order , linear time-invariant system; while actual breathing and heart rate recovery systems may be more complex , sampling τ ( as opposed to other time points ) is our best prediction for maximally detecting recovery differences . Because not all pups survived the full 4-asphyxic-bout sequence ( Figure 4B and D ) , we also separately analyzed characteristics of the last-survived bout for each pup ( Figure 4C and E ) . Notably , there were some cases in which the fatal bout was nonetheless associated with a transient recovery of HR that met the τ conditions ( HR ≥63% of pre-bout baseline sustained for ≥3 s ) , thus these values were used in calculating the mean τHR for the particular bout ( Figure 4D , their inclusion being denoted by gray filled symbols at the top of the plot ) , and were included in the scatter plot of last-recorded τHR for each pup ( Figure 4E , gray filled symbols ) . In contrast to HR , fatal bouts were never found associated with a transient recovery of breathing f that met the τ conditions ( f ≥ 63% of pre-bout baseline sustained for ≥3 s ) . Analysis of τf across asphyxia bouts showed a significant lengthening between the first and fourth bouts for Pet1-Di-CNO pups , but not for control-Di-CNO pups ( Figure 4B , two-way ANOVA interaction p=0 . 027 post hoc Tukey's multiple comparison test p<0 . 0001 ( Pet1-Di-CNO ) and p=0 . 3063 ( control-Di-CNO ) ) . This τf prolongation associated with repeated apneic challenges in Pet1-Di-CNO pups , as calculated , is an underestimation given that animals that died ( infinite attempted breathing rate recovery , if you will ) could not be included . This further emphasizes the importance of Pet1 neurons in enabling a rapid respiratory recovery response . Moreover , the mean τf in Pet1-Di-CNO pups during bout #4 was significantly prolonged by comparison to that of control-Di-CNO mice ( Figure 4B post hoc Tukey's multiple comparison test p=0 . 003 ) , indicating that , notwithstanding survival , the Pet1-Di-CNO pup response was abnormal . In analyzing the last-recovered bout for all Pet1-Di-CNO pups , τf was again found to be prolonged ( Figure 4C , p=0 . 04 . Note that one of the surviving Pet1-Di-CNO pups , during bout 4 , did not reach 63% of their pre-bout f within the allotted recording time ( 330 s ) and thus was assigned a τf of 331 s – again , leading to an underestimation of the Pet1-Di-CNO effect and thus of the importance of Pet1 neuron activity to respiratory recovery . Unlike τf , τHR recovery did not show an interaction with asphyxia bout for Pet1-Di-CNO pups nor control-Di-CNO pups , ( Figure 4D , two-way ANOVA interaction p=0 . 335 ) . Additionally , in contrast to the strong effects on τf observed in Pet1-Di-CNO pups , the τHR was not significantly different from that of control-Di-CNO pups at any point during the assay , including the last-recovered bout ( Figure 4E p=0 . 255 ) . Of the seven Pet1-Di-CNO pups that died during the assay , three nonetheless reached 63% of their baseline HR during the fatal bout before succumbing to cardiac failure , and thus could be included in the τHR analysis for that bout ( indicated as gray-filled symbols at the top of Figure 4D and in the scatter plot of 4E ) . The other four did not ( indicated as black-filled symbols at the top of 4D and plotted in 4E ) , and thus their τHR values were necessarily excluded , again underemphasizing the effect of Pet1-neuron function in HR recovery . The one control-Di-CNO pup that died also reached 63% of baseline HR during the terminal bout and thus was included in the τHR analyses ( gray-filled , blue outlined symbols in Figure 4D and E ) . Next , we plotted τf against τHR for each pup across all bouts to examine their relationship , given that Pet1-neuron silencing appeared to differentially affect breathing versus heart rate recovery . Applying a linear regression model to values generated from control-Di-CNO pups identified a nonrandom , linear relationship between τHR and τf , whether analyzing all asphyxic bout recoveries ( Figure 5A blue and extracted blue plot shown separately in 5A’ , Run's test linear deviation p=0 . 36 , R2 = 0 . 84 , and non-zero slope p<0 . 0001 ) or the recovery response to each individual asphyxia bout ( Figure 5B–E blue and extracted blue plot shown separately in 5B’−5E’ ) . These control findings suggest that the cardiorespiratory response characteristics of P8 mouse pups interact in a direct , linearly correlated fashion , reflecting a well-coordinated breathing and heart rate recovery , likely important for maintaining adequate perfusion . By contrast in Pet1-Di-CNO pups , the linear nature of this relationship appears disordered ( Figure 5A red and extracted red plot shown separately in 5A’ , Run's test linear deviation p<0 . 0001 , R2 = 0 . 73 , and non-zero slope p<0 . 0001 ) , with HR recovery proceeding in cases without the commensurate degree of breathing frequency f increases seen in controls . Appearing to drive a portion of these differences are the Pet1-Di-CNO pups that go on to die ( black-filled red symbols ) . Owing to their future mortality , they drop out from subsequent bouts and as such are under sampled when all asphyxic bouts are analyzed . Given this , we also applied the linear regression model to each asphyxic-bout recovery response ( Figure 5B–E” ) . Interestingly , during the initial asphyxia bout recovery , we found that the Pet1-Di-CNO pups have a weaker correlation between τHR and τf ( Figure 5B red and extracted red plot shown separately in 5B’ correlation p=0 . 08 , R2 = 0 . 17 , linear deviation p=0 . 68 ) , which differed from control-Di-CNO pups ( Figure 5B blue and extracted blue plot shown separately in 5B’ Run’s test linear deviation p=0 . 76 , R2 = 0 . 93 , and non-zero slope p<0 . 0001 ) . This was further evidenced by the slope differences ( Figure 5B p=0 . 001 ) . Similar to the recovery analysis that includes all bouts , our analysis of just the first-bout recovery responses showed that the Pet1-Di-CNO pups that go on to die ( Figure 5B and B” , black-filled symbols ) had the greatest decoupling of HR and breathing f recovery kinetics . Interestingly , this decoupling is less apparent in the recovery response to bouts #2 and #3 but pronounced again in bout #4 . Overall , perturbation of Pet1 neurons may result in a decoupling of the cardiac and respiratory components central to a robust autoresuscitation response ( sample illustrative tracings in Figure 5—figure supplement 1 ) . Because Pet1-Di-CNO pups were less able to recover from repeated apneas , we sought to determine whether particular cardiorespiratory responses around an apnea tracked with later mortality . We examined the initial induced apnea ( Figure 4A filled window a ) so as to focus on characteristics independent of later size effects associated with repeated apneas . We found that Pet1-Di-CNO pups that died , by comparison to control-Di-CNO siblings who survived ( 14 of 15 ) , exhibited a more disordered gasp response ( Figure 6A ) characterized by a smaller first gasp ( Figure 6B , One-way ANOVA p=0 . 047 , post hoc Tukey's multiple comparisons test p=0 . 04 for Pet1-Di-CNO dies and control-Di-CNO survives ) , a longer latency to that first gasp ( Figure 6C , One-way ANOVA p=0 . 013 , post hoc Tukey's multiple comparisons test p=0 . 02 for Pet1-Di-CNO survives versus Pet1-Di-CNO dies , and p=0 . 02 for Pet1-Di-CNO dies versus control-Di-CNO ) , and prolonged inter-gasp intervals ( Figure 6D , two-way ANOVA Genotype/survival p<0 . 0001 , post hoc Tukey’s multiple comparison test of means p<0 . 0001 and p<0 . 0001 for Pet1-Di-CNO dies versus control-Di-CNO survives , and Pet1-Di-CNO dies versus Pet1-Di-CNO survives , respectively ) during the recovery from the first asphyxia-induced apnea . As an indirect measure of gasping effectiveness , we examined the characteristic transient increase in HR between each gasp that is required for homeostasis recovery . Here too we observed Pet1-Di-CNO pups who went on to die had a lower average HR between each gasp as the gasp train progressed compared to Pet1-Di-CNO surviving pups and control-Di-CNO pups ( Figure 6E , two-way ANOVA genotype/survival p=0 . 012 and interaction p=0 . 043 , post hoc Tukey’s multiple comparison test p=0 . 042 and p=0 . 067 ( gasp interval III ) and p=0 . 044 and p=0 . 036 ( gasp interval IV ) for Pet1-Di-CNO dies versus control-Di-CNO survives , and Pet1-Di-CNO dies versus Pet1-Di-CNO survives , respectively ) . These findings suggest that the gasp characteristics of Pet1-Di-CNO pups that went on to die were not as effective at raising HR . Thus , disordered gasping after the first apnea bout associated with an increased risk for future mortality around subsequent asphyxic-apnea challenges . To determine if particular baseline homeostatic characteristics increased the risk likelihood of future mortality when confronted by asphyxic challenge in the face of acute Pet1-neuron inhibition , we performed logistic regression analyses using as the independent variable either the baseline HR , V˙E , V˙O2 , or V˙E/V˙O2 ( data points obtained from the time point indicated by open window a , Figure 1G ) and autoresuscitation outcome – death versus survival – as dependent variables . This approach would allow us to account for the state of the homeostatic network before acute perturbation . Of these input variables , only V˙E/V˙O2resulted in a statistically significant odds ratio ( Table 2 , p=0 . 027 , odds ratio of 1 . 399 ) , with higher values for V˙E/V˙O2 correlating with increased risk of future death in Pet1-Di-CNO pups . No differences were found between V˙E/V˙O2 mean and variance values between Pet1-Di-CNO and control-Di-CNO pups ( p=0 . 128 and p=0 . 392 , respectively ) . When Pet1-Di-CNO pups are separated by mortality , we similarly observe that Pet1-Di-CNO pups that go on to die have a significantly higher V˙E/V˙O2when compared to control-Di-CNO pups who survive ( Figure 6F , one-way ANOVA p=0 . 018 with Tukey’s multiple comparisons test p=0 . 01 ) . Thus , inhibition of Pet1 neurons in mouse pups whose baseline V˙E/V˙O2 value resides at the higher end of the V˙E/V˙O2 distribution , increases the probability that they will go on to die when confronted by repeated asphyxia-induced apneas . We also explored variability of baseline breathing ( Figure 6—figure supplement 1 ) and HR ( Figure 6—figure supplement 2 ) pre- versus during CNO exposure ( prior to any asphyxic-apnea challenge ) for Pet1-Di and control-Di pups by evaluating the standard deviation ( SD ) of the interbreath and interbeat intervals , the SDxSD axis of Poincare first return plot for interbreath interval and interbeat interval , and the root mean square of successive differences ( RMSSD ) . Significant findings were limited to interbreath interval parameters , specifically for Pet1-Di pups pre- versus during-CNO exposure that would go on to die in the assay ( Figure 6—figure supplement 1 ) , with CNO exposure ( and thus Pet1 neuron perturbation ) associating with a decrease in SD and decrease in estimated area of the Poincare first return plot , both suggesting a decrease in interbreath interval variability in the time domain .
Cardiorespiratory homeostasis involves central and peripheral neural circuits working in concert to sense and respond to tissue conditions of hypercapnia , hypoxia , and acidosis . These circuits in neonates are newly engaged for ex utero life , including recovery from apneas , which occur more frequently during infancy . Based on recent genetic mouse models implicating Pet1-expressing serotonergic neurons in cardiorespiratory homeostasis ( Barrett et al . , 2016; Brust et al . , 2014; Cummings et al . , 2011a; Cummings et al . , 2013; Erickson and Sposato , 2009; Erickson et al . , 2007; Ray et al . , 2011 ) , we hypothesized that in neonates too Pet1 neurons play an important , real-time role , including in the recovery response to apneas . To test this free of ambiguity associated with chronic , developmental perturbations and potentially hidden compensatory events ( Barrett et al . , 2016; Erickson et al . , 2007; Erickson and Sposato , 2009 ) , we applied an inducible , acute neuronal inhibition strategy involving targeted expression of hM4Di in Pet1 neurons , as achieved in double transgenic Pet1-Flpe , RC-FDi pups , with the cognate ligand CNO injected intraperitoneally at P8 to trigger Pet1-neuronal inhibition . In the presence or absence of this acute perturbation , we assayed cardiorespiratory function at baseline and during apnea induction and recovery , the latter allowing for exploration of the gasp response and the ability to rapidly restore HR and eupneic breathing . Significant findings include the following: ( 1 ) Repeated asphyxia-induced apneas during CNO exposure resulted in a greater frequency of failed autoresuscitation in Pet1-Di-CNO pups ( 7 of 22 Di-expressing pups ) as compared to control-Di-CNO pups ( 1 of 15 non-Di-expressing , RC-FDi-harboring pups ) . ( 2 ) Baseline room air cardiorespiratory function ( V˙E and HR ) was modestly but statistically significantly altered following acute , CNO-Di-mediated perturbation of Pet1 neurons in P8 pups ( Pet1-Di-CNO pups ) ; this contrasts with CNO-treated sibling controls ( control-Di-CNO pups ) which showed no detectable changes in these baseline properties . Findings suggest that Pet1 neurons may normally enable greater cardiorespiratory dynamic range , which narrows upon Pet1-neuron inhibition . ( 3 ) Pet1-Di-CNO pups during their last recovered asphyxic apnea-inducing bout , whether the assay-concluding fourth bout ( for pups that survived ) or the bout just prior to the fatal failed bout , took significantly longer to recover to 63% of the pre-bout baseline eupneic breathing as compared to controls . ( 4 ) In contrast to the impaired respiratory recovery characterizing Pet1-Di-CNO pups , the time to recover HR to 63% of the pre-bout baseline was indistinguishable from that of control-Di-CNO pups . ( 5 ) A linear relationship between HR and breathing ƒ recovery was observed in the autoresuscitation response of control-Di-CNO pups , but was decoupled in Pet1-Di-CNO pups . ( 6 ) The gasp response to the initial , survived apneic challenge was disordered in the Pet1-Di-CNO pups that would go on to die during a subsequent apnea; the first gasp was smaller , the latency to first gasp longer , inter-gasp intervals prolonged , and the HR increase became smaller as the gasp train progressed . ( 7 ) Pups exhibiting modest hyperventilation – within the high end of the distribution prior to any neuronal perturbation – had a higher risk likelihood for autoresuscitation failure when subjected to the combined stressors of acute Pet1-neuron inhibition and apneic challenge . Here we provide evidence through selective and , importantly , acute neuronal perturbation to support the hypothesis that Pet1 neurons at P8 play an active role in shaping the neonatal cardiorespiratory homeostatic set point and the capacity to mount a robust autoresuscitation response . We first explored baseline cardiorespiratory properties of P8 mouse pups and whether they changed following acute Pet1 neuron perturbation . The starting cardiorespiratory values were indistinguishable between Pet1-Di-expressing pups ( Pet1-Flpe , RC-FDi double transgenics ) and sibling controls ( RC-FDi single transgenics ) ( Table 1 ) suggesting relative neutrality around expression of the untriggered Di receptor in Pet1 neurons . Upon CNO administration , Di-expressing double transgenics exhibited cardiorespiratory changes , specifically an overall decrease , albeit subtle , in V˙E and HR ( Figure 2A , A’ and B , B’ , respectively ) . This is consistent with Pet1 neurons playing an active role in neonates in maintaining both respiratory and cardiac tone . Prior studies involving chronic , developmental disruption of Pet1 neurons showed only diminished HR by P8 , with V˙E levels indistinguishable from the control cohort ( Barrett et al . , 2016; Cummings et al . , 2011a; Cummings et al . , 2013 ) . This lack of detectable V˙E effect under conditions of chronic Pet1-neuron perturbation could reflect compensatory circuit plasticity around ventilation , but which occurs to a lesser extent around HR control . It is also possible that the inducible , acute neuronal perturbation approach offers greater sensitivity and thus capacity to uncover more extensive phenotypes: as applied here , it allowed each animal to serve as its own control , enabling within-animal comparisons across pre- versus during-perturbation measurements , minimizing between-animal variability . An additional benefit of the inducible-perturbation approach is that body weight variation among pups was negligible , lessening technical variability associated with acquiring plethysmographic measurements on especially small pups; by contrast , chronic developmental perturbations of Pet1 neurons results in impaired growth and diminished body weight ( Barrett et al . , 2016; Cummings et al . , 2011a; Cummings et al . , 2013; Erickson et al . , 2007; Pelosi et al . , 2014; Yang and Cummings , 2013 ) . Also uniquely uncovered by employing an inducible manipulation approach was the finding that the V˙E state of the animal before manipulation tracked with the size of the V˙E change during perturbation . Furthermore , we observed that perturbation of Pet1 neuron activity resulted in a regression toward a common V˙E set point . This may indicate that Pet1 neuron activity is an important component that allows the internal arousal state of the animal to alter VE in preparation for a stressor requiring higher ventilation . Collectively , our findings of decreased HR and V˙E immediately following CNO-Di-mediated perturbation of Pet1 neurons provides evidence that , even without an external stressor like asphyxia-induced apnea or exposure to CO2 , neonatal mice use Pet1 neurons to shape a homeostatic set point . Baseline V˙O2 measurements were less straightforward , showing in both genotypes a subtle decrease following CNO and return to the plethysmograph chamber . We speculate that this non-specific V˙O2 effect reflects relaxation in and habituation to the chamber at this advancing time point in the assay resulting in a subtle lowering of metabolic rate . It could also reflect the very real challenge in accurately measuring V˙O2 ( as compared to the other cardiorespiratory parameters ) for such tiny mouse pups and/or reflect a modest effect of CNO itself ( independent of Di expression [Gomez et al . , 2017; Manvich et al . , 2018] ) on P8 pup metabolic rate . Not only were baseline cardiorespiratory properties affected in P8 neonates upon acute disruption of Pet1-neuron activity , but also and more strikingly the capacity to autoresuscitate from repeated asphyxia-induced apneas . Mortality was increased significantly and autoresuscitation prolonged across the assay chain of apneic challenges . Thus , the neonatal homeostatic recovery response appears to actively require Pet1-neuron function for life-supporting robustness , with death in ~32% of the Pet1-Di-CNO pups as compared to ~7% in control-Di-CNO pups . Prior chronic developmental perturbation studies ( Barrett et al . , 2016; Cummings et al . , 2011a ) , while resulting in similar increases in mortality at P8 , were unable to resolve temporally whether Pet1-neuron activity was required during embryogenesis for the broader establishment of autoresuscitation circuitry or actually required at P8 as a fundamental participant in the homeostatic response . Present findings indicate active participation at P8 , revealing an active component in neonatal homeostatic circuitry and informing possible intervention inroads for mitigating risk of autoresuscitation failure . An additional striking finding in Pet1-Di-CNO pups was that deficits in the respiratory recovery response alone , without similarly early and severe dysfunction in cardiac recovery , were sufficient to increase mortality after an apnea ( Figures 4 and 5 ) . This decoupling of the heart rate recovery response and the respiratory recovery response suggests segregation in the underlying circuitry and its modulation by Pet1 neurons at P8 , building upon the segregation of phenotypes between chronic versus acute Pet1-neuron perturbation described above . Disruption of this link between respiratory and cardiac responses likely contributes to the increased mortality . Notably , continued HR elevation in the unsupportive setting of unresolved hypoxia has been shown in other models to increase mortality ( Scremin et al . , 1980 ) ; similar conditions may arise in Pet1-Di-CNO pups when productive gasping and ventilation recovery lags behind and out of sync with the HR recovery . While autoresuscitation failure or prolongation ultimately characterized most Pet1-Di-CNO pups , in the initial asphyxia-apnea bout nearly all recovered , which allowed us to ask if there were features of that first recovery response that were specific to pups that ultimately died during one of the subsequent apnea challenges . Indeed , we found discrete measurable differences in the gasping response to the first apnea challenge , as compared to controls: a smaller first gasp , a prolonged latency to first gasp , longer inter-gasp intervals . One interpretation is that an initial suboptimal gasp response , with the associated extended conditions of poor oxygenation despite ultimate apnea recovery , may set in motion some cellular deficiency that then predisposes to mortality during a forthcoming apnea . Further , our findings indicate that these impaired gasps are unable to trigger the same type of heart rate recovery during the initial apnea-recovery attempt even though the overall timing to recover 63% of HR is not impaired . We suggest that the initial suboptimal gasp response may be indicative of an intrinsic failing of the broader respiratory response , not just the gasp response , with this broader failing the result of acute Pet1-neuron perturbation , which ultimately increases mortality risk upon exposure to asphyxic-apnea stress . Given the high frequency of apneas among human infants ( Daily et al . , 1969; Kelly et al . , 1985; Southall et al . , 1980 ) , measuring apnea recovery ( gasp size , timing , HR between gasps , and breathing and heart rate coupling ) could be a potential indicator of raphe neuron dysfunction and a need for targeted infant monitoring during the peak age for SIDS risk of 2–4 months . In addition to identifying initial gasp response features that distinguished pups that died on exposure to multiple apneas , we also identified the homeostatic state of modest hyperventilation – V˙E/V˙O2 at the higher end of the distribution – as increasing the risk likelihood of death in pups challenged with the combined stressors of acute Pet1-neuron dysfunction and repeated apneas . When at the high end of the V˙E/V˙O2 distribution , the normally strong drivers of gasping , heart rate , and breathing rate recovery – i . e . increased PCO2 and decreased PO2 – are countered because these pups at baseline are always in a state of modest overbreathing and thus relatively hypocapnic and hyperoxic such that more severe apneic conditions would be required to trigger the respiratory response . This then could contribute to the insufficient gasp response , which despite the pup starting out with an elevated V˙E/V˙O2 could ultimately lead to more severe hypoxemia due to HR and breathing rate decoupling in the setting of Pet1-neuron dysfunction , resulting in a decreased ability to withstand future apneas ( Figure 7 ) . In control , wild-type pups , redundancy in the respiratory response may permit sufficiently robust homeostatic responses despite such modest hyperventilation so as to withstand repeated apneas; however , if simultaneous with Pet1-neuron dysfunction , it may no longer be possible to mount the necessary life-sustaining response ( Figure 7C ) . Such vulnerability may be exacerbated during active versus quiet sleep , as recently suggested by studies in Tph2-/- rat pups ( Magnusson and Cummings , 2018; Young et al . , 2017 ) . As well , it is worth noting that medullary Pet1 neurons project to brainstem centers relevant to cardiorespiratory control , such as the pre-Bötzinger complex with its role in respiratory rhythm and chemosensory processing , the nucleus ambiguous housing cardiovagal neurons , and the nucleus of the solitary tract , an important site of integration of the chemoreflex ( peripheral and central ) ( Brust et al . , 2014; Feldman et al . , 2003; Hennessy et al . , 2017; Hodges and Richerson , 2010; Wang and Richerson , 1999 ) . In summary , we found that Pet1 neurons play an active role in P8 mouse pups in maintaining cardiorespiratory tone and in supporting robust , life-sustaining autoresuscitation responses to asphyxia-induced apneas . When Pet1 neurons were compromised acutely , V˙E decreased , HR slowed , and autoresuscitation failure increased . Respiratory more than cardiac recovery was impaired , causing a disorganization of the normal linear relationship between breathing ƒ and HR . Moreover , early gasping abnormalities distinguished the Pet1-neuron-compromised pups that went on to die during subsequent apneas , as did modest baseline hyperventilation . Collectively , these findings shed new light on cardiorespiratory control systems and , more specifically , support a potential pathoetiological role for the SIDS-associated finding of postmortem brainstem 5-HT neuron abnormalities . Further , they suggest that gasp features might potentially help define a physiological profile associated with a higher risk likelihood for SIDS .
All experimental protocols were approved at Harvard Medical School ( HMS ) and the Geisel School of Medicine at Dartmouth by the respective Institutional Animal Care and Use Committees ( IS00000231-3 and 2035 , respectively ) and the HMS Committee on Microbiological Safety ( 15-225 ) , and were in accordance with the animal care guidelines of the National Institutes of Health . For acute , chemogenetic perturbation of Pet1 neurons in vivo , double transgenic mouse pups of the genotype Pet1-Flpe , RC-FDi ( referred genotypically as Pet1-Di ) were generated via crossing Gt ( ROSA ) 26Sortm ( CAG-FSF-CHRM4* ) Dym ( denoted in short-hand as RC-FDi [Ray et al . , 2011] ) homozygous females to hemizygous Pet1-Flpe ( Jensen et al . , 2008 ) males . Here Flpe expression mediates FRT recombination of the RC-FDi allele resulting in expression of the inhibitory , synthetic G protein-coupled receptor Di exclusively in neurons with current or a lineal history of Pet1 expression , thus enabling acute , CNO-triggered Di inhibition of serotonergic neurons upon CNO i . p . injection . We refer to these CNO-treated double transgenic mice as Pet1-Di-CNO mice . Littermate controls were of the single transgenic RC-FDi genotype and thus devoid of Di expression but of comparable genetic background ( predominantly C57BL/6J , minor 129 ) thus serving as controls ( CNO-treated controls referred as control-CNO ) . Mice were obtained from nine independent litters , yielding 22 double transgenic Pet1-Flpe , RC-FDi pups ( 11 males , 11 females ) and 15 single transgenic RC-FDi pups ( five males , 10 females ) . Mice were housed in a temperature-controlled environment on a 12:12 hr light-dark cycle in an external housing environment with ad libitum access to standard rodent chow and water . Past experiments with similar genotypes and physiological measures demonstrated that a group size of n ≥ 15 would provide sufficient statistical power to detect differences between experimental and control groups ( Barrett et al . , 2016; Brust et al . , 2014 ) . Genotypes were determined as previously described ( Brust et al . , 2014 ) . Briefly , DNA isolates from tail tip biopsies from P3-5 pups were subjected to PCR amplification using Taq DNA polymerase ( New England BioLabs Inc . ) and the following primer sequences ( Invitrogen , Carlsbad , CA ) diagnostic for Flpe ( 800 bp amplicon ) or hM4Di ( 268 bp amplicon ) : 5’-GCATCTGGGAGATCACTGAG-3’ ( Flpe forward primer ) ; 5’-CCCATTCCATGCGGGGTATCG-3’ ( Flpe reverse primer ) ; 5’-CGAATTCGGAAACATAACTTCG-3’ ( FDi forward primer ) ; 5’-GGCAATGAAGACTTTCCACCG-3’ ( FDi reverse primer ) . PCR amplification consisted of an initial 5 min denaturation at 94 ˚C , followed by 35 cycles , each consisting of 1 min at 94 ˚C , 1 . 5 min at 60 ˚C , and 1 min at 72 ˚C , followed by a final 10 min extension at 72 ˚C . For preparation of double transgenic Pet1-Flpe , RC-FDi tissue , postnatal day eight mice were briefly anesthetized with ice and immediately perfused intracardially with phosphate buffered saline ( PBS ) followed by 4% paraformaldehyde ( PFA ) in PBS . Brains were extracted , soak-fixed for 2 hr in 4% PFA at 4°C , cryoprotected in 30% sucrose/PBS for 48 hr , and subsequently embedded in OCT compound ( Tissue-Tek ) . Coronal sections were cryosectioned at 20 µm and mounted onto glass slides , were then rinsed three times with PBS for 10 min and permeabilized with 0 . 5% Triton X-100 in PBS for 1 hr , and blocked in 5% normal donkey serum ( NDS , Jackson ImmunoResearch ) , 1% BSA , 0 . 5% Triton X-100 in PBS for 1 hr at room temperature ( RT ) . Sections were rinsed three times with antibody buffer ( 5% NDS , 0 . 5% Triton X-100 in PBS ) for 10 min each , followed by incubation for 72 hr at 4°C with the primary antibodies in the same buffer . Primary antibodies: rabbit polyclonal anti-Tph2 ( 1:1000; NB100-74555; Novus Biological ) , rat monoclonal anti-HA ( 1:200; 7C9; Chromotek ) , goat polyclonal anti-5-HT ( 1:1000 , ab66047; Abcam ) . Sections were then washed with antibody buffer three times for 10 min and incubated with secondary antibodies for 2 hr at RT . Secondary antibodies: donkey anti-rabbit IgG-Alexa Fluor 488 ( 1:500 , ThermoFisher Scientific . ) , donkey anti-rat IgG-Alexa Fluor 594 ( 1:500 , ThermoFisher Scientific ) , donkey anti-goat IgG-Alexa Fluor 647 ( 1:500 , ThermoFisher Scientific ) . DAPI ( 4’ , 6-diamidino-2-phenylindole ) was used for nuclear counterstaining . Images were collected on a Zeiss LSM 780 inverted point scanning confocal microscope with a Zeiss LD LCI Plan-Apochromat 25x/0 . 8 N . A . multi-immersion objective for overview images and a Zeiss Plan Apochromat 63x/1 . 4 N . A . oil-immersion objective for higher magnification images . Laser settings were adjusted for each sample , but kept constant throughout image collection within the same areas between Pet1-Di and Control-Di sections . The images were imported to and processed with ImageJ ( Fiji distribution ) for brightness and contrast adjustment , which were also kept constant between Pet1-Di and Control-Di sections . Physiological measurements were obtained as described previously ( Barrett et al . , 2016; Cummings et al . , 2011a ) . Briefly , ventilation was measured using a head-out plethysmograph system consisting of a body chamber and a head chamber . The body chamber ( volume =~60 ml; diameter = 3 cm , length = 8 . 5 cm ) was made from a water-jacketed glass cylinder with inlet and outlet ports that were connected to a water bath , allowing for continuous circulation of water around the chamber to maintain pup body temperature . The ambient temperature ( TA ) of the body chamber and thus the body temperature ( TB ) of the mouse pup were controlled by adjusting the temperature of the water circulating around the glass chamber . Both the TA and TB were continuously monitored with a thermistor probe and a fine thermocouple , respectively ( Omega Engineering Inc , Stamford , CT ) . The head chamber ( volume =~3 ml ) was made from the bottom of a 50 ml plastic syringe tube ( Terumo Medical , Somerset , NJ ) with a piece of vinyl glove covering the larger of the two openings . A rubber gasket ( Terumo Medical , Somerset , NJ ) was used to hold the piece of vinyl glove in place and to secure the head chamber into the anterior end of the body chamber . A small hole was made in the center of the vinyl glove , where the snout of the mouse pup was inserted and the hole was sealed with Impregum F polyether impression material ( 3M , St . Paul , MN ) . The head chamber had an outlet port connected downstream to a pump ( S-3A/I , AEI Technologies , Pittsburgh , PA ) that pulled air through the head chamber at a rate of 140 ml/min . This high flow rate was chosen to prevent accumulation of CO2 in the head chamber and to ensure rapid delivery of the experimental gas to the animal . The air exiting the head chamber was passed through a Nafion drying tube ( PerkinElmer , Waltham , MA ) before being sampled by oxygen ( O2 ) analyzers ( S-3A/I , AEI Technologies , Pittsburgh , PA ) in order to monitor oxygen consumption ( ) . A pneumotach connected to the open end of the head chamber was attached to a differential pressure transducer ( Validyne Engineering Corp , Northridge , CA ) in order to measure respiratory activity . Experimental gases were delivered to the head chamber via the open end of a 50 ml syringe tube that was connected to the gas cylinder and then placed over the pneumotach . The pneumotach was calibrated by withdrawing and injecting 0 . 02 ml of air into the head chamber and the pressure signal associated with injection of this volume was integrated to determine the volume . Heart rate ( HR ) was monitored with a telemetric device ( CTA-F40 , DSI , Inc . , St . Paul , MN ) that consisted of 2 ECG leads that were placed on the surface of the pup’s chest and held in place with a vest made from a cohesive flexible bandage ( Andover , Salisbury , MA ) . Using the LabChart Heart Rate Variability module , the previously defined 30 s segments of ECG and breathing traces before and during CNO exposure were analyzed . Initial peak detection used the built-in algorithm along with experimenter verification using manual selection of all r waves on the ECG trace and maximal voltage deflection on the breathing trace . Using this module , the coefficient of variation for breathing f and HR was also determined . Additionally , this software was used to identify the standard deviation ( SD ) , the major and minor axes of the Poincare plot , and the root mean square of successive deviation ( RMSSD ) for the interbeat and interbreath intervals . Apnea was induced by introducing an asphyxia-mimicking gas mixture ( 97% nitrogen and 3% carbon dioxide ) to the 50 ml cylinder as previously described ( Barrett et al . , 2016; Cummings et al . , 2011a; Erickson and Sposato , 2009 ) . An initial examination of the time it took either genotype to develop an apnea once asphyxic conditions were introduced demonstrated no difference ( 23 . 4 ± 4 . 1 [standard deviation] ( s ) for Pet1-Di-CNO and 23 . 4 ± 4 . 9 [standard deviation] ( s ) control-CNO respectively ) . After the asphyxic exposure , we measured the heart rate recovery , and breathing rate recovery . Time to recover heart rate or breathing frequency to 63% of baseline ( τHR and τf , respectively ) was determined to be the time between the end of asphyxia to the time it took an animal to recover and sustain for at least 3 s their heart rate or breathing frequency to 63% of their baseline heart rate or breathing frequency immediately prior to that asphyxic bout ( see Figure 4A ) . The τ for each asphyxic bout was obtained using the new baseline measured immediately before each asphyxic bout ( see Figure 4A [open window a’-d’ for a-b , respectively] ) . Of the animals that died , some ( 3 of 7 Pet1-FDi-CNO and 1 of 1 control-CNO pups ) nonetheless recovered heart rate to greater than 63% of the pre-bout baseline for a short period prior to death , thus data from that fatal apneic bout ( referred to as the fatal bout ) was included in our analyses ( see Figure 4D and E gray-filled symbols ) . The other 4 Pet1-FDi-CNO pups that succumbed to an apneic bout never recovered to the 63% pre-bout breathing and heart rate levels during the fatal apneic bout , thus only data from bouts recovery prior to the fatal bout were included in the graphical representations ( Figure 4B–E black-filled symbols ) . One Pet1-Di-CNO pup , which did not die , nonetheless failed over the course of the assay to recover breathing rate to 63% of its final/4th-bout baseline , thus we assigned as recovery time the full 331 s assay duration , albeit an underestimate . One ECG lead malfunction precluded THR measurements from a single Pet1-Di-CNO animal for asphyxia bout 1 and 2 , but which was corrected for bouts 3 and 4 . For the correlation and linear regression analysis , necessarily only data from asphyxia bout recoveries with both a heart rate and breathing rate recovery were included . The first gasp was defined as the first sharp inhalation after apnea onset and during which the O2 was rising determined by the O2 sensor value , ~5 s , ensuring chamber O2 could support autoresuscitation . Gasp size was determined as the integral of the first gasp with a voltage change greater than 0 . 002 V , empirically distinguishing it from smaller pressure transducer changes reflecting instead body movement artifacts . Time between gasps was calculated as the time between these ‘steeple-like’ voltage deflections ( see Figure 6A ) . HR during gasp intervals was calculated as the mean heart rate between steeple deflections . The data are presented as the mean ±SD . The effects of gender and genotype on body weight ( BW ) , V˙E , V˙O2 , V˙E/V˙O2and HR parameters at baseline before CNO injection ( Figure 1G , open window a ) were assessed by a two-tailed Student t-test comparisons . To assess the effects of CNO on the experimental and control groups , paired two-tailed Student t-tests were performed on the baseline homeostatic characteristics ( HR , V˙E , V˙O2 , V˙E/V˙O2 , and heart rate and breathing rate variability characteristics ) of each group before vs . during CNO exposure ( Figure 1G open window a and 1 H bar b’ , respectively ) . To test the hypothesis of increased mortality in Pet1-FDi-CNO pups , we applied a one-tailed Fisher Exact test with Lancaster’s mid-p correction ( Biddle and Morris , 2011 ) due to the previous assumption that Pet1 raphe neuron disruption would increase mortality ( Barrett et al . , 2016 ) . The odds ratio for pup death as an outcome of asphyxic apnea in the face of Pet1-Di-CNO versus control-Di-CNO was calculated as follows: ( A/C ) / ( B/D ) , where A = Pet1 Di-CNO pups that died , B = Pet1 Di-CNO pups that survived , C = Control Di pups that died , and D = Control Di pups that survived . OpenEpi version three was used to perform the one-tailed Fisher Exact test ( Sullivan et al . , 2009 ) . To analyze breathing and heart rate recoveries across asphyxia bouts , we used a two-way analysis of variance ( ANOVA ) with asphyxia bout and genotype as variables , with Tukey’s test for multiple comparison correction post hoc and unpaired t-tests . A linear regression model was run to assess the relationship between BR versus HR recovery . Additionally , a Runs test was used to assess whether these data had a nonrandom linear relationship . To compare first-gasp responses , a one-way ANOVA test was applied . To assess the relationship between gasp interval duration and genotype survival between gasp intervals , we applied a repeated measures two-way ANOVA with gasp interval and genotype/survival as variables; post hoc analyses employed Tukey’s test for multiple comparisons to determine the effect of asphyxia on the different groups . We used the same analysis to assess the relationship between HR during gasp interval and genotype survival . To test for independent effects of homeostatic characteristics on mortality , a logistic regression model was fit with mortality ( yes/no ) as the outcome and homeostatic characteristics as predictors , controlling for genotype . Variance of V˙E/V˙O2 measurement was analyzed using an F test . All graphs and all other statistical analyses were performed using GraphPad Prism version 7 . 0 c for Mac OS X , GraphPad Software , La Jolla California USA , www . graphpad . com . | Our survival depends on our heart and lungs working together to supply our cells with oxygen and remove carbon dioxide waste . The brain coordinates this process by controlling the activity of the heart and lungs . Yet sometimes a person may experience an event called an apnea and briefly stop breathing . If this happens , oxygen levels in the body fall while carbon dioxide levels rise . This in turn triggers a recovery process called autoresuscitation , which includes a series of large breaths or gasps , and each gasp is accompanied by increased heart rate due to specialized parts of the nervous system . This response usually restores normal breathing . Failure of autoresuscitation may underlie many cases of sudden infant death syndrome , or SIDS ( also known as “cot death” or “crib death” ) . SIDS is the leading cause of death in young infants in the western world , and many infants who die from SIDS show abnormalities in the brain cells that produce a chemical called serotonin . Evidence suggests that serotonin helps control breathing . This raised the question: does the autoresuscitation recovery response rely on serotonin-producing neurons ? To find out , Dosumu-Johnson et al . used one-week-old mouse pups that had been genetically engineered to respond to an injected drug by rapidly inhibiting their serotonin neurons . These animals are about the same age in mouse terms as infants at greatest risk for SIDS ( ~2-4 months of age ) . Inhibiting serotonin neurons made it harder for the mouse pups to recover from artificially induced apneas . Although their heart rate showed largely normal recovery – at least at first – their breathing did not . They took fewer gasps , and were more likely to die following such episodes . These findings shed new light on how young animals control their breathing and heart rate when mounting an autoresuscitation recovery from an apnea . The observed uncoupling of breathing and heart rate recovery responses suggests that different brain cells and circuits control the two . The results also suggest that abnormalities in the activity of serotonin neurons may make infants more susceptible to SIDS . As well as offering a possible explanation to families who have lost a child to SIDS , these findings could be used to develop screening tools to identify other infants at risk . They also point to potential cellular targets for drugs that could ultimately help prevent further cases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | Acute perturbation of Pet1-neuron activity in neonatal mice impairs cardiorespiratory homeostatic recovery |
Antennae are often considered to be the nostrils of insects . Here , we sequenced the transcriptome of the pheromone gland-ovipositor complex of Helicoverpa assulta and discovered that an odorant receptor ( OR ) gene , HassOR31 , had much higher expression in the ovipositor than in antennae or other tissues . To determine whether the ovipositor was involved in odorant detection , we co-expressed HassOR31 and its co-receptor , HassORco , in a Xenopus oocyte model system , and demonstrated that the OR was responsive to 12 plant odorants , especially Z-3-hexenyl butyrate . These odorants elicited electrophysiological responses of some sensilla in the ovipositor , and HassOR31 and HassORco were co-expressed within ovipositor sensilla . Two oviposition preference experiments showed that female moths lacking antennae still preferentially selected oviposition sites containing plant volatiles . We suggest that the expression of HassOR31 in the ovipositor of H . assulta helps females to determine precise egg-laying sites in host plants .
The most important functions of adult insects are to find optimal mates and suitable habitats for survival and success of their offspring . The insect olfactory system plays a key role in these processes , and antennae are often considered to be the nostrils of insects . However , some species also use other cephalic organs , such as maxillary palps and proboscis , to detect volatile compounds ( Haverkamp et al . , 2016; Di et al . , 2017; Lu et al . , 2007 ) . Olfactory sensilla distributed on these organs are multiporous hair-like structures innervated by the dendrites of olfactory sensory neurons ( OSNs ) ( Benton and Dahanukar , 2011; Stocker , 1994 ) . Odorant receptors ( ORs ) are atypical , 7-transmembrane domain proteins , which are located on the dendritic membrane of OSNs , selectively bind to volatile ligands in the environment and are the primary determinants of the detection spectrum of OSNs ( Dobritsa et al . , 2003; Goldman et al . , 2005; Störtkuhl and Kettler , 2001; Wicher et al . , 2008 ) . These ligand-binding ORs are considered to form a heteromultimeric complexes with a co-receptor ( ORco ) and to function as non-selective cation channel ( Butterwick et al . , 2018; Larsson et al . , 2004; Neuhaus et al . , 2005; Sato et al . , 2008; Vosshall et al . , 1999 ) . Some ionotropic receptors ( IRs ) and gustatory receptors ( GRs ) are also recognized as odorant-detecting receptors ( Benton et al . , 2009 ) . Several recent studies have challenged the hypothesis that only cephalic organs are involved in the detection of volatile compounds in insects . Olfactory receptor genes are expressed not only in antennae and maxilla , but also in ovipositors of a variety of moth species , including Heliothis virescens ( Widmayer et al . , 2009 ) , Sesamia nonagrioides ( Glaser et al . , 2013 ) , Chilo suppressalis ( Xia et al . , 2015 ) , and Manduca sexta ( Klinner et al . , 2016 ) . However , the function of olfactory receptor genes expressed in the ovipositor remains unknown . Moth ovipositors are always connected to pheromone glands , and several lines of evidence from different moth species suggest that functional olfactory receptors may be present in the sensilla located on moth ovipositors . For example , a pheromone receptor is expressed in the ovipositor of H . virescens , suggesting a possible role of the ovipositor in feedback regulation of biosynthesis and emission of the female sex pheromone from the sex pheromone gland ( Widmayer et al . , 2009 ) . Two ORs are expressed in the pheromone gland and ovipositor of the grassland moth , C . suppressalis , but no ORco was detected there ( Xia et al . , 2015 ) . Sensilla with a multiporous surface are observed in Monopis crocicapitella and Homoeosoma nebulella ( Faucheux , 1991; Faucheux , 1988 ) . Finally , a subset of sensilla located on the ovipositor of M . sexta exhibit electrophysiological responses to a large array of volatile organic compounds , and expression of ORco and the ionotropic co-receptors IR8a and IR25a was detected in the ovipositor ( Klinner et al . , 2016 ) . However , the function of ORs expressed in the ovipositor of moths and whether these ORs are co-expressed with ORco remains unclear . The Oriental tobacco budworm , Helicoverpa assulta , is a serious crop pest with a narrow host plant range which includes only Solanaceae such as tobacco , hot pepper , and several Physalis species ( Wang et al . , 2004 ) . Previous antennae transcriptome studies showed that more than 60 ORs were expressed in the antennae of H . assulta ( Xu et al . , 2015; Zhang et al . , 2015 ) . The function of the pheromone receptors and several ORs responding to plant volatiles were characterized ( Cao et al . , 2016; Wu et al . , 2019; Yang et al . , 2017 ) . To determine whether any ORs were expressed in the ovipositor , we sequenced the transcriptome of the pheromone gland-ovipositor complex and found one OR with a very high expression level . Next , we functionally analyzed the response spectra of this OR to a wide range of host plant-related odorants using the Xenopus oocyte expression system and two-electrode voltage-clamp recording . In situ hybridization , scanning electron microscopy and electrophysiology studies indicated that this OR is expressed in some multiporous sensilla on the moth ovipositor . Together with the results of oviposition experiments , we suggest that the OR expressed in the ovipositor helps H . assulta females to find precise egg-laying sites on their host plants .
We conducted next-generation transcriptome sequencing analyses using a cDNA library constructed from pheromone gland-ovipositors ( PG-OVs ) of female H . assulta using the Illumina HiSeq 2000 platform . We downloaded amino acid sequences of odorant receptors , gustatory receptors , antennal ionotropic receptors , ionotropic glutamate receptors , odorant binding proteins , chemosensory proteins , general odorant binding proteins , pheromone binding proteins and sensory neuron membrane proteins of H . assulta and H . armigera from NCBI to construct a local database ( Liu et al . , 2014; Liu et al . , 2018; Tillman et al . , 1999; Xu et al . , 2015; Zhang et al . , 2015 ) , and then performed a BlastX search against this database to identify the cDNA sequences of chemosensory receptors from the transcriptome . The transcripts of 22 ORs , 6 GRs , 13 IRs , and 9 iGluRs were detected ( Figure 1A ) . Figure 1—figure supplement 1 and Supplementary file 1 shows the identified putative chemosensory related genes expressed in PG-OVs and their values of TPM ( Transcripts Per Kilobase of exon per Million mapped reads ) . HassOR31 and HassiGluR7 had approximately the same TPM value which was the highest among those of all the chemosensory receptor genes . Most crucially , HassORco was also detected but its TPM value was much lower than that of HassOR31 . Therefore , we further analyzed the tissue expression profiles of HassOR31 and HassORco by qRT-PCR . The qRT-PCR results indicated that HassOR31 had the highest expression level in PG-OV , which was 7 times higher than that in female antennae and 15 times higher than that in male antennae . HassORco was also expressed in PG-OV , but the expression level was much lower than that in the antennae of both sexes ( Figure 1C ) . In PG-OV , when Ct value of Hass18S was 20 . 41 , the Ct values of HassOR31 and HassORco were 23 . 18 , 28 . 78 , respectively . These results were in line with the RNA-seq data ( Figure 1B ) . To further determine which cells were expressing HassOR31 and HassORco in the ovipositor , two-color double in situ hybridization experiments were performed with Dig-labeled HassOR31 and Bio-labeled HassORco . In some cases , HassOR31 and HassORco were co-expressed within the cells beneath certain sensilla with a large size ( Figure 2A , Figure 2—figure supplement 1 , Video 1 ) , while in most cases HassOR31 was expressed alone in the cells beneath small hairs , but a few in the cells of sensilla with a large size ( Figure 2B and C , Figure 2—figure supplement 1 , Video 1 ) , which is consistent with the high expression of HassOR31 and the low expression of HassORco in PG-OV of H . assulta . A Xenopus oocyte expression system with two-electrode voltage-clamp recording was used to characterize the function of HassOR31 . A panel of 51 chemicals with behavioral or electrophysiological activities to Helicoverpa species as listed in Supplementary file 3 were used to screen the ligands of HassOR31 ( Di et al . , 2017 ) . They were classified into four categories , green leaf volatiles ( GLVs ) , terpenoids , aromatics , and aliphatics . The oocytes containing co-expressed HassOR31/HassORco were tuned to 12 odorants , including Z-3-hexenyl-butyrate , myrcene , citral , and Z-3-hexenyl acetate , with Z-3-hexenyl-butyrate as the most effective ligand ( Figure 3A , B ) . The oocytes in which only HassOR31 was expressed had no positive responses to these odorants . Since HassiGluR7 had a high expression level comparable to HassOR31 in the pheromone gland - ovipositor , we also co-expressed HassOR31 and HassiGluR7 in the oocytes , but no positive responses were detected ( Figure 3A ) . To further determine the putative chemosensory sensilla which may bury HassOR31/HassORco on the ovipositor of H . assulta , we examined the surface of the anal papillae using scanning electron micrography ( Figure 4 ) , and found four types of sensilla ( Figure 4A , B ) . The first type is long sensilla with smooth non-porous surfaces , sharp tips and raised sockets , which may have a potential function in mechanosensation ( Figure 4C , D ) . The second type is shorter sensilla with non-porous surfaces and sunken sockets , which may also have a mechanical function ( Figure 4E , F ) . The third type sensilla is of similar length with the second type . It is morphologically like trichoid sensilla in the antennae , with pores on the surface and a single pore at the tip ( Figure 4G , H ) , and therefore may have a function in olfaction and/or taste . They are mainly distributed on the middle part of the ovipositor . Finally , the fourth type sensilla are morphologically similar to basiconic sensilla in the antennae , with pores on the surface and a large pore at the tip and may also have a function in olfaction and/or taste ( Figure 4I , J ) . They are located near the ovipore and five to seven of them are distributed on each papilla . Moreover , the papillae are covered with short , poreless microtrichia . We determined whether these identified putative chemosensory sensilla served an olfactory function using single sensillum recordings . Seventeen odorants , including the major ligands of HassOR31/HassORco and sex pheromone components of H . assulta , were used as stimuli ( Supplementary file 3 ) . Most tested sensilla showed baseline spiking activity , several sensilla responded to the tested odorants ( Figure 5A ) . Z-3-Hexenyl-butyrate , the most effective ligand of HassOR31/HassORco , gave a dose-dependent response curve ( Figure 5B , C ) . Based on the above results , we hypothesize that the ovipositor may play a role in oviposition site detection of H . assulta female . To verify this , we designed two oviposition choice tests ( Figure 6—figure supplement 1 ) . Intact , mated females preferred to lay eggs on the areas of gauze exposed to host plant volatiles ( n = 7; p=0 . 0008 ) ( Figure 6A ) . The closer to the odor source , the higher the density of eggs on the gauze ( Figure 6—figure supplement 1 ) . Mated females whose antennae had been removed showed a reduced , but still significant preference for oviposition on volatile-treated areas ( n = 7; p=0 . 0088 ) ( Figure 6A , Figure 6B ) . Intact ( n = 5; p=0 . 0053 ) and antennectomized ( n = 5; p=0 . 0149 ) females both preferred to lay eggs on Z-3-hexenyl butyrate treated fake leaves ( Figure 6C ) , while they showed no significant difference on the oviposition preference index ( Figure 6D ) .
HassOR31 was relatively widely tuned to a spectrum of plant volatiles . The twelve GLV and terpenoid compounds are widely present in leaves , flowers , and fruits of plants , and are strongly attractive to herbivorous insects ( Gregg et al . , 2010 ) . The ortholog of HassOR31 , HarmOR31 , has a similar wide spectrum in polyphagous Helicoverpa armigera ( Di et al . , 2017 ) . Nevertheless , the most effective ligand of HassOR31 was Z-3-hexenyl butyrate while that of HarmOR31 was myrcene . The expression levels of the two orthologs in the ovipositors of the two species were also different: HassOR31 was highly expressed while HarmOR31 was poorly expressed ( Supplementary file 1 ) . This may have significance in the divergence of the host plant range between the two closely related species: H . assulta is a specialist on hot pepper , tobacco , and some Physalis plants in Solanaceae , while H . armigera is a typical generalist with a host plant range of over 300 species belonging to 68 plant families ( Pearce et al . , 2017; Sun et al . , 2012; Wang et al . , 2004 ) . As a specialist , H . assulta lays their eggs singly , preferably on or near the flowering or fruiting parts of these host plants ( Sun et al . , 2012; Wu , 1990 ) . The headspace collections of tobacco flowers were analyzed by GC-MS and GC-EAD previously . Tobacco flowers release volatiles including E-β-ocimene , Z-3-hexenyl acetate , nonanal , Z-3-hexenyl 2-methyl butyrate , linalool , and Z-3-hexenyl butyrate , and blends of them are attractive to H . assulta females ( Sun et al . , 2012 ) . Most odorants in the response spectrum of HassOR31/HassORco can be detected in headspace collections from tobacco . As a general plant volatile and most effective ligand of HassOR31 , Z-3-hexenyl butyrate is released by many plant fruits and has a green , fruity , and somewhat buttery aroma . It is also present in the headspace collections of tobacco flowers ( Sun et al . , 2012 ) and hot pepper fruits ( Forero et al . , 2009 ) . Therefore , Z-3-hexenyl butyrate is very likely to be used by females of H . assulta as a signal for choosing oviposition sites rather than a signal for searching host plants . The behavioral assay with hot pepper volatiles ( a complex cue ) showed that antennectomized females still preferred to oviposit on sites treated with hot pepper volatiles , but their oviposition preference index was significantly decreased . Another behavioral assay with Z-3-hexenyl butyrate showed that antennectomized and normal females both prefer to lay eggs on the sites treated with the ligand for the receptor in the ovipositor but had no significant difference in oviposition preference index . These results suggest that detecting host volatile blends in general oviposition is governed by both antennae and ovipositor but detecting for Z-3-hexenyl butyrate seems to be mainly governed by the ovipositor . The expression levels of HassOR31 and other ORs expressed in the ovipositor and antennae may explain such behavioral responses . We speculate that a gravid female moth takes two steps to find an oviposition site: firstly , she smells the plant volatiles mainly by using antennae to search for a host plant , and secondly when she comes near to or land on the host plant , she integrates the information from olfactory sensilla as well as mechanical and contact chemosensory sensilla on the ovipositor to determine the precise oviposition sites on the host plants . Z-3-hexenyl butyrate and other ligand compounds of HassOR31 are expected to play a major role in the second step . It is believed that ORs cannot function in the absence of ORco ( Sato et al . , 2008 ) . Recent Cryo-EM structure of the insect olfactory receptor ORco supports a model in which ORco and OR subunits assemble into a heterotetramer with a central shared ion-conduction pathway ( Butterwick et al . , 2018 ) . One unresolved question from this study relates to the large difference in expression levels between HassOR31 and HassORco in the ovipositor . We first assumed that HassOR31 played a role alone without HassORco . To test this idea , we injected HassOR31 cRNA alone into the Xenopus oocytes and found that oocytes expressing HassOR31 alone had no response to the tested compounds . We then hypothesized that some co-receptor-like proteins might replace HassORco to function in cooperation with HassOR31 . In addition to HassOR31 , the ionotropic glutamate receptor , HassiGluR7 , was also highly expressed in the ovipositors of H . assulta and had a similar tissue expression pattern to HassOR31 . To test if these two receptors function together , we co-expressed HassOR31 and HassiGluR7 in Xenopus oocytes , but no response was detected . In the electrophysiological experiments , we discovered only a few sensilla on the ovipositor that responded to Z-3-hexenyl butyrate , the most effective ligand of HassOR31/HassORco , which explains the low expression of HassORco in the ovipositor . However , it does not explain the presence of high expression levels of redundant HassOR31 . We speculate that in addition to cooperating with HassORco and functioning in olfactory sensation , HassOR31 might function in mediating cell responses to endogenous signaling molecules , regulating neural development , or play other non-chemosensory roles . The similar situation is also found in testes of A . gambiae , where some AgORs are abundantly expressed , but AgORco transcript is present at a very low level . It is proved that the AgORs and AgORco are localized to the flagella of A . gambiae spermatozoa where Orco-specific agonists , antagonists , and other odorant ligands robustly activate flagella beating in an Orco-dependent process ( Pitts et al . , 2014 ) . Studies on mammals have found that many ORs are present in tissues outside the olfactory system and have diverse functions in several physiological contexts beyond odor recognition ( Wu et al . , 2017 ) . For example , a human testicular OR , hOR17-4 was found to control cellular motility and chemotaxis in sperm cells ( Spehr et al . , 2003 ) . In addition to the ORs , we also found some GRs and IRs present in the ovipositors of H . assulta . The expression of HassGR9 , the ortholog of fructose receptor HarmGR4 ( Jiang et al . , 2015 ) , suggests that sugar taste sensilla are also present in the ovipositor of H . assulta . Sugar taste sensilla have also been found on the ovipositor of cotton leaf worm Spodoptera littoralis ( Seada et al . , 2016 ) . The co-receptors of IR families ( HassIR8a , HassIR25a ) were also expressed , indicating that IRs may also play chemosensory roles in the ovipositor . There is no evidence in the literature of any ionotropic glutamate receptor other than antennal IRs involved in the chemical sensation of insects . Previous morphology and electrophysiology studies shown that there were various types of sensilla including mechanical , contact chemosensory , and putative olfactory sensilla distributed on the ovipositor of lepidopteran insects ( Faucheux , 1991; Klijnstra and Roessingh , 1986; Maher and Thiery , 2004; Seada et al . , 2016; Waladde , 1983; Yamaoka et al . , 1971 ) . On H . assulta ovipositors we identified two types of chemosensory sensilla with shapes similar to the trichoid and basiconic sensilla on the antenna , respectively . However , unlike traditional olfactory sensilla , these multiporous sensilla also have a large terminal pore , which is characteristic of contact chemosensory sensilla . Therefore , we infer that these two types of sensilla may perform both olfactory and taste functions . The electrophysiological results indicated that some sensilla on the ovipositor responded to the tested odorants though their firing frequency was lower than that on the antennae . A 100-microgram equivalent of the tested compound stimulated around 20 spikes per second . The putative olfactory sensilla identified in the ovipositor of M . sexta have similar characteristics ( Klinner et al . , 2016 ) . In summary , this was the first study to functionally characterize an OR , HassOR31 , expressed in H . assulta ovipositors . We also characterized sensilla related to the ovipositor OR with a supplemental function to antennal olfactory sensilla . These sensilla were involved in detecting GLV and terpenoid compounds in host plants , and most likely played a role in oviposition site selection of this oligophagous herbivore . Future researches should focus on knocking out HassOR31 or destroying the related sensilla in the ovipositor and then measuring how insect behaviors are affected . The CRISPR-Cas9 genome editing method provides a good opportunity for us to knock out HassOR31 to verify its function . Moreover , the pathway of OSN-expressed HassOR31/HassORco projecting into the terminal abdominal ganglion and the function of superfluous HassOR31 in the ovipositor also require further investigation .
H . assulta were originally collected as larvae in tobacco fields in Zhengzhou , Henan Province , China , and successive generations were maintained in the laboratory under a 16 L: 8 D photoperiod cycle at 26 ± 1°C and 55–65% relative humidity . The larvae were reared on an artificial diet mainly constituted of wheat germ , yeast , and chili . Pupae were sexed , and males and females were put into separate cages for eclosion . After emergence , moths were fed with 10% honey in water . Two- to three-day old virgin females were used in the experiments . Xenopus laevis frogs were kindly provided by Prof . Qinghua Tao’s laboratory in School of Life Sciences , Tsinghua University , Beijing , China , and reared with pig livers as food in our laboratory at 20 ± 1°C . Xenopus laevis were anesthetized by submersion in frozen water , and the oocytes were surgically collected before the related experiments . All procedures were approved by the Animal Care and Use Committee of the Institute of Zoology , Chinese Academy of Sciences for the care and use of laboratory animals . One hundred PG-OVs were dissected from virgin female H . assulta during the 5th-8th hour of the scotophase and stored in −80°C freezer until RNA extraction . Total RNA was isolated using RNeasy Plus Universal Mini Kit ( Qiagen , Hilden , Germany ) , in which genomic DNA was removed by gDNA Eliminator . RNA concentration was determined using an ND-2000 spectrophotometer ( Nanodrop , Wilmington , DE , USA ) . RNA integrity was verified on Agilent 2100 BioAnalyzer ( Agilent , USA ) , and mRNA was isolated by magnetic beads with Oligo ( dT ) from ten µg of total RNA using Dynabeads mRNA purification kit ( Invitrogen , USA ) . In the next step , paired-end RNA-seq libraries were prepared by following Illumina’s library construction protocol . The libraries were sequenced on Illumina HiSeq2000 platform ( Illumina , USA ) in the Beijing Institutes of Biological Sciences , Chinese Academy of Sciences . FASTQ files of raw-reads were produced and sorted by barcodes for further analysis . Prior to assembly , 2 × 100 bp paired-end raw reads from each cDNA library were processed to remove adaptors , low quality sequences ( Q < 20 ) , and reads contaminated with microbes using Trimmomatic package ( Bolger et al . , 2014 ) . The FastQC package was used to verify the quality of resulting trimmed and filtered reads . The clean reads were de novo assembled to produce contigs using Grabherr et al . , 2011 ( https://github . com/trinityrnaseq/trinityrnaseq/ ) , the short reads assembling program using default parameters . We downloaded amino acid sequences of ORs of H . armigera from NCBI ( https://www . ncbi . nlm . nih . gov/ ) to construct local database . We then used BlastX with an E-value cut-off of 1e-5 to search the database to identify putative OR transcripts from the transcriptome we sequenced . To evaluate transcript expression abundances , the RSEM ( Li et al . , 2017a; http://deweylab . github . io/RSEM/ ) package was applied for calculation of the normalized gene expression value FPKM and TPM . We also sequenced transcriptomes of H . assulta’s male antennae , female antennae , proboscis , and forelegs . The clean reads from the above libraries were assembled with H . assulta’s pheromone gland-ovipositor transcriptome as described in Li et al . , 2017b . FPKM and TPM values of all candidate genes from the different tissues were calculated to indicate the tissue abundance distribution of identified genes . PCR sequencing was used to verify authentication of the identified genes . PCR experiments were conducted in a 25 µL reaction system with Q5 High-Fidelity DNA Polymerase ( NEB ) by using a thermal cycler . The thermal cycling conditions were set as follows: 98°C for 30 s; 30 cycles of 98°C for 10 s , 50°C for 30 s , and 72°C for 90 s; and 72°C for 2 min . PCR products were analyzed on 1 . 2% agarose gels and then verified by DNA sequencing . Primers for expression analysis were designed according to the sequencing results . To illustrate and compare the expression of HassOR31 in different tissues and organs , semi-quantitative reverse transcription PCR ( RT-PCR ) and real-time quantitative PCR ( qRT-PCR ) were conducted . Male antennae , female antennae , female head without antennae , female thorax , female abdomen , female legs , female wings , and pheromone gland–ovipositor were separately collected from 3 to 100 individuals , depending on the size of the organ , and then stored at −80°C . Total RNA was isolated as described above . cDNA was synthesized with M-MLV reverse transcriptase ( Promega , Madison , WI , USA ) from the total RNA . The synthesized cDNA was used as a template in RT-PCR reactions with gene-specific primers . PCR was performed under the following conditions: 94°C for 3 min , 30 cycles of 94°C for 30 s , 57°C for 30 s , 72°C for 30 s , and 72°C for 5 min . PCR amplification products were run on a 1 . 2% agarose gel and verified by DNA sequencing . An actin gene fragment was used as the reference to adjust the initial amount of cDNA used in the PCR procedure . qRT-PCR was conducted using Mx3005P qPCR System ( Agilent Technologies , CA , USA ) . All reactions were performed in triplicate in a total volume of 20 µL containing 10 µL SYBR Premix Ex TaqII ( TaKaRa , Otsu , Japan ) and 0 . 4 mM of each primer under the following conditions: 95°C for 30 s followed by 40 cycles of 95°C for 5 s , 60°C for 34 s , and 72°C for 30 s , 1 cycle 95°C for 15 s , 60°C for 1 min , 95°C for 15 s . Expression levels of all detected genes were calculated using the 2-ΔCt method , with 18S gene transcript as an internal control for sample normalization . All experiments were repeated three times using three independent RNA samples . The primer sequences are listed in Supplementary file 2 . Two-color double in situ hybridizations were performed following protocols reported previously ( Ning et al . , 2016 ) . Primers were designed to synthesize the gene-specific probe sequences from an open reading frame ( Supplementary file 2 ) . The product sizes of HassOR31 and HassORco were 846 and 834 nucleotides , respectively . Both digoxin ( Dig ) -labeled HassOR31 probe and biotin ( Bio ) -labeled HassORco probe were synthesized with DIG RNA labeling kit version 12 ( SP6/T7 ) ( Roche , Mannheim , Germany ) , with Dig-NTP or Bio-NTP ( Roche , Mannheim , Germany ) labeling mixture , respectively . Antisense and sense probes were generated from linearized recombinant pGEM-T vector using the T7/SP6 RNA transcription system ( Roche , Basel , Switzerland ) following recommended protocols . RNA probes were subsequently fragmented to an average length of about 300 bp by incubation in carbonate buffer . Ovipositors were dissected from three-day-old female moths , embedded in JUNG tissue freezing medium ( Leica , Nussloch , Germany ) and frozen at −80°C until use . Sections ( 12 µm ) of ovipositors were then mounted on SuperFrost Plus slides ( Boster , Wuhan , China ) . After a series of fixing and washing procedures , 100 µL hybridization solution ( Boster , Wuhan , China ) containing both Dig and Bio probes was placed onto the tissue sections . After adding a coverslip , slides were incubated in a humid box at 55°C overnight . After hybridization , slides were washed twice for 30 min in 0 . 1 × saline sodium citrate ( SSC ) at 60°C , treated with 1% blocking reagent ( Roche ) in TBST for 30 min at room temperature , and then incubated for 60 min with anti-digoxigenin ( Roche , Mannheim , Germany ) and Streptavidin-HRP ( PerkinElmer , Boston , USA ) . Visualization of hybridization signals was performed by incubating the sections first for 30 min with HNPP/Fast Red ( Roche , Mannheim , Germany ) , followed by three 5 min washes in TBS , with 0 . 05% Tween-20 ( Tianma , Beijing , China ) at room temperature with agitation . Then sections were incubated with Biotinyl Tyramide Working Solution for 8 min at room temperature followed by the TSA kit protocols ( PerkinElmer , USA ) . Sections were then washed three more times for 5 min each in TBS with 0 . 05% Tween-20 at room temperature with agitation . Finally , sections were mounted in Antifade Mounting Medium ( Beyotime , Beijing , China ) . Images were taken using a Carl Zeiss LSM710 confocal microscope ( Zeiss , Oberkochen , Germany ) and processed using ZEN 2012 software . Adobe Illustrator ( Adobe systems , San Jose , CA ) was used to arrange figures and the images were only altered to adjust the brightness or contrast . We expressed full length coding sequences of HassOR31/HassORco , HassOR31 , HassOR31/HassiGluR7 in X . laevis oocytes and analyzed the oocytes using two-electrode voltage clamping , as previously described ( Jiang et al . , 2014 ) . Oocytes were challenged with 51 chemicals with a concentration of 10−4 M , including host plant volatile compounds and two principal sex pheromone components ( listed in Supplementary file 3 ) . Two-electrode voltage clamping was used to detect the whole cell current . Total RNA and cDNA of ovipositors were obtained as described above . To obtain the full-length coding sequences , PCR was carried out using gene-specific primers with Kozak consensus sequence and Restriction Enzyme cutting site based on the mRNA sequences of HassOR31 , HassORco , and HassiGluR7 . The primer sequences are provided in Supplementary file 2 . The PCR program included initial denaturation at 94°C for 2 min , followed by 35 cycles of 30 s at 94°C , 30 s at 55°C , 90 s at 72°C; and a final extension step of 8 min at 72°C . Then , the coding sequences of HassOR31 , HassORco , and HassiGluR7 were cloned into pGEM-T easy vector ( Promega , Madison , WI , USA ) , then subcloned into pCS2+ vector . The pCS2+ vectors were linearized by using NotI ( Takara Shuzo , Shiga , Japan ) , cRNAs were synthesized from the linearized pCS2+ vectors with mMESSAGE mMACHINE SP6 ( Ambion , Austin , TX , USA ) . The cRNAs were dissolved in RNase-free water and stored at −80°C . Mature , healthy oocytes were treated with 2 mg/mL of collagenase type I ( Sigma-Aldrich ) in Ca2+-free saline solution ( 82 . 5 mM NaCl , 2 mM KCl , 1 mM MgCl2 , and 5 mM HEPES , pH 7 . 5 ) for 1–2 hr at room temperature . Each oocyte was microinjected with 23 . 6 nL ( 50 ng ) of the mixture of HassOR31 and HassORco ( or HassiGluR7 ) cRNA at a ratio of 1:1 or HassOR31 cRNA alone . Oocytes injected with RNAase-free water was used as a negative control . Injected oocytes were incubated for 3–4 days at 17°C in the bath solution ( 96 mM NaCl , 2 mM KCl , 1 mM MgCl2 , 1 . 8 mM CaCl2 , and 5 mM HEPES , pH 7 . 5 ) supplemented with 100 µg/mL gentamycin and 550 µg/mL sodium pyruvate . Whole-cell currents were recorded with a two-electrode voltage clamp . Intracellular glass electrodes were filled with 3 M KCl and presented resistances of 0 . 2–2 . 0 MΩ . Signals were amplified with an OC-725C amplifier ( Warner Instruments , Hamden , CT , USA ) at a holding potential of −80 mV , low-pass filtered at 50 Hz and digitized at 1 kHz . Sodium bicarbonate ( NaHCO3 ) was diluted in Ringer’s solution before being introduced to the oocyte recording chamber using a perfusion system . Data acquisition and analysis were carried out with Digidata 1322A and pCLAMP software ( Axon Instruments Inc , Foster City , CA , USA ) . Dose-response data were analyzed using GraphPad Prism 7 . The ovipositors of H . assulta were carefully dissected and placed into phosphate buffer ( PBS , 0 . 1 M , pH = 7 . 2 ) containing 2 . 5% glutaraldehyde and fixed at 4°C for 4 hr . They were then flushed in PBS buffer three times for 10 min each time . The ovipositors were placed in increasing gradient ethanol solutions for dehydration . The ethanol concentrations were 10% , 30% , 50% , 70% , 90% , and 100% , respectively . For each concentration , the ovipositors were rinsed for 30 min . Then , they were ultrasonically cleaned in 100% ethanol for 20 s . Then the ovipositors were treated in isoamyl acetate for 30 min . All the samples were dried using a CO2 critical point dryer ( model HCP-2 , Hitachi , Tokyo , Japan ) . The dried samples were stuck to the sample stage in different orientations . The samples were coated with gold using a Hitachi Sputter Ionizer ( model S-4800 , Hitachi ) for 60–90 s . Finally , the samples were photographed using a Hitachi S-4800 scanning electron microscope . A single moth was placed into a 1 mL Eppendorf pipette tip with the narrow end cut off . The moth was gently pushed until its abdomen protruded from the cut end . The ovipositor was extended by gently pressing onto the abdomen , then fixed with dental wax and wrapped tightly with Parafilm . The reference electrode was inserted into the abdomen of the insect , and the sharpened tungsten recording electrode was inserted into the base of the sensilla housed in the ovipositor ( Video 2 ) . The recorded signals were then amplified through an IDAC interface amplifier ( IDAC-4 , Syntech , Germany ) . The software Autospike ( Syntech , Germany ) was used to store and analyze data . A continuous stream of purified and humidified air was directed onto the ovipositor ( 12 . 5 mL/s ) from the outlet of a steel tube ( i . d . 6 mm , length 15 cm ) , positioned 1 cm from the ovipositor . Test odorants were injected into the air stream using a stimulus flow controller ( CS-55 , Syntech , Germany ) , which generated 200 ms air pulses through the odor cartridge at a flow rate of 10 mL/s , while a compensating air flow was provided to keep a constant current . The odorants supplied during single sensillum recordings are listed in Supplementary file 3 . All odorants were diluted to a final concentration of 10 µg/µL ( 1% w/v ) in mineral oil . Ten µL of the diluted odors were pipetted onto a small piece of filter paper ( 2 . 5 cm ×0 . 7 cm ) and placed inside a glass Pasteur pipette . Newly emerged female and male moths were mixed at a sex ratio of 1:1 . 3 in 26 cm ×26 cm × 26 cm cubic cages covered with gauze for 3 days to ensure the female moths were fully mated , and then the mated female moths were randomly separated into two groups . One group of insects had their antennae removed just before the experiments , and another group of insects were kept as intact females . Two choice tests were performed . Choice test 1: Ten females from each group were put into a cylinder cage ( diameter 24 cm , height 26 cm ) . Only the top side of the cage was covered with gauze for females to lay eggs , and all the other sides were covered with black cloth . The gauze side was equally divided into four areas . From the beginning of the scotophase , two fresh hot pepper fruit discs of 1 . 5 cm diameter were positioned above each of two opposite areas of the gauze , and no pepper discs were put above the other two areas . The pepper discs were supported by a stainless net shelf to avoid the moths inside the cage directly contacting the pepper ( Figure 6—figure supplement 1A ) . After 24 hr the number of eggs on each section of gauze was counted . After counting , the gauze was replaced by a new one . The number of eggs laid on each part of the gauze were counted every day for 4 days and the mean number of eggs was calculated . Seven replications were run . Choice test 2: Oviposition preference of female H . assulta to Z-3-hexenyl butyrate was performed in screened cages ( 1m × 1 m×1 m ) as described in Wu et al . , 2019 , fifteen females from each group were put into each cage . In each cage , four fake green plants were respectively placed at the four corners . Two fake plants treated with Z-3-hexenyl butyrate were put in one diagonal , and the other two fake plants with paraffin oil were put in the other diagonal and used as control . Z-3-Hexenyl butyrate were dissolved in paraffin oil and then dropped into a rubber head , which was placed on the leaf of the fake plant ( Figure 6—figure supplement 1B ) . The dosage of Z-3-hexenyl butyrate used in the experiment was 100 μg . The number of eggs laid on each fake leaf in 48 hr were counted . The oviposition preference index was calculated as ( T−C ) / ( T+C ) . T is the number of eggs on the leaves with Z-3-hexenyl butyrate , and C is the number of eggs on the leaves with paraffin oil . Five replications were run . Electrophysiological response values ( currents and spikes per second ) are indicated as mean ± SEM . One-way ANOVA and Turkey’s multiple comparisons test were used to compare the responses of HassOR31/HassORco to different tested compounds and single sensillum recording data . Paired t tests were performed to analyze numbers of eggs in oviposition choice tests . When comparing oviposition preference indexes , we used unpaired t tests and differences were considered significant when p<0 . 05 . All data were analyzed using the GraphPad Prism version 7 . 00 for Mac OS X , GraphPad Software , La Jolla California USA , www . graphpad . com . | When most insects reproduce they lay eggs that hatch into juveniles known as larvae . To provide good sources of food for the larvae , the adult insects have to carefully select where to lay the eggs . Host plants produce specific sets of chemicals known as odorants that the adult insects are able to smell using proteins called odorant receptors . It is generally thought that odorant receptors in the antennae on the head are responsible for guiding adult insects to good egg-laying sites . However , recent studies have reported that odorant receptors are also present in the egg-laying organs of several different species of moth . It remains unclear what role these odorant receptors may play in egg-laying . The oriental tobacco budworm ( Helicoverpa assulta ) is considered a serious pest in agriculture . The adult moths lay their eggs on a narrow range of plants in the nightshade family including tobacco and hot pepper . Li et al . have now investigated the odorant receptors of H . assulta and found that one gene for an odorant receptor called HassOR31 was expressed much more in the egg-laying organs of the moths than in the antennae . Further experiments showed that this receptor was tuned to respond to 12 odorants that also stimulated responses in the egg-laying organ of H . assulta . Together these findings suggest that this odorant receptor in the egg-laying organ helps the moths find suitable host plants to lay their eggs on . The work of Li et al . may help us understand how H . assulta evolved to lay its eggs on specific members of the nightshade family and lead to new methods of controlling this pest . An insect’s sense of smell guides many other behaviors including finding food , mates and avoiding enemies . Therefore , these findings may inspire researchers to investigate whether odorant receptors in the antennae or other organs guide these behaviors . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"ecology"
] | 2020 | A moth odorant receptor highly expressed in the ovipositor is involved in detecting host-plant volatiles |
The Roundabout ( Robo ) guidance receptor family induces axon repulsion in response to its ligand Slit by inducing local cytoskeletal changes; however , the link to the cytoskeleton and the nature of these cytoskeletal changes are poorly understood . Here , we show that the heteropentameric Scar/Wave Regulatory Complex ( WRC ) , which drives Arp2/3-induced branched actin polymerization , is a direct effector of Robo signaling . Biochemical evidence shows that Slit triggers WRC recruitment to the Robo receptor’s WRC-interacting receptor sequence ( WIRS ) motif . In Drosophila embryos , mutants of the WRC enhance Robo1-dependent midline crossing defects . Additionally , mutating Robo1’s WIRS motif significantly reduces receptor activity in rescue assays in vivo , and CRISPR-Cas9 mutagenesis shows that the WIRS motif is essential for endogenous Robo1 function . Finally , axon guidance assays in mouse dorsal spinal commissural axons and gain-of-function experiments in chick embryos demonstrate that the WIRS motif is also required for Robo1 repulsion in mammals . Together , our data support an essential conserved role for the WIRS-WRC interaction in Robo1-mediated axon repulsion .
The brain is the most complex organ in the body , with trillions of specific synapses whose formation depends on the precise targeting of axons and dendrites during nervous system development . Axons are guided to their appropriate targets by a number of conserved guidance cues and their receptors , which enable neurons to form specific connections and establish functional neural circuits . The axon guidance receptors that mediate axonal guidance and targeting are tightly regulated to achieve a controlled balance between attractive and repulsive signaling , and disruption of this process results in a number of movement disorders and other neurological deficits ( Bosley et al . , 2005; Depienne et al . , 2011; Jen et al . , 2004 ) . Specifically , the Roundabout ( Robo ) family of repulsive axon guidance receptors has been implicated in many neurodevelopmental disorders like autism spectrum disorder , dyslexia , horizontal gaze palsy , and others ( Anitha et al . , 2008; Hannula-Jouppi et al . , 2005; Jen et al . , 2004; Suda et al . , 2011 ) . Elucidating the mechanisms by which these guidance receptors function is crucial for understanding the formation of neural circuits both during development and in disease pathogenesis . The Drosophila midline is analogous to the vertebrate spinal cord and serves as an intermediate target for commissural axons that cross from one side of the body to the other ( Klämbt et al . , 1991; Seeger et al . , 1993 ) . The Drosophila ventral nerve cord has a ladder-like structure consisting of 13 repeated segments , each containing an anterior commissure and a posterior commissure into which commissural neurons extend their axons to cross the midline . Midline glial cells secrete a number of guidance cues that act on their cognate receptors present on axon growth cones to induce attraction toward or repulsion away from the midline . Slit is secreted by midline glia and acts as a repulsive ligand for the Robo family of receptors ( Brose et al . , 1999; Kidd et al . , 1999; Kidd et al . , 1998 ) . There are three Robo receptors in Drosophila and four in vertebrates . The Robo receptors are transmembrane proteins with an ectodomain consisting of five immunoglobulin-like domains and three fibronectin repeats , and an intracellular domain containing short , highly conserved cytoplasmic ( CC ) motifs ( Bashaw et al . , 2000; Kidd et al . , 1998 ) . Robo1 induces repulsion in growth cones of navigating axons primarily by modulating the actin cytoskeletal network . Previous work has identified some downstream effectors for Robo1 including Ena , an uncapping protein for actin filaments ( Bashaw et al . , 2000 ) , and Son of Sevenless ( SOS ) , a GEF for Rac1 ( Yang and Bashaw , 2006 ) . However , downstream signaling of Robo1 is not completely understood , especially in relation to effectors that directly link Robo1 to the actin cytoskeleton and the nature of cytoskeletal changes orchestrated by Robo1 . While it seems intuitive for repulsive signaling to induce depolymerization of the actin network , a recent study reports that dorsal root ganglion axons first extend actin-rich filopodia toward a source of Slit before retracting away from it ( McConnell et al . , 2016 ) . This challenges the prevailing notion that repulsive signaling primarily relies on actin depolymerization and suggests that the actin rearrangements occurring downstream of Robo1 are more nuanced and complex than previously thought . Indeed , several of the well-known downstream effectors of Robo1 signaling , namely Ena and Rac1 , are documented enhancers of actin polymerization ( Barzik et al . , 2005; Ridley et al . , 1992 ) . The Scar or WAVE regulatory complex ( WRC ) is a heteropentameric complex consisting of five different proteins: Scar/WAVE , CYFIP/Sra1 , Kette/Nap1 , HSPC300/Brick1 , and Abi ( Eden et al . , 2002 ) . Scar or WAVE contains a VCA ( verprolin homology , cofilin homology , acidic ) region and serves as a nucleation-promoting factor for Arp2/3 , thereby driving branched actin polymerization . While mammals have multiple orthologs of these proteins , Drosophila has single homologs of all five members of the complex , making it a simpler , more tractable model system for studying the WRC . The WRC has been previously implicated in axon guidance and targeting in Drosophila and Caenorhabditis elegans ( Shakir et al . , 2008; Stephan et al . , 2011; Xu and Quinn , 2012 ) ; however , if and how it is recruited and activated downstream of guidance receptors is not known . Recent work identified a unique binding site for the WRC known as the WRC-interacting receptor sequence ( WIRS ) motif ( Chen et al . , 2014a ) . The WIRS motif is a short six amino acid peptide sequence characterized by a bulky hydrophobic residue at position 1 and a threonine or a serine at position 3 , followed by a phenylalanine at position 4 . The WIRS motif is present in a number of transmembrane proteins including Robo1 ( Chen et al . , 2014a ) . Robo1 has a WIRS motif between its CC0 and CC1 domains that is conserved across species , including humans . Previously , the WIRS motif has been shown to be important for recruitment of the WRC by neuroligins and SYG-1 in synapse formation ( Chia et al . , 2014; Xing et al . , 2018 ) and for neogenin function in maintaining the stability of adherens junctions ( Lee et al . , 2016 ) . To our knowledge , this study is the first to demonstrate a role for the WIRS-WRC interaction in axon guidance . Here , we show that the WRC is required for Slit-Robo1 repulsive signaling at the Drosophila midline . We present evidence that Robo1 interacts with the WRC partially via its WIRS motif and that this interaction is enhanced in the presence of Slit . We show that the WIRS motif in Robo1 is important for its ability to induce ectopic repulsion in vivo . Using rescue assays , we show that Robo1 also requires its WIRS motif to mediate repulsion in ipsilateral axons in vivo . In addition , using CRISPR-Cas9-mediated mutagenesis , we show that the WIRS motif is important for endogenous Robo1 function as mutating the endogenous WIRS motif results in the complete loss of Robo1 repulsion at the midline . Finally , we use mouse dorsal spinal cord explants and growth cone collapse assays in mouse commissural neurons , together with gain-of-function experiments in chick embryos , to demonstrate that the WIRS motif is also important for vertebrate Robo1 repulsive signaling . We propose a model in which Slit binding induces recruitment of the WRC to the WIRS motif of Robo1 where it functions in Robo1-mediated repulsion at the midline .
WRC members are enriched in the Drosophila ventral nerve cord during embryonic stages 12–17 , encompassing the developmental window when midline crossing decisions are being made ( Schenck et al . , 2004 ) . To confirm these previously published observations , we examined the expression of Scar by immunofluorescence and observed strong axonal staining throughout embryonic stages when midline axon guidance occurs ( Figure 1—figure supplement 1A ) . To investigate the potential role of the WRC in Slit-Robo repulsion at the midline , we tested for genetic interactions between cyfip and hspc300 , two members of the WRC , and the Slit-Robo signaling pathway . In wild-type embryos , FasII-positive ipsilateral axons project longitudinally and never cross the midline ( Figure 1A ) . In robo1 mutants , axons in the medial most Fas-II bundle frequently cross and re-cross the midline , resulting in a very strong ectopic crossing phenotype ( Kidd et al . , 1998; Figure 1B ) . In slit , robo1/+ embryos , where the slit and robo1 gene dosage is reduced by half , the phenotype is milder ( Figure 1D ) . This represents a sensitized background in which we can detect enhancers or suppressors of the Slit-Robo pathway ( Chance and Bashaw , 2015; Coleman et al . , 2010; Fan et al . , 2003; Hsouna et al . , 2003 ) . While we see no crossing errors in FasII-positive axons in hspc300 mutants alone ( Figure 1C ) , in the slit , robo1/+ sensitized background , hspc300 mutants exhibit a significant enhancement of the ectopic crossing defects ( Figure 1E ) . These interactions are dosage sensitive as removing one copy of hspc300 results in a moderate enhancement of crossing errors while removing both copies of hspc300 results in a much stronger phenotype ( Figure 1F ) . Similarly , we see almost no crossing errors in FasII-positive axons in cyfip mutants alone ( Figure 1G ) ; however , in the slit , robo1/+ sensitized background ( Figure 1H ) , cyfip mutants show a strong dose-dependent enhancement of the ectopic crossing defects ( Figure 1I , J ) . Strikingly , removing both copies of cyfip in this background results in a very strong phenotype with ectopic crossing defects in nearly 100% of segments , similar to the robo1 mutant phenotype ( Figure 1B , J ) . These ectopic crossing defects can be significantly rescued by the transgenic expression of UAS-CYFIP using the pan-neuronal elav-Gal4 driver ( Figure 1K , L ) . This suggests that the neuronal function of CYFIP is important for Slit-Robo-mediated repulsion at the midline . It is important to note that zygotic hspc300 and cyfip mutants , like mutants for all other members of the WRC , still have significant amounts of the protein remaining due to maternal deposition ( Schenck et al . , 2004; Zallen et al . , 2002 ) . This likely explains why these zygotic mutants have no phenotype on their own . This can be seen in scar zygotic mutants where the overall Scar protein level is significantly reduced but there is still a considerable amount of Scar protein remaining in central nervous system ( CNS ) axons ( Figure 1—figure supplement 1B , C ) . To determine whether CYFIP is required cell-autonomously , we examined a more restricted subset of ipsilateral axons , the apterous ( ap ) axons . Just like FasII axons , ap axons are sensitive to a partial loss of repulsion . Reducing the slit and robo1 gene dosage by half in slit , robo1/+ embryos results in a mild phenotype where ectopic midline crossing of ap axons is seen in approximately 40% of segments ( Figure 1N , Q ) . Homozygous cyfip mutants in this sensitized background show a strong enhancement of the ectopic ap crossing defects with 85% of segments exhibiting ectopic crossing ( Figure 1O ) . We also visualized all CNS axons using HRP and observed abnormal thickening and fusion of the commissures , a phenotype that bears strong resemblance to robo1 mutants . Importantly , ap-specific expression of UAS-CYFIP significantly rescues the ectopic ap crossing defects but not the pan-neuronal HRP defects ( Figure 1P , Q ) providing strong support for a cell-autonomous role for CYFIP in Slit-Robo1 signaling . Together , these genetic data suggest that the WRC functions in the Slit-Robo1 pathway at the Drosophila midline . Previous work has identified Rac1 as an important effector of Robo1 signaling in both Drosophila and mouse ( Fan et al . , 2003; Wong et al . , 2001 ) . SOS is a Rac-GEF that activates Rac1 downstream of Robo1 and is required for Robo1-mediated midline repulsion ( Chance and Bashaw , 2015; Yang and Bashaw , 2006 ) . Since Rac1 is a well-known activator of the WRC ( Chen et al . , 2017; Chen et al . , 2010; Eden et al . , 2002; Ismail et al . , 2009 ) , we reasoned that Rac1 might be responsible for activating the WRC downstream of Robo1 , and that Rac1 and the WRC would function cooperatively in the same pathway to regulate Robo1-mediated repulsion . Thus , we predicted that the simultaneous reduction of CYFIP and the Rac-GEF , SOS , would greatly impair Robo1-mediated repulsion , resulting in axons ectopically crossing the midline . As SOS is also maternally deposited ( Yang and Bashaw , 2006 ) , zygotic sos mutants show very mild ectopic crossing defects in approximately 15% of segments ( Figure 2A ) . In contrast , double mutants for sos and cyfip show a striking phenotype in which FasII-positive axons ectopically cross the midline in over 80% of segments ( Figure 2B , C ) , a phenotype that bears strong resemblance to the robo1 mutant phenotype . In addition to examining the phenotype with FasII immunostaining , we also visualized all CNS axons using HRP and observed frequent thickening and fusion of the anterior and posterior commissures , which again bears strong resemblance to robo1 mutants ( Figure 2B ) . Thus , cyfip genetically interacts with sos to give a strong ectopic crossing phenotype very similar to that seen in robo1 mutants , supporting the idea that Rac1 and the WRC act cooperatively to regulate midline repulsion . In Drosophila embryos , both Robo1 and , to a lesser extent , Robo2 contribute to midline repulsion in response to Slit ( Rajagopalan et al . , 2000; Simpson et al . , 2000 ) . Indeed , on their own robo2 mutants exhibit only mild phenotypes; however , robo1 , robo2 double mutants exhibit a complete collapse of all CNS axons at the midline , phenocopying the slit mutant phenotype . Therefore , mutations in genes that contribute to robo1 repulsion would be expected to strongly enhance the mild phenotype observed in robo2 mutants . In robo2 mutant embryos , FasII-positive axons ectopically cross the midline in approximately 17% of segments ( Figure 2D ) . In robo2 , cyfip double mutant embryos , ectopic crossing defects are greatly increased to approximately 75% of segments ( Figure 2E , F ) and axon commissures are thicker and frequently fused , providing additional support for a role for the WRC in midline repulsion . Taken together , these genetic interaction results strongly suggest that the WRC functions in Slit-Robo1-mediated repulsive signaling at the midline . The cytoplasmic tail of Robo1 contains a WIRS motif , which is conserved in vertebrates ( Figure 3A ) . The purified cytoplasmic tail of human Robo1 directly interacts with the WRC in pulldown assays via its WIRS motif ( Chen et al . , 2014a ) . To determine if this WIRS-dependent interaction with the WRC is conserved in Drosophila Robo1 , we performed co-immunoprecipitation assays in Drosophila embryonic S2R+ cells ( DGRC , #150 ) using tagged constructs of Robo1 and HSPC300 . The relatively small size of HSPC300 facilitated consistent levels of expression and reduced trial-to-trial variability . We found that Robo1 immunoprecipitated with HSPC300 , indicating that Drosophila Robo1 , like human Robo1 , can also interact with the WRC ( Figure 3C ) . Next , we introduced point mutations into the WIRS motif of Robo1 ( Robo1ΔWIRS; Figure 3B ) and found a significant decrease in the amount of Robo1 that immunoprecipitated with HSPC300 ( Figure 3C , E ) . Thus , mutating the WIRS motif substantially disrupts the binding of Robo1 to the WRC , indicating that Robo1 interacts with the WRC partly via the WIRS motif . In contrast , the previously published interaction data for human Robo1 ( Chen et al . , 2014a ) showed that mutating the WIRS motif completely abolishes binding to the WRC . We speculate that there may be a small amount of indirect binding of Robo1 to the WRC via Ena or DOCK , which are known interactors of Robo1 ( Bashaw et al . , 2000; Fan et al . , 2003 ) . Previous work has identified interactions between Ena and Abi ( Chen et al . , 2014b ) and between the DOCK homolog Nck and Nap1 ( Kitamura et al . , 1996 ) . Both Abi and Nap1 are members of the WRC . As the pulldown assay with human Robo1 was done using purified proteins , any indirect binding will not be detected . Support for this notion comes from our co-immunoprecipitation results of Robo2 and HSPC300 . Drosophila Robo2 is structurally similar to Robo1 except that it lacks the CC motifs CC2 and CC3 present in Robo1 that serve as the interaction sites for Ena and DOCK ( Bashaw et al . , 2000; Fan et al . , 2003; Figure 3—figure supplement 1A ) . Indeed , we find that Robo2 can also interact with HSPC300 though mutating the WIRS motif of Robo2 almost completely abolishes this interaction ( Figure 3—figure supplement 1B , C ) . This result is consistent with the idea that there might be indirect binding of the WRC to Robo1 via its interaction with other WRC partners but not to Robo2 that lacks any such interactions . Next , we wanted to test whether the Robo1-WRC interaction is regulated by the Robo ligand Slit . We treated S2R+ cells with bath application of Slit-conditioned media ( Slit-CM ) and found a substantial increase in the interaction between Robo1 and HSPC300 as compared to cells treated with mock-CM ( Figure 3D , F ) . By contrast , Robo1ΔWIRS shows no significant increase in binding to HSPC300 upon Slit-CM treatment . As there is significant variability in the activity of Slit-CM with each preparation , we see different levels of enhancement in binding obtained with each Slit treatment . Nevertheless , Slit application consistently increases the interaction between Robo1 and HSPC300 . These results suggest that upon Slit binding the WRC is recruited to Robo1 via its WIRS motif . Finally , to test whether this interaction occurs in vivo , we performed co-immunoprecipitation assays using Drosophila embryonic protein lysates . We generated transgenic flies using the GFP-tagged HSPC300 construct and HA-tagged Robo1 constructs . The pan-neuronal elav-Gal4 driver was used to drive expression of UAS-HSPC300-GFP either alone or with the wild-type UAS-HA-Robo1 or UAS-HA-Robo1ΔWIRS transgenes in Drosophila embryos . While wild-type Robo1 co-immunoprecipitates with HSPC300 , mutating the WIRS motif results in a significant decrease in this binding ( Figure 3G , H ) . These results indicate that Robo1 interacts with the WRC in vivo as well and that this interaction is partly dependent on the WIRS motif . To test whether this interaction with the WRC is required for Robo1 function in vivo , we compared the gain-of-function and rescue phenotypes of wild-type Robo1 and Robo1ΔWIRS in specific neuronal subsets in the Drosophila ventral nerve cord . We generated transgenic flies with wild-type UAS-Robo1 or UAS-Robo1ΔWIRS constructs . Both the transgenes are tagged with an HA epitope and inserted into the same genomic locus . Immunostaining for HA shows that both transgenes are expressed at comparable levels ( Figure 4D , E ) . Using the eg-Gal4 driver , we expressed these transgenes in eagle neurons , a subset of commissural neurons . Eagle neurons , visualized here using a GFP reporter , consist of two populations: the EG population , which extends its axons in the anterior commissure of a segment , and the EW population , which extends axons in the posterior commissure ( Figure 4A ) . Overexpression of wild-type Robo1 in these neurons causes ectopic repulsion from the midline , resulting in a strong gain-of-function phenotype where almost all EW axons fail to cross the midline ( Figure 4B ) . In contrast , overexpression of Robo1ΔWIRS results in a significantly weaker gain-of-function phenotype where EW axons in approximately 70% of segments fail to cross the midline ( Figure 4C , F ) . Thus , mutating the WRC interaction site on Robo1 hampers its ability to induce ectopic repulsion in vivo . Next , we assessed the ability of Robo1ΔWIRS to rescue the ectopic crossing defects of FasII-positive axons seen in robo1 mutant embryos . Unlike in wild-type embryos , where FasII axons never cross the midline ( Figure 4G ) , in robo1 mutants , axons in the medial most fascicle freely cross and recross the midline in 100% of segments ( Figure 4H ) . Re-expressing wild-type Robo1 with the pan-neuronal driver elav-Gal4 restores the ipsilateral projection pattern in most of the segments , lowering the frequency of ectopic crossing to 25% of segments ( Figure 4I ) . In contrast , re-expression of Robo1ΔWIRS fails to rescue the crossing defects in 70% of segments ( Figure 4J , K ) . This indicates that in the absence of a functional WIRS motif Robo1 is not nearly as effective at restoring repulsive signaling in ipsilateral axons in vivo . Altogether , these results suggest a role for the WIRS motif in Robo1 repulsive signaling at the midline . Our in vivo results obtained so far have relied on misexpression or overexpression of Robo1 that likely is not subject to the adequate spatial and temporal regulation that is critical for guidance receptor function . Further , such unregulated high levels of Robo1 expression on the cell surface could potentially mask dysfunction in receptor activity . We see this especially for the rescue experiments with our UAS-Robo1 transgenes . While the difference in rescue activity between 5XUAS-Robo1 and 5XUAS-Robo1ΔWIRS is around 50% ( Figure 4K ) , performing this rescue assay with 10XUAS-Robo1 and 10XUAS-Robo1ΔWIRS transgenes , which have double the number of UAS enhancer sites and express much higher levels of the Robo1 variants , gives a much more modest difference of 13% ( Figure 4—figure supplement 1A–E ) . Indeed , in rescue experiments using 10XUAS-Robo1 transgenes , we see strong gain-of-function effects that lead to both rescue of abnormal crossing of FasII-positive axons , as well as ectopic repulsion of commissural axons ( Figure 4—figure supplement 1F–J ) . Notably , the ectopic repulsion of commissural axons induced by the 10XUAS-Robo1ΔWIRS transgene is significantly weaker than the ectopic repulsion induced by the wild-type receptor ( Figure 4—figure supplement 1H–J ) . Given these caveats , we sought to analyze the function of the WIRS motif in Robo1 signaling in a more endogenous context . First , we performed a rescue assay with an HA-tagged genomic rescue construct of robo1 that contains upstream and downstream regulatory regions of Robo1 in addition to the Robo1 coding sequence ( Brown et al . , 2015 ) . Transgenics created with this construct show a Robo1 expression pattern that closely resembles that of endogenous Robo1 ( Brown et al . , 2015 ) . We mutated the WIRS motif in this robo1 genomic rescue construct and inserted the transgene into the same genomic site as the wild-type construct . Both transgenes show comparable levels of Robo1 expression upon HA immunostaining ( Figure 5—figure supplement 1A , B ) . We tested the ability of these transgenes to rescue the robo1 mutant phenotype in FasII-positive axons ( Figure 5B ) . One copy of the wild-type robo1 genomic rescue construct ( genRobo ) was able to rescue ectopic crossing of FasII-positive axons in almost all segments with only 6% still showing defects ( Figure 5C ) while robo1ΔWIRS genomic rescue construct ( genRoboΔWIRS ) was unable to rescue ectopic crossing defects in over 70% of segments ( Figure 5D , E ) . Similarly , for HRP stained axons , the frequent thickening and fusion of the anterior and posterior commissures in robo1 mutants ( Figure 5B ) can be rescued with the wild-type genRobo but not with genRoboΔWIRS ( Figure 5C , D ) . These results , in more physiologically relevant contexts , demonstrate a marked decline in Robo1 function upon disruption of the WRC binding site . Finally , using the CRISPR-Cas9 system , we mutated the WIRS motif in the endogenous robo1 locus . We used a single-guide RNA that targets the endogenous WIRS motif and a single-stranded oligonucleotide template to introduce point mutations in the WIRS motif ( Figure 5—figure supplement 2A ) . We sequenced the regions surrounding the WIRS motif to verify that we had successfully mutated the WIRS motif without introducing any unwanted frameshift mutations or deletions . While we found no frameshifts , we did notice that our strategy had resulted in an unexplained loss of the smaller intron 16 ( Figure 5—figure supplement 2A ) . Since the genRobo constructs and the previously used robo swap alleles ( Spitzweck et al . , 2010 ) that can restore Robo1 function fully do not contain any intronic sequences , we believe that it is extremely unlikely that the loss of this intron affects Robo1 function . Next , we analyzed the phenotypes of both HRP and FasII-positive axons in these roboΔWIRS CRISPR embryos . We see a surprisingly strong ectopic crossing phenotype in these embryos with defects in almost 100% of segments , showing that they fully phenocopy the robo mutant embryos ( Figure 5B , F ) . We were able to achieve a near perfect rescue with the introduction of one copy of genRobo , indicating that this phenotype is not a result of any off-target effects arising from Cas9-mediated cleavage ( Figure 5G , H ) . This result also supports our interpretation that the loss of intron 16 in our CRISPR allele has no effect on Robo1 function since the genRobo construct does not include any introns . As an additional control , we also tested whether the roboΔWIRS CRISPR mutations disrupt normal Robo1 expression . To investigate this , we immunostained for Robo1 expression using a monoclonal Robo1 antibody . Unlike the robo mutants in which no Robo1 protein can be detected ( Figure 5—figure supplement 2C , F ) , we see substantial Robo1 staining in the roboΔWIRS CRISPR mutants , suggesting that the phenotype is not due to a failure in protein production ( Figure 5—figure supplement 2D , G ) . Unlike in wild-type embryos where Robo1 expression is seen primarily on longitudinal tracts and is downregulated on commissures ( Figure 5—figure supplement 2B , E ) , in the roboΔWIRS CRISPR mutant embryos , we see Robo1 also being expressed on commissures ( Figure 5—figure supplement 2D , G ) . While interesting , this observation is not necessarily surprising to us as this altered Robo1 localization on commissures has also been noted in previous studies when Robo1 signaling is disrupted ( Coleman et al . , 2010 ) . Altogether , our genomic Robo rescue assays and roboΔWIRS CRISPR mutant phenotypes strongly suggest an important role for the WIRS motif in Robo1 repulsive function in vivo . We have shown that the WRC is an important component of the Slit-Robo1 repulsive pathway at the midline . But what happens after the WRC is recruited to Robo1 ? Is the WRC acting via Arp2/3 to promote branched actin polymerization downstream of Robo1 ? To address this question , we tested for genetic interactions between arpc2 , a member of the Arp2/3 complex and the Slit-Robo pathway . Similar to members of the WRC , arpc2 mutants alone have no ectopic crossing phenotype in FasII axons; however , when introduced into the slit , robo/+ sensitized background , arpc2 homozygous mutants show a significant enhancement of the ectopic FasII crossing defects ( Figure 6A–C ) , suggesting that the Arp2/3 complex functions in the Slit-Robo repulsive pathway . Additionally , when we remove one copy of arpc2 together with one copy of cyfip , we again observe a significant enhancement of the slit , robo/+ ectopic crossing defects ( Figure 6—figure supplement 1A–C ) . This genetic interaction between arpc2 and cyfip suggests a cooperative effect of the WRC and Arp2/3 in the Slit-Robo1 signaling pathway at the midline . Next , we overexpressed Robo1 in eagle neurons , which results in a strong gain-of-function phenotype where almost all EW neurons fail to cross the midline . In contrast , overexpressing Robo1 in arpc2 mutants results in a small but significant suppression of this phenotype ( Figure 6—figure supplement 1D–F ) that is similar to the suppression seen in cyfip mutants ( Figure 6—figure supplement 1G–I ) , demonstrating a reduction in Robo1’s ability to induce ectopic repulsion . Together , these genetic data strongly suggest that the Arp2/3 complex functions in the Slit-Robo1 repulsive pathway . To determine whether the Arp2/3 complex can physically interact with Robo , we performed co-immunoprecipitation assays in Drosophila embryonic S2R+ cells using tagged constructs of Robo1 and Arp3 , another component of the Arp2/3 complex . We found that Robo immunoprecipitated with Arp3 , suggesting that the Arp2/3 complex can physically interact with Robo ( Figure 6D ) . We reasoned that if the Arp2/3 complex was being recruited by the WRC to Robo , we would expect that mutating the WIRS motif would disrupt this interaction between Arp2/3 and Robo . Indeed , we found a significant decrease in the amount of RoboΔWIRS that immunoprecipitated with Arp3 as compared to wild-type Robo ( Figure 6D , E ) . Furthermore , we can detect an increase in the interaction between Robo and Arp3 in the presence of Slit-CM as compared to mock-CM , suggesting that similar to the WRC , the Arp2/3 complex is also recruited to Robo in response to Slit . By contrast , RoboΔWIRS shows no significant increase in binding to Arp3 in the presence of Slit , demonstrating that the WIRS motif is important for this Slit-dependent recruitment of the Arp2/3 complex to Robo . Together , these observations support the model that upon Slit binding the WRC is recruited to the WIRS motif of Robo and activated , which is in turn responsible for the recruitment of the Arp2/3 complex to facilitate cytoskeletal remodeling downstream of Robo . One possible outcome of initiating localized actin polymerization is the endocytosis and recycling of transmembrane receptors . Indeed , both Drosophila Robo as well vertebrate Robo1 have been previously shown to undergo endocytosis following Slit stimulation ( Chance and Bashaw , 2015; Kinoshita-Kawada et al . , 2019 ) . Furthermore , the WRC has been shown to play a role in initiating receptor endocytosis ( Basquin et al . , 2015; Xu et al . , 2016 ) . Thus , to further evaluate the mechanism of WRC function in Slit-Robo signaling , we investigated whether mutating the WIRS motif in Robo could disrupt signaling by preventing Robo endocytosis . To address this question , we tested whether RoboΔWIRS displays increased surface localization compared to wild-type Robo in both Drosophila embryonic neurons and in cultured dorsal commissural neurons from mice . First , we tested whether Drosophila embryos expressing the genomic HA-tagged Robo rescue transgenes display any difference in surface localization . We dissected embryos live and visualized surface expression of Robo by staining the N-terminal HA tag before fixation and permeabilization ( Figure 6—figure supplement 2A ) . Surface Robo was quantified as the mean fluorescence intensity of HA normalized to HRP . We observed no significant difference in the surface expression of Robo and RoboΔWIRS ( Figure 6—figure supplement 2B ) . We next cultured E12 mouse dorsal commissural neurons that were electroporated with either wild-type MYC-tagged human Robo1 ( hRobo1 ) or MYC-tagged hRobo1ΔWIRS . Following a 30 min bath application with Slit , we visualized surface expression of hRobo1 by staining the N-terminal MYC tag before fixation and permeabilization ( Figure 6—figure supplement 2C ) . Surface hRobo1 was quantified as the mean fluorescence intensity of MYC , and the analysis was limited to Robo3-positive commissural neurons . Here again , we observed no significant difference in the surface localization of hRobo1 and hRobo1ΔWIRS ( Figure 6—figure supplement 2D ) , suggesting that the WIRS motif has no detectable effect on Robo1 surface levels . Together , these observations point to a non-endocytic role for the WRC in promoting Robo repulsion . The WIRS motif in the Robo1 receptor is conserved in vertebrates , raising the possibility for a potential role in vertebrate Robo1 signaling . Indeed , the cytoplasmic tail of human Robo1 can bind to the WRC via its WIRS motif ( Chen et al . , 2014a ) . Thus , to address the question of whether the WIRS motif is important for vertebrate Robo1 signaling , we introduced point mutations into the WIRS motif of hRobo1 and performed gain-of-function experiments with wild-type hRobo1 and hRobo1ΔWIRS constructs . We electroporated E12 mouse spinal cords with wild-type hRobo1 or hRobo1ΔWIRS , along with RFP as a reporter for efficiency of electroporation and cultured dorsal spinal cord explants next to mock 293 T cell aggregates or cell aggregates expressing Slit ( Figure 7A ) . We observe poor penetration of the anti-MYC antibody in explants and hence use RFP as a measure of electroporation efficiency . We observe comparable levels of RFP staining in explants ( Figure 7—figure supplement 1A ) . Explants cultured adjacent to mock cell aggregates show uniform outgrowth on all sides of the explant ( Figure 7B ) . In contrast , explants cultured adjacent to Slit-expressing aggregates show decreased outgrowth on the side proximal to the Slit-expressing aggregate as compared to the distal side ( Figure 7C ) . Explants electroporated with wild-type hRobo1 show an increased repulsive response to Slit with even less outgrowth on the proximal side and a significantly lower proximal/distal outgrowth ratio ( Figure 7D , F ) . In contrast , explants electroporated with hRobo1ΔWIRS show no such gain-of-function response to Slit and have a proximal/distal outgrowth ratio similar to that seen for RFP electroporated explants ( Figure 7E , F ) , suggesting that the WIRS motif is important for the Slit-induced repulsive response of vertebrate Robo1 . Next , to assess whether the WIRS motif is also important for the collapsing activity of Robo1 in response to Slit , we performed Slit-induced collapse assays using dissociated E12 mouse dorsal spinal commissural neurons ( Figure 7G , H ) . In our control cultures , 38% of Robo3-positive commissural axons show collapsed growth cones ( Figure 7I ) . Following a 30 min treatment with recombinant Slit2 , we see an increase in the collapse rate to 62% . Neurons electroporated with wild-type MYC-tagged hRobo1 show a further increase in collapse rate with 77% of Robo3- and MYC-positive axons ending in collapsed growth cones . In contrast , we saw no increase in the number of collapsed growth cones in neurons electroporated with MYC-tagged hRobo1ΔWIRS ( Figure 7I ) , suggesting that the WIRS motif is also important for the Slit-induced collapsing activity of Robo1 . The hRobo1 variants show comparable levels of MYC staining in neurons ( Figure 7—figure supplement 1B , C ) . To study the function of the Robo1 WIRS motif in an in vivo context , we examined its role in commissural axon guidance in the embryonic chicken spinal cord . We reasoned that unilateral expression of Robo1 in pre-crossing commissural neurons would prevent their axons from crossing the floor plate by inducing a premature responsiveness to midline-secreted Slits ( Brose et al . , 1999; Long et al . , 2004 ) . To do this , we used in ovo electroporation to introduce a GFP expression plasmid either alone ( Control ) or with MYC-tagged wild-type human Robo1 or human Robo1ΔWIRS expression constructs into pre-crossing commissural neurons at Hamburger–Hamilton ( HH ) stage 14 ( Hamburger and Hamilton , 1951 ) . At HH stage 22–23 , a ‘crossing index’ was calculated by measuring GFP and MYC signal in the contralateral side of the spinal cord as a fraction of GFP and MYC signal on the electroporated side ( Figure 8D ) . We found that ectopic expression of wild-type Robo1 and GFP resulted in a GFP crossing index of 0 . 21 ± 0 . 13% ( mean ± SD , n = 6 ) , which was significantly less than that of GFP alone ( Control ) , with a crossing index of 1 . 8 ± 1 . 1% ( n = 6 , p=0 . 004 ) , indicating that Robo1 expression was sufficient to block commissural crossing ( Figure 8A , B , E ) . Robo1ΔWIRS and GFP overexpression resulted in a GFP crossing index of 0 . 68 ± 0 . 60% ( n = 8 ) , which was not significantly different from that of wild-type Robo1 ( p=0 . 472; Figure 8C , E ) . However , quantification of the crossing index based on the MYC tag fused to the wild-type Robo1 and Robo1ΔWIRS constructs resulted in a significantly higher MYC crossing index of Robo1ΔWIRS-expressing neurons ( 1 . 7 ± 0 . 97% , n = 8 ) compared to that of wild-type Robo1-expressing neurons ( 0 . 53 ± 0 . 36% , n = 6 , p=0 . 013; Figure 8F ) . The disparity between the effects of the WIRS mutation calculated using GFP and MYC-based quantification may reflect a greater efficiency of GFP plasmid transduction and expression compared to the Robo1 expression constructs . These data demonstrate a significant reduction in Robo1’s ability to prevent spinal commissural crossing in the absence of the WIRS motif . Altogether , the results from mouse dorsal spinal cord explants and dissociated neuron cultures along with the in vivo experiments in chick embryos show that while overexpression of wild-type hRobo1 is able to enhance the repulsive response to Slit , mutating the WIRS motif in hRobo1 abolishes this gain-of-function response . These observations indicate that the WIRS motif is important for vertebrate Robo1 signaling and suggest an evolutionarily conserved role for the WIRS motif in Robo1 repulsive signaling .
In this article , we have documented a conserved role for the WRC in Slit-mediated Robo1 repulsive signaling . Using the developing Drosophila embryonic CNS , we demonstrate a series of dose-dependent genetic interactions between components of the WRC and Slit-Robo1 signaling , which show that the WRC functions in vivo to regulate Robo1 repulsive signaling at the midline . Biochemical experiments in cultured cells show that Robo1 can bind to the WRC partially via its WIRS motif and that Slit stimulation can induce recruitment of the WRC to Robo1 . Further , we present several lines of evidence to demonstrate that the WIRS motif is important for Robo1 function in vivo . First , mutating the WIRS motif results in a significantly weaker gain-of-function phenotype when Robo1 is misexpressed in commissural axons . Second , the Robo1 variant with mutations in its WIRS motif fails to rescue the robo1 mutant phenotype as effectively as wild-type Robo1 . Finally , mutating the WIRS motif in the endogenous robo1 locus using the CRISPR-Cas9 system results in embryos with severe ectopic crossing defects that phenocopy robo1 mutants . These data demonstrate a severe decline in Robo1 function upon disruption of the WRC binding site . Together , our observations support the model that Slit stimulation results in recruitment of the WRC to the WIRS motif in Robo1 , which is vital to Robo1-mediated repulsive signaling at the midline ( Figure 9 ) . Further , using genetic and biochemical approaches , we show that the Arp2/3 complex functions in the Slit-Robo signaling pathway and undergoes a WIRS-dependent recruitment to Robo1 in response to Slit . We propose that downstream of Robo1 the WRC functions to recruit the Arp2/3 complex to initiate localized cytoskeletal remodeling . We also present several lines of evidence that support an evolutionarily conserved role for the WIRS motif in vertebrate Robo1 signaling . First , we show that Robo1∆WIRS is less effective at mediating repulsion in response to Slit in explants from the mouse dorsal spinal cord . In addition , Robo1∆WIRS is less responsive to the collapsing activity of Slit in dissociated spinal commissural axons . Finally , we show that mutating the WIRS motif in human Robo1 results in a reduced ability to induce ectopic repulsion in embryonic chicken commissural axons as compared to wild-type human Robo1 . These data highlight a vital conserved role for the WIRS motif in Robo1 function . In this study , we used a series of complementary approaches to evaluate the importance of the WIRS motif for Robo1 repulsive signaling . While it is generally assumed that the high expression levels resulting from the Gal4/UAS system are unlikely to reflect normal spatial and temporal regulation , it remains unclear to what extent this might confound comparisons between different mutant variants of a given protein . Our results with Robo1 indicate that Gal4-UAS/directed expression significantly hinders the detection of critical structural elements of the receptor . For example , when using a pan-neuronal driver to reintroduce Robo1 into the robo1 mutant embryos , we observe only very modest differences between the wild-type and WIRS mutant forms of overexpressed Robo1 . In contrast , we see much more severe phenotypes when the WIRS motif is disrupted under conditions that more closely match the endogenous robo1 levels using the robo1 genomic rescue constructs or when mutating the WIRS motif in the endogenous robo1 locus using the CRISPR-Cas9 system . These direct comparisons between different assays to measure protein function suggest that rescue experiments using the Gal4/UAS system must be interpreted with caution . This also suggests that our results comparing the gain-of-function effects of Robo1 and Robo1∆WIRS in vertebrate systems are likely to underestimate the significance of the WIRS motif and recruitment of the WRC for Robo1 repulsion . Here , we have shown that the WRC is an important component of the Slit-Robo1 repulsive pathway at the midline . But what happens once the WRC is recruited to Robo1 ? The VCA region of Scar/Wave is sequestered within the complex until activation , which triggers a conformational change releasing the VCA domain ( Chen et al . , 2010; Ismail et al . , 2009 ) . Rac1 is an important activator of the WRC and has been previously found to be activated downstream of Robo1 in both Drosophila and mouse ( Fan et al . , 2003; Wong et al . , 2001 ) . The genetic interaction between sos and cyfip suggests that these proteins function cooperatively to regulate midline repulsion . We propose a model where the WRC is recruited to the WIRS motif in response to Slit where it is activated by increased local Rac1 signaling . Active WRC can then promote Arp2/3-meditated actin assembly . Such a direct interaction with the WRC via the WIRS motif would allow for localized WRC activity at desired subdomains to achieve tighter spatiotemporal control and support directional actin changes . At first glance , the initiation of actin polymerization downstream of Robo1 might seem paradoxical; however , many repulsive guidance cues recruit downstream effectors that enhance actin polymerization . One recent study demonstrates that dorsal root ganglion neurons initially extend filopodia toward a source of Slit before retracting ( McConnell et al . , 2016 ) . The McConnell study highlights the nuanced and complex actin rearrangements that occur downstream of guidance cues , which potentially contribute to sensing of the environment for improved resolution of a guidance gradient . The WRC might play an important role in the generation of these Slit-induced filopodia by initiating the formation of branched actin filaments that are subsequently rebundled to form filopodia as suggested by the convergent elongation model that supports a role for Arp2/3 in filopodia formation in neurons ( Yang and Svitkina , 2011 ) . Ena/VASP proteins , downstream effectors of Robo1 , are important for these Slit-induced filopodial extensions ( McConnell et al . , 2016 ) and on account of their actin bundling activity are perfectly poised to orchestrate this actin reorganization in order to drive filopodia formation . The WRC has also been shown to be important for receptor endocytosis ( Basquin et al . , 2015; Xu et al . , 2016 ) , and another possible outcome of initiating localized actin polymerization is the endocytosis and recycling of Robo1 . Indeed , previous work has demonstrated that endocytosis of Drosophila Robo1 upon Slit stimulation is essential for Robo1 repulsive signaling ( Chance and Bashaw , 2015 ) and that vertebrate Robo1 also undergoes endocytosis and recycling following Slit stimulation ( Kinoshita-Kawada et al . , 2019 ) , suggesting a conservation of this regulatory mechanism . However , our data suggests that mutating the WIRS motif has no detectable effect on the surface localization of Robo1 either in Drosophila or in cultured mouse commissural neurons . Future studies are needed to decipher the exact nature and function of the local actin remodeling induced by the Rac-WRC-Arp2/3 complex signaling axis downstream of Robo1 . Additionally , other known downstream effectors of Robo1 like Ena and Abl have also been shown to influence WRC activity ( Chen et al . , 2014b; Leng et al . , 2005 ) . It would thus be interesting to dissect how this inter-regulation between these different components of Robo1 signaling contributes to fine-tuning of WRC activity to generate a specific output for Slit-Robo1 repulsion . Previously , the WIRS motif has been shown to be important in Neuroligins and Syg-1 for proper synapse formation ( Chia et al . , 2014; Xing et al . , 2018 ) as well as in Neogenin for the maintenance of adherens junctions ( Lee et al . , 2016 ) . In these contexts , it is apparent that the WRC reinforces the F-actin network at these membrane junctions . However , it is unclear if the WIRS-WRC interactions are subject to regulation by the respective ligands or if the WRC performs more of a scaffolding function . In the context of axon guidance , our work demonstrates a ligand-dependent recruitment of the WRC to the WIRS domain of Robo1 suggestive of both spatial and temporal specificity of WRC activation . To our knowledge , this study is the first to demonstrate that the WRC can be recruited to a guidance receptor in response to a ligand . In addition to functioning as part of the actin-regulating complex , WRC members can have functions independent of the complex as well . For example , CYFIP proteins can interact with fragile X mental retardation protein ( FMRP ) to regulate mRNA localization and protein translation ( Abekhoukh and Bardoni , 2014; Schenck et al . , 2003; Schenck et al . , 2001 ) and Abi can interact with WASP and Diaphanous to regulate F-actin ( Bogdan et al . , 2005; Ryu et al . , 2009 ) . While we cannot entirely rule out a role for WRC-independent functions of these proteins in Robo1 signaling , several lines of evidence point to the involvement of the WRC as a whole downstream of Robo1 . First , two separate WRC members , cyfip and hscp300 , show genetic interactions with the Slit-Robo pathway . Second , the physical interaction between Robo1 and HSPC300 is dependent on the WIRS motif , which requires a binding interface formed by CYFIP and Abi that comes together only in the fully assembled WRC ( Chen et al . , 2014a ) . Additionally , the strong midline crossing phenotypes we see upon manipulating the WIRS motif suggests that it is indeed this interaction with the fully assembled WRC that is important for Robo1 signaling in vivo . Finally , we tested whether the Drosophila homolog of FMRP , dfmr1 , genetically interacts with the Slit-Robo1 pathway . In contrast to cyfip and hspc300 , completely removing dfmr1 has no effect on the slit , robo transheterozygous phenotype ( Figure 6—figure supplement 1J ) . This data further supports a WRC-dependent function for cyfip in Slit-Robo signaling and suggests that CYFIP/dFMR interactions are not important for Robo repulsion . Drosophila Robo1 has numerous functions in development outside of its role in midline repulsion , and the robo1ΔWIRS CRISPR mutants generated here also provide an opportunity to discern which developmental functions of Robo1 require the WRC in future studies . Robo1 regulates the migration of chordotonal sensory neurons ( Kraut and Zinn , 2004 ) and mesodermal migration for muscle patterning ( Kramer et al . , 2001 ) . Embryos lacking robo1 show defects in heart lumen formation ( Qian et al . , 2005 ) and tracheal migration ( Englund et al . , 2002 ) . In mammals , Robo1 also plays important roles outside of the nervous system , including the formation of blood vessels ( Rama et al . , 2015 ) and organs like the heart ( Mommersteeg et al . , 2013 ) and the mammary glands ( Macias et al . , 2011 ) , and it can also regulate stem cell proliferation ( Ballard et al . , 2015 ) . Finally , the Slit-Robo pathway has been shown to regulate tumor angiogenesis along with tumor cell migration and metastasis ( Tong et al . , 2019 ) . Mis-regulation of Slit-Robo signaling has been implicated in multiple types of tumorigenesis , making it a promising target for cancer treatments ( Koohini et al . , 2019 ) . Such therapeutic avenues require a comprehensive understanding of Slit-Robo signaling in specific cancers , highlighting the importance of investigating the WRC as a downstream effector of Robo in disease contexts as well . In addition to Robo1 , other Robo receptors also contain WIRS motifs in their cytoplasmic domains . Drosophila Robo2 plays a minor role in midline repulsion and , together with Robo3 , also regulates lateral positioning of the longitudinal fascicles ( Evans and Bashaw , 2010; Rajagopalan et al . , 2000 ) . As Robo2 and Robo3 do not contain CC2 and CC3 domains , to which most of the known Robo1 effectors bind , very little is known about their downstream signaling . Vertebrate Robo3 can induce repulsive signaling in response to a recently identified ligand , NELL2 ( Jaworski et al . , 2015 ) Vertebrate Robo3 also contains a WIRS motif in its cytoplasmic domain , raising the possibility that these Robo receptors could share a common cellular mechanism for repulsion , despite responding to distinct ligands . Additionally , attractive axon guidance receptors like Fra and its vertebrate ortholog DCC also contain WIRS motifs in their cytoplasmic domains . Unsurprisingly , many core actin-modifying proteins that act downstream of repulsive cues like Ena/VASP and Abl kinase also function in attractive signaling ( Forsthoefel et al . , 2005; Gitai et al . , 2003 ) . We can speculate that the WRC might also function in both repulsion and attraction by regulating different actin-based processes like membrane trafficking versus growth cone advancement . Alternatively , as other studies have shown the importance of an initial growth cone extension toward repulsive cues , it is likely that tight spatiotemporal activation along with regulation by other effector molecules can result in fine-tuning of WRC activity to contribute to distinct cytoskeletal outputs downstream of different guidance receptors .
Further information and requests for resources and reagents should be directed to the lead contact , Greg J . Bashaw ( gbashaw@pennmedicine . upenn . edu ) . The following Drosophila strains were used: w1118 , roboGA285 , slit2 , sos4G , robo2x123 , scarΔ37 , arpc2KG04658 , apGal4 , egGal4 , UAS-CD8GFP II , UAS-TauMycGFP III , 10XUAS-HA-Robo1 86F8 , and 5XUAS-HA-Robo1 86F8 . Fly strains hspc300Δ54 . 3 , cyfipΔ85 . 1 , and UAS-CYFIP were a kind gift from A . Giangrande . The fmr13 strain was a kind gift from T . Jongens . The genomic robo1 rescue strain robo1::HArobo1 28E7 was a kind gift from T . Evans . The following transgenic stocks were generated: 10UAS-HA-RoboΔWIRS 86F8 , 5UAS-HA-RoboΔWIRS 86F8 , robo1::HArobo1ΔWIRS 28E7 , 10UAS-HSPC300-GFP 86F8 . Transgenic flies were generated by BestGene Inc ( Chino Hills , CA ) using ΦC31-directed site-specific integration into landing sites at cytological position 86F8 ( For UAS-Robo constructs ) or 28E7 ( for genomic robo1 rescue constructs ) . Genomic robo1::HArobo1ΔWIRS 28E7 rescue transgene was introduced onto a roboGA285 chromosome via meiotic recombination , and the presence of theroboGA285 mutation was confirmed in all recombinant lines by DNA sequencing . The CRISPR line robo1ΔWIRS was generated by cloning a guide targeting the WIRS motif into a pCFD3-dU6:3 backbone ( Addgene , #49410 ) and sending positive clones to BestGene Inc for injection . Flies were screened by PCR and restriction digest followed by DNA sequencing . All crosses were carried out at 25°C . Timed pregnant female CD-1 mice were obtained from Charles River . All animal work was approved by the Institutional Animal Care and Use Committee ( IACUC ) of the University of Pennsylvania . Embryos of both sexes were randomly used for spinal cord explants and primary dissociated neuron cultures . All animal experiments were carried out in accordance with the Canadian Council on Animal Care guidelines and approved by the IRCM Animal Care Committee and the McGill University Animal Care Committee . Fertilized chicken eggs ( FERME GMS , Saint-Liboire , QC , Canada ) were incubated ( Lyon Technologies , model PRFWD ) at 39°C according to standard protocols . Primary commissural neuron cultures were prepared as described previously ( Langlois 2010 ) and maintained at 5% CO2 in a humidified incubator . Briefly , commissural neurons were isolated from E12 . 5 dorsal spinal cords and plated on acid-washed , poly-D-lysine ( Sigma , #P6407 ) and 2 μg/ml N-cadherin ( R&D , #1388-NC ) coated coverslips . Neurons were cultured in Neurobasal medium supplemented with 10% heat-inactivated FBS ( Gibco , #10437-028 ) and 1X penicillin/streptomycin/glutamine ( Gibco , #10378-016 ) . After ~20 hr , the medium was replaced with Neurobasal supplemented with 1X B-27 ( Thermo , #A3582801 ) and the neurons were used for experiments 1 hr later . Dorsal spinal cord explants from E12 . 5 embryos were dissected and cultured in collagen gels as described previously ( Serafini 1994 ) . Briefly , explants were cultured in 50% OptiMEM ( Gibco , #31985-070 ) and 45% Ham’s F-12 ( Gibco , #11765-054 ) media supplemented with 5% horse serum ( HS , Gibco , #16050122 ) , 0 . 75% glucose ( Thermo , #D16-500 ) , and 1X penicillin/streptomycin/glutamine for 48 hr with 500 ng/ml Netrin-1 ( R&D , #1109-N1/CF ) . Drosophila S2R+ cells ( DGRC , Cat#150 ) were maintained at 25°C in Schneider’s media ( Life Technologies , #21720024 ) supplemented with 10% ( vol/vol ) FBS and a mixture of 1% penicillin and streptomycin . Morphology and doubling time were used for validation of the cell line . The cells grow as a loose semi-adherent monolayer with a doubling time of about 40 hr . 293 T cells ( ATCC CRL-3216 ) were maintained at 37°C and 5% CO2 in a humidified incubator in DMEM ( Gibco , #11965084 ) supplemented with 10% ( vol/vol ) FBS and a mixture of 1% penicillin and streptomycin . Cells were authenticated by STR profiling using ATCC Cell Line Authentication services . Mycoplasma testing was negative for both cell lines . For synthesizing the guide RNA to target the WIRS motif in the endogenous robo1 locus , the following sense and antisense oligonucleotides were used: GTCGGCGTACGGCGTGGGATTAT and AAACATAATCCCACGCCGTACGC . This guide RNA was selected with zero predicted off-target effects using http://targetfinder . flycrispr . neuro . brown . edu . The oligos were annealed and cloned into a BbsI-digested pCFD3-dU6:3 vector . A single-stranded oligonucleotide template was designed to introduce point mutations into the WIRS motif . These mutations also destroy the gRNA target sequence and the PAM sequence to prevent subsequent cleavages by Cas9 . An MfeI site is mutated , which was used for screening potential CRISPR flies . The sequence of the template used is: CAATCCAACTACAATAACTCCGATGGAGGAACCGATTATGCAGAAGTTGACACCCGTAATGCTACCGCCGCCTACGCTTGTCGCAAGGTGAGGATCATATGAATTGCATCACACAACAATTTC . The template along with the pCFD3 vector containing the guide RNA was sent to BestGene Inc for injection . The progeny from these flies were crossed to balancer stocks to generate stable lines . Flies from these lines were then screened with MfeI following genomic DNA extraction , and positive hits were sent for DNA sequencing . S2R+ cells were transiently transfected with Effectene transfection reagent ( Qiagen , Valencia , CA , #301425 ) and induced 24 hr later with 0 . 5 mM copper sulfate . 24 hr after induction , cells were lysed in TBS-V ( 150 mM NaCl , 10 mM Tris pH-8 , 1 mM ortho-vanadate ) supplemented with 0 . 5% Surfact-AMPS NP40 ( Thermo , Waltham , MA , #85124 ) and 1x Complete Protease Inhibitor ( Roche , #11697498001 ) for 20 min at 4°C . Soluble proteins were recovered by centrifugation at 15 , 000 × g for 10 min at 4°C . Lysates were pre-cleared with 30 μl of a 50% slurry of protein A ( Invitrogen , #15918-014 ) and protein G agarose beads ( Invitrogen , #15920-010 ) by incubation for 20 min at 4°C . Pre-cleared lysates were then incubated with 0 . 7 μg of rabbit anti-GFP antibody for 2 hr at 4°C to precipitate HSPC300-GFP . After incubation , 30 μl of a 50% slurry of protein A and protein G agarose beads was added and samples were incubated for an additional 30 min at 4°C . The immunocomplexes were washed three times with lysis buffer , boiled for 10 min in 2x Laemmli SDS sample buffer ( Bio-Rad , #1610737 ) , and analyzed by western blotting . Proteins were resolved by SDS-PAGE and transferred to nitrocellulose membrane ( Amersham , #10600032 ) . Membranes were blocked with 5% dry milk and 0 . 1% Tween 20 in PBS for 1 hr at room temperature and incubated with primary antibodies overnight at 4°C . Following three washes with PBS/0 . 1% Tween 20 , membranes were incubated with the appropriate HRP-conjugated secondary antibody at room temperature for 1 hr . Signals were detected using Clarity ECL ( Bio-Rad , #1705061 ) according to manufacturer’s instructions . For preparation of Slit-CM , cells were transfected with a pUAST-Slit vector and a PMT-Gal4 vector using Effectene transfection reagent . Gal4 was induced 24 hr later with 0 . 5 mM copper sulfate . 24 hr after induction , Slit-CM was collected and concentrated using Amicon filters ( Amicon Ultracel 30K , Millipore , #UFC903096 ) . For CM treatment , cells were incubated with control-CM ( prepared using an empty pUAST vector ) or Slit-CM on an orbital shaker at room temperature for 12 min , then lysed for immunoprecipitation as described above . Antibodies used: for immunoprecipitation , rabbit anti-GFP and for western blot , rabbit anti-GFP ( 1:500 , Invitrogen , #a11122 ) , mouse anti-MYC ( 1:1000 , DSHB , #9E10-C ) , mouse anti-Slit ( 1:50 , DSHB , #C555 . 6D ) , HRP goat anti-rabbit ( 1:10 , 000 , Jackson Immunoresearch , #111-035-003 ) , and HRP goat anti-mouse ( 1:10 , 000 , Jackson Immunoresearch , #115-035-146 ) . For co-immunoprecipitation assays in Drosophila embryos , embryonic protein lysates were prepared from approximately 100 μl of embryos overexpressing UAS-HSPC300-GFP alone or with the HA-tagged UAS-Robo1 variants in all neurons . Embryos were lysed in 0 . 5 ml TBS-V ( 150 mM NaCl , 10 mM Tris pH 8 . 0 , 1 mM ortho-vanadate ) supplemented with 1% Surfact-AMPS NP40 and protease inhibitors by manual homogenization using a plastic pestle . Homogenized samples were incubated with gentle rocking at 4°C for 10 min and centrifuged at 15 , 000 × g for 10 min in a pre-chilled rotor . Supernatants were collected after centrifugation , and immunoprecipitations and western blotting were performed as described above . Antibodies used: for immunoprecipitation , rabbit anti-GFP ( 1:500 , Invitrogen , #a11122 ) and for western blot , rabbit anti-GFP ( 1:500 , Invitrogen , #a11122 ) , mouse anti-HA ( 1:1000 , BioLegend , #901502 ) , mouse anti-beta tubulin ( 1:1000 , DSHB , #E7 ) , HRP goat anti-rabbit ( 1:10 , 000 , Jackson Immunoresearch , #111-035-003 ) , and HRP goat anti-mouse ( 1:10 , 000 , Jackson Immunoresearch , #115-035-146 ) . Dechorionated , formaldehyde-fixed Drosophila embryos were fluorescently stained using standard methods . The following antibodies were used: rabbit anti-GFP ( 1:250 , Invitrogen , #a11122 ) , mouse anti-HA ( 1:500 , BioLegend , #901502 ) , chick anti-beta gal ( 1:500 , Abcam , #ab9361 ) , mouse anti-Scar ( 1:50 , DSHB , #P1C1 ) , mouse anti-Robo ( 1:50 , DSHB , #13C9 ) , Alexa647 goat anti-HRP ( 1:500 , Jackson Immunoresearch , #123-605-021 ) , Alexa488 goat anti-rabbit ( 1:500 , Invitrogen , #A11034 ) , Alexa488 goat anti-mouse ( 1:500 , Invitrogen , #A11029 ) , Alexa488 goat anti-chick ( 1:500 , Invitrogen , #A11039 ) , Cy3 goat anti-mouse ( 1:500 , Jackson Immunoresearch , #115-165-003 ) , Cy3 goat anti-Chick ( 1:500 , Abcam , #ab97145 ) , and 647 goat anti-HRP ( 1:1000 , Jackson Immunoresearch , #123-605-021 ) . Embryos were filleted and mounted in 70% glycerol/1× PBS . Surface staining of the HA-tagged genomic Robo rescue transgenes in Drosophila embryos was carried out as previously described ( Bashaw , 2010 ) . Briefly , embryos were dissected live , blocked with in 5% normal goat serum ( NGS ) in PBS for 15 min at 4°C , and stained with mouse anti-HA ( 1:500 , BioLegend , #901502 ) in PBS for 30 min at 4°C . Following washes with PBS , embryos were fixed in 4% paraformaldehyde ( Electron Microscopy Services , #15710 ) for 15 min at 4°C . Following washes with PBS , fixed embryos were then permeabilized with 0 . 1% Triton X-100 in PBS ( PBT ) for 10 min and stained with 647 goat anti-HRP ( 1:1000 , Jackson Immunoresearch , #123-605-021 ) in 5% NGS in PBT overnight at 4°C . Following washes with PBT , secondary antibody consisting of Alexa488 goat anti-mouse ( 1:500 , Invitrogen , #A11029 ) diluted in 5% NGS in PBT was added and incubated for 1 hr at room temperature . Embryos were then washed with PBT and mounted in Aquamount . Dissociated spinal commissural neurons were fixed for 20 min in 4% paraformaldehyde ( Electron Microscopy Services , #15710 ) at room temperature and washed three times with PBS . Fixed neurons were then permeabilized with 0 . 1% Triton X-100 in PBS ( PBT ) for 10 min and blocked with 2% HS in PBT for 30 min at room temperature . The blocking solution was replaced with primary antibody diluted in 2% HS in PBT and incubated overnight at 4°C . Following three washes with PBT , secondary antibody diluted in 2% HS in PBT was added and incubated for 1 hr at room temperature . Neurons were then washed three times with PBT and the coverslips were mounted in Aquamount . The following antibodies were used: mouse anti-MYC ( 1:500 , DSHB , #9E10-C ) , goat anti-Robo3 ( 1:200 , R&D Systems , #AF3076 ) , Cy3 donkey anti-goat ( 1:500 , Jackson Immunoresearch , #705-165-147 ) , and 488 donkey anti-goat ( 1:500 , Jackson Immunoresearch , #715-545-150 ) . For surface staining of MYC-tagged hRobo1 variants in dissociated commissural neurons , neurons were treated with recombinant hSlit2-N ( R&D , #5444-SL-050 ) at 2 μg/ml for 30 min at 37°C , following which neurons were blocked in 5% NGS in PBS for 15 min at 4°C and stained with mouse anti-MYC ( 1:500 , DSHB , #9E10-C ) in PBS for 30 min at 4°C . Following two washes with PBS , neurons were fixed in 4% paraformaldehyde ( Electron Microscopy Services , #15710 ) for 20 min at 4°C . Following washes with PBS , fixed neurons were then permeabilized with 0 . 1% Triton X-100 in PBS ( PBT ) for 10 min and stained with goat anti-Robo3 ( 1:200 , R&D Systems , #AF3076 ) in 5% NGS in PBT overnight at 4°C . Following washes with PBT , secondary antibody consisting of 488 donkey anti-goat ( 1:500 , Jackson Immunoresearch , #715-545-150 ) and Cy3 donkey anti-goat ( 1:500 , Jackson Immunoresearch , #705-165-147 ) diluted in 5% NGS in PBT was added and incubated for 1 hr at room temperature . Neurons were then washed with PBT and mounted in Aquamount . Collagen-embedded explants were fixed in 4% paraformaldehyde overnight at 4°C and washed three times for 10 min in PBS . Fixed explants were then blocked in 2 . 5% NGS in PBT for 2 hr at room temperature and incubated with primary antibody diluted in blocking solution overnight at 4°C . Explants were washed six times for 1 hr with PBT and incubated with secondary antibody diluted in blocking solution overnight at 4°C . After six 1 hr washes with PBT , explants were mounted on cavity slides . The following antibodies were used: mouse anti-MYC ( 1:500 , DSHB , #9E10-C ) , mouse anti-beta tubulin ( 1:300 , DSHB , #E7 ) , rabbit anti-dsRed ( 1:200 , Takara , #632496 ) , Alexa488 goat anti-mouse ( 1:500 , Invitrogen , #A11029 ) , and Cy3 goat anti-rabbit ( 1:500 , Jackson Immunoresearch , #111-165-003 ) . Fixed samples of Drosophila embryo nerve cords , mouse-dissociated commissural neurons , and mouse dorsal spinal cord explants were imaged using a spinning disk confocal system ( Perkin Elmer ) built on a Nikon Ti-U inverted microscope using a Nikon ×40 objective ( for nerve cords and neurons ) and a ×10 objective ( for explants ) with Volocity imaging software . Images were processed using NIH ImageJ software . E12 . 5 embryos were electroporated ex utero by injecting 100 ng/μl DNA in electroporation buffer ( 30 mM HEPES pH 7 . 5 [Thermo , #BP299-1] , 300 mM KCl [Thermo , #BP366-1] , 1 mM MgCl2[Thermo , #BP214-500] , and 0 . 1% Fast Green FCF [Thermo , #F99-10] ) into the central canal of the neural tube . A BTX ECM 830 electroporator ( BTX Harvard Apparatus , #45-0662 ) was used for bilateral electroporation into spinal cord neurons ( five 30 V pulses , each of 50 ms duration for each half of the spinal cord ) . Following electroporation , dorsal spinal cords were dissected out and cut into explants for the explant outgrowth assay or used for preparation of dissociated neuronal cultures . For neuron culture , dissected spinal cords were washed in Hanks’ Balanced Salt Solution ( HBSS , Gibco , #14175-079 ) and digested with 0 . 05% trypsin ( Gibco , #25300054 ) for 7 min at 37°C . 1 μl of DNase I ( NEB , #M0303L ) and 0 . 15% MgSO4 ( Thermo , #7487-88-9 ) was added for an additional minute , and the samples were centrifuged at 400 × g for 4 min . Samples were washed with pre-warmed HBSS , and a small fire-polished Pasteur pipette was used to triturate the tissue and dissociate it into single cells . Cells were plated on acid-washed , poly-D-lysine and N-cadherin-coated coverslips and cultured in plating media ( Neurobasal [Gibco , #21103-049] medium supplemented with 10% heat-inactivated FBS and 1X penicillin/streptomycin/glutamine ) . Dorsal spinal cord explants form E12 . 5 mouse embryos were dissected and cultured in collagen gels as previously described ( Serafini 1994 ) . Briefly , explants were embedded in rat tail collagen ( Corning , #354249 ) gels at a distance of one explant diameter away from a mock 293 T cell aggregate ( ATCC , CRL-3216 ) or a cell aggregate expressing Slit ( pSecTagB-hSlit2-MYC , kind gift from A . Chedotal ) . Explants were grown in 50% OptiMEM and 45% Ham’s F-12 media supplemented with 5% HS , 0 . 75% glucose , and 1X penicillin/streptomycin/glutamine for 48 hr with 500 ng/ml Netrin-1 . Explants were subsequently fixed and stained as described above . For preparation of 293 T cell aggregates , cells were trypsinized and resuspended in a rat tail collagen solution , drawn into a glass Pasteur pipette , and allowed to polymerize . The collagen-embedded cells were released from the pipette using a rubber bulb and the aggregates cut into 1 mm clusters . Dissociated commissural neurons from E12 . 5 mouse embryos were cultured in plating media ( Neurobasal medium supplemented with 10% heat-inactivated FBS and 1X penicillin/streptomycin/glutamine ) for 1 day in vitro . Plating media was replaced with Neurobasal supplemented with 1X B-27 for 1 hr . Neurons were treated with recombinant hSlit2-N ( R&D , #5444-SL-050 ) at 2 μg/ml for 30 min at 37°C . Neurons were fixed immediately and immunostained for Robo3 ( a marker for commissural neurons ) and MYC to identify commissural neurons that had been successfully electroporated with the hRobo1-MYC or hRobo1ΔWIRS-MYC expression constructs . Chicken in ovo electroporations were carried out as previously described ( Croteau et al . , 2019 ) at HH stage 14 and embryos were harvested at HH stage 22–23 . Chicken embryos were electroporated with either pCAGGS , pCAGGS-hRobo1 wild-type or pCAGGS-hRobo1 WIRS-deletion constructs combined with pN2-eGFP at a 4:1 DNA weight ratio . Spinal cords were sectioned and stained with DAPI , anti-GFP ( A11122 , 1:5 , 000 , Thermo Fisher ) and anti-MYC ( 9E10 , 1:200 , DSHB ) antibodies . For analysis of Drosophila nerve cord phenotypes , image analysis was conducted blind to the genotype . Data are presented as mean values ± SEM . For statistical analysis , comparisons were made between two groups using the Student’s t-test . For multiple comparisons , significance was assessed using one-way ANOVA with Tukey’s post-hoc tests . Differences were considered significant when p<0 . 05 . For quantitation of Scar intensity or surface HA intensity in Drosophila embryos , mean gray value for Scar or HA was obtained using ImageJ and normalized to the mean gray value for HRP . Data are presented as mean values ± SEM . For statistical analysis , comparisons were made between two groups using the Student’s t-test . Differences were considered significant when p<0 . 05 . For the collapse assay , only Robo3-positive ( and MYC-positive for neurons electroporated with hRobo1 variants ) axons were imaged and analyzed . Growth cones were defined by the presence of lamellipodia and/or filopodia . Three trials were conducted and at least 30 neurons per condition were scored in each trial . Data are presented as mean values ± SEM . For statistical analysis , comparisons were made between groups using one-way ANOVA with Tukey’s post-hoc tests . Differences were considered significant when p<0 . 05 . To measure MYC signal intensity for total hRobo1 or surface hRobo1 quantitation in dissociated neurons , Robo3-positive neurons were carefully traced in ImageJ and integrated signal density in the traced region was obtained . Background signals were subtracted and mean fluorescence intensity calculated as integrated signal density per area is presented in graphs . Data are presented as mean values ± SEM . For statistical analysis , comparisons were made between two groups using the Student’s t-test . Differences were considered significant when p<0 . 05 . For the explant outgrowth assay , explants images were converted to black-and-white composites using the Threshold function . Each experimental set was quantified using the same threshold parameters . Explant quadrants were delineated by placing a right-angled crosshair at the center of each explant with the proximal quadrant directly facing the cell aggregate . The total area of black pixels was measured for the proximal and distal quadrants using the Analyze Particles function . The particles showing axonal outgrowth were then erased using the Eraser tool , and the total area of black particles was measured again . The difference was recorded as total area of axonal outgrowth . Next , the length of each quadrant was measured by tracing the border of the quadrant using the Freehand Line tool . Values for total area of outgrowth were normalized to length of the quadrant , and these final values were used to obtain the proximal/distal ratios for each explant . The measurements for each explant in a set were averaged , and the ratios of experimental conditions compared with control condition were calculated . Data are presented as mean ± SEM . Total number of explants for RFP control , RFP Slit , hRobo1 control , hRobo1 Slit , hRobo1ΔWIRS control , and hRobo1ΔWIRS Slit is 29 , 39 , 33 , 39 , 29 , and 41 , respectively ( from three independent experiments ) . For statistical analysis , comparisons were made between groups using one-way ANOVA with Tukey’s post-hoc tests . Differences were considered significant when p<0 . 05 . For western blots , densitometric analysis was performed and band intensities of co-immunoprecipitating proteins in the immunoprecipitates were normalized to band intensities of HSPC300 in the immunoprecipitates as well as to lysate levels of the co-immunoprecipitating proteins . For each independent experiment , values were compared with wild-type Robo1 normalized values . Data are presented as mean ± SEM . For statistical analysis , comparisons were made between two groups using the Student’s t-test . For multiple comparisons , significance was assessed using one-way ANOVA with Tukey’s post-hoc tests . Differences were considered significant when p<0 . 05 . For analysis of crossing index in chicken embryos , fluorescence intensities were generated by pixels above threshold using ImageJ . Five sections of each embryo were analyzed . For GFP crossing index , statistical significance was calculated using one-way ANOVA with post-hoc multiple comparisons . And for MYC crossing index , unpaired t-test was used . For all graphs , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . | The brain is the most complex organ in the body . It contains billions of nerve cells , also known as neurons , with trillions of precise and specific connections , but how do these neurons know where to go and which connections to make as the brain grows ? Neurons contain a small set of proteins known as guidance receptors . These receptors respond to external signals that can be attractive or repulsive . They instruct neurons to turn towards , or away from , the source of a signal . During embryonic development , neurons use these signals as guideposts to find their way to their destination . One such guidance receptor-signal pair consists of a receptor called Roundabout , also known as Robo , and its cue , Slit . Robo , which is located on the neuron’s surface , responds to the presence of Slit in the environment , by initiating a set of signalling events that instruct neurons to turn away . Neurons make the turn by rearranging their internal scaffolding , a network of proteins called the actin cytoskeleton . How Robo triggers this rearrangement is unclear . One possibility relies on a group of proteins called the WAVE regulatory complex , or the WRC for short . Researchers have already linked the WRC to nerve cell guidance , showing that it can trigger the growth of new filaments in the actin cytoskeleton . Proteins can activate the WRC by binding to it using a set of amino acids called a WRC-interacting receptor sequence , or WIRS for short , which Robo has . Chaudhari et al . used fruit flies to find out how Robo and the WRC interact . The experiments revealed that when Slit binds to Robo on the outside of a nerve cell , the WRC binds to Robo via its WIRS sequence on the inside of the cell . This attracts proteins inside the cell involved in rearranging the actin cytoskeleton . Disrupting this interaction by mutating either WRC or WIRS leads to severe errors in pathfinding , because when the WRC cannot connect to Robo , neurons cannot find their way . Experiments in mouse and chicken embryos showed that vertebrates use the WIRS sequence too , indicating that evolution has conserved this method of passing signals from Robo to the cytoskeleton . The fact that Slit and Robo work in the same way across fruit flies and vertebrates has implications for future medical research . Further work could explain how the brain and nervous system develop , and what happens when development goes wrong , but Slit and Robo control more than just nerve cell pathfinding . Research has linked disruptions in both proteins to many types of cancer , so a better understanding of how Robo interacts with the WRC could lead to new developments in different fields . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2021 | Robo recruitment of the Wave regulatory complex plays an essential and conserved role in midline repulsion |
MFN2 encodes mitofusin 2 , a membrane-bound mediator of mitochondrial membrane fusion and inter-organelle communication . MFN2 mutations cause axonal neuropathy , with associated lipodystrophy only occasionally noted , however homozygosity for the p . Arg707Trp mutation was recently associated with upper body adipose overgrowth . We describe similar massive adipose overgrowth with suppressed leptin expression in four further patients with biallelic MFN2 mutations and at least one p . Arg707Trp allele . Overgrown tissue was composed of normal-sized , UCP1-negative unilocular adipocytes , with mitochondrial network fragmentation , disorganised cristae , and increased autophagosomes . There was strong transcriptional evidence of mitochondrial stress signalling , increased protein synthesis , and suppression of signatures of cell death in affected tissue , whereas mitochondrial morphology and gene expression were normal in skin fibroblasts . These findings suggest that specific MFN2 mutations cause tissue-selective mitochondrial dysfunction with increased adipocyte proliferation and survival , confirm a novel form of excess adiposity with paradoxical suppression of leptin expression , and suggest potential targeted therapies .
Adipose tissue is critical for metabolic homeostasis , and either global excess ( obesity ) or pathologic deficiency ( lipodystrophy ) lead to metabolic disease . Anatomically distinct adipose depots vary in embryological origins , gene expression profiles and patterns of hormonal regulation ( Tchkonia et al . , 2013 ) , but many questions about the determinants of adipose depot differences , and their importance for metabolic regulation , remain unanswered . In studies of human metabolic disease , attention focussed for many years on visceral white adipose tissue , and on femorogluteal subcutaneous white adipose tissue , however interest has recently increased sharply in upper body subcutaneous adipose tissue , especially in interscapular and supraclavicular regions , where both obligate and facultative thermogenic adipocytes have been shown to exist under some circumstances ( Betz and Enerbäck , 2015 ) . Since the mid 19th century a devastating but rare disease characterised by a selective , massive and sometimes life-threatening increase in upper body adipose tissue has been known ( Brodie , 1846; Herbst , 2012; Madelung , 1888 ) . This disease is called either multiple symmetrical lipomatosis ( MSL ) or Madelung’s disease . Although the degree of adipose overgrowth seen may be spectacular , it varies in distribution ( Herbst , 2012 ) and the underlying pathogenesis , which is currently poorly understood , is likely to be heterogeneous . Previous studies have suggested hyperplasia of adipocytes which manifest impaired catecholamine-induced lipolysis ( Enzi et al . , 1977 ) , and later structural studies of pathologically increased adipose tissue have suggested that it is composed of adipocytes with some morphological and metabolic features of dysfunctional brown adipocytes ( Enzi et al . , 2015 ) . However MSL is not precisely defined , and phenotypic heterogeneity is apparent . People with MSL frequently exhibit variable additional features including peripheral neuropathy , autonomic neuropathy , non-alcoholic fatty liver disease and/or alcoholic liver disease , dyslipidaemia , and type 2 diabetes ( Herbst , 2012 ) . In some patients upper body adipose overgrowth contrasts with concomitant lipodystrophy in the femorogluteal region ( Herbst , 2012 ) . Physiological studies have described elevated resting energy expenditure , altered lipolytic rates and perturbed adipokine levels , however these findings have tended to vary , consistent with pathogenic heterogeneity ( Herbst , 2012 ) . Sporadic MSL has been strongly linked to chronic alcohol excess ( Enzi et al . , 2015 ) , and in rare cases is seen in the mitochondrial myoclonus epilepsy and ragged red fibers ( MERRF ) syndrome , accounted for in most cases by the mitochondrial tRNA ( Lys ) A8344G mutation ( Chong et al . , 2003 ) . These observations led to the suggestion that mitochondrial dysfunction may be an important pathogenic factor in MSL . In 2015 , a homozygous missense mutation , p . Arg707Trp , was identified in MFN2 , or mitofusin 2 , in three patients with MSL and neuropathy from two families ( Sawyer et al . , 2015 ) . Mitofusin 2 is a nucleus-encoded mitochondrial outer membrane protein involved in mitochondrial fusion , mitochondrial interactions with the endoplasmic reticulum ( Naon et al . , 2016 ) , and other aspects of cellular metabolism ( de Brito and Scorrano , 2008; Zorzano et al . , 2015 ) . Arg707 lies in the carboxy-terminal coiled coil domain HR2 of mitofusin 2 that is believed to be critical for homotypic and heterotypic interactions of mitofusin 2 ( Franco et al . , 2016; Koshiba et al . , 2004 ) with counterparts on other organelles , leading to the suggestion that the p . Arg707Trp mutation may compromise outer mitochondrial membrane fusion ( Sawyer et al . , 2015 ) . MFN2 p . Arg707Trp is the only reported cause of Mendelian MSL to date , however the nature of the associated adipose pathology , and the mechanism linking this mutation to MSL is unknown . We now describe four further patients with MSL associated with biallelic MFN2 mutations , in all cases including at least one p . Arg707Trp allele , investigate key aspects of their intermediary metabolism , and examine the microscopic and transcriptomic derangements in affected adipose tissue . We thereby provide a monogenic example of primary , apparently tissue-selective mitochondrial dysfunction leading to insulin resistance , and describe a form of leptin deficiency in the face of severe upper body obesity . Our findings yield novel insights into the role of mitofusin 2 in adipocyte energy metabolism and proliferation in humans in vivo , and suggest targeted treatments for MSL that warrant further assessment .
Four patients of European ancestry from three families were studied . The first patient ( P1 ) presented to medical attention at 5 years old with increasing upper body adiposity . Despite repeated attempts to reduce food intake this progressed throughout childhood in tandem with loss of lower body adipose tissue . She also developed axonal peripheral neuropathy and secondary foot contractures , requiring corrective surgery at the age of 17 years . By the age of 18 years , her body mass index ( BMI ) was 33 kg/m2 due to massive upper body adiposity , which contrasted with near absence of femorogluteal , leg and distal arm adipose tissue ( Figure 1A ) . She also had primary amenorrhea with clinical , biochemical and imaging-based evidence of partial hypogonadotropic hypogonadism including retarded bone age , delayed development of secondary sexual characteristics , and a small uterus , all in the context of relatively low luteinizing hormone ( LH ) , follicle stimulating hormone ( FSH ) and estradiol levels ( Table 1 ) . 10 . 7554/eLife . 23813 . 003Figure 1 . Multiple symmetrical lipomatosis associated with biallelic mutations in MFN2 . ( A ) Profile of P1 , showing upper body adipose overgrowth and loss of adipose tissue and muscle from legs ( left ) and T1-weighted coronal MRI image of P1 , highlighting excess upper body fat ( right ) . ( B ) Profile of P4 , showing upper body adipose overgrowth ( left ) and T1-weighted MRI images highlighting excess lower abdominal adipose tissue and a paucity of adipose tissue at the mid femoral region . ( C ) Linear depiction of human MFN2 , highlighting regions contributing to key structural domains . ( D ) Structure of a GTP-bound , extended tetramer ( dimer of dimers ) of Bacterial Dynamin-Like Protein ( BDLP ) as a model of human MFN2 . Monomers are distinguished in yellow and blue and HR2 domain in green . The MFN2 mutations are shown in red . ( E ) Homology model of antiparallel homodimer of HR2 domains rendered in surface presentation with the individual monomers coloured in yellow and green and the charged atoms of residues R707 and D742 in blue and red . ( F ) Plasma non-esterified fatty acid ( NEFA ) concentrations in P1 ( at age 18 ( 1 ) and 20 ( 2 ) years ) , P2 and P3 before and after a 75g oral glucose load . Data from 770 obese adolescents and 22 obese adolescents with type 2 diabetes ( Hershkop et al . , 2016 ) are shown as controls . ( G ) Quantitative real-time PCR of interscapular adipose biopsies from P2 and P3 , expressed as fold change from five female controls . ( H ) Immunoblots from patient and control adipose lysates for adiponectin ( ADIPOQ ) , perilipin 1 ( PLIN1 ) , perilipin 2 ( PLIN2 ) , and mitofusin 2 ( MFN2 ) . TATA-binding protein ( TBP ) and β-actin ( ACTB ) were used as loading controls . ( I ) Assessment of adipocyte size in interscapular adipose tissue . Representative haematoxylin and eosin-stained interscapular adipose tissues from a patient and a control are shown on top . Scale bars = 100 µm . The bottom left chart shows adipocyte size distribution analysis with bars indicating percentages of cells in the specified area range . The chart on the right shows cumulative frequencies . ≥500 adipocytes analyzed per biopsy . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 00310 . 7554/eLife . 23813 . 004Figure 1—figure supplement 1 . Pedigree diagram for P2 , P3 . Only P2 , P3 and their parents were clinically assessed and genotyped as part of this study . MFN2 genotypes are indicated in red . Phenotypes of non-genotyped relatives were determined from family reports , with supporting photographs . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 00410 . 7554/eLife . 23813 . 005Figure 1—figure supplement 2 . GTP/GDP-dependent conformations of MFN2 modelled on homology to Bacterial Dynamin-Like Protein . R343 and R707 are indicated in red . ( A ) Global view of MFN2 in the GDP-bound , lipid-free , folded conformation . Colours: blue - GTPase domain; yellow - ‘neck’; green - ‘stalk’; magenta - lipid binding domain; light purple - membrane anchor , black lines - boundaries of heptad repeat 2 ( HR2 ) . ( B ) Surface presentation of the view in the previous Figure A . HR2 is distinguished in green . Note that the side chain of R707 does not participate in the interface packing in contrast to the GTPase-independent conformation where it stabilises the cis-dimer ( Figure 1E ) . ( C ) Surface presentation of the position of HR2 in the GTP-bound , extended tetramer at the interface of dimer of dimers ( global view in the inset and in Figure 1D ) . Monomers are distinguished in yellow and blue and HR2 in green . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 00510 . 7554/eLife . 23813 . 006Figure 1—figure supplement 3 . Expression of perilipin in affected adipose tissue . Perilipin immunostaining in paraffin-embedded adipose tissue sections . Control adipose tissue is shown on the top row and representative patient adipose tissue on the bottom . Low magnification images are shown on the left and high magnification images on the right . All control adipocytes showed perilipin 1 immunopositivity , however staining was heterogeneous in patient adipose tissue , with many adipocytes showing no staining ( some indicated by asterisks ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 00610 . 7554/eLife . 23813 . 007Table 1 . Biochemical profiles of probands P1-4 . MRS = Magnetic Resonance Spectroscopy; USS = Ultrasound scanning . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 007P1P2P3P4Reference rangeAge at time of assessment - years1817 . 816 . 717Age at first presentation - yearsc . 5c . 12c . 13c . 13Height – m1 . 631 . 581 . 621 . 63Weight – kg88 . 461 . 653 . 672BMI - kg/m233 . 024 . 720 . 427 . 0Liver fat26% by MRS ( normal < 5 . 5 ) ----increased ( USS and biopsy ) Triglyceride - mmol/L7 . 11 . 00 . 97 . 7<1 . 7HDL-Cholesterol - mmol/L0 . 91 . 40 . 90 . 6>1 . 0Total-Cholesterol - mmol/L5 . 84 . 54 . 24 . 0<5 . 2Insulin - pmol/L8996152264<60Glucose - mmol/L4 . 54 . 65 . 24 . 5<6 . 1Glycated haemoglobin -mmol/mol27303235<53Leptin - µg/L0 . 6 ( 14 . 9–60 . 2 ) 2 . 4 ( 2 . 4–24 . 4 ) 2 . 2 ( 0 . 4–8 . 3 ) <0 . 6 ( 8 . 6–38 . 9 ) *Adiponectin - mg/L0 . 7 ( 2 . 6–14 . 9 ) 0 . 8 ( 4 . 4–17 . 7 ) 1 . 4 ( 2 . 6–12 . 6 ) 2 . 0 ( 5 . 0–28 . 8 ) *Lactate – mmol/L7 . 3----4 . 8LH – U/L2 . 010 . 15 . 33 . 01 . 3-8 . 4‡FSH – U/L3 . 48 . 05 . 01 . 02 . 9-8 . 4‡Estradiol – pmol/L105107--35 ng/dl72-529‡Testosterone – nmol/L----7 . 693 ng/dl9-22†*Sex , BMI and assay-specific reference ranges are indicated in brackets for each patient . †Adult male range . ‡Follicular phase . The parents of P1 ( aged 43 and 51 years ) were clinically unaffected . Exome sequencing in the family trio was undertaken and directly interrogated to look for mutations in MFN2 , given the known association of MFN2 with axonal neuropathy , and its recently reported association with MSL ( Sawyer et al . , 2015 ) . This revealed compound heterozygous mutations ( p . Arg707Trp and p . Arg343del ) in MFN2 ( MIM*608507 ) in P1 , whereas her parents were single heterozygotes . Patients P2 and P3 were more mildly affected . These patients , a 17-year-old female ( P2 ) and a 16-year-old male ( P3 ) , were siblings from a multiply consanguineous family of Irish ancestry . Only the two affected siblings and their unaffected parents were available for genetic study , however multiple family members across three generations of the family were reported to have developed progressive overgrowth of adipose tissue in the upper back and chest , head and neck , from early in their second decades . Two older relatives required long term tracheostomies due to airway occlusion by overgrown neck adipose tissue , with MRI in one demonstrating diffuse fatty infiltration of the tongue and a lipomatous mass in the supraclavicular fossa . A paternal uncle was reported to have died peri-operatively after airway problems related to adipose overgrowth in his fourth decade , while three further relatives were reported to show a similar pattern of head and neck adipose overgrowth . One of these had been demonstrated to have severe sensorimotor axonal neuropathy ( Figure 1—figure supplement 1 ) . In keeping with the familial pattern of overgrowth , P2 and P3 developed striae , or stretch marks , over their upper back , together with adipose overgrowth and concomitant loss of fat from the limbs around the age of 13 years old . Examination of P2 at 16 years old showed mild loss of adipose tissue in the forearms , with abnormal adipose tissue accumulation around the neck , upper back , and upper arms . P3 , examined at the age of 15 years old was more mildly affected , but showed dorsal striae , mild adipose tissue loss from the forearms , a prominent dorsal fat pad or ‘buffalo hump’ and increased adipose tissue in the anterior neck . Neither sibling showed any other clinical abnormalities , with age-appropriate progression through puberty , and normal nerve conduction studies . Given parental consanguinity for all affected family members , an autosomal recessive mode of inheritance was hypothesized . Exome sequencing was undertaken for P2 , P3 and their parents , and the p . Arg707Trp mutation in MFN2 was identified in homozygous form in both P2 and P3 , and in heterozygous form in both parents . The final patient , P4 , was evaluated at 17 and 37 years old . She first reported increased head and neck adiposity with increased muscularity of her extremities around 13 years old . This progressed over ensuing years , with striking accumulation of upper back and anterior neck adipose tissue associated with severely increased appetite and frequent food craving behaviours . Other symptoms included male pattern hair growth , an irregular menstrual cycle , and acanthosis nigricans , all suggestive of insulin resistance , as well as leg cramps after exercise , and burning and tingling sensations in hands and feet . On evaluation at 17 years old she was overweight with highly dysmorphic accumulation of truncal , head and neck adiposity ( Figure 1B ) , extending to the pubic region . This contrasted with a paucity of adipose tissue on the limbs , which were muscular with prominent superficial veins . Adipose tissue removed surgically for cosmetic purposes was noted to be highly vascular and lobulated . She was subsequently lost to follow up until 37 years old . At that stage her dominant problems were distressing upper body adipose accumulation with back pain exacerbated by scoliosis , with leg cramps as well as burning and tingling in hands and feet . Menstrual bleeds were irregular , but she had had one successful pregnancy . Abnormally increased appetite persisted , but was less striking than during adolescence . Sanger sequencing of MFN2 was undertaken in P4 and her unaffected father , although her mother ( not knowingly related to her father ) was unavailable . P4 was found to be homozygous for the MFN2 p . Arg707Trp mutation , while her father was heterozygous . Arg707 is exquisitely conserved among MFN2 orthologues , and lies in the second heptad repeat ( HR2 ) coiled-coil domain ( Figure 1C , D , Figure 1—figure supplement 2 ) . The p . Arg707Trp variant is found with an allele frequency of 0 . 055% in Europeans in the ExAC dataset . Literature review revealed a further patient with MSL and the p . Arg707Trp mutation co-inherited with a second , different loss-of-function allele featuring deletion of exons 7 and 8 ( Carr et al . , 2015 ) . Three further siblings were identified with p . Arg707Trp together with the p . Gly109Arg mutation . All three had severe early onset axonal neuropathy , but no adipose phenotype was reported when the oldest of the siblings was 25 years old ( Table 2 ) . No simple heterozygote for the p . Arg707Trp allele has been described with MSL , and most of those heterozygotes reported do not have clinically manifest neuropathy . No other HR2 domain mutations ( at or distal to amino acid residue 646 ) were identified in either homozygous or compound heterozygous forms . 10 . 7554/eLife . 23813 . 008Table 2 . Genotypes and phenotypes of all patients identified with biallelic MFN2 mutations including p . Arg707Trp; MSL = Multiple Symmetrical Lipomatosis . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 008IDAge , yearsSexMFN2 genotypePhenotype ( s ) ( age of onset , years ) CommentReferenceP118Fp . Arg707Trp R343delMSL ( <10 ) Axonal neuropathy ( <10 ) ↑↑ lactate ↓↓ leptin ↓↓ adiponectinThis studyP217Fp . Arg707Trp p . Arg707TrpMSL ( c . 13 ) No neuropathy ( 17 ) ↓ adiponectinThis studyP316Mp . Arg707Trp p . Arg707TrpMSL ( c . 13 ) No neuropathy ( 16 ) ↓ adiponectinThis studyP437Fp . Arg707Trp p . Arg707TrpMSL ( c . 13 ) Likely neuropathy* Renal tubular dysfunction† ( c . 13 ) ↑↑ lactate ↓↓ leptin ↓↓ adiponectinThis studySawyer et al Patient 161Mp . Arg707Trp p . Arg707TrpMSL ( 20 s ) Axonal neuropathy ( 50 s ) Diabetes ( 50 s ) ↑ lactate ↓↓ leptin ↓ adiponectin ( Sawyer et al . , 2015 ) Sawyer et al Patient 263Mp . Arg707Trp p . Arg707TrpMSL ( 40 s ) Axonal neuropathy ( 40s ) ↑ lactate ( Nicholson et al . , 2008; Sawyer et al . , 2015 ) Sawyer et al Patient 360Fp . Arg707Trp p . Arg707TrpMSL ( 20 s ) Axonal neuropathy ( c . 2 ) ↑ lactate ( Sawyer et al . , 2015 ) Carr et al Patient 4/II-144Mp . Arg707Trp MFN2 ex7-8delMSL ( not reported ) Axonal neuropathy ( 2 ) ( Carr et al . , 2015 ) Calvo et al Patient 619Mp . Arg707Trp p . Gly108ArgAxonal neuropathy ( <10 ) ( Calvo et al . , 2009 ) Calvo et al Patient 724Fp . Arg707Trp p . Gly108ArgAxonal neuropathy ( <10 ) ( Calvo et al . , 2009 ) Calvo et al Patient 825Mp . Arg707Trp p . Gly108ArgAxonal neuropathy ( <10 ) ( Calvo et al . , 2009 ) *Based on clinical signs and symptoms . †increased renal protein , magnesium , potassium and calcium with low serum potassium and magnesium; still apparent at 37 years old . The structure of a full-length mitofusin is not yet known . Sequence homology searching using HHPred ( Söding , 2005 ) identified mitofusins as the closest mammalian homologues of the bacterial dynamin-like protein ( BDLP ) , whose structure has been solved ( Low et al . , 2009 ) . Homology modeling suggests that mitofusins retain all key structural domains of BDLP ( Figure 1C ) , and so mitofusin 2 may exist , like BDLP , in two conformations , one GDP-bound , lipid-free and compacted ( Figure 1—figure supplement 2A ) and one GTP-bound , extended and membrane-associated ( Figure 1D ) . The extended conformation is expected to cis-dimerise upon GTP-binding , with dimers further cis-oligomerising on the cell membrane ( Low and Löwe , 2010 ) . In-frame deletion of Arg343 destroys the helix next to the hinge between the GTPase and stalk domains ( Figure 1D ) and is predicted to disrupt the tip of the stalk and seriously impair transition between closed and extended conformations . Membrane fusion depends on trans-dimerization of mitofusins on different membranes . This requires a distinct GTPase-independent extended conformation featuring an anti-parallel dimer between their extended C-terminal domains HR2 ( Franco et al . , 2016 ) . The X-ray structure of a MFN1 HR2 fragment in a homodimer is known ( Koshiba et al . , 2004 ) and the structure of the corresponding MFN2 homodimer ( Figure 1E ) and MFN1/MFN2 heterodimer ( not shown ) can readily be modelled . The HR2 domain is involved in different interfaces in GDP-bound monomers ( Figure 1—figure supplement 2B ) , in GTP-bound cis-oligomers ( in which there are two different interfaces between dimers and between dimer of dimers [Figure 1D] ) , in GTPase-independent trans-dimers ( Figure 1E ) and in MFN2/MFN1 trans heterodimers . Arg707 contributes differently to the stability of these conformations: whereas it makes an ion pair with Asp742 in the trans antiparallel homodimer ( Figure 1E ) or with Glu728 of MFN1 in the heterodimer , it appears on the surface in all other conformations and is not directly involved in intra- or intermolecular helix packing . Its mutation will thus almost certainly destabilise trans-dimerisation by destroying the ion pair , but will influence other conformations to much lesser degrees . Given the anatomically obvious increase in upper body adipose tissue in all four patients studied , the function of that adipose tissue was next assessed in vivo and ex vivo . Fasting blood tests showed P1 and P4 but not P2 and P3 to have elevated plasma triglycerides , while P1 and P4 also had severe hepatic steatosis ( Table 1 ) . Initial oral glucose tolerance testing revealed normal glucose tolerance with elevated insulin levels in all four patients , consistent with moderate to severe insulin resistance , with fasting hyperglycaemia seen in P1 and P4 at 20 and 37 years respectively ( Table 3 ) . P1 and P4 also showed failure of the normal suppression of plasma non-esterified fatty acids ( NEFA ) in the face of increased plasma insulin . This was assessed by OGTT at two ages in P1 , in whom impaired NEFA suppression contrasted both with healthy controls and with obese adolescents with overt type 2 diabetes ( Hershkop et al . , 2016 ) ( Figure 1F ) . In P4 it was assessed by frequently sampled intravenous glucose tolerance testing using a previously published protocol ( Sumner et al . , 2004 ) , and basal NEFA of 786 µmol/L showed maximal suppression of only 18% , occurring 27 min after the glucose challenge . These findings suggest impaired suppression of lipolysis by insulin in the overgrown adipose tissue and/or the depleted lower limb adipose depots of P1 and P4 . Blood lactate concentration was elevated at baseline and on two occasions in P1 , rising further after an oral glucose load , suggesting impairment of mitochondrial function in an insulin-responsive tissue , although whether adipose tissue , muscle or both were responsible for this could not be discerned ( Table 3 ) . Lactate levels were not determined in P2 and P3 . 10 . 7554/eLife . 23813 . 009Table 3 . Oral glucose tolerance test results for patients P1 , P2 and P3 . P1 and P4 were tested on two occasions . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 009Time ( minutes ) 0306090120P1 ( 18 years old ) Glucose ( mmol/L ) 4 . 66 . 46 . 44 . 34 . 6Insulin ( pmol/L ) 1091010996249190P1 ( 20 years old ) Glucose ( mmol/L ) 6 . 99 . 39 . 89 . 79 . 3Insulin ( pmol/L ) 207630609677708Lactate ( mmol/L ) 6 . 88 . 39 . 69 . 99 . 9P2 ( 17 . 8 years old ) Glucose ( mmol/L ) 4 . 67 . 18 . 96 . 06 . 2Insulin ( pmol/L ) 96595962478566P3 ( 16 . 7 years old ) Glucose ( mmol/L ) 5 . 79 . 88 . 47 . 05 . 5Insulin ( pmol/L ) 199204817961888773P4 ( 17 years old ) Glucose ( mmol/L ) 4 . 55 . 86 . 95 . 86 . 3Insulin ( pmol/L ) 2641160247913611514P4 ( 37 years old ) Glucose ( mmol/L ) 6 . 6--------Insulin ( pmol/L ) 688-------- Serum adipokine profiles were highly abnormal in all four patients . Serum leptin was undetectable in P1 and P4 on initial evaluation ( Table 1 ) despite increased upper body adiposity , as evidenced by body mass indices of 33 and 27 kg/m2 and whole body adipose content determined by DXA of 31% and 27 . 4% respectively . Serum leptin remained undetectable in P4 at 37 years old . In P2 and P3 circulating leptin concentrations were also low , but overlapped with levels seen in healthy controls , possibly reflecting the relatively early stage of evolution of the pathological adipose overgrowth . In all patients serum adiponectin levels were suppressed to levels usually only seen in generalised lipodystrophy or extreme , non receptoropathy insulin resistance ( Table 1 ) ( Semple et al . , 2006 ) . Leptin secretion from adipose explants taken from the neck and abdominal subcutaneous adipose tissue of P1 was moreover undetectable , and leptin mRNA expression in the same sample was commensurately suppressed ( data not shown ) . Surgical biopsies of overgrown interscapular subcutaneous adipose tissue were independently obtained from P2 and P3 , and from five age-matched otherwise healthy female controls undergoing elective surgery for idiopathic scoliosis . Leptin mRNA expression , and adiponectin mRNA and protein expression were severely suppressed in P2 and P3 compared to controls , while mitofusin 2 protein levels were unchanged ( Figure 1G , H ) . In view of the reduced suppression of lipolysis seen in P1 , perilipin 1 and 2 expression were determined , with levels of perilipin 1 found to be reduced , whereas perilipin 2 expression was preserved ( Figure 1H ) . In all three patients from whom subcutaneous adipose tissue biopsies were obtained , histological examination revealed the expanded adipose tissue to feature exclusively unilocular adipocytes ( Figure 1I ) . UCP1 immunostaining was negative ( data not shown ) . There was no difference in adipocyte size among patients and controls ( Figure 1I ) , and in keeping with low perilipin expression in affected adipose tissue detected by immunoblotting , immunostaining for perilipin was also reduced , with marked heterogeneity seen in patient but not control adipose tissue ( Figure 1—figure supplement 3 ) . In contrast to the nearly normal appearance of affected adipose tissue by light microscopy , electron microscopy of adipose tissue from P2 and P3 revealed highly abnormal adipocyte and preadipocyte ultrastructure . Control adipocytes exhibited scattered ovoid mitochondria with narrow cristae ( Figure 2A ) , while adipocytes from P2 and P3 showed thickening of the cytoplasmic rim and proliferation of round , enlarged mitochondria , with fragmented cristae on cross sectional imaging ( Figure 2B ) , reminiscent of those reported in different cell types and tissues in the context of Mfn2 deficiency ( e . g . [Lee et al . , 2012] ) . Membrane-bound structures enclosed in double membranes were seen in many cells ( Figure 2B , C , D ) , potentially consistent with mitophagy , although no examples of morphologically distinct mitochondria enclosed in such membranes were seen . In occasional cells apparent extrusion of membrane-enclosed damaged mitochondria was seen ( Figure 2D ) . In both P2 and P3 , 3–4 preadipocytes were seen among approximately 30 adipocytes studied by electron microscopy , while no preadipocytes were seen in control samples . Preadipocytes were associated with capillaries , showed characteristic small lipid droplets and glycogen granules , but also showed evidence of proliferation of fragmented mitochondria , many with deranged cristae , as seen in mature adipocytes ( Figure 2D ) . mRNA extracted from affected adipose tissue from three patients was subjected to transcriptomic analysis , using RNAseq in the case of P2 , P3 and five matched controls , and a microarray-based approach in a different laboratory in the case of P1 , from whom subcutaneous adipose samples were independently studied from both neck and abdomen and compared to abdominal subcutaneous adipose tissue from ten female controls . Primary analysis was undertaken of P2 and P3 compared to controls , with the more limited data from the comparison of P1 with controls subsequently aligned . Affected adipose tissue was transcriptionally distinct from control adipose tissue ( Figure 3A ) , and unsupervised pathway analysis identified the most highly significant perturbation of gene expression to be related to mitochondrial dysfunction and oxidative phosphorylation ( Figure 3B ) . The mitochondrial signal was accounted for by downregulation of 66 of 68 transcripts derived from mitochondria-encoded genes , contrasting with upregulation of 83 of 91 nucleus-encoded mitochondrial genes ( Figure 3C ) . Despite this upregulation , however , immunoblotting for a panel of nucleus-encoded oxidative phosphorylation proteins showed universally strongly reduced levels , with citrate synthase , a marker of mitochondrial mass , also sharply reduced ( Figure 3D ) . Adipose mitochondrial DNA was modestly reduced in P2 and P3 , but was not assayed in P1 ( Figure 3E ) . These findings suggest that the increased numbers of morphologically abnormal mitochondria seen on electron microscopy in P2 and P3 are severely hypofunctional , and that the increased transcript levels of nucleus-encoded mitochondrial genes represents a compensatory response that is insufficient to normalise function . Indeed , transcript levels of PGC1A , PGC1B , and NFE2L2 , encoding transcription factors driving mitochondrial biogenesis were upregulated 3 . 2-fold , 1 . 8-fold , and 3 . 8-fold , respectively ( Figure 3F ) . 10 . 7554/eLife . 23813 . 010Figure 2 . Ultrastructural appearances of interscapular subcutaneous adipose tissue in individuals with the homozygous MFN2 p . Arg707Trp mutation . ( A ) Transmission electron micrograph ( TEM ) of interscapular subcutaneous adipose tissue obtained from a control volunteer , showing a thin cytoplasmic rim , and , in the panel below , ovoid mitochondria with thin cristae . ( B ) TEM of interscapular subcutaneous adipose tissue obtained from P2 , showing a thickened cytoplasmic rim containing increased numbers of round , enlarged mitochondria . Lower panel and inset: Mitochondria showed evidence of degeneration with disorganised cristae and complex inclusions ( pointed by red arrows ) ; L= lipid droplet . ( C ) TEM of affected adipose tissue showing mitochondria at different degree of degeneration ( m ) and double membrane-bound structures suggestive of mitophagy ( inset in lower panel shows a concentric double membrane ) . N: adipocyte nucleus . ( D ) TEM of juxtacapillary cell representative of several pre-adipocytes ( Pa ) identified in P2 and P3 but in no controls . Early lipid accumulation and glycogen granules can be seen . In the inset , another image suggestive for active mitophagy is seen in the adjacent adipocyte . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01010 . 7554/eLife . 23813 . 011Figure 3 . Gene expression profiling of hyperplastic adipose tissue . ( A ) Multidimensional Scaling Plot showing distinct patterns of gene expression of interscapular tissue samples from P2 and P3 , compared to five age-matched controls . Three separate tissue samples from each of the patients and two to three separate tissue samples from the controls were used in the analysis . ( B ) Bar chart showing the 10 most significantly altered pathways identified by Ingenuity Pathway Analysis ( IPA ) . Canonical pathways are shown in red to indicate increased expression; non-canonical pathways , to which no directionality was assigned , are shown in grey . The red line represents the p-value cut-off of <0 . 0001 . ( C ) Heat maps showing decreased expression of genes in the mitochondrial genome and increased expression of nuclear-encoded oxidative phosphorylation ( OxPhos ) genes in P2 and P3 . ( D ) Immunoblots of control and patient adipose tissue for OXPHOS subunits ATP synthase subunit alpha ( V ) , Complex III subunit Core 2 ( III ) , Complex II subunit 30 kDa ( II ) , and the mitochondrial marker citrate synthase ( CS ) . TATA-binding protein ( TBP ) and β-actin ( ACTB ) levels were used as loading controls ( as also shown in Figure 1H ) . ( E ) Reduced levels of mtDNA in affected adipose tissue determined by Taqman real-time PCR . Means of three independent experiments ± SEM are shown . ( **p<0 . 01 ) . ( F ) Increased expression of PGC1A , PGC1B , and NEF2L2 assessed by real-time qPCR and expressed as fold change in expression relative to five controls . Shown are means of 3 replicates . Error bars represent SEM . ( G ) Heat map showing the top 32 ‘disease and biology’ pathways from IPA . A word/tag weighted cloud of component pathways is shown . ( H ) Venn Diagrams comparing the top 100 genes up and down with an FDR cut-off of <0 . 00001 among patients . Genes shared between all three patients’ samples are shown in expanded lists with genes noted in the text highlighted in red . ( I ) Transcript levels of DDIT3 , HSPA5 , ATF5 , TRIB3 , PSAT1 , and CIDEA in P3 and P2 , expressed as fold change relative to controls determined by Taqman quantitative real-time PCR . Shown are means of eight replicates of different dilutions with error bars calculated from R-squared values of fitted line . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01110 . 7554/eLife . 23813 . 012Figure 3—source data 1 . Differential expression analysis data . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01210 . 7554/eLife . 23813 . 013Figure 3—source data 2 . Ingenuity Pathway Analysis results for P1 . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01310 . 7554/eLife . 23813 . 014Figure 3—source data 3 . Ingenuity Pathway Analysis results for P2 . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01410 . 7554/eLife . 23813 . 015Figure 3—source data 4 . Ingenuity Pathway Analysis results for P3 . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01510 . 7554/eLife . 23813 . 016Figure 3—source data 5 . Data used to generate the heat maps shown in Figure 3C . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01610 . 7554/eLife . 23813 . 017Figure 3—source data 6 . Ingenuity Pathway Analysis ‘Disease and Biology’ data used to generate the heat map shown in Figure 3G . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01710 . 7554/eLife . 23813 . 018Figure 3—figure supplement 1 . Expression of human brown , ‘beige’ , and white adipocyte markers in affected adipose tissue . Expression of previously suggested markers of human adipocyte subclasses in adipose tissue from P1 , P2 , and P3 . Results are represented as fold changes from control adipose tissue from the matched volunteers described . Data for P2 and P3 were derived from RNAseq analysis , and for P1 from microarray analysis . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 018 Other genesets related to cellular stress responses were also highly significantly altered , including oxidative stress and unfolded protein response pathways , with additional evidence of mTOR pathway activation ( Figure 3B ) . Despite the strong transcriptional suggestion of mitochondrial dysfunction and secondary activation of stress signalling pathways , analysis of disease-associated transcriptional pathways showed that signatures of cell and tissue death were suppressed , while signatures related to cell and tissue proliferation and survival were strongly upregulated ( Figure 3G ) . The top 100 upregulated and downregulated genes were then compared across all three patients studied . This revealed 22 shared significantly upregulated genes , which featured genes involved in the unfolded protein response , including the key transcription factors DDIT3 ( better known as CHOP ) and ATF5 , the ER chaperone HSPA5 ( better known as BiP ) , and TRB3 , which has been shown in different contexts to inhibit insulin signalling ( Du et al . , 2003 ) and to drive apoptosis ( Ohoka et al . , 2005 ) among other roles ( Figure 3H , I ) . BiP , ATF5 , and TRB3 are known to be direct transcriptional targets of CHOP and ATF4 ( Han et al . , 2013 ) , which was also highly significantly upregulated , while ATF5 has also recently been suggested to serve as a direct and critical mediator of the mammalian mitochondrial unfolded protein response ( Fiorese et al . , 2016 ) , analogous to ATFS-1 in C . elegans . Strong transcriptional downregulation of both leptin and adiponectin was confirmed in all three patients , with other downregulated genes including the brown fat lipid droplet-associated gene CIDEA , and the pro-adipogenic transcription factor KLF4 ( Figure 3H , I ) . To address the type of adipose tissue in affected depots , expression of genes suggested in recent human studies to be markers for white , beige and brown fat were examined ( Basse et al . , 2015; Carey et al . , 2014; Lee et al . , 2016 ) . No consistent pattern of change in gene expression was seen , with BAT- and beige adipose tissue-selective genes unchanged or lower in affected tissue than in controls , with the exception of CITED1 , proposed as a marker of beige fat , which was upregulated in all three patients . WAT genes showed no consistent changes in expression from controls ( Figure 3—figure supplement 1 ) , consistent with the overgrown adipose tissue studied being white adipose tissue . Dermal fibroblasts were cultured from the site of adipose biopsy in P1 , P2 , and P3 . Like affected adipose tissue , cultured dermal fibroblasts showed normal MFN2 expression , however quite unlike in adipose tissue , expression of citrate synthase ( Figure 4A ) and a panel of oxidative phosphorylation proteins was normal ( Figure 4B ) , while mitochondrial DNA content was mildly lower than in controls ( Figure 4C ) . Furthermore , immunocytochemistry showed appropriate localisation of MFN2 expression to the mitochondrial network ( Figure 4D , Figure 4—figure supplement 1 ) , which in patient cells appeared morphologically indistinguishable from control cells both at baseline ( Figure 4E ) , and after treatment with the small molecule mitochondrial fission inhibitor mdivi-1 ( Figure 4—figure supplement 2 ) . Transmission electron microscopy of dermal fibroblasts from P1-3 revealed normal mitochondrial ultrastructure , with no evidence of the striking fragmentation and dysmorphism seen in adipose biopsies ( Figure 4—figure supplement 3 ) . Diffusion of photoactivated GFP within the network was also normal when assessed in living cells ( Figure 4F ) , although this diffusion was impaired in wild type cells by hydrogen peroxide treatment to fragment the network ( data not shown ) . Mitochondrial membrane potential was unchanged in patient cells , as assessed by a mitochondrial membrane potential sensor JC-1 ratio assay ( Figure 4G ) . 10 . 7554/eLife . 23813 . 019Figure 4 . Preserved mitochondrial network in dermal fibroblasts from patients with MFN2-related MSL . ( A ) Immunoblots of dermal fibroblast lysates for mitofusin 2 ( MFN2 ) , mitofusin 1 ( MFN1 ) , and the mitochondrial marker citrate synthase ( CS ) . γ-tubulin was used as loading control . ( B ) Immublotting of OXPHOS subunits using an optimized pre-mixed cocktail of monoclonal antibodies specific to Complex I subunit NDUF88 ( CI ) , Complex II subunit 30 kDa ( CII ) , Complex III subunit Core 2 ( CIII ) , Complex IV subunit ( IV ) , and ATP synthase subunit alpha ( CV ) . ( C ) Reduced levels of mtDNA in patient dermal fibroblasts determined by Taqman real-time PCR . Means of three independent experiments ± SEM are shown . ( *p<0 . 05 ) . ( D ) Immunofluorescence confocal microscopy of endogenous MFN2 in fixed control and patient dermal fibroblasts showing preserved MFN2 expression and localization to mitochondria ( See also Figure 4—figure supplement 1 ) . Mitochondria were visualized with MitoTracker Orange ( MT ) . Images in the bottom rows show digitally magnified detail of the boxed areas in the top rows . Scale bars: 10 µm ( top rows ) ; 5 µm ( bottom rows ) . ( E ) Normal mitochondrial morphology in patient dermal fibroblasts demonstrated using the mitochondrial dye MitoTracker Orange . Cells were classified as having mostly fragmented , short tubular , and long tubular mitochondria as indicated in the representative confocal microscopy images shown . Asterisks indicate regions digitally magnified to show mitochondrial morphology in detail ( insets ) . Scale bars , 10 µm . Data from four independent experiments ( >50 cells per experiment ) are presented as means ± SD . ( F ) Visualization and quantification of mitochondrial fusion using photoactivatable mitochondria matrix-targeted GFP ( mito-PA-GFP ) : a dermal fibroblast from P1 expressing both mito-PA-GFP and mito-DsRed is shown to illustrate the in vivo mitochondria fusion assay . Photoconversion of PA-GFP was restricted to the area marked by the white dashed box . The activated mito-PA-GFP molecules , in time , redistributed along the area within the green dashed lines . The percentage of fluorescence remaining in the boxed region was measured , normalized to mito-DsRed signal , and plotted against time . Data are represented as means ± SD of single cell time-lapse measurements ( 10 randomly selected cells expressing both mito-DsRed and mito-PA-GFP per cell line were imaged ) . Scale bar , 10 µm . ( G ) Quantification of mitochondrial membrane potential ( △Ψ ) using the fluorescent sensor JC-1 . Data show ratio of 590 nm ( red ) and 535 nm ( green ) fluorescence signals . Data from three independent experiments are represented as mean ± SD . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 01910 . 7554/eLife . 23813 . 020Figure 4—figure supplement 1 . Mitochondrial network morphology and mitofusin 2 expression in dermal fibroblasts from P1 and P2 . Representative immunofluorescence microscopy images of endogenous MFN2 in fixed dermal fibroblasts from P1 and P2 . Mitochondria were visualized with MitoTracker Orange ( MT ) . Representative micrographs from Control and P3 dermal fibroblasts are shown in Figure 4D . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 02010 . 7554/eLife . 23813 . 021Figure 4—figure supplement 2 . Response of dermal fibroblasts to mitochondrial fission inhibitor mdivi-1 . Representative confocal microscopy images of fixed dermal fibroblasts . MitoTracker Orange-labelled mitochondria are shown in green . DAPI-stained nuclei are shown in blue . Asterisks indicate regions digitally magnified and shown in insets . Scale bars: 10 µm ( 5 µm in insets ) . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 02110 . 7554/eLife . 23813 . 022Figure 4—figure supplement 3 . Transmission electron micrographs of dermal fibroblasts obtained from A . a control volunteer , B . P3 , and C . P1 , showing normal mitochondrial morphology in all cases . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 022 Neither adipose tissue nor dermal fibroblasts were available for study from P4 , however electron microscopy of a liver biopsy obtained at 17 years old was reported to show increased lipid droplets and pericellular collagen but neither abnormal mitochondrial size nor ultrastructure . Together with dermal fibroblast findings this is consistent with the abnormalities seen in adipose tissue in this study being tissue-selective .
MSL is a phenotypically dramatic condition that has been known since the nineteenth century . Understanding of its pathogenesis is likely to yield insights into poorly understood aspects of adipose tissue function that are critical to pandemic diseases , including the determinants of the differing metabolic behaviours of different adipose depots . Study of MSL has been hampered , however , by its clinical heterogeneity , and lack of knowledge of its cause ( Herbst , 2012 ) . The recent report of homozygosity for the MFN2 p . Arg707Trp mutation in three individuals with MSL ( Sawyer et al . , 2015 ) suggested the first germline Mendelian cause of the disorder , presenting a critical opportunity to further understanding of the underlying pathogenesis , however the pathology of the overgrown adipose tissue and the systemic metabolic consequences were not specifically studied . We now describe four more affected patients , confirming the genetic association , demonstrate that increased adipose mass is accounted for by hyperplasia of UCP1-negative , unilocular adipocytes , and provide evidence of mitochondrial dysfunction in affected adipose tissue . We also show that depot-selective adipose hyperplasia is discordant with suppressed leptin levels and impaired insulin-mediated suppression of lipolysis in severely affected patients , while insulin resistance is seen in all . These findings offer a new window into the extraordinary depot-selective adipose overgrowth of MSL , and offer new opportunities for novel therapy , and to gain insights into fundamental aspects of adipocyte behaviour crucial to pandemic disease . Bioenergetic efficiency and ATP-generating capacity of mitochondria are heavily influenced by the dynamic architecture of the mitochondrial network ( eg . [Liesa and Shirihai , 2013; Mishra and Chan , 2016] ) . This is regulated by specialised proteins including mitofusin 2 ( Schrepfer and Scorrano , 2016 ) , which is important not only in mitochondrial fusion , but also in mitochondrial-ER tethering ( Naon et al . , 2016 ) , apoptosis ( de Brito and Scorrano , 2008 ) , mitophagy ( Chen and Dorn , 2013 ) and regulation of the unfolded protein response ( Pablo Muñoz et al . , 2014 ) . Global knockout of Mfn2 in mice is embryonically lethal ( Chen et al . , 2003 ) , but knockout specific to cardiac or skeletal muscle ( Chen and Dorn , 2013; Sebastián et al . , 2016 ) , to dopaminergic neurones ( Pham et al . , 2012 ) , to the renal proximal tubule ( Gall et al . , 2015 ) , or to POMC neurones of the hypothalamus ( Schneeberger et al . , 2013 ) is tolerated . Nevertheless in each case cell or organ dysfunction is seen . Together with evidence that MFN2 expression is lowered in key metabolic tissues in disease states ( Schneeberger et al . , 2013; Zorzano , 2009 ) , this has led to keen interest in mitofusin 2 as a contributor to the pathogenesis of pandemic obesity-related diseases . No adipose-specific knockout has been described . Despite this largely murine evidence for important roles of MFN2 in several tissues , MFN2 loss-of-function mutations in humans have until recently been implicated solely in axonal sensorimotor neuropathy ( Züchner et al . , 2004 ) , which is only weakly recapitulated in mouse models ( Detmer et al . , 2008; Strickland et al . , 2014 ) . Our findings , together with those of Sawyer et al . ( 2015 ) , now highlight a second tissue-selective phenotype in humans , namely severe upper body adipose overgrowth or ‘Multiple Symmetrical Lipomatosis’ ( MSL ) . All cases of MFN2-related MSL have featured biallelic mutations including at least one MFN2 p . Arg707Trp allele , and many , but not all , patients have shown early-onset neuropathy in addition . Our structural analysis suggests that the mutation may have differential effects on different conformations of MFN2 , particularly trans-dimerisation and thus membrane fusion . It may be important that MFN2 homodimers will be most affected ( Figure 1E ) as both surfaces at the interface have a mutation , while in heterodimers only one surface is mutated . The second mutation we report in P1 , who is severely affected , is predicted significantly to impair or to abrogate MFN2 function by destroying the helical support of the hinge between the GTPase and stalk domains ( Figure 1C ) . Although the association of the MFN2 p . Arg707Trp mutation specifically with MSL is striking , existing human genetic data do not exclude the possibility that other HR2 domain mutations may also lead to the phenotype when both MFN2 alleles are affected . The overgrown adipose tissue is unable to execute its key metabolic functions normally , as evidenced by insulin resistance , and reduced suppression of circulating free fatty acid concentrations after a glucose load . One caveat to this assertion is that it is possible that the ‘lipodystrophic’ or triglyceride-depleted lower limb adipose depots might also be contributing to the perturbed suppression of lipolysis and insulin resistance . Most strikingly , despite dramatic upper body adipose overgrowth in MFN2–related MSL , adipose tissue expression of the archetypal adipocyte hormones leptin and adiponectin was extremely low , with P1 , P4 and one of the older , more severely affected patients previously described having undetectable plasma leptin despite high fat mass ( Sawyer et al . , 2015 ) . We hypothesize that the differences observed in leptin concentrations among the patients we describe reflects the evolution of the disease , with all adipose depots affected either by overgrowth ( upper body ) or dystrophy ( lower body ) in P1 and P4 , while in P2 and P3 some normal adipocytes remain . We suggest that advanced MFN2-related MSL represents a fourth form of severe leptin deficiency in humans , adding to congenital LEP mutations , lack of adipocytes in generalised LD , and ‘empty’ adipocytes in extreme leanness or starvation . We note that this phenotype may not be present in all MSL cases , as others have reported normal leptin levels in MSL cases of undefined cause ( Haap et al . , 2004 ) . Suppressed leptin expression could be explained if the pathologically expanding cell lineage had naturally low levels of leptin expression , as reported for brown adipocytes , or alternatively may reflect perturbation of a cellular mechanism sensing mitochondrial function and cellular energy state and transducing it into leptin gene expression . Although we demonstrate that increased adipose mass is accounted for by hyperplasia of cells with morphologic and transcriptional characteristics of UCP1-negative , unilocular white adipocytes , we cannot rule out the possibility that these cells are dysfunctional , ‘whitened’ thermogenic adipocytes , and indeed mitochondrial depletion has previously been shown to be involved in ‘whitening’ of beige adipocytes ( Altshuler-Keylin et al . , 2016 ) . Evidence of impaired mitochondrial function in MFN2-related MSL is seen in strikingly elevated blood lactate concentrations in P1 , P4 and previously described patients ( Sawyer et al . , 2015 ) , reduced expression of citrate synthase and respiratory chain components , and highly abnormal mitochondrial ultrastructure in affected adipose depots . The reduction in mitochondria-encoded transcripts , contrasting with increased nucleus-encoded mitochondrial transcripts and mitochondrial biogenesis factors suggests that compensatory transcriptional programmes are activated to try to restore cellular mitochondrial function . In contrast , gene expression , and mitochondrial morphology and fusion of primary dermal fibroblasts is normal , confirming mitochondrial dysfunction in MFN2-related MSL to be tissue-selective , and suggesting that MFN2 Arg707Trp remains able to mediate mitochondrial fusion in some contexts . Affected adipose tissue shows strong and consistent evidence of activation of mitochondrial stress responses , with notably strong upregulation of genes known to be positively regulated by ATF4 and CHOP ( e . g . TRB3 , ATF5 , PSAT1 , CHAC1 , HSPA5 ( BiP ) , SLC7A5 , SLC7A11 ) ( Han et al . , 2013 ) or mTORC1 signaling . ATF4 and CHOP activities and expression are upregulated by PERK , which mediates a key arm of the ER unfolded protein response , and has also been shown to interact directly with MFN2 at ER-mitochondrial junctions ( Naon and Scorrano , 2014; Pablo Muñoz et al . , 2014 ) , while ATF5 has recently been suggested to be the mammalian counterpart of the C . elegans mitochondrial unfolded protein sensor ATFS-1 ( Fiorese et al . , 2016 ) . mTORC1 may be activated either downstream or upstream from the UPR ( Appenzeller-Herzog and Hall , 2012 ) , driving protein synthesis , especially of nucleus-encoded mitochondrial genes ( Morita et al . , 2013 ) . Increased cell proliferation or survival in the face of mitochondrial dysfunction is not unprecedented: The phenomenon of mitohormesis ( Yun and Finkel , 2014 ) , whereby mild mitochondrial dysfunction exerts pro-survival effects on cells and organisms , is well established in model systems . MSL is also associated with other forms of genetic or acquired mitochondrial insult , including a common mitochondrial lysyl tRNA mutation ( Chong et al . , 2003 ) , HIV infection treated with highly active anti-retroviral therapy ( Giralt et al . , 2011 ) , and excess alcohol intake ( Enzi et al . , 2015 ) . This raises the possibility that in MFN2-related MSL mitochondrial dysfunction falls in a narrow window of severity that drives mitohormetic upper body preadipocyte differentiation and/or reduced adipocyte clearance . Whether lower body adipocytes are relatively less unaffected , and thus depleted of triglyceride stores as a ‘bystander’ effect of pathologically expanding upper body adiposity , or whether they are more severely affected and thus die in the face of the same genetic insult , remains to be determined . Arguing against a critical role for a general reduction in mitochondrial function as the critical driver of MSL , the overwhelming majority of known forms of monogenic mitochondrial dysfunction , of varying severity , are not associated with it , suggesting an alternative , or additional , specific effect of the MFN2 mutations observed . This is plausible as MFN2 deficiency has been shown in cellular contexts to suppress apoptosis ( Pablo Muñoz et al . , 2014; Shen et al . , 2007 ) , and in ischemic hearts and kidneys to reduce ischaemic tissue ( Hall et al . , 2016; Pablo Muñoz et al . , 2014 ) damage and cell death . MFN2-related MSL could thus reflect a dyscoordinated cellular stress response to severe mitochondrial dysfunction , with MFN2 mutations preventing cell death , allowing survival and proliferation of functionally impaired cells . Finally , our findings have potentially important implications for therapy . Hypogonadism and hyperphagia related to absolute leptin deficiency in P1 may respond to subcutaneous leptin therapy , reducing the burden imposed on overgrown adipose tissue by chronic positive energy balance , while the effect of inhibiting the upregulated anabolic pathways detected transcriptomically , such as mTOR signalling , using clinically available inhibitors , is worthy of consideration . This is in keeping with the promise shown by sirolimus in murine mitochondrial disease ( Johnson et al . , 2013 ) .
Written informed consent was obtained from all participants or their parents if under 18 years old . The research was approved by relevant UK , US , or Danish Research Ethics Committees and was conducted in accordance with the Declaration of Helsinki . Biochemical evaluations were performed in accredited diagnostic laboratories . Age- and BMI-matched reference ranges for adiponectin and leptin were derived from the 5-95th percentiles of the MRC Ely study ( Williams et al . , 1995 ) . Oral glucose tolerance tests ( OGTTs ) were performed with 75 g glucose following an overnight fast . IVGTT was performed as previously described ( Sumner et al . , 2004 ) . Body composition was assessed using Lunar Prodigy dual-energy X-ray absorptiometry ( DEXA , GE Lunar ) . Hepatic triglyceride was assessed in P1 using proton magnetic resonance spectroscopy on a Siemens 3T Verio scanner using methods previously described ( Semple et al . , 2009 ) , and quantified as the ratio of methylene to combined methylene and water signals corrected for spin-spin relaxation . Adipose distribution was determined by magnetic resonance imaging , using T1-weighted turbo spin echo , coronal images , also generated by a Siemens 3T Verio scanner . Exome sequencing of lymphocyte DNA was undertaken in separate laboratories for families of P1-3 , in both cases using SureSelect capture ( Agilent ) followed by HiSeq2000 sequencing ( Illumina ) . Exome-wide sequencing , sequence alignment and variant calling and filtering are described separately below for P1 and P2 , 3 . Sanger sequencing was used for primary genetic diagnosis in P4 , and for confirmation of MFN2 mutations in P1-3 . Lymphocyte DNA from P1 and her parents was used for exome- sequencing , employing SureSelect capture ( Agilent ) followed by HiSeq2000 sequencing ( Illumina ) . Sequence data were aligned to the human genome assembly hg19 using the software bwa23 ( version 0 . 7 . 5a-r405 ) ( Li and Durbin , 2009 ) with default parameters . Duplicates were removed , and the mate-pair information was defined using picard tools ( http://picard . sourceforge . net ) ( version 1 . 95 ) . Subsequently , local realignment surrounding known indel regions was performed , and the quality score for each read was recalibrated with gatk24 ( version 2 . 4–7-g5e89f01 ) ( McKenna et al . , 2010 ) using the same resource files as in the 1000 Genomes Project ( Abecasis et al . , 2010 ) . Finally genotypes were called using SAMtools25 ( version 0 . 1 . 19–44428 cd ) http://samtools . sourceforge . net/; SAMTOOLS , RRID:SCR_002105 ) ( Li , 2011 ) and bcftools with standard parameters ( -A-c -e -g-v ) . Variants were annotated with Ensembl Variant Effect Predictor against Ensembl release 84 ( McLaren et al . , 2010 ) and the NCBI dbSNP database build 144 ( ftp://ftp . ncbi . nlm . nih . gov/snp/organisms/human_9606/ ) . The resulting mutation list was interrogated to look for rare sequence variants in MFN2 based on the strong pretest probability of a mutation based on the published association of MSL , neuropathy and MFN2 mutations . Lymphocyte DNA from P2 , P3 and their parents was isolated using the QIAamp and the Autopure kits . Exome Sequencing was undertaken at the Genomic Analysis Facility at the Duke Centre for Human Genome Variation , USA . Exons were captured using the Agilent exon capture platform and sequencing undertaken on the Illumina HiSeq 2000 system . Reads were aligned to the reference sequence using Novoalign alignment tool ( http://www . novocraft . com/products/novoalign/; NovoAlign , RRID:SCR_014818 ) . Depth of capture and completeness of coverage of the exome was assessed using analytical tools built around the BED tools package . Quality filtering was undertaken using the Sam tools and Picard tools packages . Single nucleotide substitutions and small indels were called using the Sam tools package ( http://samtools . sourceforge . net/; SAMTOOLS , RRID:SCR_002105 ) ( Li et al . , 2009 ) . All identified variants were annotated with respect to open reading frames and cross referenced with publically available variant databases and internal control samples using scripts built around the Annovar tool ( http://wannovar . usc . edu/index . php; wANNOVAR , RRID:SCR_000565 ) ( Yang and Wang , 2015 ) . Filtering for novel variants was performed by comparison to dbSNP132 and 1000 Genomes SNP calls and patient variants were compared to variants identified in 250 control exomes sequenced and analyzed in a similar manner . 30 rare variants ( in coding sequence or splice sites ) with an allele frequency <0 . 01 in control populations ( 1000 genome and exome variant server ) were identified . Of these only the MFN2 variant was recorded in ClinVar ( ClinVar , RRID:SCR_006169 ) where is was classified as Pathogenic ( https://www . ncbi . nlm . nih . gov/clinvar/RCV000002369/ ) . The Exome Variant Server ( http://evs . gs . washington . edu/EVS/ ) and EXaC Browser ( http://exac . broadinstitute . org/gene/ENSG00000116688 ) datasets were accessed in June 2015 . The sequences of human MFN2 and Bacterial Dynamin-Like Protein ( BDLP ) from Nostoc punctiforme were aligned using HHpred and PROMALS ( Pei and Grishin , 2007 ) algorithms . PDB coordinates of monomer ( 2J69 , [Low and Löwe , 2006] ) , dimer ( 2W6D , [Low et al . , 2009] ) , and tetramer ( provided by Dr . Harry Low ) were used to display the MFN2 regions of interest in Figure 1D and Figure 1—figure supplement 2 . Model of MFN2 HR2 dimer ( Figure 1E ) and HR1/HR2 heterodimer were built using coordinates of HR2 NFN1 dimer ( 1T3J , [Koshiba et al . , 2004] ) using Modeller software ( MODELLER: Program for Comparative Protein Structure Modelling by Satisfaction of Spatial Restraints , RRID:SCR_008395 ) ( Sali and Blundell , 1993 ) . Subcutaneous adipose tissue biopsies were taken from the interscapular region under general anaesthesia in the fasting state for P2 , P3 , and five otherwise healthy volunteers undergoing corrective surgery for idiopathic scoliosis ( all female , median age 14 . 9 ( 12 . 9–18 . 9 ) , median B . M . I . 21 . 9 kg/m2 ( 19 . 0–28 . 4 ) ) . Adipose tissue was dissected from surrounding connective and vascular tissue on dry ice and either frozen in liquid nitrogen for later protein or RNA extraction , fixed in 4% PFA/PBS and paraffin embedded using standard procedures for light microscopy , or fixed for 24 hr in 2% glutaraldehyde/2% formaldehyde in 0 . 05 M sodium cacodylate at pH 7 . 4 , followed by storage in 0 . 05 M sodium cacodylate , for electron microscopy . Overlying skin from P1 , P2 , P3 , and two otherwise healthy volunteers biopsies was used to culture dermal fibroblasts using standard procedures . A third dermal fibroblast control cell line was obtained from ATCC ( ATCC Cat# CRL-2707 , RRID:CVCL_2373 ) . DNA was extracted using QIAamp DNA Micro Kit ( Qiagen ) . For RNA extraction , three frozen adipose aliquots per subject were transferred into 1 mL QIAzol ( Qiagen ) and homogenized on thawing with a hand-held TissueRuptor ( Qiagen ) before centrifugation at 13 , 000 g for 3 min . Infranatant beneath the fatty layer was mixed with 1 mL 100% ethanol before binding to RNeasy columns and purifying using the RNAeasy kit according to manufacturer’s instructions . Frozen adipose tissue samples were homogenized manually with a pestle in modified RIPA buffer ( 150 mM NaCl , 1% Nonidet P-40 , 0 . 25% sodium deoxycholate , 50 mM Tris-Cl , pH 7 . 8 ) supplemented with Complete Protease Inhibitor cocktail ( Roche ) . Tissue extracts were then cleared by centrifugation . Debris and the top layer of fat were discarded . Subcutaneous adipose biopsies were also obtained in a separate centre under local anaesthetic from the neck and abdomen of P1 . RNA was extracted from these biopsies , and from surgical biopsies obtained from abdominal adipose tissue from ten women undergoing elective abdominal surgery ( Median age 68 . 5 years ( range 45–83 ) ; median B . M . I . 27 . 2 kg/m2 ( 22 . 2–41 . 0 ) ) . Relative levels of mitochondrial DNA ( mtDNA ) and nuclear DNA ( nDNA ) in total DNA were quantified by singleplex quantitative real-time PCR with primers targeting the mitochondria-encoded ND1 and the nucleus-encoded B2M genes followed by evaluation of Ct values using the comparative Ct ( 2-ΔΔCt ) method . Primers and fluorogenic probes used are detailed in Table 4 . Differences between means were analyzed by two-tailed Student’s t test using GraphPad Prism 5 . 0b software ( RRID:SCR_002798 ) . 10 . 7554/eLife . 23813 . 023Table 4 . Primers and Probes used for mRNA and mtDNA quantification . DOI: http://dx . doi . org/10 . 7554/eLife . 23813 . 023GeneTaqMan GE assayTaqman GE primers and probePPARGC1AHs01016719_m1NAPPARGC1BHs00991676_m1NANFE2LHs00975961_g1NADDIT3Hs00358796_g1NAHSPA5Hs00607129_gHNAATF5Hs01119208_m1NALEPHs00174877_m1NATRIB3Hs01082394_m1NAPSAT1Hs00795278_mHNACIDEAHs00154455_m1NAADIPOQHs00605917_m1NATBPHs99999910_m1NAUBCHs01871556_s1NAYWHAZHs03044281_g1NAND1NAForward: 5'-CCCTAAAACCCGCCACATCT-3' Reverse: 5'-GAGCGATGGTGAGAGCTAAGGT-3' Probe: [6FAM]−5'-CCATCACCCTCTACATCACCGCCC-3'-[TAM]B2MNAForward: 5'-CCAGCAGAGAATGGAAAGTCAA-3' Reverse: 5'-TCTCTCTCCATTCTTCAGTAAGTCAACT-3' Probe: [6FAM]−5'- ATGTGTCTGGGTTTCATCCATCCGACA-3’-[TAM] Images were captured from sections using an upright Olympus BX-41 microscope ( Olympus ) equipped with a 20 x objective and a colour ColorView CCD camera ( Olympus ) . A semi-automated workflow in CellP software ( Olympus ) was used to detect cell membranes of individual adipocytes and determine the cell area . Briefly , images were background subtracted , filtered , thresholded , binarised , prior to ROI inspection and manual correction , cell labelling for traceability , and measurement . GraphPad Prism ( RRID:SCR_002798 ) was used to calculate relative frequency and cumulative distributions . More than 500 adipocytes were measured per condition . For RNAseq , RNA was concentrated with the RNeasy MinElute Cleanup Kit ( Qiagen ) and subject to assessment of quantity and quality using an Agilent BioAnalyzer 2100 . RNA samples were whole transcriptome-amplified with NuGEN RNA-Seq System V2 and converted to indexed libraries with NuGEN Ovation Rapid DR Multiplex Library System following the manufacturer’s instructions . Libraries were sequenced on an llumina HiSeq 4000 system ( 50 bp , single read sequencing ) . Cuffdiff from the Samtools suite was used to align reads , FeatureCount to count fragments overlapping annotated human genes , and the resulting data were analysed with the EdgeR package ( edgeR , RRID:SCR_012802 ) : normalization used weighted trimmed means of M-values ( TMM ) and a design matrix comparing affected with unaffected samples . Generalized linear modeling ( glmLRT ) was used for differential expression analysis . The resulting gene list was analysed using Ingenuity Pathway Analysis ( IPA ) software . Data were deposited in GEO under accession number GSE97156 ( http://www . ncbi . nlm . nih . gov/geo/query/acc . cgi ? acc=GSE97156 ) . Confirmation of selected RNAseq findings was undertaken using TaqMan quantitative real-time PCR using the primer/probesets indicated in Table 4 . TBP , UBC , and YWHAZ were chosen as endogenous controls based on RNAseq analysis . Serial dilutions of all controls and samples were run , and Ct-values were compared using an efficiency-corrected comparative Ct method based on the geometric mean of the three endogenous controls . Subcutaneous abdominal RNA and neck adipose RNA from P1 was separately analysed in duplicate in the same batch as subcutaneous abdominal RNA from 10 female controls . RNA was purified using the miRNeasy Mini Kit from Qiagen and mRNA expression measured using the Illumina HT-12 v4 BeadChip . Fold changes in mRNA expression levels were calculated by comparing the median value of P1 with the median value of the 10 female controls . Proteins were resolved using NuPAGE Novex 4–12% Bis-Tris gels electroblotted onto PVDF membranes using the iBlot system ( Invitrogen ) . Antibodies specific to the following proteins were used: Citrate synthase ( Abcam Cat# ab129095 RRID:AB_11143209 ) , MFN1 ( Mitofusin-1 ( D6E2S ) , Cell Signaling Techonology ) , MFN2 ( Mitofusin-2 ( D2D10 ) , Cell Signaling Technology ) , Total Human OXPHOS cocktail ( Mitosciences , Abcam ) , γ-tubulin ( Sigma-Aldrich Cat# T5326 RRID:AB_532292 ) , TATA binding protein ( Abcam Cat# ab818 RRID:AB_306337 ) , β-actin ( Sigma-Aldrich Cat# A5316 RRID:AB_476743 ) , PLIN1 ( GP33 , PROGEN Biotechnik ) , PLIN2 ( GP47 , PROGEN Biotechnik ) , adiponectin ( Abcam Cat# ab13881 RRID:AB_2221613 ) . Dermal fibroblasts were cultured in DMEM ( Invitrogen ) supplemented with 10% fetal bovine serum ( Hyclone ) , 1% penicillin-streptomycin and 2 mM L-glutamine ( Invitrogen ) , in an incubator at 37°C in 5% CO2/95% O2 atmosphere . All dermal fibroblast lines were routinely tested as negative for mycoplasma contaminations using VenorGem Classic Mycoplasma Testing PCR Kit ( Minerva Biolabs , Cat# 11–1050 ) . Dermal fibroblast lysates were prepared using M-PER Mammalian Protein Extraction Reagent ( Thermo Scientific ) supplemented with protease inhibitors . Whole cell extracts were cleared by centrifugation . For confocal studies of MFN2 localisation , dermal fibroblasts grown on glass coverslips were labeled with MitoTracker Orange CMTMRos dye ( Molecular Probes ) and then fixed with 4% paraformaldehyde for 15 min at room temperature , permeabilized in 0 . 05% ( v/v ) Triton X-100/PBS , and immunostained with antibodies specific to MFN2 in a BSA/PBS blocking solution . Coverslips were mounted in ProLong Gold Antifade Reagent with DAPI mounting medium ( Molecular Probes ) . To inhibit mitochondrial fission dermal fibroblasts were incubated at 37°C with 50 µM mdivi-1 ( SIGMA ) for 3 hr . To assess mitochondrial fusion cells stably expressing photoactivatable-GFP ( mitoPA-GFP ) and matrix-targeted DsRed ( mitoDsRed ) were first generated . Lentivector particles were produced by transfecting HEK293T cells with either lentiviral mitochondria matrix-targeted PA-GFPmt ( a gift from Orian Shirihai , Addgene plasmid # 19989 ) or lentiviral mitoDsRed ( a gift from Pantelis Tsoulfas , Addgene plasmid # 44386 ) , and the lentivirus packaging vectors psPAX2 and pMD2 . G using CalPhos Mammalian Transfection Kit ( Takara ) following the manufacturer’s guidelines . Dermal fibroblasts were simultaneously transduced with both lentivector particles in the presence of 8 µg/mL polybrene ( SIGMA ) . Resulting cells expressing mitoDsRed and mitoPA-GFP were plated in glass-bottomed microwell dishes ( Matek ) and imaged live on a SP8 confocal microscope ( Leica ) equipped with a cage incubator for temperature and CO2/O2 control ( Life Imaging Services ) . PA-GFP was photoactivated in regions of interest within mito-DsRed-expressing mitochondria with a 405 nm laser . The activated fluorescent signal was then collected in confocal image stacks every 5 min for 30 min . Average fluorescence intensities of mitoPA-GFP at each time point was measured in the region of photoconversion , normalized to the mito-DsRed signal , and plotted against time . Images were analysed with the Fiji software package ( https://fiji . sc; Fiji , RRID:SCR_002285 ) . To obtain transmission electron micrographs of mitochondria in dermal fibroblasts , cells were grown on Thermanox coverslips ( Nunc ) and fixed in 2% glutaraldehyde in 0 . 1 M phosphate buffer , pH7 . 4 . Cells were then scraped , harvested by centrifugation , and post-fixed in a solution of 1% osmium tetroxide and 1% potassium hexacyanoferrate ( II ) , dehydrated in acetone and finally epoxiresin embedded . Semi-thin sections were stained with toluidine blue . Thin sections were mounted on copper grids , stained with lead citrate , and examined using a Tecnai G2 80-200kv transmission electron microscope . | Obesity and the diseases associated with it are among the biggest healthcare problems in developed countries . The word obesity means , simply , the accumulation of too much fat tissue in the body , but this ignores growing evidence that fat tissue is highly complex . Fat tissue is important for “mopping up” and storing excess calories safely , but also sends messages to the brain and other organs to report how full the body’s energy stores are . Understanding how fat tissues perform these roles will aid the development of strategies to treat or prevent obesity . A hormone called leptin acts as a signal of the status of the body’s fat stores . High levels of leptin in the blood tell the brain that the body has plenty of fat stored . On the other hand , if the levels of leptin in the blood become very low it tells the brain to prioritize finding food and shut down any nonessential processes . This helps to prevent the body from starving . It is not clear how the production of leptin is controlled , in part because fat tissues in different parts of the body behave very differently . Individuals who have a particular rare genetic mutation accumulate large amounts of fat tissue in their upper bodies and gradually lose fat tissue in their arms and legs . Despite accumulating a lot of fat tissue in the upper body , these individuals have extremely low levels of leptin in their blood . To investigate this genetic condition , Rocha et al . studied two children with the mutation and their healthy parents . The experiments show that this mutation alters a protein called mitofusin 2 , which is found in cell compartments called mitochondria . Mitofusin 2 helps the mitochondria to bind to each other and to other parts of the cell , which is important for the mitochondria to generate the energy needed for vital cell processes . The mitochondria in the fat cells of the children are less closely linked to each other and have an unusual appearance compared to the mitochondria in the parents’ fat cells . Further experiments showed that some genes , including the one that produces leptin , are less active in the children compared to their parents – while other genes that are involved in starvation or stress responses are more active . This work suggests that mitochondria play an important role in regulating the production of leptin . Furthermore , it suggests that leptin or drugs that switch off stress-related genes may have the potential to be used to treat individuals with this particular mutation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology"
] | 2017 | Human biallelic MFN2 mutations induce mitochondrial dysfunction, upper body adipose hyperplasia, and suppression of leptin expression |
Angiopoietin-like proteins ( angptls ) are capable of ex vivo expansion of mouse and human hematopoietic stem and progenitor cells ( HSPCs ) . Despite this intriguing ability , their mechanism is unknown . In this study , we show that angptl2 overexpression is sufficient to expand definitive HSPCs in zebrafish embryos . Angptl1/2 are required for definitive hematopoiesis and vascular specification of the hemogenic endothelium . The loss-of-function phenotype is reminiscent of the notch mutant mindbomb ( mib ) , and a strong genetic interaction occurs between angptls and notch . Overexpressing angptl2 rescues mib while overexpressing notch rescues angptl1/2 morphants . Gene expression studies in ANGPTL2-stimulated CD34+ cells showed a strong MYC activation signature and myc overexpression in angptl1/2 morphants or mib restored HSPCs formation . ANGPTL2 can increase NOTCH activation in cultured cells and ANGPTL receptor interacted with NOTCH to regulate NOTCH cleavage . Together our data provide insight to the angptl-mediated notch activation through receptor interaction and subsequent activation of myc targets .
Human hematopoietic stem and progenitor cells ( HSPCs ) are defined as cells with the ability to self-renew and differentiate into all blood lineages . They provide tremendous therapeutic potential for bone marrow transplantation in the treatment of hematologic malignancy , inherited blood disorders , and cancer chemotherapy . An important goal in studying stem cells is to identify factors that can expand HSPCs in vitro , while maintaining their self-renewal capacity . Scientists have turned to clues during embryonic hematopoietic development to identify novel factors that regulate this expansion . Developmental hematopoiesis occurs in at least two distinct phases , in which an initial transient wave produces mainly primitive erythrocytes and myeloid cells , followed by a definitive wave which produces long-term HSPCs ( Orkin and Zon , 2008 ) . Ontogeny studies across species have identified the aorta-gonad-mesonephros ( AGM ) region as the site from which definitive HSPCs arise ( Dieterlen-Lievre , 1975; Medvinsky and Dzierzak , 1996; Tavian et al . , 1996; Jaffredo et al . , 1998 ) . Recent studies using time-lapse imaging in live zebrafish embryos and live mouse thick tissue sections revealed that HSPCs bud off from the endothelium lining the ventral wall of the developing dorsal aorta ( DA ) prior to entering circulation ( Bertrand et al . , 2010; Boisset et al . , 2010; Kissa and Herbomel , 2010 ) . Subsequent to their birth in the AGM , HSPCs migrate to the fetal liver where they undergo significant expansion in vivo ( Ema and Nakauchi , 2000 ) before colonizing the bone marrow and provide life-long supply of all blood cells . Angiopoietin-like proteins ( ANGPTLs ) were recently identified as growth factors capable of expanding mouse ( Zhang et al . , 2006; Zheng et al . , 2011; Farahbakhshian et al . , 2014 ) and human ( Zhang et al . , 2008; Khoury et al . , 2011; Ventura Ferreira et al . , 2013; Fan et al . , 2014 ) HSPCs in culture . ANGPTLs are secreted proteins that closely resembled the Angiopoietins , which are important vascular regulators , but despite structural similarities , they do not bind to TIE-2 or TIE-1 ( Kim et al . , 1999 , 2000; Oike et al . , 2004 ) . They also exert much wider functions outside of the vasculature such as regulation of lipid , glucose and energy metabolism ( Hato et al . , 2008 ) , inflammation ( Tabata et al . , 2009 ) , and cancer ( Zhu et al . , 2011 ) . Notch receptors ( NOTCH1-4 ) are single-pass type I transmembrane receptors implicated in various developmental and disease processes including the activation of the hematopoietic program . They are synthesized as ∼300 kD full-length precursor proteins that undergo a series of proteolytic cleavages in order to become fully activated . Prior to translocating to the cell surface , NOTCH is first cleaved by furin-like convertases in the trans-Golgi compartment , resulting in a heterodimer composed of the N-terminal extracellular domain and the C-terminal-transmembrane/intracellular domain , bound through noncovalent linkage . Canonical NOTCH activation requires its obligatory interaction with NOTCH ligands belonging to the Delta/Serrate/Jagged/LAG-2 ( DSL ) family ( Fortini , 2009; Kopan and Ilagan , 2009 ) . Several studies support a model in which ligand binding followed by endocytosis creates a mechanical force that alters the conformation of a juxtamembrane NOTCH negative regulatory region ( Kopan and Ilagan , 2009 ) . This permits cleavage of NOTCH by ADAM/TACE ( a disintegrin and metalloprotease/tumor necrosis factor α converting enzyme ) at Site 2 ( S2 ) , generating the membrane-anchored NOTCH extracellular truncation ( NEXT ) fragment . NEXT is a substrate for the γ-secretase complex , which cleaves Site 3 ( S3 ) , releasing the NOTCH intracellular domain ( NICD ) to freely translocate into the nucleus to interact with DNA binding protein CSL ( CBF1/Su ( H ) /Lag-1 ) /RBPjκ and initiate transcription of target genes . While the role of NOTCH in adult HSPC homeostasis still remains controversial , NOTCH is irrefutably important during embryonic hematopoiesis . Several Notch receptors and ligands are expressed in the mouse AGM and deletion of Notch1 , Jagged1 , or CSL resulted in impaired intra-embryonic hematopoiesis ( Kumano et al . , 2003; Robert-Moreno et al . , 2005 , 2008 ) . Notch target genes such as Gata2 ( Minegishi et al . , 2003 ) , Runx1 ( North et al . , 2002 ) and those belonging to the Hairy and Enhancer-of-split related basic helix-loop-helix transcription factors , Hes1 , Hey1 , and Hey2 , are also expressed in the AGM ( Robert-Moreno et al . , 2005 , 2008; Guiu et al . , 2013 ) . Previous studies from our lab demonstrated that HSPC fate is dictated by the notch-runx1 pathway , in which overexpression of runx1 mRNA in the notch mutant mindbomb ( mib ) can partially restore the loss of HSPCs normally observed in mib ( Burns et al . , 2005 ) . Furthermore , recent studies demonstrated an even earlier role for notch in which somite-derived signals such as wnt16 ( Clements et al . , 2011 ) or physical intracellular contacts between the jam1a/2a adhesion proteins ( Kobayashi et al . , 2014 ) can regulate notch signaling in HSC precursors . Because of their potential in hematological applications and therapy , it is important to decipher the molecular pathways on which these ANGPTLs act . Here , we utilized zebrafish genetics to help provide insights into the mechanism by which ANGPTLs can expand adult HSPCs . We found that angptls1 and 2 are indispensible for zebrafish definitive hematopoiesis and that they genetically interacted with notch signaling . To further uncover potential mechanisms for this interaction , we utilized cultured human cells and found that ANGPTL2 mediates NOTCH receptor cleavage/activation , occurring at the level of ANGPTL receptor binding to NOTCH . Our novel findings that angptls can induce notch activation provide an additional layer of regulation of canonical notch signaling .
Angptl2 and 3 are highly expressed in the mouse fetal liver during hematopoietic expansion ( Zhang et al . , 2006 ) but it is not known whether they are important prior to this . To determine the role of zangptls during zebrafish hematopoiesis , we first generated a stable heatshock-inducible transgenic ( Tg ) zebrafish overexpressing full-length zangptl2 cDNA , Tg ( hsp70:zangptl2 ) . Heatshocked embryos had increased zangptl2 mRNA after 2 hr ( Figure 1—figure supplement 1A ) . Definitive hematopoiesis in zebrafish embryos is assessed at 36 hr post-fertilization ( hpf ) , when emerging HSPCs develop in the AGM marked by cmyb and runx1 transcripts ( Burns et al . , 2005; North et al . , 2007 ) . We observed significantly higher number of cmyb- and runx1-positive HSPCs in heatshocked Tg embryos compared to their non-heatshocked or non-Tg siblings by whole mount in situ hybridization ( WISH ) ( Figure 1A ) . Increased cmyb- and runx1-positive HSPCs were also found in the caudal hematopoietic tissue ( or CHT , akin to mammalian fetal liver ) and the pronephros , which matures into the kidney marrow , the mammalian bone marrow equivalent ( Figure 1A ) . Rag1-positive differentiated thymic T-cells that derive from definitive HSPCs were increased in heatshocked Tg siblings ( Figure 1A ) . Together , these results indicate that overexpression of zangptl2 is sufficient to increase zebrafish definitive hematopoiesis in vivo , recapitulating the initial finding that ANGPTL2 can expand HSPCs ex vivo ( Zhang et al . , 2006 ) . 10 . 7554/eLife . 05544 . 003Figure 1 . Angptls are sufficient and required for definitive hematopoiesis . ( A ) Heatshocked Tg ( hsp70:zangptl2 ) embryos have increased cmyb- and runx1-positive HSPCs in the AGM ( black arrows , 36hpf ) , CHT ( white brackets , 3 days post-fertilization , dpf ) , larval kidney ( white arrows , 4dpf ) and rag1-positive T-cells in the thymus ( white arrows , 5dpf ) compared to control no heatshock or non-Tg siblings . ( B ) Angptls-MO morphants ( 2 ng ) had decreased cmyb- and runx1-positive HSPCs ( black arrows ) in the AGM at 36hpf and ( C ) severe disruption to vascular development with loss of kdrl-positive ISVs ( black arrowheads ) , loss of arterial efnb2a and ectopic expression of venous flt4 in the DA ( red arrowheads ) in addition to PCV ( green arrowheads ) at 28hpf . Scale bars: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 00310 . 7554/eLife . 05544 . 004Figure 1—figure supplement 1 . Zangptl2 overexpression in Tg ( hsp70:zangptl2 ) embryos and endogenous zangptl2 expression . ( A ) qPCR analysis of zangptl2 mRNA levels in Tg ( hsp70:zangptl2 ) embryos that have been heatshocked for 1 hr and collected at the indicated times post-heatshock . Heatshocked embryos ( red bars ) overexpressed zangptl2 mRNA at least 100-fold in excess compared to non-heatshocked siblings ( blue bars ) . Error bars denote S . E . M . , *p < 0 . 05 , **p < 0 . 01 compared to 0 hr , one way ANOVA . ( B ) WISH of endogenous zangptl2 at 23hpf ( the highest of all timepoints observed ) is mostly restricted in the yolk sac extension , spinal cord , and head region . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 00410 . 7554/eLife . 05544 . 005Figure 1—figure supplement 2 . Angptls-MO morphants had no defect in primitive hematopoiesis . Top panels: somite matched embryos are scored by double staining with myod ( orange , staining somite boundaries ) and scl ( purple , for early blood and vascular progenitor cells in the anterior ( A ) and posterior ( P ) bilateral stripes of the lateral plate mesoderm ( LPM ) , black arrowheads , 10–12 ss ) . Middle and bottom panels: gata1-positive proerythroblast formation was also unaffected in these morphants in the posterior LPM ( black arrowheads , 10–12 ss ) and intermediate cell mass ( black arrows , 24hpf ) . Scale bars: 200 μm ( for top and middle panels ) ; 50 μm ( for bottom panels ) . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 005 Previous studies demonstrated that zangptl1 and 2 act cooperatively in zebrafish ( Kubota et al . , 2005 ) . We next performed anti-sense knockdown experiments using previously established morpholinos ( MOs ) ( Kubota et al . , 2005 ) and found that while single zangptl1-MO or angptl2-MO can decrease cmyb-positive HSPCs in the AGM ( data not shown ) , knocking down both zangptls ( angptls-MO ) led to a near complete absence of cmyb- and runx1-positive AGM HSPCs at 36hpf ( Figure 1B ) , indicating that zangptl1 and 2 are required for definitive HSPCs formation . In zebrafish , HSPCs arise from specialized kdrl ( mammalian KDR/Flk1 orthologue ) -positive hemogenic endothelial cells in the dorsal aorta ( DA ) ( Bertrand et al . , 2010; Kissa and Herbomel , 2010 ) . Because we observed the highest endogenous expression of zangptl2 at ∼23hpf ( Figure 1—figure supplement 1B ) , before the onset of definitive hematopoiesis , we examined the morphant vasculature at this time point . We found that angiogenic sprouting of kdrl-positive intersegmental vessels ( ISVs ) in control-MO injected ( data not shown ) or uninjected siblings was absent in angptls-MO morphants ( Figure 1C ) . Axial vessel specification was also severely disrupted with decreased arterial efnb2a in the DA and ectopic expression of venous flt4 , also in the DA , normally restricted to the posterior cardinal vein ( PCV ) by 28hpf ( Figure 1C ) . These results suggest that zangptls regulation of definitive HSPC development may occur through an early specification of a patent and functional hemogenic endothelium . To assess whether zangptls can act even earlier during primitive hematopoiesis , we examined angptls-MO morphants at 10–12 ss ( somite stage , equivalent to ∼12–14hpf ) for defects in the bilateral stripes of the lateral plate mesoderm ( LPM ) . Stage and somite-matched uninjected and angptls-MO injected siblings ( marked by myod ) had similar expression of early blood/vascular progenitor transcription factors ( Davidson and Zon , 2004; Dooley et al . , 2005; Zhu et al . , 2005 ) such as scl ( Figure 1—figure supplement 2 ) , lmo2 , and fli1 ( data not shown ) . Furthermore , gata1-positive primitive erythrocytes ( Detrich et al . , 1995 ) also appeared to be unchanged in the posterior LPM or intermediate cell mass at 24hpf ( Figure 1—figure supplement 2 ) , indicating that zangptls1 and 2 are dispensable for primitive hematopoiesis . The angptls-MO morphant phenotypes closely resembled that of the notch mutant , mib ( Lawson et al . , 2001; Itoh et al . , 2003; Burns et al . , 2005 ) , which also exhibited defective definitive hematopoiesis and vascular specification ( Figure 2—figure supplement 1 ) . Mib encodes for the highly conserved E3 ubiquitin ligase important for endocytic processing of notch ligands ( Itoh et al . , 2003 ) . To determine whether angptls and notch genetically interact , we injected angptls-MO into a notch reporter line , Tg ( Tp1bglob:eGFP ) um14 where eGFP is expressed under the control of a notch-responsive element consisting twelve RBPjκ binding sites ( Parsons et al . , 2009 ) . Seen in Figure 2A , angptls-MO morphants had reduced eGFP expression , suggesting that zangptls are required for notch signaling . We then evaluated whether forced expression of constitutively active notch could rescue angptls-MO morphants . We observed that angptls-MO injected into heatshock-inducible double Tg , Tg ( hsp70:Gal4;UAS:NICD ) ( Scheer et al . , 2001; Burns et al . , 2005 ) restored cmyb- and runx1-positive HSPCs ( Figure 2B and data not shown ) . Surprisingly , mib embryos , normally devoid of HSPCs , crossed to Tg ( hsp70:zangptl2 ) also rescued cmyb- and runx1-positive HSPCs ( Figure 2C and data not shown ) at 36hpf upon heatshock . Together , these genetic relationships between zangptls and notch place zangptl signaling downstream of mib but upstream of notch activation/NICD formation during zebrafish definitive hematopoiesis . 10 . 7554/eLife . 05544 . 006Figure 2 . Angptls genetically interact with notch . ( A ) Notch eGFP Tg reporter embryos injected with angptls-MO ( 2 ng ) had decreased notch activity in ISV ( black arrowheads ) and DA ( red arrowhead ) at 28hpf . ( B ) Overexpression of constitutively active notch in angptls-MO injected ( 2 ng ) Tg ( hsp70:Gal4;UAS:NICD ) can restore cmyb-positive HSPCs ( arrows ) at 36hpf after heatshock . ( C ) Mib had no cmyb-positive HSPCs ( arrows ) at 36hpf compared to WT or het siblings . Overexpression of zangptl2 in mib by crossing Tg ( hsp70:zangptl2 ) into mib can restore this defect upon heatshock . Scale bars: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 00610 . 7554/eLife . 05544 . 007Figure 2—figure supplement 1 . Notch mutant mib has hematopoietic and vascular defects . ( A ) Mib has no cmyb- or runx1-positive HSPCs ( arrows ) in the AGM at 36hpf . ( B ) Mib embryos also display vascular defects including disorganized ISV sprouting ( black arrowheads ) , loss of arterial marker , efnb2a in the DA ( red arrowheads ) , and ectopic expression of the venous marker , flt4 in the DA in addition to the PCV ( green arrowheads ) at 28hpf . Scale bars: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 007 While we know that ANGPTLs has direct effects on several mammalian cell types in culture ( Kubota et al . , 2005 ) but because they are secreted factors , the lack of their endogenous receptor identification in zebrafish limits our ability to assess which endogenous cell population they act upon . The mammalian ANGPTL receptor was identified to belong to a large superfamily of leukocyte immunoglobulin-like receptors ( Zheng et al . , 2012 ) . Our attempts at establishing the zebrafish orthologue did not yield any likely candidate , due to the high similarity between the structural and functional domains for all members within this family and is thus beyond the scope of this paper . Despite this , we can extrapolate from whether notch , through which zangptls signal , can exert in a cell autonomous fashion on the endothelial cells or HSPCs during hematopoiesis . To do this , we utilized a transient overexpression system to image the AGM in live embryos by time-lapse spinning disk confocal microscopy . Because HSPCs in the AGM originate from the once kdrl-positive hemogenic endothelium , we first addressed whether notch has an effect on endothelial cells . We microinjected transposon-based vectors ( Tol2 ) containing a draculin ( drl ) promoter driving eGFP and a 6 . 4-kb kdrl promoter driving expression of either constitutively active NOTCH ( NICD , derived from human NOTCH1 ) or eGFP ( as control ) into Tg ( kdrl:Hras-mCherry ) ;casper embryos . The latter is a pan-vascular Tg line ( Chi et al . , 2008 ) that expresses membrane-bound mCherry in all endothelial cells and is bred into the transparent casper ( White et al . , 2008 ) background for ease of imaging without interference from developing melanocytes . Drl is expressed very early ( ∼3 ss ) and marks all blood , vascular and cardiac lineages during zebrafish development ( Mosimann C et al . , submitted ) . Its sustained expression well before and throughout the onset of hematopoiesis would hence serve as a marker for all cells expressing the Tol2 vectors ( GFP+ ) . We screened embryos with similar degree of GFP mosaicism to image from ∼28hpf for 24 hr in the AGM and scored only the number of budding GFP+ cells that lined the ventral floor of the DA ( also mCherry+ ) . In the single frames ( Figure 3A ) of AGM budding time-lapse videos ( Videos 1 and 2 ) , we very rarely observe any AGM budding from GFP+/mCherry+ hemogenic endothelium in the control embryos ( Video 2 ) . In contrast , there was significantly more AGM budding events in embryos injected with constitutively active NOTCH ( Video 1 ) ( ranges for eGFP: 0–2; NICD: 0–6 ) , all of which are depicted in the Figure 3C graph . These results highly suggest that NOTCH can exert a cell autonomous effect on hemogenic endothelial cells destined to become HSPC . 10 . 7554/eLife . 05544 . 008Figure 3 . Notch cell autonomously increase definitive hematopoiesis . ( A ) Time-lapse sequence ( hours:minutes post-28hpf ) of HSPCs emerging from the ventral wall of the DA . Hemogenic endothelial cells ( red ) that have incorporated the injected transgene ( green ) are marked with numbers and white arrows . Each injected embryo was scored for 24 hr and tabulated in ( C ) . Scale bar: 25 μm . ( B ) Still images of the CHT from 72hpf embryos . Runx1-positive HSPCs ( red ) that have incorporated the injected transgene ( green ) were scored for double positivity ( yellow , examples marked by white arrows ) and tabulated in ( D ) . Boxed areas in the top panels are magnified and split into double or single fluorescent panels below . Scale bars: 50 μm . A: Anterior , P: Posterior , D: Dorsal; V: Ventral . Error bars denote S . E . M . , **p < 0 . 01 , compared to eGFP injected controls , Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 00810 . 7554/eLife . 05544 . 009Video 1 . Time-lapse video of AGM budding in embryos transiently expressing constitutively active NOTCH . Representative video depicting the HSPCs budding from the AGM in Tg ( drl:eGFP;kdrl:NICD ) -injected Tg ( kdrl:Hras-mCherry ) ;casper . Hemogenic endothelial cells ( red ) lining the ventral wall of the dorsal aorta that express the injected transgene ( green ) in the top panel depict at least three cell budding events ( numbered with white arrow in the bottom panel showing GFP alone ) . Time-lapse imaging was captured at 10 min/frame and rendered at 3 frames/s . Each frame is a maximum projection of confocal z-stack , cropped from the entire AGM . Scale bar: 25 μm . DA: dorsal aorta; PCV: posterior cardinal vein . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 00910 . 7554/eLife . 05544 . 010Video 2 . Time-lapse video of AGM budding in control injected embryo . Representative video depicting the DA of control Tg ( drl:eGFP;kdrl:eGFP ) injected Tg ( kdrl:Hras-mCherry ) ;casper . In this example , the two cells that contain the control transgene never budded off from the AGM during the 24 hr imaged . Videos were captured on the same day as NICD-injected embryos . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 010 To further delineate whether notch also influences HSPCs , we did similar experiments as above by microinjecting Tol2 vectors containing drl:eGFP and a HSPC-specific Runx1+23 enhancer to drive NICD or eGFP into Tg ( Runx1+23:NLS-mCherry ) ;casper . The Runx1+23 enhancer element was identified in mouse to be highly expressed in the hematopoietic clusters in the AGM but unlike endogenous Runx1 , it is expressed in only a very small subset of the underlying endothelial cells ( Nottingham et al . , 2007 ) . The Tg ( Runx1+23:NLS-mCherry ) Tg line expresses a nuclear localized ( NLS ) mCherry mainly restricted to HSPCs ( Tamplin et al . , 2015 ) . We also detected nearly no mCherry mRNA expression in the hemogenic endothelial cells of Tg ( Runx1+23:NLS-mCherry ) embryos at 36hpf by WISH ( data not shown ) justifying its use to drive expression of our transgenes specifically in budded HSPCs . Newly born HSPCs colonize the CHT starting at ∼48hpf and expand tremendously until 80hpf in vivo . For these reasons , we scored the total number of transgene-positive ( GFP+ ) and mCherry+ cells in the CHT at 72hpf to determine whether NOTCH could cell autonomously expand budded HSPCs . Shown in Figure 3B , we can readily observe significantly higher number of double positive HSPCs in which NOTCH is constitutively expressed compared to the control eGFP transgene injected siblings . These results are tabulated in Figure 3D and strongly indicated that NOTCH can exert cell autonomous effects on HSPCs . From these results in zebrafish , even though we cannot conclude whether zangptls themselves act cell autonomously on both endothelial cells and HSPCs , their downstream effector notch act cell autonomously . To further assess the molecular signaling downstream of zangptls and notch , we examined a known target of notch , myc , and its relationship with respect to zangptls . The MYC proto-oncogene has been previously shown to be a direct NOTCH target in T-cell acute lymphoblastic leukemia ( T-ALL ) ( Palomero et al . , 2006; Weng et al . , 2006 ) . Moreover , Myc is also critical in the regulation of hematopoietic/vascular development ( Wilson et al . , 2004; Laurenti et al . , 2008 ) . Conditional cMyc knockout in hematopoietic lineages resulted in severe cytopenia at E11 . 5 and lethality at E12 . 5 in mice ( He et al . , 2008 ) , and Myc-deficient HSPCs are functionally defective , unable to engraft in recipient mice ( Wilson et al . , 2004 ) . To first establish the role of myc during zebrafish definitive hematopoiesis , we performed morpholino knockdown of myc and observed a significant reduction in cmyb-positive HSPCs in the AGM ( Figure 4A ) , suggesting its requirement . Next , to determine whether myc is downstream of notch during this process , we overexpressed myc in mib embryos and found partially restored cmyb-positive HSPCs in the AGM ( Figure 4B ) . Our observed genetic interaction between notch and angptls in Figure 2 prompted us to performed parallel myc rescue experiments in angptl-MO-injected morphants . We found similar restoration of cmyb-positive HSPCs in the AGM ( Figure 4C ) , thereby implying that myc maybe one of the underlying effectors downstream of zangptl signaling . Together , these results further strengthened our hypothesis that angptls may exert their effects through notch signaling to converge upon myc during zebrafish definitive hematopoiesis . 10 . 7554/eLife . 05544 . 011Figure 4 . Myc is downstream of angptl and notch signaling . ( A ) Myc MO ( 6 ng ) injected embryos had decreased cmyb-positive HSPCs ( arrows ) at 36hpf . ( B ) Mib embryos injected with myc mRNA ( 100 pg ) had restored cmyb-positive HSPCs ( arrows ) at 36hpf compared to siblings injected with control mCherry mRNA ( 100 pg ) . ( C ) Overexpression of myc mRNA ( 100 pg ) in angptls-MO ( 2 ng ) morphants also had restored cmyb-positive HSPCs ( arrows ) at 36hpf compared to control siblings injected with mCherry mRNA ( 100 pg ) . Control-MO: non-targeting MO ( 2 ng ) . ( D ) Representative GSEA plot that positively correlate ANGPTL2-stimulated human CD34+ gene expression and MYC targets . ( E ) GSEA plot showing significant correlation between MYC target genes from NOTCH ChIPseq and expression data of ANGPTL2-stimulated CD34+ cells . NES , Normalized Enrichment Score , FDR q value , False Discovery Rate q value . Scale bars: 50 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 01110 . 7554/eLife . 05544 . 012Figure 4—figure supplement 1 . NOTCH ChIP-seq in ANGPTL2-stimulated CD34+ cells . ( A ) Enriched transcription factor motifs at the peaks of NOTCH binding sites from NOTCH ChIP-seq . Parentheses indicate E values for the significance of the motif . An embedded RBPjκ ( ( consensus motif ( red box ) was found within the ZNF143 motif as previously described ( Wang et al . , 2011 ) . ( B ) Examples of MYC-related categories from GREAT analysis of NOTCH ChIP-seq . ( C ) IPA network analysis of NOTCH ChIP-seq bound genes revealed a dense network of target genes centered on MYC . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 012 Our genetic data in zebrafish point to the possibility that zangptls signaling interacts with notch . We next carried out experiments in cultured cells to help dissect the mechanisms through which zangptls may act . First , to examine the molecular signaling downstream of ANGPTL , we performed gene expression studies on ANGPTL2-stimulated human CD34+ cells using the Human Exon 1 . 0 ST Arrays . More than 3700 genes were differentially regulated with significant q value <0 . 05 ( False Discovery Rate ( FDR ) , Benjamini-Hochberg ) . In order to assess the biological significance of this signature , it was compared to 5562 gene sets in the Broad Molecular Signature Database using Gene Set Enrichment Analysis ( GSEA ) . 284 gene sets showed a highly significant positive enrichment score ( ES ) with FDR<0 . 05 , indicating that these sets contained genes that positively correlated with our expression phenotype , that is , upregulated in ANGPTL2-treated cells ( Supplementary file 1 ) . The top 20 gene sets were then chosen to perform a Leading Edge Analysis , in which the most significant subset of contributing genes from each gene set was cross-compared . We identified several clusters of genes/gene sets that contained direct MYC targets , contained a MYC signature , contained MYC/MAX binding motif , or were upregulated when MYC was upregulated . Despite a wide range of disease state responses or experimental parameters represented in these 284 gene sets , the strongest correlation was found with those relating to MYC signaling , an example of which is shown in Figure 4D . In concordance with the GSEA analysis , the ANGPTL2 signature was analyzed using Ingenuity Pathway Analysis ( IPA ) , and MYC was again predicted to be one of the top upstream regulators in an active state ( p-value of 2 . 89E-16 ) . The compelling results from these bioinformatics analyses provided independent evidence to corroborate with our genetic data above , placing myc downstream of zangptls signaling during zebrafish hematopoiesis ( Figure 4C ) . To examine the effect of MYC downstream of NOTCH signaling , we performed chromatin immunoprecipitation using a NOTCH-specific antibody ( Wang et al . , 2011 ) followed by sequencing ( ChIP-seq ) in human CD34+ cells stimulated with ANGPTL2 . Similar to the previously published ChIP-seq data ( Wang et al . , 2011 ) , we found enrichment for ETS , RUNX1 , and ZNF143 motifs , the latter containing an embedded consensus sequence for CSL/RBPjκ ( indicated by the red box in Figure 4—figure supplement 1A ) , in the sites bound by NOTCH . GREAT ( Genomic Regions Enrichment of Annotations Tool ) analysis revealed in some instances , that significantly more regions in our NOTCH ChIP-seq data set fall within the regulatory domains of MYC target genes or genes that are involved in the regulation of MYC targets ( Figure 4—figure supplement 1B ) . Further pathway analyses using IPA revealed MYC-related genes forming the second highest ranking network as well as placing MYC as a significant upstream regulator in an active state ( Figure 4—figure supplement 1C ) , again implying that those gene regions that were bound by NOTCH are indicative of MYC being positively regulated . GSEA analysis comparing the ANGPTL2 microarray signature to the NOTCH ChIP-bound MYC targets showed a strong and significant enrichment ( Figure 4E ) , suggesting that the gene expression of MYC targets identified from the ANGPTL2 microarray overlapped significantly with those bound by NOTCH . Together with our genetic data and the fact that these informatics analyses revealed converging signaling on MYC activation by ANGPTL2 and NOTCH points to the likelihood that ANGPTL2 signaling may affect NOTCH activation . To examine whether ANGPTLs have direct effects on NOTCH , we stimulated human CD34+ progenitor cells with ANGPTL2 and observed a rapid increase in NOTCH receptor cleavage , generating the product NICD without affecting total levels of full-length NOTCH receptor ( Figure 5A ) . Upon NOTCH cleavage , NICD can translocate into the nucleus to initiate transcription of target genes such as HES1 , RUNX1 , and cMYC . First , to assess transcriptional activity of HES1 , we treated cells that were transfected with the NOTCH-responsive luciferase plasmid , HES1-luc , and saw a significant increase in luciferase activity upon ANGPTL2 stimulation ( Figure 5—figure supplement 1A ) . Furthermore , we found significant increase in HES1 mRNA level by qPCR during ANGPTL2-stimulation time course ( Figure 5B blue bars ) . Pretreating these cells with DAPT was able to suppress these increases ( Figure 5B red bars ) , implying that this is NOTCH-dependent . Similarly , we also detected significant increases in RUNX1 and cMYC mRNA upon ANGPTL2 treatment that are again , NOTCH dependent ( Figure 5—figure supplement 1B ) . These results indicate that ANGPTL2 can rapidly activate NOTCH as a result of induced NOTCH receptor cleavage to initiate downstream transcription of the NOTCH target genes . 10 . 7554/eLife . 05544 . 013Figure 5 . ANGPTL2 activates NOTCH by receptor cleavage . ( A ) Human CD34+ cells were stimulated with ANGPTL2 ( 1 μg/ml ) or PBS vehicle and Western blotted for S3 cleaved ( NICD ) and full-length NOTCH receptor . β-actin was used as loading control . Ratios indicated below the Western blot represent densitometry of the band intensity normalized to loading control . ( B ) Human endothelial cells pre-treated for 4 hr with DMSO vehicle or DAPT ( 20 μM ) assayed for HES1 by qPCR . Error bars denote S . E . M . , p < 0 . 0001 , two way ANOVA; ***p < 0 . 001 , ****p < 0 . 0001 compared to 0 hr and ††p < 0 . 0001 compared to DMSO , Bonferroni post-hoc test . ( C ) Human endothelial cells pre-treated with DMSO , DAPT ( 20 μM ) , TAPI-2 ( 20 μM ) , or SB-3CT ( 30 μM ) prior to stimulating with ANGPTL2 ( 1 μg/ml ) or PBS ( 1 hr ) and Western blotted for S2 cleaved ( NEXT ) , S3 cleaved ( NICD ) , or full-length NOTCH . β-actin was used as loading control . Ratios indicated below the Western blots represent densitometry of the band intensity normalized to loading control . ( D ) NICD degradation experiment: post-treatment of ANGPTL2-stimulated cells ( 1 μg/ml for 1 hr ) with DMSO or DAPT for 1 , 2 , or 4 hr to prevent de novo formation of NICD product , assayed by Western blotting . All experiments were repeated at least three times . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 01310 . 7554/eLife . 05544 . 014Figure 5—figure supplement 1 . ANGPTL2 activates NOTCH signaling . ( A ) K562 cells assayed for HES1 transcriptional activity by measuring the transfected HES1-responsive firefly luciferase activity normalized to the co-transfected internal control of Renilla luciferase activity after ANGPTL2 ( 1 μg/ml ) stimulation . Error bars denote S . E . M . , *p < 0 . 05 compared to 0 hr ( white bar ) , Student's t test . ( B ) Similar qPCR performed as in Figure 5B for RUNX1 or MYC mRNA . Error bars denote S . E . M . , *p < 0 . 05 , **p < 0 . 01 compared to PBS DMSO vehicle control , Student's t test . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 014 NOTCH ectodomain shedding at S2 is highly regulated and thought to be key in NOTCH activation . Because the rate-limiting S2 cleavage step gives rise to the transient product , NEXT , which is then very rapidly cleaved by γ-secretase at S3 to generate the terminal fragment , NICD , we sought to determine whether ANGPTL2 can regulate S2 cleavage . To better detect NEXT accumulation , we treated ANGPTL2-stimulated human endothelial cells with the γ-secretase inhibitor , DAPT to block S3 cleavage to allow NEXT buildup . Endothelial cells have a more robust NOTCH response and because we hypothesized that angptls can regulate hemogenic endothelial cells in zebrafish , we felt that using human endothelial cells maybe more suitable to use in our studies . As observed in Figure 5C , ANGPTL2 stimulated an increase in NICD in endothelial cells ( as in CD34+ cells , Figure 5A ) in vehicle ( DMSO ) -treated cells and this increase was blocked by DAPT . Blocking upstream of S3 at S2 using two different inhibitors of S2 cleavage ( TAPI-2 , a broad spectrum ADAM and TACE inhibitor; and SB-3CT , a non-competitive inhibitor for TACE ) results in inhibiting NEXT and NICD formation . In control DMSO-treated cells , ANGPTL2 also stimulated the generation of NEXT fragment and this was further enhanced by treatment with DAPT , thus establishing that ANGPTL2 may regulate NOTCH activation via S2 cleavage . The activation of Notch receptors triggers a cascade of downstream signaling and requires highly regulated mechanisms to efficiently turn off via proteasomal degradation of NICD in order to avoid deleterious effects ( Kopan and Ilagan , 2009 ) . Thus , to eliminate the possibility that the ANGPTL2-mediated increase in NICD products is due to obstructed NICD degradation , we post-treated ANGPTL2-stimulated cells with DAPT to prevent further cleavage of NOTCH receptor . If the increase in NICD product following ANGPTL2 treatment is a result of NICD degradation blockade and not a result of NOTCH receptor cleavage as we postulated above , it follows that post-ANGPTL2 treatment with DAPT should inhibit any de novo NOTCH receptor cleavage resulting in sustained levels of NICD . In Figure 5D , we first measured NICD protein levels and found that ANGPTL2 stimulated an increase in NICD as previously observed ( Figure 5A , C ) . At 1 hr post-ANGPTL2 treatment , cells were treated with either DMSO ( blue bar ) or DAPT ( red bar ) . Though NICD level continued to remain high in the DMSO group , it started to decrease starting at 2 hr in the DAPT group suggesting that blocking de novo S3 cleavage was sufficient to bring ANGPTL2-stimulated NOTCH activation down to a basal state ( at 4 hr of DAPT ) . We can conclude from this study that the rise in NICD levels induced by ANGPTL2 is a result of NOTCH receptor cleavage as opposed to interference with NICD degradation . The human leukocyte immunoglobulin ( Ig ) -like receptor B2 ( LILRB2 ) was identified to be the ANGPTLs receptor ( Zheng et al . , 2012 ) . Typical of the Ig superfamily , of which includes more than 850 members , LILRB2 contains 4 Ig domains in its extracellular domain ( ECD ) , a transmembrane domain ( TM ) , and 3 immunoreceptor tyrosine-based inhibitory motifs ( ITIMs ) in its intracellular domain ( ICD ) . Since proper compartmentalization of membrane receptors with signaling molecules is critical for coordinating and eliciting downstream cascades , we hypothesized that for ANGPTL2 to facilitate NOTCH cleavage , this interaction needs to occur proximal to NOTCH . To test this , we first transfected HEK293T cells with full-length LILRB2 and/or NOTCH1 and cell lysates were immunoprecipitated ( IP ) for NOTCH or LILRB2 . Shown in Figure 6A , LILRB2 is bound to NOTCH and this interaction is increased upon ANGPTL2 stimulation . To assess endogenous interaction between LILRB2 and NOTCH , we performed co-IPs in human CD34+ cells and observed similarly ( Figure 6B , C ) . To control for specificity of our co-IPs , we immunoprecipitated the highly abundant transmembrane protein , VE-Cadherin , in human endothelial cells and do not observe neither NOTCH nor LILRB2 association ( Figure 6—figure supplement 1A , B ) . 10 . 7554/eLife . 05544 . 015Figure 6 . ANGPTL2 receptor , LILRB2 , interacts with NOTCH in cis . ( A ) Cell lysates from ANGPTL2-stimulated HEK293T cells transfected with GFP control , full-length LILRB2 and/or NOTCH1 plasmids were IP for NOTCH1 or LILRB2 . Note the interaction between both receptors and this association increased upon ANGPTL2 ( 1 μg/ml , 1 hr ) stimulation . Whole cell lysates were used for loading control . ( B and C ) Cell lysates from human CD34+ cells were co-IP for endogenous NOTCH1 , endogenous LILRB2 or IgG isotype and Western blotted for associated LILRB2 or NOTCH1 , respectively . ( D ) Cell lysates from NOTCH1 and HA-tagged LILRB2 truncation mutant ( schematics ) transfected cells were co-IP for NOTCH1 and Western blotted for HA . NOTCH1 interaction was only observed with full-length LILRB2 or LILRB2-ECD . ( E ) Stable lentiviral knockdown of LILRB2 using two different sequences . Partial knockdown of LILRB2 was observed compared to the control scrambled shRNA ( sh-CT ) . The interaction between LILRB2 and NOTCH1 was also decreased ( top panels ) . The ANGPTL2-induced NOTCH receptor cleavage in sh-CT cells was abolished in sh-LILRB2 cells . Whole cell lysates were done for loading control . All experiments were repeated at least 3 times . Ratios indicated below the Western blots represent densitometry of the band intensity . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 01510 . 7554/eLife . 05544 . 016Figure 6—figure supplement 1 . NOTCH interacts with LILRB2 in endothelial cells . Angptl2 ANGPTL2-stimulated ( 1 μg/ml ) human endothelial cell lysates were immunoprecipitated for ( A ) LILRB2 , ( B ) NOTCH , IgG isotype , or VE-Cadherin ( VEC , a highly abundant surface transmembrane protein ) and Western blotted for NOTCH , LILRB2 , and VEC to demonstrate antibody specificity for co-IP . Experiment repeated at least three times . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 01610 . 7554/eLife . 05544 . 017Figure 6—figure supplement 2 . NOTCH and LILRB2 interact in cis . Similar experiments as those performed in Figure 6D except cells were either singly transfected ( left and right lanes ) or doubly transfected ( middle lanes ) with LILRB2 ECD , LILRB2 ICD , and/or NOTCH1 . Those that are singly transfected were then trypsinized before mixing and replating to exclude co-expression of LILRB2 and NOTCH within a single cell ( right lanes ) . Cell lysates were co-IP for NOTCH and Western blotted for HA . Only the doubly transfected LILRB2 ECD and NOTCH showed interaction between the two ( middle lanes ) . Whole cell lysates were used as loading controls . All experiments repeated at least 3 times . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 017 Next , to elucidate whether this interaction occurs in cis or in trans , that is , whether LILRB2 interacts with NOTCH on the same cells or in neighboring cells , we made either N-terminal or C-terminal HA-tagged full-length LILRB2 , LILRB2 extracellular domain ( ECD ) , or LILRB2 intracellular domain ( ICD ) mutants , all of which still retained their TM domain for proper membrane localization ( Figure 6D schematics ) . They are then singly or doubly transfected with NOTCH receptor into HEK293T cells prior to co-IP with NOTCH antibody and subsequently Western blotted for HA . We found that only the full-length or LILRB2 ECD can interact with NOTCH suggesting that this interaction occurs extracellularly ( Figure 6D ) . Moreover , we also performed a replating experiment where singly transfected HEK293T cells ( LILRB2 ECD or ICD truncations , or NOTCH ) were trypsinized one day after transfection , and each plate of LILRB2 truncation cells was mixed with NOTCH transfected cells before being replated and allowed to grow to confluency . These cells were then processed for the same co-IP/Western blotting . We could not detect any interaction after replating ( right lanes ) while still being able to detect LILRB2 ECD interaction with NOTCH in double transfections ( middle lanes ) , thus suggesting that the LILRB2 and NOTCH interaction likely occur in cis , that is , within the same cells extracellularly ( Figure 6—figure supplement 2 ) . Now that we have established physical interaction between LILRB2 and NOTCH , we investigated whether ANGPTL2-mediated LILRB2 activation is responsible for NOTCH receptor cleavage . We made 2 stable knockdown cell lines that were lentivirally transduced for short hairpin RNA ( shRNA ) against LILRB2 and performed co-IP for LILRB2 . Even though we could only observe a modest knockdown of LILRB2 ( Figure 6E , second blot , with sh-LILRB2-2 being more effective than sh-LILRB2-1 ) , we saw that in contrast to the control scrambled shRNA transduced cells ( sh-CT ) , which still retained the ANGPTL2-mediated NOTCH cleavage , sh-LILRB2 transduced cells had lost this response with no ANGPTL2-induced NICD formation ( Figure 6E , third blot ) . Furthermore , the interaction between LILRB2 and NOTCH receptors also decreased in sh-LILRB2 cells ( Figure 6E , first blot ) . Together , these studies demonstrate that ANGPTL2-induced NOTCH activation may occur through a ligand-regulated recruitment of LILRB2 to NOTCH into the same microdomain . Furthermore , this novel finding provides a plausible means by which ANGPTL2 can assist in NOTCH cleavage . The cleavage leads ultimately to an activation of MYC target genes that are involved in stimulating endothelial and hematopoietic cells .
Previous studies have demonstrated that notch is indispensable during the developmental specification of HSPCs ( Burns et al . , 2005; Clements et al . , 2011 ) . In our present study , we defined the essential role of zangptl1 and 2 during developmental hematopoiesis ( Figure 1 ) , likely through regulation of notch signaling in the hemogenic endothelium . To our knowledge , this is the first time zangptls has been shown to activate notch . The current model for notch specification of HSPCs during development has mostly been suggested to be a cell-autonomous one ( Hadland et al . , 2004; Burns et al . , 2005; Robert-Moreno et al . , 2005 ) . Our live imaging data from transient Tg ( drl:eGFP;kdrl:NICD ) -injected embryos resulting in significantly more budding HSPCs in the AGM of DA corroborate with this . More recently , studies have revealed an additional and separate role for somite-derived notch signaling during HSPC specification that occurs much earlier than previously thought ( Clements et al . , 2011; Kobayashi et al . , 2014 ) . Based on our genetic data in zebrafish in which we observed a tight genetic interaction between zangptls and notch ( Figure 2 ) , we believe that zangptls exert their effects on definitive hematopoiesis during the vascular specifications of the axial vessels including the DA , when it coincides with the highest endogenous expression of zangptl2 ( Figure 1—figure supplement 1B ) . Though it remains to be determined , information on the spatial/temporal expression pattern of the yet unidentified zangptl receptor would solidify this premise . There are receptors in zebrafish with similar domains as the mammalian receptor , but there is no obvious orthologue . Nevertheless , our current proposed model ( Figure 7A ) extrapolated from our genetic data places zangptl2 downstream of mib as Tg ( hsp70:zangptl2 ) crossed into mib can rescue its hematopoietic defect ( Figure 2C ) . Zangptl2 acts upstream of notch activation , that is , generation of NICD , as forced expression of NICD from Tg ( hsp70:Gal4;UAS:NICD ) can rescue angptls-MO morphant phenotype ( Figure 2B ) . Interestingly , loss of notch signaling in angptls-MO-injected notch reporter embryos ( Figure 2A ) indicates that zangptls are required for notch signaling but also raises questions as to whether notch ligands needed to be present for zangptls to regulate notch activation . One of the notch ligands , deltaC , has considerable overlapping expression both spatially and temporally with notch in the developing DA , suggesting that it may play a role during the formation of the hemogenic endothelium and the subsequent generation of HSPCs . DeltaC expression , normally absent in the DA of mib was similarly rescued in Tg ( hsp70:zangptl2 ) ;mib ( data not shown ) . This restoration of notch ligand in the mib by zangptl2 overexpression was intriguing , and it remains to be established whether this restored notch ligand is functional in the absence of mib . 10 . 7554/eLife . 05544 . 018Figure 7 . Proposed models for ANGPTL-mediated NOTCH signaling . ( A ) Notch signaling in zebrafish is believed to be important for HSPC specification in the hemogenic endothelium on the ventral side of the developing dorsal aorta ( DA ) . A simplified view of canonical notch signaling starts with mib endocytic processing of notch ligands to potentiate their ability to activate notch , depicted in the wild-type ( WT ) model . Notch receptors interact with ligands in the neighboring cell and this leads to subsequent cleavage of notch to release NICD to translocate into the nucleus and initiate transcription of target gene like myc . In the mib mutant however , the lack of signal from notch ligands prevents further downstream signaling from the receptor and HSPCs are not formed . From our genetic interaction studies , we believe that zangptl2 is downstream of mib ( Figure 2C ) but upstream of NICD generation ( Figure 2B ) . ( B ) Using cultured cells to further explore the mechanism by which ANGPTL2 can regulate NOTCH signaling , we found that ANGPTL2 can stimulate NOTCH receptor cleavage at S2 and S3 , leading to transcription of target genes . This was dependent on the ANGPTL receptor , LILRB2 , as ANGPTL2 can induce recruitment of LILRB2 to NOTCH . DOI: http://dx . doi . org/10 . 7554/eLife . 05544 . 018 To examine potential downstream signaling from angptls and notch , we looked at the previously identified target of notch , myc . The relationship between NOTCH and MYC has been best studied in T-ALL cells whereby MYC transduces growth and survival signals to many NOTCH-dependent T-ALL cell lines and forced ectopic expression of MYC can restore the leukemogenic signals for these cells in instances when NOTCH is inhibited ( Palomero et al . , 2006; Weng et al . , 2006; Chan et al . , 2007 ) . However , myc has not been studied in zebrafish hematopoiesis . We found that not only is myc required during HSPC formation in the AGM ( Figure 4A ) , overexpression of exogenous myc in embryos lacking zangptls ( angptls-MO ) or notch ( mib ) signaling could similarly rescue AGM defects ( Figure 4C , B ) . This puts myc downstream of zangptls and notch signaling . Through independent gene expression analysis , we found MYC to be highly activated upon ANGPTL2 stimulation in human CD34+ progenitor cells ( Figure 4D ) . Furthermore , we ran GSEA comparing our ANGPTL2-induced gene expression signature to the set of MYC target genes whose regulatory elements were bound by NOTCH from our ChIP-seq data and found significant overlap ( Figure 4E ) . This implies that the set of MYC targets downstream of ANGPTL2 and NOTCH signaling is highly similar , supporting our hypothesis that ANGPTL may signal partially through NOTCH . To test this hypothesis , we utilized cultured mammalian cells to directly investigate the mechanism by which ANGPTL2 may act on NOTCH . We first discovered that purified ANGPTL2 ligand added to cells can rapidly induce NOTCH activation , measurable using three different assays: a luciferase reporter assay ( Figure 5—figure supplement 1A ) , qPCR of known NOTCH target genes ( Figure 5B and Figure 5—figure supplement 1B ) , and Western blotting for the NOTCH cleavage product NICD ( Figure 5A ) . Of particular note , when we used different drugs that block certain cleavage events , we found that ANGPTL2-stimulated S2 cleavage of NOTCH ( Figure 5C ) . Interestingly , this step occurs at the cell surface when NOTCH engages with its ligands . Although the precise mechanism in which ANGPTL2 mediates NOTCH cleavage is subject to further investigation , our assumption that this level of regulation must occur in close proximity led to our surprising discovery that human ANGPTL2 receptor , LILRB2 , can interact with NOTCH ( Figure 6A–D and Figure 6—figure supplement 1 , 2 ) . These preliminary results from co-IPs certainly suggest that ANGPTL signaling occurs in very close proximity to NOTCH . Such co-IP experiments do not definitively signify that there is direct physical interaction and do not address the stoichiometry of this potential interaction . Nevertheless , ANGPTL2-mediated NOTCH cleavage appears to be dependent on LILRB2 signaling , as sh-LILRB2 knockdown resulted in diminished NICD generation ( Figure 6E ) . This demonstrates a functional interaction . A more thorough mapping of the interactions between these two surface receptors will shed light on the mechanism by which ANGPTL2 can regulate NOTCH activation . An attractive speculation as to why this interaction may occur would be to potentially facilitate or even poise the NOTCH receptor for more efficient cleavage , as depicted in the model proposed in Figure 7B . This also raises an intriguing possibility of whether LILRB2 can physically alter NOTCH interaction with its ligand or whether ANGPTL2 signaling can influence NOTCH ligand availability . Indeed several studies have suggested that Notch receptors oligomerize at the cell surface at sites of DSL ligand contact from neighboring cells ( Luty et al . , 2007; Nichols et al . , 2007 ) . The recruitment of LILRB2 to NOTCH may facilitate this clustering process although the stoichiometry of LILRB2 to NOTCH remains to be established . Various alternative non-DSL type ligands have been identified to activate NOTCH , such as the adhesion molecule F3/Contactin ( Hu et al . , 2003 ) , EGF-repeat factor DNER ( Eiraku et al . , 2005 ) , EGF-like domain 7 ( Schmidt et al . , 2009 ) , and Microfibrillar associated glycoproteins ( MAGP ) 1 and 2 ( Miyamoto et al . , 2006 ) . Interestingly , a related family member to the latter , MFAP4 was found to be capable of expanding mouse HSPCs ex vivo along with ANGPTLs 2 and 3 ( Zhang et al . , 2006 ) . Our study does not propose ANGPTL2/LILRB2 as a non-canonical NOTCH ligand but offers an additional layer of regulation on NOTCH activation which may not be previously appreciated . Previous report on the discovery of ANGPTL receptor , LILRB2 , suggested downstream activation of CAMK ( calcium/calmodulin-dependent protein kinase ) -2 and -4 and subsequent recruitment of SHP-2 ( a Src homology 2 domain containing non-transmembrane protein tyrosine phosphatase ) in freshly isolated mouse spleen cells ( Zheng et al . , 2012 ) . CAMK2 has been shown to activate NOTCH in human prostate cancer cells ( Mamaeva et al . , 2009 ) whereas SHP-2 has also been shown to genetically interact with notch signaling in Drosophila ( Oishi et al . , 2006 ) . Although the answer as to whether CAMKs and SHP-2 may directly or indirectly regulate NOTCH is still yet to be determined , this suggests that LILRB2 signaling may extend beyond the models presented in our studies . The ability of ANGPTLs to regulate the local events at which NOTCH cleavage occurs creates new opportunity for therapeutic intervention . The role of Notch during HSPC renewal at homeostasis is controversial ( Bigas et al . , 2010 ) . Recent studies demonstrate that endothelial-derived Notch signal can stimulate murine cKit+Sca1+Lin− cells expansion ( Butler et al . , 2010 ) . Delta1ext−IgG , in which the engineered activating NOTCH ligand is tethered to the Fc domain of human IgG1 , was developed for ex vivo expansion of human CD34+ cord blood ( CB ) progenitors ( Delaney et al . , 2005 ) . More recently , Delta1ext−IgG-expanded HSPCs showed great clinical promise in rapid hematopoietic engraftment with shortened time to neutrophil recovery ( Delaney et al . , 2010 ) . In support of an interaction between the angptls and notch , preliminary analysis of human CD34+ CB progenitors treated with ANGPTL2 or Deltaext−IgG had comparable expansion in CD34+/CD90 ( Thy-1 ) lo populations . Furthermore , these expansions were effectively abrogated when cells were treated with a combination of NOTCH1 and NOTCH2 blocking antibodies , suggesting that they were at least partially NOTCH mediated . This supports that the ANGPTLs and NOTCH can interact in another cell type such as human cord blood CD34+ cells . In conclusion , our studies of ANGPTL-mediated activation of NOTCH presented here have uncovered an important genetic , physical , and functional interaction of these two signaling pathways that are critical for hematopoiesis .
Zebrafish were maintained in accordance to Boston Children's Hospital Animal Research Guidelines . The following mutants and transgenic zebrafish were used in the study: mindbombta56b ( Jiang et al . , 1996 ) , Tg ( UAS:NICD ) ( Scheer et al . , 2001 ) , Tg ( hsp:70:Gal4 ) ( Scheer et al . , 2001 ) , Tg ( Tp1bglob:eGFP ) um14 ( Parsons et al . , 2009 ) , Tg ( kdrl:Hras-mCherry ) ( Chi et al . , 2008 ) , or Tg ( Runx1+23:NLS-mCherry ) ( Tamplin et al . , 2015 ) crossed into the casper ( White et al . , 2008 ) background and Tg ( hsp70:zangptl2;cmlc2:DsRed2 ) . The latter was generated by co-injecting linearized plasmids: one with a 1 . 5 kb heatshock protein 70 promoter ( pzhsp70/4prom ) driving full-length D . rerio angptl2 ( provided by Y Kubota and T Suda ) and the other containing the 5 . 1 kb cardiac myosin light chain promoter 2 ( Rottbauer et al . , 2002 ) driving nuclear DsRed2 . A stably integrated transgenic line containing both transgenes was maintained , and transgenic embryos were identified with DsRed2-positive heart . To induce heatshock overexpression of zangptl2 , embryos at the 10–12 ss were immersed in E3 fish water and heated to 38°C for 30 min to 1 hr . Similarly , to induce heatshock overexpression of NICD , embryos from Tg ( UAS:NICD ) and Tg ( hsp:70:Gal4 ) crosses were heatshocked at 39°C for 20 min . Embryos were quickly washed with room temperature E3 to stop heatshock and allowed to develop normally until the appropriate stage before fixation with 4% paraformaldehyde and processed for WISH . Embryos from at least three separate breeding clutches were scored for each experiment and tabulated together for the observed phenotype . The results are represented as a single ratio of those depicted in each panel to the total number of embryos scored . In instances where a genetic cross were done , we scored within each breeding clutch , the different genotypes first , then the observed phenotype . The denominator in the ratios indicated in each panel represents the total number of embryos scored within that genotype . After the WISH pictures were taken , all transgenic and mutant embryos had their genotype confirmed by PCR . Genotyping PCR primers for mib-F: GGTGTGTCTGGATCGTCTGAAGAAC , mib-R: GATGGATGTGGTAACACTGATGACTC , UAS:NICD-F: CATCGCGTCTCAGCCTCAC , UAS:NICD-R: CGGAATCGTTTATTGGTGTCG , hsp:70:Gal4-F: GCAATGAACAGACGGGCATTTAC , hsp:70:Gal4-R: CTTCAGACACTTGGCGCACTTCGG , hsp:70:zangptl2-F: CAGAGAAACTCAACCGAAGAGAAGCGAC , hsp:70:zangptl2-R: GCTCCTGTAACCTTCTGCTGGGGTA , cmlc2:DsRed2-F: TGTATTTAGGAGGCTCTGGGTGTC , and cmlc2:DsRed2-R: CTTCTTGTAGTCGGGGATGTCG . Each experiment was repeated at least twice . WISH was performed as described ( Thisse and Thisse , 2008 ) . Previously published morpholinos ( Gene Tools , Philomath , OR ) used to knockdown angptl1 and angptl2 were targeted to the start codons or 5′ UTR ( Kubota et al . , 2005 ) . Additional morpholinos used to confirm knockdown was generated as splicing morpholinos with the sequences: angptl1: 5′-CCTGTGGAAAATGCAGAGAAATGCA-3′ and angptl2: 5′-GAGGTTTTTCTTGTGGCTCACCTTA-3′ . Control non-targeting morpholino sequence was 5′-CCTCTTACCTCAGTTACAATTTATA-3′ . Myc morpholino sequence was 5′-GTGGTAAAAGCTGAATGAACACTGA-3′ . mRNA for D . rerio myc ( provided by A Gutierrez ) was synthesized using the mMESSAGE mMACHINE kit ( Life Technologies , Grand Island , NY ) per manufacturer's protocol . For every experiment , equal amounts of morpholinos and/or mRNAs are always controlled for by injecting the non-targeting control morpholinos and/or mCherry mRNA into sibling embryos at 1 cell stage as previously described ( Burns et al . , 2005 ) . RNA was extracted from whole embryos ( average of 50 embryos per condition , three clutches of embryos per condition ) or cells ( average of 106 cells , triplicates per experiment ) using TRIzol ( Life Technologies , Grand Island , NY ) or RNeasy miniprep ( Qiagen , Germantown , MD ) respectively , followed by DNaseI ( Qiagen , Germantown , MD ) digestion and RNeasy ( Qiagen , Germantown , MD ) cleanup . cDNA was synthesized using SuperScript VILO cDNA Synthesis Kit ( Life Technologies , Grand Island , NY ) . Between 1 and 5 ng of cDNA was used per qPCR reaction with 200 μM primers using the iQ SYBR Green Supermix ( Bio-Rad , Hercules , CA ) on CFX384 Real-Time PCR Detection System ( Bio-Rad , Hercules , CA ) . qPCR primers used: zangptl2-F: TCAGAGTGGGCCGTTATCATGGAA , zangptl2-R: TGATAACGACTGCGGTAATGCCCT , HES1-F: ATAGCTCGCGGCATTCCAAGC , HES1-R: CCAGCACACTTGGGTCTGTGCT , RUNX1-F: AGGAAGACACAGCACCCTGGA , RUNX1-R: ACGTGCATTCTGAGGGCTGTCA , MYC-F: CGACTCTGAGGAGGAACAAG , MYC-R: GTGCGCACCTCGGTATTAAC , GAPDH-F: CCTGCACCACCAACTGCTTA , GAPDH-R: CCATCACGCCACAGTTTCC ( GAPDH used as normalizing gene ) . The 6 . 4-kb regulatory region upstream of drl was cloned into the backbone of the Tol2 destination vector , #394 pDESTTol2pA2 ( from the Tol2kit [Kwan et al . , 2007] ) driving eGFP . The resultant vector was then recombined using the Multisite Gateway technology ( Life Technologies , Grand Island , NY ) using 5′ entry clones containing the kdrl promoter ( p5E-kdrl , provided by C . B . Chien ) or the Runx1+23 enhancer ( p5E-Runx1+23 provided by OJ Tamplin ) ; middle entry clones containing eGFP ( #383 pME-eGFP , Tol2kit ) or constitutively active NOTCH1 ( pME-NICD ) ; and 3′ entry clones containing SV40 late pA ( #302 p3E-pA , Tol2kit ) . All vectors were sequence verified prior to injection into 1 cell stage Tg ( kdrl:Hras-mCherry ) or Tg ( Runx1+23:NLS-mCherry ) embryos at 25 ng/embryo . 15 pg of Tol2 ( transposase enzyme ) mRNA was co-injected per embryo . Dechorionated and staged embryos were imaged beginning at 28hpf and 72hpf for the AGM and CHT , respectively . To control for variability in injection , embryos were pre-screened for a similar degree of mosaicism . Selected embryos were mounted in glass bottom 6-well plates ( MaTek , No . 1 . 5 cover glass ) in 1% LMP agarose in E3 embryo medium containing 0 . 02% Tricaine . Time-lapse imaging was performed using a Yokogawa CSU-X spinning disk and Andor iXon EM-CCD cameras mounted on an inverted Nikon Eclipse Ti microscope ( Andor Technology , South Windsor , CT ) . 488 nm and 561 nm lasers were used to image GFP and mCherry , respectively . Control eGFP and NICD-injected embryos were imaged together in separate wells of the same 6-well plate on the same day to control for imaging variability . All imaging was done within a 28 . 5°C humidified chamber . Confocal z-stacks were collected every 10 min using a motorized stage piezo , acquiring 5 µm z-steps across 120 µm , tiling three fields of view for the AGM , or 5 µm z-steps across 60 µm , tiling five fields of view for the CHT . 3D time-lapse video files were rendered as maximum image projections using Imaris ( Bitplane , South Windsor , CT ) . Human endothelial cells ( either human umbilical vein endothelial cells [primary or EA . hy926] or human aortic endothelial cells ) or the leukemia cell line K562 were cultured per manufacturer's protocol . Human hematopoietic CD34+ progenitor cells isolated from peripheral blood of GCSF-mobilized healthy volunteers were obtained from the Fred Hutchinson Cancer Research Center . These cells were expanded for 6 days in StemSpan SFEM ( Stem Cell Technology , Canada ) supplemented with CC100 cytokine mix and 2% Pen/Strep . We have previously found that these cells will remain at least 75–80% CD34+ and greater than 90% CD38− after this cultured period ( Trompouki et al . , 2011 ) . Though they do not truly represent HSCs after culture , they nevertheless retain the capacity to differentiate and thus are utilized as hematopoietic progenitors in our studies . Prior to stimulation with purified FLAG-tagged ANGPTL2 ligand , cells were either serum starved ( endothelial or K562 cells ) or cytokine starved ( CD34+ cells ) for 2–16 hr at which time they were drug treated or vehicle treated ( DMSO ) . FLAG-tagged murine ANGPTL2 ( vector provided by Y Oike ) was expressed in HEK293T cells for 72 hr and cell supernatant was purified as previously described ( Zhang et al . , 2006 ) . Briefly , cell supernatant was centrifuged to remove cell debris and supplemented to a final concentration of 150 mM NaCl . Complete Protease Inhibitor Cocktail ( Roche , Indianapolis , IN ) was added at 1 tablet/50 ml . Cell supernatant was pre-cleared using unconjugated Sepharose beads prior to incubation with the anti-FLAG M2 affinity Gel ( Sigma , St . Louis , MO ) overnight at 4°C . Gel column was washed with TBS 8–10 times and eluted using FLAG peptide at 150 μg/ml , dissolved in TBS . Eluent was dialyzed against 4 changes of PBS and analyzed on SDS-PAGE to determine protein concentration . The Notch firefly luciferase reporter plasmid pGL2-Hes1 reporter ( Hes1-luc , provided by JC Aster ) and pRL-TK ( Promega , Madison , WI ) encoding the Renilla luciferase plasmid were co-transfected into K562 cells using AMAXA nucleofector according to the manufacturer protocols ( Lonza , Hopkinton , MA ) . Cells were lysed and assayed using the Dual-Luciferase Reporter Assay System ( Promega , Madison , WI ) . Firefly luciferase activity was normalized to Renilla luciferase activity and plotted as a ratio of the two . Full-length human LILRB2 and intracellular ( ICD ) truncation mutant were obtained from CC Zhang . Full-length LILRB2 contains four extracellular ( ECD ) IgG domains , a transmembrane domain ( TM ) and a short ICD . The LILRB2 ECD truncation mutant includes the entire ECD of LILRB2 , the transmembrane domain and the first six amino acids from the ICD . The LILRB2 ICD truncation mutant includes 41 amino acids from the end of ECD , TM and the entire ICD . The HA-tag ( TACCCATACGATGTTCCAGATTACGCT ) followed by a linker sequence ( GGAGGCTCAGGGGGTTCC ) was cloned 5′ to the ATG of full-length LILRB2 or LILRB2 ECD to make the N-terminal-tagged mutants . Likewise , the linker sequence followed by the HA tag was cloned 3′ to the last amino acid before STOP of full-length LILRB2 and LILRB2 ECD to make the C-terminal-tagged mutants . To generate N- or C-terminal HA-tagged LILRB2-ICD , we performed site-directed mutagenesis to insert the HA and linker sequences ( Q5 Site-Directed Mutagenesis Kit , NEB , Ipswich , MA ) . Lentiviral shRNA in the pLKO . 1-puromycin vector were obtained from the Sigma Mission shRNA library . Two lentiviral shRNA constructs targeted against human LILRB2 were obtained with the following sequences: sh-LILRB2-1 ( TRCN0000416153: CCGGGAAGTAAGAATGTGCTTTAAACTCGAGTTTAAAGCACATTCTTACTTCTTTTTTG ) and sh-LILRB2-2 ( TRCN0000416926: CCGGTGACGTTGGCTTTCGTATAAGCTCGAGCTTATACGAAAGCCAACGTCATTTTTTG ) . An shRNA with a scrambled sequence was used as a non-targeting control ( sh-CT ) . HEK293 cells were cultured in Dulbecco's modified Eagle's Medium supplemented with 10% fetal bovine serum and 2% penicillin-streptomycin ( Life Technologies , Grand Island , NY ) . They were transfected at 80% confluence using branched polyethylenimine ( Sigma , St . Louis , MO ) with psPAX2 and VSV-G lentiviral plasmids as well as the relevant shRNA plasmid . Medium was refreshed 16–24 hr after transfection . Medium containing viral particles was collected at 48 and 72 hr after transfection and concentrated by ultracentrifugation . Stable EA . hy926 cells were infected with each virus with 8 μg/ml protamine sulfate ( Sigma , St . Louis , MO ) overnight at 37°C and selected for 2 weeks in 1 μg/ml Puromycin ( Sigma , St . Louis , MO ) . Subsequent to cells stimulated with purified ANGPTL2 , cells were washed with ice-cold PBS prior to lysing in RIPA buffer ( Life Technologies , Grand Island , NY ) supplemented with 20 mM NaF , 1 mM Na4P2O7 , 0 . 3 mg/ml Pefabloc SC ( Roche , Indianapolis , IN ) , Complete Protease Inhibitor Cocktail ( Roche , Indianapolis , IN ) , and 1 mM Na3VO4 . Normalized protein lysates were diluted to 1 mg/ml prior to pre-clearing using Protein G-conjugated Dynal beads ( Life Technologies , Grand Island , NY ) . Pre-cleared lysates were immunoprecipitated with the appropriate antibodies overnight at 4°C . Protein G-conjugated Dynal beads were used to pull down immunocomplexes and washed with IP buffer ( Life Technologies , Grand Island , NY ) , supplemented with the above mentioned protease/phosphatase inhibitors , three times prior to eluting in SDS sample buffer and loading onto SDS-PAGE for Western blotting . Antibodies used to perform co-IPs and Western blotting are Notch1 ( Santa Cruz , Dallas , TX , sc-6014R or Cell Signaling , Danvers , MA , #3608 ) , Cleaved Notch Val1744 ( NICD , Cell Signaling , Danvers , MA , #4147 ) , Cleaved Notch ( NEXT , provided by S Blacklow ) , LILRB2 ( Santa Cruz , Dallas , TX , sc-33454 ) , β-actin ( Sigma , St . Louis , MO , #A2228 ) , HA ( Sigma , St . Louis , MO , #H9658 ) , and VE-Cadherin ( Santa Cruz , Dallas , TX , sc-6458 ) . All co-IPs and Westerns were repeated at least three times with similar results . The densitometries of Western blots are measured using ImageJ v1 . 48 and indicated below the blots as a ratio of band intensity normalized to the loading control β-actin or the protein used for IP . Samples for microarray analysis were obtained by isolating RNA from three biological triplicates each of human CD34+ progenitor cells that have been stimulated with ANGPTL2 ( 1 μg/ml ) or PBS vehicle using TRIzol ( Life Technologies , Grand Island , NY ) extraction followed by DNaseI ( Qiagen , Germantown , MD ) digestion and RNeasy ( Qiagen , Germantown , MD ) cleanup . RNA was prepared for hybridization to the Affymetrix Human Exon 1 . 0 ST Arrays . Raw CEL files were analyzed using the Genepattern software suite ( Broad Institute , Cambridge , MA ) . Raw CEL files were converted into GCT format using the ExpressionFileCreator Module , with RMA and quantile normalization . The GCT file was then used in the PreprocessDataSet module , where threshold and ceiling values were set ( floor 20 ceiling 20 , 000 , fold = 1 . 5 , delta = 1 ) , and any value lower/higher than the threshold/ceiling were reset to the threshold/ceiling value , and data were processed to discard any invariant genes . Comparisons between groups were performed using the ComparativeMarkerSelection modules , using two-sided t-test with Benjamini-Hochberg multiple hypothesis testing . The data for these microarrays are available on NCBI Gene Expression Omnibus database under GEO accession number GES51652 . The data generated from the Affymetrix arrays were used to query the Broad Molecular Signature Database . The GCT files generated by Genepattern were used as input to GSEA ( version 2 , Broad Institute ) and queried against the c2 curated gene sets , c3 motif gene sets , and c5 GO gene sets . GSEA parameters used for this analysis: scoring_scheme=weighted , metric=Signa2Noise , permute_type=gene_set , permutations=1000 . Genesets with a FDR<0 . 25 were considered significant , and the top 20 sets ( all with FDR = 0 . 00 ) were used for Leading Edge analysis . The complete list of GSEA results can be found in Supplementary File 1 . Samples were prepared as previously described ( Trompouki et al . , 2011 ) . Briefly , 108 of the ANGPTL2-stimulated human peripheral blood CD34+ progenitors were cross-linked with fresh formaldehyde , quenched with Glycine and lysed . The nuclear extracts were sonicated ( Bioruptor ) for 36 cycles of 30-s followed by 1-min rest prior immunoprecipitation with 100 μg of NOTCH antibody ( Wang et al . , 2011 ) and 100 μg of Protein G Dynal beads overnight at 4°C . 50 μl of sonicated lysates prior to antibody addition was reserved as input sample . Beads are washed , eluted , and reverse cross-linked before RNase and Proteinase K treatment and DNA extracted by Phenol/Chloroform . ChIP or input DNA overhangs were blunted and purified using PCR purification kit ( Qiagen , Germantown , MD ) . Single A in the 3′ end was added and purified with the MinElute PCR purification kit ( Qiagen , Germantown , MD ) . Illumina adaptor oligos were added , followed by PCR purification again . The samples were amplified by PCR ( 18 cycles ) to add linker sequence to the fragments to prepare for annealing to the Genome Analyzer flow-cell . Amplified samples were separated on a 2% agarose gel , and products between 150 and 350 bp were excised and gel extracted using Gel Extraction kit ( Qiagen , Germantown , MD ) . Polony generation and sequencing were done as previously described ( Trompouki et al . , 2011 ) . Sequences obtained from the Illumina/Solexa sequencer were processed through the bundled Solexa image extraction pipeline as previously described ( Trompouki et al . , 2011 ) . Briefly , sequences were aligned using ELAND software to NCBI Build 36 ( UCSC hg18 ) of the human genome . Only sequences that mapped uniquely to the genome with zero or one mismatch were used for our analysis . Analysis methods were based on previously published methods ( Johnson et al . , 2007; Guenther et al . , 2008; Marson et al . , 2008 ) . Each read was extended 200 bp toward the interior of the sequence fragment , based on the strand of the alignment . Across the genome the number of ChIP-seq reads was tabulated in 10 bp bins , and the genomic bins that contained statistically significant ChIP-seq enrichment were identified by comparison to a Poissonian background model . Assuming background reads are spread randomly throughout the genome , the probability of observing a given number of reads in a 1-kb window can be modeled as a Poisson process in which the expectation can be estimated as the number of mapped reads multiplied by the number of bins into which each read maps , divided by the total number of bins available . Enriched bins within 200 bp of one another were combined into regions . The Poissonian background model assumes a random distribution of background reads . However , significant deviations from this expectation have been observed . Some of these non random events can be detected as sites of apparent enrichment in negative control DNA samples creating false positives . To remove these false-positive regions , negative control DNA from whole-cell extract ( Input DNA sample ) was sequenced . Enriched bins and enriched regions were defined as having greater than fivefold density in the experimental sample compared with the control sample when normalized to the total number of reads in each data set . This served to filter out genomic regions that are biased to having a greater than expected background density of ChIP-seq reads . Enriched regions within 5 kb upstream or downstream of the body of the gene were called bound . Additionally , data files that contain genome browser tracks showing genome-wide ChIP-seq density and enriched regions for all experiments are available on NCBI Gene Expression Omnibus database under GEO accession number GES63010 . Non-repeat sequences from NOTCH ChIP-seq were uploaded onto MEME-ChIP ( http://meme . nbcr . net/meme/tools/meme-chip ) for motif analysis . Enriched regions from the NOTCH ChIP-seq were imported into Genomic Regions Enrichment Annotations Tool ( GREAT ) ( McLean et al . , 2010 ) and genomic regions that are associated with putative genes were used to generate terms that include gene ontology , phenotype data and human disease , pathway data , gene expression , regulatory motifs , and gene families . Each term is determined by the region–gene association settings and significance is computed from the enriched regions of the NOTCH ChIP-seq falling in the regulatory domains of genes involved in a particular function compared to random . Both the enriched genes from the NOTCH ChIP-seq and the gene lists from the microarrays were imported into Ingenuity Pathway Analysis ( IPA ) . The functional analysis identified the biological functions and/or disease states that are most significant to the data set . Of note , the Upstream Regulator analysis was used to predict which molecules ( in its active or repressed state ) are likely to give rise to the observed expression data . MYC was identified as the top active molecule . | Bone marrow contains types of stem cell that can produce new blood and immune cells . Transplanting bone marrow from a healthy person can be used to treat people with certain disorders of the blood and immune system , by providing a new supply of regenerating bone marrow stem cells . Bone marrow transplants are also critical for individuals who had their own bone marrow stem cells destroyed by cancer or by toxic anti-cancer therapies like chemotherapy . Acquiring bone marrow for transplants can be a difficult process . Sometimes doctors can use the patient's own stem cells for a transplant . But in circumstances where the patient lacks healthy bone marrow , a bone marrow donor must be found . Donors and recipients must be carefully matched: certain proteins on the donor's bone marrow cells must be very similar to the proteins on the recipient's bone marrow cells , or the recipient's immune system will attack and kill the new cells . If scientists could learn to grow the stem cells found in bone marrow in laboratories , they could circumvent some of the challenges associated with bone marrow donation . To do that , scientists must first understand the precise molecular mechanisms that allow the blood cell-producing stem cells to regenerate themselves and produce new blood cells . Angiopoietin-like proteins are commonly used to help stem cells grow in the laboratory , and Lin et al . have now looked in detail at how these proteins work . This revealed that angiopoietin-like proteins also cause the stem cells that produce blood cells to grow in zebrafish embryos and are necessary for the embryos' vascular system to develop properly . Zebrafish embryos lacking angiopoietin-like proteins develop a similar set of developmental problems as zebrafish embryos with a mutation in another protein called Notch . Through a series of experiments , Lin et al . show that angiopoietin-like proteins interact with Notch and help transform Notch into its active form , which is necessary for blood stem cell growth . Lin et al . also found that both angiopoietin-like proteins and Notch affect the same signaling molecules . This suggests that the two proteins may work together as part of the same molecular pathway . The work suggests an alternative method to activate Notch for blood stem cell stimulation during processes such as bone marrow or cord blood transplantation . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2015 | Angiopoietin-like proteins stimulate HSPC development through interaction with notch receptor signaling |
Lithium is widely used as a treatment for Bipolar Disorder although the molecular mechanisms that underlie its therapeutic effects are under debate . In this study , we show brain-derived neurotrophic factor ( BDNF ) is required for the antimanic-like effects of lithium but not the antidepressant-like effects in mice . We performed whole cell patch clamp recordings of hippocampal neurons to determine the impact of lithium on synaptic transmission that may underlie the behavioral effects . Lithium produced a significant decrease in α-amino-3-hydroxyl-5-methyl-4-isoxazolepropionic acid receptor ( AMPAR ) -mediated miniature excitatory postsynaptic current ( mEPSC ) amplitudes due to postsynaptic homeostatic plasticity that was dependent on BDNF and its receptor tropomyosin receptor kinase B ( TrkB ) . The decrease in AMPAR function was due to reduced surface expression of GluA1 subunits through dynamin-dependent endocytosis . Collectively , these findings demonstrate a requirement for BDNF in the antimanic action of lithium and identify enhanced dynamin-dependent endocytosis of AMPARs as a potential mechanism underlying the therapeutic effects of lithium .
Lithium was initially reported for treatment as a mood stabilizer over 60 years ago ( Cade , 1949 ) and is still widely used for the treatment of Bipolar Disorder ( Mitchell , 2013; Poolsup et al . , 2000; Vieta and Valentí , 2013 ) . However , despite its extensive use for the treatment of bipolar disorder , the cellular and molecular mechanisms underlying its antimanic and antidepressant responses remain poorly understood ( Can et al . , 2014; Malhi and Outhred , 2016 ) . Putative mechanisms for the therapeutic effects of lithium include inhibition of glycogen synthase kinase-3β ( GSK3β ) ( Klein and Melton , 1996 ) , upregulation of neurotrophins , as well as their receptors , and downregulation of α-amino-3-hydroxyl-5-methyl-4-isoxazolepropionic acid receptor ( AMPAR ) expression ( Ankolekar and Sikdar , 2015; Du et al . , 2004 , 2010; Gray et al . , 2003; Seelan et al . , 2008; Wei et al . , 2010 ) among others . However , which —if any— of these effects of lithium are the primary mechanism for treatment for bipolar disorder is currently unknown . Earlier studies suggested a potential link between the action of lithium as a mood stabilizer and neurotrophins , in particular brain-derived neurotrophic factor ( BDNF ) . BDNF and aberrant signaling through its high affinity receptor tropomyosin receptor kinase B ( TrkB ) have been proposed to underlie both the pathophysiology and treatment of bipolar disorder ( Autry and Monteggia , 2012; Malhi et al . , 2013; Scola and Andreazza , 2015 ) . Previous work has shown that patients with bipolar disorder have decreased peripheral BDNF mRNA in blood lymphocytes and monocytes in comparison to healthy controls ( D'Addario et al . , 2012 ) . Moreover , both manic and depressive states in patients with bipolar disorder have been associated with significantly decreased BDNF blood serum levels compared to patients in euthymic states and healthy controls ( Fernandes et al . , 2015; Tunca et al . , 2014 ) . Post-mortem analysis of hippocampal tissue has also revealed reduced BDNF protein levels in patients with bipolar disorder ( Knable et al . , 2004 ) . In rodents , experimental interventions that cause manic-like ( Frey et al . , 2006; Fries et al . , 2015; Jornada et al . , 2010 ) and depressive-like behaviors ( Smith et al . , 1995; Tsankova et al . , 2006; Ueyama et al . , 1997 ) have also been shown to result in decreased BDNF mRNA and protein levels in the hippocampus . In contrast , lithium treatment is associated with increased BDNF protein levels in the serum of patients with bipolar disorder ( Cunha et al . , 2006; de Sousa et al . , 2011; Tramontina et al . , 2009 ) . In rodent models , chronic lithium treatment has been shown to increase BDNF mRNA and protein expression in hippocampus , cortex , and amygdala as well as cortical neurons in culture ( Fukumoto et al . , 2001; Jornada et al . , 2010; Yasuda et al . , 2009 ) . Lithium treatment has also been shown to increase TrkB activity in neuronal cultures suggesting an increase in BDNF-TrkB signaling ( Hashimoto et al . , 2002 ) . While previous work has demonstrated that BDNF is necessary for the antidepressant response of conventional antidepressants ( Adachi et al . , 2008; Ibarguen-Vargas et al . , 2009; Monteggia et al . , 2004 , 2007 ) , as well as the rapid antidepressant effects of ketamine ( Autry et al . , 2011; Lepack et al . , 2014 ) , it is currently unknown whether BDNF or the TrkB receptor are required for lithium’s antidepressant or antimanic effects . In this study , we establish the necessity of BDNF-TrkB signaling in the antimanic-like but not antidepressant-like response to lithium . We also identify a direct effect of lithium treatment on AMPAR trafficking , where AMPA receptor surface expression is decreased due to a sustained increase in dynamin mediated AMPAR endocytosis . This process is dependent on BDNF-TrkB function and results in synaptic downscaling where unitary postsynaptic responses are decreased in a manner proportional to their relative strengths . Collectively , our findings reveal a requirement for neurotrophic signaling in the behavioral and cellular effects of lithium . These data provide novel mechanistic insight into the action of lithium which may underlie its therapeutic effect for the treatment of bipolar disorder .
Previous studies have used mice to examine the impact of lithium treatment on molecular signaling . However , typically the lithium serum concentration is not determined in mice following lithium treatment leaving it unclear whether reported changes occur within the therapeutic range . We established a lithium protocol in which C57BL/6 mice were given 0 . 2% lithium chloride ( LiCl ) chow for four days , followed by 0 . 4% LiCl chow for the remainder of the treatment and testing period , which lasted a total of 11–17 days depending on the experiment ( Figure 1A ) . This protocol resulted in a lithium serum concentration in treated mice that was within the therapeutic range of 0 . 5–2 mM ( Amdisen , 1980 ) , with an average concentration of 1 mM ( Figure 1B ) . Control ( CTL ) mice were given the same chow without lithium . Following 11 days of lithium treatment , mice were sacrificed and the hippocampus rapidly removed . To confirm our lithium protocol produced the expected molecular effects , we examined GSK3 inhibition ( Klein and Melton , 1996; Stambolic et al . , 1996 ) , by measuring the phosphorylation of GSK3β at serine nine with western blot analysis and found a significant increase relative to total GSK3β ( Figure 1C ) . We next examined whether the lithium treatment regulated the expression of BDNF in the hippocampus . Using Q-PCR targeted to the coding exon of Bdnf , we found lithium treated mice had a significant increase in Bdnf mRNA expression ( Figure 1D ) . We also found a significant increase in BDNF protein expression ( Figure 1E ) in lithium treated mice relative to CTL mice . 10 . 7554/eLife . 25480 . 003Figure 1 . Lithium serum concentration within the therapeutic range increases BDNF expression in mice . ( A ) Timeline of chronic lithium chloride ( LiCl ) treatment and behavioral testing . ( B ) ( Left ) Example standard curve of lithium counts used to calculate lithium concentration in blood serum . ( Right ) . Average lithium concentration in blood serum in control and treated mice ( n= 9–12 per group ) . ( C ) Chronic lithium treatment in mice produces a significant decrease in immobility in the FST compared to the control group ( Student’s unpaired t test *p<0 . 0001 , n = 10 mice per group ) . ( D ) Chronic lithium treatment caused a significant increase in Bdnf mRNA expression in the hippocampus ( Student’s unpaired t test *p=0 . 04 , n = 8–9 mice per group ) . ( E ) Chronic lithium treatment caused a significant increase in BDNF protein expression in the hippocampus ( Student’s unpaired t test *p=0 . 002 , n = 10 mice per group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 003 Since lithium increases the expression of BDNF mRNA and protein levels , we investigated whether BDNF was necessary for lithium’s behavioral action . Inducible Bdnf KO mice , in which Bdnf is deleted ~70–80% in broad forebrain regions ( Monteggia et al . , 2004 ) , were given lithium treated chow following our protocol ( Figure 1A ) . To determine whether this lithium treatment produced the expected behavioral effects , we tested mice in the forced swim test ( FST ) , a test with predictive value for antidepressant responses ( Porsolt et al . , 1977 ) . Littermate control ( CTL ) mice treated with lithium showed a significant reduction in immobility time that was suggestive of an antidepressant-like response compared to CTL mice that did not receive lithium ( Figure 2A ) . In agreement with previous data we found that loss of Bdnf in the inducible Bdnf KO mice did not alter immobility in the FST compared to CTL mice ( Monteggia et al . , 2004 ) . We also found that Bdnf KO mice administered lithium showed a significant reduction in immobility comparable to KO mice that did not receive lithium ( Figure 2A ) suggesting that BDNF is not necessary for the antidepressant-like effect of lithium in the FST . 10 . 7554/eLife . 25480 . 004Figure 2 . BDNF expression is required for the antimanic-like effect of lithium . ( A ) Chronic lithium treatment caused a significant decrease in the immobility time in the FST in littermate CTL and inducible Bdnf KO mice in comparison to vehicle chow treated mice ( ANOVA F3 , 90 = 18 . 1 *p<0 . 0001 , Dunnett’s multiple comparisons Veh CTL vs LiCl CTL *p<0 . 0001 , Veh Bdnf KO vs LiCl Bdnf KO *p<0 . 0001 , n = 19–28 mice per group ) . ( B ) Acute amphetamine injection caused a significant increase in locomotor activity in control and Bdnf KO mice in comparison to the saline injected mice . Lithium treatment resulted in a nonsignificant increase in locomotor activity of CTL and Bdnf KO mice compared to vehicle treated mice . Chronic lithium treatment blunted the increased locomotor activity following acute amphetamine injection in the CTL littermate mice , but not in the Bdnf KO mice ( ANOVA F7 , 98 = 12 . 75 p<0 . 0001 , Tukey’s multiple comparisons CTL Veh-Sal vs CTL Veh-Amph *p<0 . 0001 , CTL LiCl-Sal vs CTL LiCl-Amph p=0 . 97 , CTL Veh-Sal vs CTL LiCl-Sal p=0 . 23 , Bdnf KO Veh-Sal vs Bdnf KO Veh-Amph *p<0 . 0001 , Bdnf KO LiCl-Sal vs Bdnf KO LiCl-Amph *p=0 . 01 , Bdnf KO Veh-Sal vs Bdnf KO LiCl-Sal p=0 . 44 , n = 11–15 mice per group ) . ( C ) Chronic lithium treatment resulted in a 20% reduction in the I/O curve in the CTL mice chronically treated with LiCl chow in comparison to Veh CTL mice , which is plotted as the slope of the fEPSP is plotted as a function of the presynaptic fiber volley . Linear fit slopes were calculated for Veh CTL ( 5 . 53 ± 0 . 2 ) vs CTL LiCl ( 4 . 656 ± 0 . 03825 ) . There were no significant differences in the fEPSP slope between Veh Bdnf KO slices ( 5 . 85 ± 0 . 178 ) and LiCl treated Bdnf KO ( 5 . 899 ± 0 . 1703 ) . ( Two-way ANOVA Interaction F1 , 29 = 3 . 24 p=0 . 08 , Row ( Genotype ) F1 , 29 = 5 . 09 *p=0 . 032 , Column ( Diet ) F1 , 29 = 12 . 61 *p=0 . 001 . Sidak’s multiple comparisons Veh CTL vs LiCl CTL *p=0 . 029 , Veh Bdnf KO vs LiCl Bdnf KO p=0 . 999 , Veh CTL vs Veh Bdnf KO p=0 . 739 , LiCl CTL vs LiCl Bdnf KO *p=0 . 007 n = 6–10 recordings per group ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 004 To investigate whether BDNF was required for lithium’s antimanic-like effects , we used the amphetamine hyperlocomotor test in which lithium blunts the increased locomotor activity that occurs following acute amphetamine injection , a commonly used assay for the antimanic-like effects of lithium in mice ( Flaisher-Grinberg and Einat , 2010; Gould et al . , 2007 ) . The amphetamine induced hyperlocomotor test was performed subsequent to the FST in the same cohorts of mice . Littermate CTL mice treated with acute amphetamine have the expected significant increase in the total number of horizontal beam breaks compared to vehicle treated mice ( Figure 2B ) . CTL mice receiving lithium treatment showed a small non-significant increase in the total number of beam breaks compared to vehicle treated CTL mice , and were indistinguishable from chronic lithium treated mice receiving acute amphetamine , indicating that lithium treatment could prevent the hyperactivity induced by amphetamine ( Figure 2B ) . In contrast , while inducible Bdnf KO mice receiving acute amphetamine showed a significant increase in the number of beam breaks compared to vehicle treated Bdnf KOs , and a slight increase with chronic lithium treatment , the chronic lithium treated Bdnf KO mice receiving acute amphetamine still had a significant increase in the number of beam breaks comparable to amphetamine treatment alone . These data demonstrate that lithium treatment does not block the hyperlocomotor effects of amphetamine in inducible Bdnf KO mice and thus suggests that BDNF is required for the antimanic-like effect of lithium . We next assessed whether the loss of BDNF impacted lithium’s effect on synaptic efficacy from hippocampal slices . Inducible Bdnf KO and littermate control mice were given lithium treated chow or untreated chow ( Figure 1A ) . The slope of the input/output ( I/O ) curve , which was plotted as the presynaptic volley versus the fEPSC slope , was taken as a measure of synaptic efficacy associated with synaptic density or strength of individual synapse . We found that in CTL mice , chronic lithium treatment resulted in a significant reduction in the slope of the I/O curve in hippocampal slices in comparison to mice not treated with lithium ( Figure 2C ) , indicating that chronic lithium treatment reduces hippocampal synaptic strength . Inducible Bdnf KO mice that did not receive lithium had a similar slope of the I/O curve compared to untreated littermate CTL mice ( Figure 2C ) showing that the loss of BDNF did not impact synaptic efficacy . We also found that the slope of the I/O curve in the hippocampal slices from the inducible Bdnf KO mice treated with lithium was also unchanged in comparison to the untreated CTL and Bdnf KO mice ( Figure 2C ) , suggesting that BDNF expression is required for the lithium-mediated reduction of synaptic strength . Previous work has shown that lithium treatment produces a significant decrease in AMPAR miniature excitatory postsynaptic current ( mEPSC ) amplitudes in cultured neurons , which has been suggested to underlie the antimanic effects of lithium ( Ankolekar and Sikdar , 2015; Du et al . , 2008; Wei et al . , 2010 ) . We therefore incubated dissociated C57BL/6 hippocampal neurons with 1 mM LiCl or 1 mM NaCl , to control for changes in osmolarity , for 11–15 days and then recorded AMPAR mEPSCs ( Figure 3A ) . Lithium treatment resulted in a significant decrease in AMPAR mEPSC amplitudes in comparison to NaCl treated or untreated CTL neurons ( Figure 3B–D ) . To examine whether lithium’s effect on AMPAR mEPSC amplitudes was due to synaptic scaling , we plotted in rank order the amplitudes from each condition with a linear fit equation . We found that lithium treatment resulted in a 41% and 26% decrease in the slope compared to untreated and NaCl treatment , respectively ( Figure 3E ) . These data show lithium produces a downward scaling of all AMPAR mEPSC amplitudes and impacts post-synaptic homeostatic plasticity . Notably , lithium did not produce generalized effects on synaptic measures as there were no changes in mEPSC frequency compared to untreated or NaCl treated neurons ( Figure 3F ) . Since lithium’s effects on AMPAR mEPSC amplitudes were due to synaptic scaling , which can be influenced by AMPAR trafficking , we examined the surface expression of the AMPAR subunit , GluA1 . We performed surface biotinylation experiments and found lithium treatment results in a significant decrease in GluA1 surface expression relative to total GluA1 in comparison to untreated or NaCl treated neurons ( Figure 3G ) . 10 . 7554/eLife . 25480 . 005Figure 3 . Chronic lithium treatment causes a significant decrease in synaptic scaling and surface GluA1 expression . ( A ) Timeline for treatment of dissociated hippocampal neurons with LiCl and NaCl . ( B ) Example traces of AMPAR mEPSCs from CTL untreated ( top ) , 1 mM NaCl treated ( middle ) and 1 mM LiCl treated ( bottom ) dissociated hippocampal neurons . ( C ) Chronic LiCl treatment of cultured hippocampal neurons caused a trend towards decreased AMPA mEPSC average amplitudes in comparison to CTL neurons . Chronic NaCl treatment did not cause a change in average AMPA mEPSC amplitude in comparison to CTL neurons ( ANOVA F2 , 39 = 2 . 597 p=0 . 0873 , Bonferroni’s multiple comparisons CTL vs NaCl p>0 . 999 , CTL vs LiCl p=0 . 091 ) . ( D ) Cumulative probability histogram showing a significant leftward shift ( decrease ) in the amplitudes of AMPAR-mEPSCs from cells chronically treated with 1 mM LiCl in comparison to CTL untreated and 1 mM NaCl treated neurons ( Kolmogorov-Smirnov test: CTL vs 1 mM >LiCl *p=1 . 75×10−43 , D = 0 . 234 , 1 mM NaCl vs 1 mM LiCl *p=8 . 14×10−27 , D = 0 . 167 , n= 12–18 recordings per condition ) . ( E ) Rank order plot of CTL untreated AMPAR mEPSC amplitudes vs 1 mM LiCl AMPAR mEPSC amplitudes revealed a 41% decrease . Rank order plot of 1 mM NaCl AMPAR mEPSC amplitudes vs 1 mM LiCl AMPAR mEPSC amplitudes showed a 26% decrease . ( CTL vs LiCl line of best fit y = 0 . 59x , CTL vs NaCl line of best fit y = 0 . 85x , Difference between NaCl and LiCl . 26 , n= 12–18 recordings per condition ) . ( F ) AMPAR-mEPSC frequency is indistinguishable between CTL untreated , NaCl treated , and LiCl treated neurons ( ANOVA F4 , 57 = 0 . 129 p=0 . 97 , n = 12–18 recordings ) . ( G ) Surface biotinylation experiments revealed that chronic lithium treatment of hippocampal neurons results in a significant decrease in the surface/total GluA1 ratio ( ANOVA F2 , 21 = 2 . 911 p=0 . 0766 , Dunnett’s multiple comparisons CTL vs NaCl p=0 . 58 , CTL vs LiCl *p=0 . 04 , n= 3 separate experiments ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 005 Since our data demonstrated that BDNF was required for lithium’s antimanic-like effects in mice and lithium-mediated decrease in synaptic strength , we examined whether chronic lithium treatment’s effects on synaptic function , which may ultimately trigger the behavioral effects , were dependent on BDNF and its high affinity receptor TrkB . BDNF is a secreted protein that has been shown to influence synaptic upscaling and downscaling , depending on the brain region examined ( Leslie et al . , 2001; Reimers et al . , 2014; Rutherford et al . , 1998 ) . We therefore examined whether lithium’s effects on AMPAR mEPSC amplitudes were dependent on BDNF . We cultured hippocampal neurons from Bdnf fl/fl mice and infected them with lentivirus expressing Cre recombinase ( Cre ) tagged with GFP to delete the gene of interest or GFP alone as a control . Previous studies from our laboratory have shown that this lentivirus construct expressing Cre tagged with GFP can efficiently knock out endogenous gene expression without triggering cell death ( Akhtar et al . , 2009; Nelson et al . , 2006 ) . As expected , GFP infected neurons treated with chronic lithium had a significant decrease in AMPAR mEPSC amplitudes in comparison to untreated or NaCl treated neurons ( Figure 4A–D ) . In BDNF deficient cultures , Bdnffl/fl neurons infected with Cre , there was no difference in AMPAR mEPSC amplitudes compared to GFP infected neurons suggesting loss of BDNF does not impact this synaptic measure ( Figure 4A–D ) . In contrast to wild-type neurons , we found chronic lithium treatment of BDNF deficient cultures induced a small but nonsignificant decrease in AMPAR mEPSC amplitudes ( Figure 4D ) suggesting a requirement for BDNF . We also examined mEPSC frequency from these six treatment groups and did not find a significant change with any condition ( data not shown ) . Taken together , these data suggest that chronic lithium’s effect on AMPAR mEPSC amplitudes is dependent on BDNF . 10 . 7554/eLife . 25480 . 006Figure 4 . BDNF is required for the lithium-mediated decrease in AMPAR mEPSC amplitude . ( A ) Example traces from GFP infected Bdnf fl/fl neurons with no treatment ( CTL , top ) or 11–17 day incubation with 1 mM NaCl ( middle ) or 1 mM LiCl ( bottom ) . ( B ) Example traces from Cre infected Bdnf fl/fl neurons with no treatment ( control , top ) or 11–17 day incubation with 1 mM NaCl ( middle ) or 1 mM LiCl ( bottom ) . ( C ) Cumulative probability histogram showing a significant leftward shift ( decrease ) in AMPAR mEPSC amplitudes between Bdnf fl/fl -GFP control untreated and Bdnf fl/fl -GFP LiCl conditions . There is also a significant leftward shift ( decrease ) between Bdnf fl/fl -GFP LiCl and Bdnf fl/fl -Cre LiCl conditions ( Kolmogorov-Smirnov test: GFP Control vs GFP LiCl p<0 . 0001 D = 0 . 309 , GFP LiCl vs Cre LiCl *p<0 . 0001 D = 0 . 19 , n = 10–13 recordings per condition ) . ( D ) Lithium caused a significant decrease in AMPAR mEPSC amplitudes in Bdnf fl/fl -GFP neurons compared to control Bdnf fl/fl -GFP neurons . However , lithium did not impact AMPAR mEPSC amplitudes between Bdnf fl/fl -Cre neurons in comparison to untreated control Bdnf fl/fl -Cre neurons ( ANOVA F5 , 59 = 5 . 694 *p=0 . 0002 , Tukey’s multiple comparisons GFP Control vs GFP LiCl *p=0 . 003 , Cre Control vs Cre LiCl p=0 . 197 , n = 10–13 recordings per condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 006 To further evaluate whether BDNF signaling is required for lithium’s effects on AMPAR mediated synaptic transmission , we explored the requirement for the TrkB receptor . We cultured hippocampal neurons from Ntrk2 fl/fl mice and infected them with either lentivirus expressing Cre or GFP . Consistent with our previous results , GFP infected neurons treated with lithium had a significant decrease in AMPAR mEPSC amplitude compared to untreated or NaCl treated neurons ( Figure 5A–D ) . We found Ntrk2 fl/fl neurons infected with CreGFP , had indistinguishable AMPAR mEPSC amplitudes compared to GFP infected neurons demonstrating that loss of TrkB does not affect this synaptic measure ( Figure 5D ) . In agreement with data from the Bdnf deficient neurons , we found lithium treatment of Ntrk2 deficient neurons did not affect AMPAR mEPSC amplitudes showing a requirement for TrkB in lithium’s effect on AMPAR mEPSC responses ( Figure 5D ) . We measured mEPSC frequency in the six treatment groups and did not observe any significant differences demonstrating a specific effect of lithium on mEPSC properties ( data not shown ) . 10 . 7554/eLife . 25480 . 007Figure 5 . Lithium-mediated decrease in AMPAR mEPSC amplitude is dependent on TrkB . ( A ) Example traces from GFP infected Ntrk2 fl/fl neurons , untreated Control ( top ) , 1 mM NaCl treatment ( middle ) , and 1 mM LiCl treatment ( bottom ) . ( B ) Examples from Cre infected Ntrk2 fl/fl neurons , untreated Control ( top ) , 1 mM NaCl treatment ( middle ) and 1 mM LiCl treatment ( bottom ) . ( C ) Cumulative probability histogram showing a significant leftward shift ( decrease ) in AMPAR-mEPSC amplitudes from Ntrk2 fl/fl -GFP neurons treated with 1 mM LiCl . ( Kolmogorov-Smirnov test: GFP control vs GFP LiCl *p<0 . 0001 , D = 0 . 181 , GFP LiCl vs Cre LiCl *p<0 . 0001 , D = 0 . 153 , n = 9–17 recordings per condition ) . ( D ) Lithium caused a significant decrease in AMPAR-mEPSC amplitudes in Ntrk2 fl/fl -GFP neurons compared to untreated control Ntrk2 fl/fl -GFP neurons . However , lithium was unable to cause any significant changes in AMPAR-mEPSCs in Ntrk2 fl/fl -Cre neurons compared to untreated Ntrk2 fl/fl -Cre control ( ANOVA F5 , 76 = 5 . 107 *p=0 . 0004 , Bonferonni multiple comparison test: GFP control vs GFP LiCl *p=0 . 002 , Cre Control vs Cre LiCl p>0 . 999 , n = 9–17 recordings per condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 007 Our data so far shows that chronic lithium treatment of hippocampal neurons leads to a significant decrease in AMPAR mEPSC amplitudes that is dependent on BDNF-TrkB signaling . The decrease in AMPAR mEPSC amplitudes and the reduced GluA1 surface membrane expression suggests lithium is augmenting AMPAR endocytosis . Therefore , we examined whether inhibiting AMPAR endocytosis would rescue the synaptic phenotype observed with lithium treatment . AMPAR endocytosis is dependent on the GTPase dynamin ( Lu et al . , 2007 ) , which is expressed by three related genes in mammals ( Raimondi et al . , 2011 ) . Dynamin 1 and 3 ( Dnm1 and Dnm3 ) are highly expressed in neurons and found in dendrites ( Calabrese and Halpain , 2015; Noda et al . , 1993 ) . Previous work has shown that constitutive Dnm1−/−/ Dnm3−/− double KO mice are not viable ( Raimondi et al . , 2011 ) . Therefore , to examine the contribution of both dynamin1 and dynamin3 on lithium mediated synaptic effects , we crossed Dnm1 fl/fl mice with Dnm3−/− mice as the constitutive deletion of Dnm3 does not impact synaptic function ( Raimondi et al . , 2011 ) . We cultured Dnm1 fl/fl / Dnm3−/− neurons and infected them with lentivirus expressing Cre to delete Dnm1 and thus generate the double KOs , or GFP as a control . The GFP infected neurons treated with lithium had a significant reduction in AMPAR mEPSC amplitudes compared to untreated neurons showing that the loss of Dnm3 does not impact the effect of lithium on synaptic function ( Figure 6A–C ) . In contrast , deletion of both Dnm1 and Dnm3 occluded the effect of chronic lithium treatment on AMPAR mEPSC amplitudes in comparison to untreated cultures demonstrating that lithium is potentiating dynamin dependent endocytosis of AMPAR ( Figure 6C ) . 10 . 7554/eLife . 25480 . 008Figure 6 . Dynamin-dependent endocytosis is required for lithium-mediated decrease in AMPAR-mEPSC amplitudes . ( A ) Example traces from GFP infected Dnm1 fl/fl/Dnm3−/− neurons , control ( top ) , chronic NaCl treatment ( 1 mM , middle ) and chronic LiCl treatment ( 1 mM , bottom ) . ( B ) Example traces from Cre infected Dnm1 fl/fl/Dnm3−/− neurons , control ( top ) , chronic NaCl treatment ( 1 mM , middle ) , and chronic LiCl treatment ( 1 mM , bottom ) . ( C ) Chronic lithium treatment caused a significant decrease in AMPAR mEPSC amplitude in Dnm1 fl/fl/Dnm3−/− neurons infected with lenti-GFP virus compared to untreated Dnm1 fl/fl/Dnm3−/− neurons . However , there was no significant difference in AMPAR-mEPSC amplitudes between untreated and lithium treated Dnm1 fl/fl/Dnm3−/− neurons infected with lenti-CreGFP virus ( ANOVA F5 , 68 = 5 . 048 *p=0 . 0006 , Bonferonni multiple comparison test: GFP control vs GFP LiCl *p=0 . 02 , Cre Control vs Cre LiCl p=0 . 455 , n = 8–12 recordings per condition ) . ( D ) Example traces from wildtype neurons recorded with DMSO-internal solution , control ( top ) , chronic NaCl treatment ( 1 mM , middle ) , and chronic LiCl treatment ( 1 mM , bottom ) . ( E ) Example traces from wildtype neurons recorded with Dyngo-internal solution , control ( top ) , chronic NaCl treatment ( 1 mM , middle ) , and chronic LiCl treatment ( 1 mM , bottom ) . ( F ) In comparison to control untreated neurons , chronic lithium treatment caused a significant decrease in AMPAR-mEPSC amplitudes . In contrast , lithium did not cause a significant change in AMPAR-mEPSC amplitudes in comparison to the untreated control when Dyngo is included in the internal pipette solution ( ANOVA F5 , 69 = 6 . 985 , *p<0 . 0001 , Bonferonni multiple comparison test: DMSO control vs DMSO LiCl *p<0 . 0001 , Dyngo Control vs Dyngo LiCl p>0 . 999 , n = 11–17 recordings per condition ) . DOI: http://dx . doi . org/10 . 7554/eLife . 25480 . 008 To control for possible compensatory effects due to the loss of Dnm1 and Dnm3 , in our next experiments we cultured hippocampal neurons from C57BL/6 mice and acutely inhibited dynamin activity in the postsynaptic neuron . We chronically treated neurons with lithium and then added the pan-dynamin inhibitor Dyngo ( McCluskey et al . , 2013 ) , or the vehicle 1% DMSO , to the internal pipette solution prior to our recordings . Consistent with previous data , chronic lithium treatment resulted in a significant decrease in AMPAR mEPSC amplitudes in neurons recorded with vehicle in the pipette solution ( Figure 6D–F ) . In contrast , the presence of Dyngo in the pipette solution rescued the decrease in AMPAR mEPSC amplitudes seen after lithium treatment ( Figure 6F ) . Taken together , these experiments demonstrate that chronic lithium treatment enhances AMPAR endocytosis to alter synaptic function in a dynamin dependent manner and the loss of dynamin , specifically Dnm1 and Dnm3 , can rescue the synaptic deficits .
In the current study , we report that lithium requires BDNF-TrkB signaling to mediate its antimanic as well as synaptic effects . Chronic lithium treatment in vivo results in a sustained increase in BDNF mRNA and protein expression . However , while BDNF is required for the antimanic effects of lithium it is not required for the antidepressant effects . We determined that chronic lithium treatment in vivo caused a significant decrease in I/O curves from hippocampal slices that were dependent on BDNF expression . We investigated the effect of chronic lithium on synaptic function and found a significant synaptic downscaling of AMPAR mEPSC amplitudes that was dependent on BDNF as well as its high affinity receptor , TrkB . The decrease in AMPAR mEPSC amplitudes was due to a decrease in postsynaptic surface expression of GluA1 driven by dynamin-dependent AMPAR endocytosis , which could be acutely countered by administration of a dynamin inhibitor . Collectively , these data demonstrate that lithium can target AMPA receptor endocytosis to alter synaptic function . In these experiments we used lithium treated chow to administer the drug to mice to more closely mimic the oral administration used by patients with bipolar disorder . In initial experiments , we established a chronic lithium regimen that resulted in clinically relevant lithium levels in serum . In each experiment we confirmed that an animal had a serum lithium level of 0 . 5–3 . 0 mM/L to ensure that the molecular and behavioral studies were performed under conditions similar to long-term treatment for Bipolar Disorder . We show that chronic lithium treatment in rodents causes a sustained increase in BDNF mRNA and protein in the hippocampus . However , BDNF was not required for the antidepressant action of lithium . Previous studies have shown that BDNF is required for the antidepressant effects of traditional antidepressants ( Adachi et al . , 2008; Monteggia et al . , 2004 ) , as well as rapid antidepressant effects of ketamine ( Autry et al . , 2011 ) . While the current data showing lithium can exert an antidepressant effect in inducible Bdnf KOs is unexpected , the finding could be due to several possibilities . First , these data may suggest that mood stabilizers mediate an antidepressant response in a manner distinct from conventional antidepressants that target the monoamine system and independent of BDNF . Second , previous work with traditional antidepressants has largely focused on changes in Bdnf mRNA , with changes in BDNF protein not as fully explored leaving it unclear the time frame that conventional antidepressants increase BDNF protein ( Duman and Monteggia , 2006 ) . The rapid antidepressant action of ketamine requires BDNF and is protein translation dependent , with BDNF protein rapidly increased but then returning to baseline ( Autry et al . , 2011 ) . In contrast , data from the current study shows that lithium treatment results in a sustained increase in Bdnf mRNA and protein expression . Taken together , these may data suggest acute manipulation of BDNF protein expression is involved in rapid antidepressant effects while a more chronic increase in BDNF expression is required for antimanic effects . Lithium produces significant effects on synaptic transmission that may underlie its antimanic properties . Previous work has consistently shown that lithium application significantly decreases AMPAR mEPSC amplitudes in cultured neurons ( Ankolekar and Sikdar , 2015; Du et al . , 2004; Gray et al . , 2003; Seelan et al . , 2008; Wei et al . , 2010 ) . We replicated these data and showed chronic lithium treatment of hippocampal neurons results in a significant decrease in AMPA mEPSC amplitudes compared to control conditions . We also found that lithium treatment did not alter AMPA mEPSC frequency demonstrating specificity to the impact of the drug on synaptic transmission . Further analysis revealed that lithium produced a significant downward multiplicative synaptic scaling of all AMPAR-mEPSC amplitudes consistent with an impact on post-synaptic homeostatic plasticity . Previous in vivo and in vitro studies have shown that lithium decreases GluA1 and GluA2 surface expression in the hippocampus and in cultured cortical neurons ( Du et al . , 2003 , 2008; Wei et al . , 2010 ) . In addition , we found that in vivo lithium treatment caused a significant decrease in I/O curves in hippocampal slices , which is reminiscent of the decreased AMPA/NMDA ratio previously seen in rats following chronic lithium treatment ( Du et al . , 2008 ) . Lithium treatment has also been shown to decrease phosphorylation of Thr840 on GluA1 , which is associated with decreased AMPAR signaling ( Szabo et al . , 2009 ) . In the current study , we found that lithium treatment of hippocampal neurons elicits synaptic downscaling due to decreased GluA1 surface expression demonstrating a previously unknown effect of lithium on post-synaptic homeostatic plasticity . The lithium triggered significant decrease on AMPAR mEPSC amplitudes and hippocampal I/O curves was occluded in neurons lacking Bdnf suggesting a requirement in lithium’s action . To confirm these data , we examined neurons with a deletion of Ntrk2 and found that lithium’s effect on AMPAR-mEPSC amplitudes was also occluded . While these data demonstrate that the effect of lithium on AMPA mEPSC amplitudes is dependent on BDNF and TrkB , the requirement for TrkB was more robust than BDNF per se . These findings may be due to low level expression of BDNF remaining after viral infection of Cre recombinase that acts on TrkB receptors or alternatively some minor effects of neurotrophin 3 ( NT3 ) or neurotrophin 4 ( NT4 ) activating the TrkB receptor to impact lithium’s effects on synaptic transmission . Regardless , these results demonstrate a rather unexpected requirement for BDNF-TrkB signaling in lithium mediated effects on synaptic scaling as increased BDNF expression is typically associated with increased AMPAR surface expression ( Jakawich et al . , 2010; Nakata and Nakamura , 2007; Nosyreva et al . , 2013 ) . However , there is precedent for BDNF to induce downscaling of AMPAR surface expression . Specifically , acute increases in BDNF have been shown to increase GluA1 and GluA2 surface expression , whereas chronic increases in BDNF decrease surface expression of GluA1 and GluA2 ( Reimers et al . , 2014 ) . Collectively , these data suggest that chronic lithium treatment through a sustained increase in BDNF expression leads to synaptic downscaling . The BDNF-TrkB dependence of lithium-mediated effects on mEPSC amplitudes was due to decreased surface expression of GluA1 , suggesting the removal of AMPARs by endocytosis . In neuronal synapses , dynamin has been shown to be a key mediator of endocytosis . To determine whether lithium promotes AMPAR endocytosis by a dynamin-dependent mechanism , we utilized a genetic approach with deletion of Dnm1 and Dnm3 and reversed the lithium-mediated decrease in AMPAR mEPSC amplitudes . To further establish that lithium triggers AMPAR endocytosis , we treated individual cells acutely with the dynamin inhibitor , Dyngo and demonstrated that lithium’s impact on AMPAR endocytosis can be rapidly countered by dynamin inhibition . Previous work has shown that AMPAR internalization/endocytic membrane trafficking of the postsynaptic neuron can be regulated in a dynamin-dependent process to control strength ( Lu et al . , 2007 ) . Taken together , our data show that lithium regulates dynamin-dependent endocytosis and rapidly removes AMPAR from the postsynaptic membrane altering synaptic efficacy . Altered glutamatergic signaling and hyperactivity have been implicated in the pathology of bipolar disorder . Mania in bipolar patients has been associated with elevated glutamate signaling ( Lan et al . , 2009; Ongür et al . , 2008 ) . In post-mortem brain tissue of individuals with bipolar disorder an increase in expression of the vesicular glutamate transporter ( VGlut1 ) mRNA and glutamate have been observed ( Eastwood and Harrison , 2010; Hashimoto et al . , 2007 ) . Human induced pluripotent stem cell ( IPSC ) derived neurons from patients with bipolar disorder were shown to have lower action potential thresholds , increased evoked and spontaneous action potentials , and larger action potential amplitudes suggesting they are hyperactive in comparison to neurons derived from healthy controls ( Mertens et al . , 2015 ) . Conversely , lithium treatment has been shown to reverse the glutamatergic hyperactivity from IPSC derived neurons from patients with bipolar disorder ( Mertens et al . , 2015 ) . In addition , recent work has suggested that lithium may alter neuronal activity by regulating G-protein gated inward rectifier K+ ( GIRK ) channels ( Farhy Tselnicker et al . , 2014 ) , which may modulate homeostatic scaling of AMPAR responses . Although , we cannot exclude lithium-mediated regulation of neuronal activity as a component of our observations , a parsimonious interpretation of our results suggests that direct biochemical effects of lithium mediated signaling is the key driver for the AMPAR downscaling we observed . Our findings showing that lithium triggers dynamin-dependent endocytosis of AMPARs may be a way in which to counteract the increased glutamatergic signaling associated with mania . Further work will be necessary to pursue this hypothesis and explore the role of lithium on dynamin-dependent endocytosis as a potential mechanism in antimanic action . Findings from the current study identify a critical role for BDNF in the antimanic and synaptic effects of lithium treatment . These results demonstrate that lithium specifically impacts excitatory neurotransmission by decreasing AMPAR-mEPSCs through dynamin-dependent endocytosis to reduce surface expression of AMPAR subunits . These data start to provide a framework to elucidate the mechanisms underlying lithium’s effect on synaptic transmission which may offer insight into novel therapeutic targets for the treatment of bipolar disorder .
Male C57BL/6 mice aged 6–8 weeks were habituated to the animal facilities for at least 7 days prior to behavior testing . The mice were maintained on a 12 hr light/dark cycle with ad libitum access to food and water , unless otherwise noted for lithium treatment groups . The Bdnffl/fl and Ntrk2 fl/fl mice were generated as previously described and maintained as homozygous crosses ( Luikart et al . , 2005; Rios et al . , 2001 ) . The inducible Bdnf knockout ( KO ) mice were generated as previously described ( Monteggia et al . , 2004 ) . The Dnm1 fl/fl and the Dnm3−/− null KO mice were generated as previously described ( Ferguson et al . , 2007; Raimondi et al . , 2011 ) . All behavioral testing was done with age and weight matched male mice that were balanced by genotype . In these experiments we utilized only male mice as previous work from our laboratory has shown that loss of BDNF can impact depression-related behaviors in a sex-dependent manner ( Autry et al . , 2009 ) . All electrophysiology and biochemistry experiments utilized male and female mice . All experiments were performed and data analyzed by an experimenter blind to drug condition and genotype . Animal protocols were approved by the Institutional Care and Use Committee at UT Southwestern Medical Center . The lithium treatment consisted of mice given 0 . 2% lithium chloride ( LiCl ) chow ( Harlan Teklad , East Millstone , NJ ) for 4 days and then switched to 0 . 4% LiCl ( Harlan Teklad ) chow for the remainder of the study , which lasted from 11 to 17 days total . All mice received water as well as a bottle of 0 . 9% sodium chloride ( NaCl ) to control for ion imbalances known to occur with lithium administration . Control mice were kept on identical chow except it did not contain lithium . Trunk blood was collected from all animals after the completion of the lithium treatment and behavioral testing . Whole blood was kept on ice until it was spun at 3000 RPM at 4°C for 10 min to separate red blood cells and serum . Lithium ion counts were made using a flame photometer ( Jenway PFP7 ) and concentration was calculated following the construction of a standard curve . Mice with serum lithium concentrations below 0 . 5 mM and above 3 mM were excluded from behavioral and biochemical analyses . Mice chronically treated with lithium ( 11–17 days ) were anesthetized with isoflurane before decapitation . The brain was removed and immersed for 2–3 min in ice-cold dissection buffer containing the following ( in mM ) : 2 . 6 KCl , 1 . 25 NaH2PO4 , 26 NaHCO3 , 0 . 5 CaCl2 , 5 MgCl2 , 212 sucrose and 10 glucose . The hippocampus was dissected out and cut with a vibratome into 400-μm-thick transverse sections in ice-cold dissection buffer continuously supplied with 95% O2 and 5% CO2 . Area CA3 was surgically removed from each slice immediately after sectioning . The sections recovered in oxygenated ACSF containing the following ( in mM ) : 124 NaCl , 5KCl , 1 . 25 NaH2PO4 , 26 NaHCO3 , 2 CaCl2 , 2 MgCl2 and 10 glucose , pH 7 . 4 ( continuously equilibrated with 95% O2 and 5% CO2 ) for 2–3 hr at 30°C . Hippocampal slices were transferred to the recording chamber and perfused with ACSF at a rate of 2–3 ml/min at 30°C . Field EPSPs ( fEPSPs ) were evoked by inserting a bipolar platinum-tungsten stimulating electrode ( Frederick Haer , Bowdoin , ME ) into Schaffer collateral/commissural afferents . The extracellular recording electrodes filled with ACSF ( resistance , 1–2 MΩ ) were placed into the CA1 just beneath the molecular layer . Baseline responses were collected every 30 s using an input stimulus intensity that induced 30–40% of the maximum response . After a 20 min stable baseline , an ascending series of stimulus input intensities ( range , 5 to 200 μA ) were applied and the input/output ( I/O ) curve was plotted by the slope of fEPSP versus afferent volley amplitude . Data are presented as mean ± SEM and analyzed using two-way ANOVA . Sidak’s multiple comparisons test were conducted after significant interaction effects were found . A p value of <0 . 05 was required for statistical significance . Dissociated hippocampal cultures were prepared as previously described ( Kavalali et al . , 1999; Reese and Kavalali , 2015 ) . Whole hippocampi were dissected from postnatal day 0–3 ( P0-P3 ) mice . The hippocampi were trypsinized ( ~10 mg/mL trypsin; Invitrogen ) , dissociated mechanically , and plated on Matrigel ( Corning Biosciences , Tewksbury , MA ) -coated coverslips for electrophysiology or directly onto tissue culture treated plates for protein collection for western blot analysis . At 1 d in vitro ( DIV ) , 4 µM cytosine arabinoside ( ARAC; Sigma ) was added . At 4 DIV , the ARAC concentration was decreased to 2 µM with a media change . Treatment with LiCl ( 1 mM ) solution ( Sigma ) or NaCl ( 1 mM ) solution ( Sigma ) was initiated at 4 DIV and lasted for 11–17 days . The Bdnffl/fl , Ntrk2 fl/fl , and Dnm1 fl/fl / Dnm3−/− cultures were infected with lentivirus expressing Cre recombinase ( Cre ) tagged with GFP or lentivirus expressing GFP alone as a control at 4 DIV . Lentivirus constructs and virus preparation from HEK293T/17 cells were prepared as previously described ( Akhtar et al . , 2009 ) . HEK293T/17 cell line was purchased from ATCC ( Cat . Number: CRL-11268 ) . All electrophysiology experiments and protein collected for western blot analysis were done on 15–21 DIV cultures . Whole-cell patch clamp recordings were performed on hippocampal pyramidal neurons as previously described ( Gideons et al . , 2014 ) . The external Tyrode’s solution contained ( in mM ) : 150 NaCl , 4 KCl , 2 CaCl2 , 1 . 25 MgCl2 , 10 glucose , and 10 Hepes ( ph 7 . 4 ) at ~300 mOsm . The pipette internal solution contained ( in mM ) : 110 K-gluconate , 20 KCl , 10 NaCl , 10 Hepes , 0 . 6 EGTA , 4 Mg-ATP , 0 . 3 Na-GTP , and 10 lidocaine N-ethyl bromide ( pH 7 . 3 ) at ~300 mOsm . Pipettes had a resistance between 3–7 MΩ . The junction potential between the internal and external solutions was ~12 mV and was subtracted from all recordings . AMPAR-mediated mEPSCs were recorded in the presence of 50 µM ( 2R ) -amino-5-phosphonovaleric acid ( AP5; Tocris , Bristol , UK ) , 1 µM tetrodotoxin ( TTX; EMD Millipore , Billerica , MA ) , and 50 µM picrotoxin ( PTX; Sigma ) . For the acute dynamin inhibitor studies , dynamin activity was inhibited in the postsynaptic neuron by adding 30 µM Dyngo-4A ( Abcam , Cambridge , MA ) to the internal pipette solution ( McCluskey et al . , 2013 ) . DMSO was added to the internal pipette solution as a control for these experiments . Data were acquired using a MultiClamp 700B amplifier and Clampex 10 . 0 software ( Molecular Devices , Sunnyvale , CA ) . Recordings were sampled at 100 µs , filtered at 2 kHz with a gain of 5 . No more than three recordings were obtained per coverslip . AMPAR-mEPSCs were analyzed from a 3–5 min recording using MiniAnalysis software by an experimenter blind to drug condition and genotype . Briefly , fresh frozen hippocampi were dissected and total RNA was extracted using Trizol ( Invitrogen ) following the manufacturer’s instruction . Conditions for cDNA synthesis , amplification , and primer sequences were previously described ( Adachi et al . , 2008 ) . Anterior hippocampal slices ( ~1 mm thick , 2–3 per mouse ) were dissected and flash-frozen following 11 days of lithium treatment or immediately following the last behavioral test depending on the experiment . Hippocampal tissue was lysed in a radio immunoprecipitation assay ( RIPA ) buffer containing: 50 mM Tris pH 7 . 4 , 1% Igepal , 0 . 1% SDS , 0 . 5% Na deoxycholate , 4 mM EDTA , 150 mM NaCl , phosphatase inhibitors ( 10 mM Na pyrophosphate , 50 mM NaF , 2 mM Na orthovanadate ) , and protease inhibitors ( cOmplete Mini tablets , Roche , Basel , Switzerland ) . Protein concentration was quantified with the Quick-Start Bradford assay ( Bio-Rad , Hercules , CA ) . Approximately 30 µg of protein was electrophoresed on SDS-PAGE gels and then transferred to nitrocellulose membranes . The membranes were incubated with primary antibodies overnight at the following dilutions: BDNF ( Abcam , Cat . Number: ab108319; RRID: AB_10862052 ) , 1:1000 , GAPDH ( Cell Signaling , Cat . Number: 2118S; RRID: AB_561053 ) 1:50 , 000 , phosphorylated GSK3β ( Ser9 , Cell Signaling , Cat . Number: 9323S ) and total GSK3β ( Cell Signaling , Cat . Number: 9315S ) 1:30 , 000 . Primary antibodies for phospho-GSK3β included 5% BSA . After washing , the membranes were incubated in anti-rabbit secondary antibodies: BDNF , 1:5000 , GAPDH , 1:10 , 000 , phospho-GSK3β and total GSK3β , 1:10 , 000 . Protein bands were detected using ECL then exposed to film . Following development of phospho-GSK3β , membranes were stripped using Restore PLUS Western Blot Stripping Buffer ( ThermoScientific , Waltham , MA ) , put in block , and then in primary antibody for total GSK3β . BDNF expression was normalized to GAPDH while a ratio of phospho- GSK3β intensity to total GSK3β was examined . The phospho-GSK3β and total GSK3β antibodies are known to recognize doublet bands ( Cell Signaling ) . Membrane biotinylation experiments were performed as previously described ( Nosyreva et al . , 2013 ) . Dissociated hippocampal cultures from C57BL/6 mice were incubated in Tyrode’s solution containing 1 mg/ml sulfo-NHS-LC-biotin ( Pierce/Thermofisher ) for 20 min on ice . The biotin reactions were quenched by incubating the cultures in Tris-buffered saline ( TBS ) with 15 mM ammonium chloride for 5 min on ice , and then washed twice with TBS for 5 min on ice . Following the second TBS wash , the cultures were lysed in RIPA buffer ( as described above ) for 10 min on ice and spun at 12 , 000 rpm for 5 min to remove non-solubilized material . Total protein concentration was quantified by Quik-Start Bradford assay ( Bio-rad ) . 100 µg of protein from each sample was incubated with 100 µL of washed UltraLink NeutrAvidin ( Pierce/Thermofisher ) immobilized beads and rotated overnight at 4°C . Beads were washed with three times with RIPA buffer , followed by three washes with TBS at 4°C . Protein was eluted from the beads with SDS-PAGE sample buffer supplemented with β-mercaptaethanol ( BME ) for 10 min at 95°C . Eluted surface protein and 20 µg of total protein in SDS-PAGE-BME sample buffer were resolved by 10% SDS-PAGE gel , transferred to nitrocellulose , and probed with anti-GluA1 antibody ( 1:1000 , Chemicon/EMD Millipore , Cat . Number: MAB2263 ) and anti-GAPDH antibody ( 1:50 , 000 ) overnight . Secondary anti-rabbit antibodies were at 1:2000 and 1:10 , 000 for GluA1 and GAPDH respectively . Surface GluA1 over total GluA1 ratio is reported . Data are reported as mean ± SEM . Statistical differences in the FST , locomotor tests , western blot , QPCR , mEPSC amplitude and frequency were assessed using unpaired two-tailed Student’s t test or one-way ANOVA when appropriate . Tukey and Bonferroni post hoc tests were used when appropriate . Differences in the cumulative probability histograms were assessed with Kolmogorov-Smirnov test . Statistical significance was defined at p≤0 . 05 . Statistical outliers were identified with the Robust regression and Outlier removal ( ROUT ) method . All t-test , ANOVAs , associated post hoc tests , and ROUT method statistical tests were performed with Prism 6 ( GraphPad , La Jolla , California ) . The Kolmogorov-Smirnov test was performed using Past 3 . 02 ( http://folk . uio . no/ohammer/past/ ) . | Nerve cells , or neurons , communicate with each other by releasing chemical messengers that bind to and activate receptor proteins on the surface of the other cells . The chemicals affect the connections between neurons , and many diseases – including bipolar disorder – are related to there being too much or too little of these chemicals in the brain . Patients with bipolar disorder experience periods of both depression and mania . During a manic episode , affected individuals typically feel elated and have more energy than usual despite needing less sleep , but also can also be irritable and impulsive . The exact cause of bipolar disorder is unknown . Patients with bipolar disorder often have low levels of a protein called brain-derived neurotrophic factor , or BDNF for short , which plays an essential role in keeping the brain healthy , and may also regulate the connections between neurons . One of the main treatments for bipolar disorder , a mood stabilizer called lithium , has also been linked to BDNF in previous studies; however , the details of the interaction were not clear . Gideons et al . studied how lithium works by feeding mice food pellets that contained lithium . After a few weeks , the mice had concentrations of lithium in their blood comparable to those of people taking the drug , as well as increased levels of BDNF in the brain . Gideons et al . then examined if BDNF was needed for the lithium’s ability to treat manic episodes . Mice exposed to another drug , amphetamine , normally move around a lot , mimicking the increased energy of someone with mania . As expected , feeding normal mice lithium blocked this effect of amphetamine , but feeding lithium to mutant mice that lack BDNF did not . This indicates that BDNF is indeed needed for the antimanic effect of lithium . Further experiments showed that BDNF is not needed for lithium’s antidepressant effect . By studying the animals’ brains , Gideons et al . went on to show that the lithium-fed mice had weaker connections between their neurons than mice that had eaten standard food . In the lithium-fed mice , many of the receptor proteins had been reabsorbed back into the neurons , lowering the ability of neurons to communicate with one another . This process depended on BDNF , suggesting that this protein is essential for lithium to suppress the connections between neurons . Taken together , these results reveal that the effects of lithium on both an animal’s brain and its behavior rely on BDNF . This knowledge should make it easier to develop new strategies and identifying new molecularly specific targets for treating bipolar disorder as well as other neuropsychiatric diseases . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | Chronic lithium treatment elicits its antimanic effects via BDNF-TrkB dependent synaptic downscaling |
Dopamine is critical for higher neural processes and modifying the activity of the prefrontal cortex ( PFC ) . However , the mechanism of dopamine contribution to the modification of neural representation is unclear . Using in vivo two-photon population Ca2+ imaging in awake mice , this study investigated how neural representation of visual input to PFC neurons is regulated by dopamine . Phasic stimulation of dopaminergic neurons in the ventral tegmental area ( VTA ) evoked prolonged Ca2+ transients , lasting ∼30 s in layer 2/3 neurons of the PFC , which are regulated by a dopamine D1 receptor-dependent pathway . Furthermore , only a conditioning protocol with visual sensory input applied 0 . 5 s before the VTA dopaminergic input could evoke enhanced Ca2+ transients and increased pattern similarity ( or establish a neural representation ) of PFC neurons to the same sensory input . By increasing both the level of neuronal response and pattern similarity , dopaminergic input may establish robust and reliable cortical representation .
The prefrontal cortex ( PFC ) plays an important role in adaptive behavior such as associative learning ( Duncan , 2001 ) . Dopaminergic input from the ventral tegmental area ( VTA ) is crucial for PFC function ( Schultz , 2007 ) . In primates , sensory cues , which are used in associative learning tasks , create specific temporal activity patterns in PFC neurons and a neural representation of the sensory cue ( Jacob et al . , 2013 ) . However , how dopamine contributes to form neural representations at the neuronal network level is largely unknown . One potential mechanism is that dopaminergic ( DA ) neurons target dopamine over a number of inhibitory and excitatory neurons via their widespread axonal arborizations ( Matsuda et al . , 2009 ) . Thus , investigating modification of neuronal activity at the population level will reveal the role of dopamine signaling in the formation of neural representation . Therefore , utilizing in vivo two-photon Ca2+ imaging in awake mice , this study investigated how neural representation in PFC neurons is developed under regulation by dopamine , in response to visual sensory input .
PFC neuronal activity was recorded through a cranial window at the secondary motor cortex ( M2 ) . The M2 is categorized as the dorsomedial PFC in rodents in some publications , based on both its anatomical features , including thalamocortical and cortical–basal ganglia connections , and its functional role in motor decision ( Sul et al . , 2011; Uylings et al . , 2003; Hoover and Vertes , 2007 ) . The M2 receives inputs from both the VTA and the secondary visual cortex lateral area ( V2L ) ( Figure 1A , B ) . Microelectrodes were implanted in these two brain areas to supply electrical stimulation ( Figure 1C , D ) . 10 . 7554/eLife . 02726 . 003Figure 1 . Ca2+ imaging setup and calcium transients in response to VTA stimulation with and without anesthesia . ( A ) Retrograde tracing . Fluorogold was injected into the prefrontal cortex ( PFC ) . Cells in the secondary visual cortex lateral area ( V2L ) were labeled . Scale bar , 250 μm . ( B ) Anterograde tracing . FITC-dextran was injected into the V2L . Labeled fibers were observed in the PFC . Scale bar , 100 μm . ( C ) Experimental setup for calcium imaging . A cranial window was opened over M2 ( PFC ) . Two electrodes were implanted in the ventral tegmental area ( VTA ) and the V2L . ( D ) Two-photon image of calcium indicator ( OGB-1 ) -labeled layer 2/3 cells in M2 ( neurons , green; sulforhodamine-101 counterstained astrocytes , red to orange ) . Approximately 50–80 cells were analyzed in each animal . Scale bar , 30 μm . ( E ) The Ca2+ transients evoked in an awake animal and in an animal under anesthesia . Population average of Ca2+ transients ( dF/F ) in response to 10 pulses at 50 Hz VTA stimulation ( n = 4 animals ) . VTA electrical stimulation was applied at the 30 s time point ( red arrowhead ) . In contrast to a clear response in awake mice ( left ) , the long-lasting Ca2+ transients were not detected in mice under anesthesia ( right , 4% isoflurane ) . The same animals were used for the ‘Awake’ and ‘Under anesthesia’ experiments . ( F ) The effect of isoflurane on Ca2+ transients . The summed values of Ca2+ transients from the 30 to 50 s time points were compared with the ‘Awake’ value ( changing ratio ) . Paired t test , *p < 0 . 05 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 003 To determine the single input response , the PFC neuronal response to VTA stimulation was first measured by recording Ca2+ transients . Most DA neurons are spontaneously active with firing patterns that range from regular tonic firing ( 1–5 Hz ) to phasic firing ( 40–50 Hz ) ( Robinson et al . , 2004; Grace and Bunney , 1984a; Grace and Bunney , 1984b ) . In addition , salient events such as experiencing a novel environment or receiving a reward-associated signal , evoke phasic firing in the VTA ( Schultz , 2007; Horvitz , 2000 ) . The PFC activity evoked by VTA stimulation in these physiological ranges was therefore examined . When low frequency VTA microstimulation was used ( 1 Hz–10 Hz , tonic range ) , Ca2+ transients in the PFC neurons were evoked and decayed within 5 s . This steep increase and short decay is comparable with the reported response to sensory input observed in cortical neurons of anesthetized animals ( Ohki et al . , 2005 ) . In contrast , high frequency VTA stimulation ( 40–50 Hz , phasic range ) evoked substantially elongated Ca2+ transients that lasted 20–30 s and then returned to the original baseline ( Figure 2 ) . In addition , these Ca2+ transients of the PFC neurons reached a peak relatively slowly , 6–7 s after stimulation . Furthermore , high frequency VTA stimulation robustly evoked Ca2+ transients in all tested animals , whereas low frequency VTA stimulation did not as only about half of the animals tested showed detectable Ca2+ transients in the PFC . Importantly , these long-lasting Ca2+ transients were only detected in awake and not in anesthetized mice ( Figure 1E , F ) , possibly because of the potential inhibition of voltage-gated calcium channels by isoflurane ( Herring et al . , 2009; Study , 1994 ) . This result highlights the advantage of a system using awake animals to elucidate dopamine regulation of neuronal responses in the PFC . 10 . 7554/eLife . 02726 . 004Figure 2 . Ca2+ transients in response to ventral tegmental area ( VTA ) stimulation . ( A ) The left panel shows the population average of Ca2+ transients across the cells of a single animal . The right panel shows the Ca2+ transients of each single neuron using a color map according to its dF/F value . A train of 10 pulses was applied at each frequency of VTA stimulation ( 1 , 5 , 10 , 20 , 30 , 40 , and 50 Hz ) . ( B ) Population average of Ca2+ transients across eight animals . VTA electrical stimulation was applied at the 30 s time point ( red arrowhead ) . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 00410 . 7554/eLife . 02726 . 005Figure 2—figure supplement 1 . Ca2+ transients in response to secondary visual cortex lateral area ( V2L ) stimulation . ( A ) The left panel shows the population average of Ca2+ transients across the cells of a single animal . The right panel shows the Ca2+ transients of each single neuron indicated in a color map according to its dF/F value . A train of 10 pulses was applied at each frequency of V2L stimulation ( 1 , 5 , 10 , 20 , 30 , 40 , and 50 Hz ) . ( B ) Population average of Ca2+ transients across eight animals . V2L electrical stimulation was applied at a 30 s time point ( red arrowhead ) . Each stimulation frequency evoked a Ca2+ response with a short decay time of 5 s . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 005 To test whether long-lasing Ca2+ transients can be evoked by other neural inputs , the same microstimulation protocol used in the VTA ( 1–50 Hz ) was applied to a region of the visual cortex , the V2L . Both stimulation frequencies only evoked short Ca2+ transients in PFC neurons ( Figure 2—figure supplement 1 ) . Input from the VTA , especially through high frequency , phasic stimulation can therefore induce a specific signal required to evoke long-lasting Ca2+ transients in PFC neurons . Phasic stimulation of the VTA is known to facilitate DA release from its terminals ( Tsai et al . , 2009; Lavin et al . , 2005 ) , therefore , it was hypothesized that DA receptors are involved in induction of the long-lasting Ca2+ transients . DA receptors are members of the G-protein coupled receptor family and are composed of two groups: D1 and D2 , which have opposing effects on intracellular signaling ( Seamans and Yang , 2004; Soltani et al . , 2013 ) . Selective DA receptor antagonists for each family were administered by intraperitoneal injection ( i . p . ) . Treatment with the D1 antagonist SCH23390 ( 1 mg/kg ) reduced the long-lasting Ca2+ transients by 50% and the decay time constant by 60% . In contrast , no detectable attenuation was observed in mice treated with the D2 antagonist eticlopride ( 0 . 5 mg/kg ) ( Figure 3A , B ) . In addition , the short Ca2+ response evoked by a 5 Hz VTA stimulation was not affected by the D1 or D2 antagonist ( Figure 3—figure supplement 1A , B ) . However , this short Ca2+ response was reduced by NMDA and AMPA antagonists ( Figure 3—figure supplement 1C , D ) . In the VTA , up to 65% of the neurons are dopaminergic and the others are GABAergic or glutamatergic ( Nair-Roberts et al . , 2008; Gorelova et al . , 2012 ) , and some DA neurons co-release glutamate ( Hnasko et al . , 2010 ) . However , glutamatergic terminals are dominant in the mesocortical pathway ( Gorelova et al . , 2012 ) . Inhibition of the short Ca2+ response by glutamate receptor antagonists suggests that glutamatergic neurons contribute to the 5 Hz responses . Initial experiments implicated the D1 receptors for the long-lasting Ca2+ transient . To exclude a potential role of glutamatergic neurons in initiating the long calcium transients , a cocktail of NMDA and AMPA receptor antagonists was used . However , the long-lasting Ca2+ transients were not affected by NMDA and AMPA antagonists ( Figure 3C , D ) . This result further suggests that the long-sustained increase in intracellular Ca2+ concentration is not mediated by glutamatergic local recurrent networks . Over all , pharmacological experiments clearly suggest D1 receptors are mainly involved in the long-lasting Ca2+ response in PFC neurons . 10 . 7554/eLife . 02726 . 006Figure 3 . Long-lasting Ca2+ transients depend on D1 receptors . ( A and B ) The effect of the D1 antagonist SCH23390 ( 1 mg/kg ) ( upper panel ) , the D2 antagonist eticlopride ( 0 . 5 mg/kg ) ( middle panel ) , and H2O , used as a control ( bottom panel ) , on the long-lasting Ca2+ transients evoked by 10 pulses at 50 Hz stimulation of the ventral tegmental area ( VTA ) ( n = 6 animals in each experimental group ) . ( C and D ) The effect of a cocktail of NMDA and AMPA antagonists ( CPP: 3 mg/kg , and CNQX: 10 mg/kg , i . p . ) on long-lasting Ca2+ transients ( n = 4 animals ) . ( A and C ) Population average of Ca2+ transients across animals in each group . VTA electrical stimulation was applied at the 30 s time point ( red arrowhead ) . ( B and D ) The changing ratio of the summed values of Ca2+ transients between the 30 and 50 s time points compared with the ‘Before’ value ( left panels ) , the peak of dF/F ( middle panels ) , and the decay time constant ( right panels ) . Paired t test with Holm's adjustment , *p < 0 . 05 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 00610 . 7554/eLife . 02726 . 007Figure 3—figure supplement 1 . Short Ca2+ transients do not depend on D1 or D2 receptors but on glutamate receptors . ( A and B ) The effect of D1 antagonist SCH23390 ( 1 mg/kg , upper panel ) , D2 antagonist eticlopride ( 0 . 5 mg/kg , middle panel ) and H2O , used as a control ( bottom panel ) , on the short Ca2+ transients evoked by 10 pulses at 5 Hz for ventral tegmental area ( VTA ) stimulation ( n = 6 animals in each experimental group ) . ( C and D ) The effect of a cocktail of NMDA and AMPA antagonists ( CPP: 3 mg/kg , and CNQX: 10 mg/kg , i . p . ) on short Ca2+ transients ( n = 4 animals ) . ( A and C ) Population average of Ca2+ transients across animals in each group . VTA electrical stimulation was applied at a 30 s time point ( red arrowhead ) . ( B and D ) To evaluate the effect of the drugs , the summed values of Ca2+ transients from the 30 to 35 s time points were compared with the ‘Before’ value ( changing ratio ) . Paired t test with Holm's adjustment , *p < 0 . 05 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 007 To resolve the potential role of dopamine-induced , long-lasting Ca2+ responses for PFC circuitry modification , the PFC neuronal response to sensory input from the V2L after combined repetitive stimulation of the V2L and VTA was investigated . An experimental paradigm composed of a pre-conditioning phase , conditioning phase , and test phase was used ( Figure 4A ) . In the pre-conditioning phase , V2L stimulations ( 20 Hz , five pulses ) were given three times to acquire the base response of PFC neurons to V2L inputs . In the conditioning phase , the VTA and V2L were simultaneously stimulated every minute for 30 min . After this conditioning phase , repetitive V2L stimulation ( 20 Hz , five pulses ) was applied at three time points: right after , 1 hr after , and 2 hr after completion of the conditioning phase ( test phase , Figure 4A ) , to determine whether or not the responses of the PFC neurons were modified . 10 . 7554/eLife . 02726 . 008Figure 4 . Combined repetitive stimulation of the secondary visual cortex lateral area ( V2L ) and the ventral tegmental area ( VTA ) causes a modification of prefrontal cortex ( PFC ) neuronal response . ( A ) Experimental design for stimulation of the V2L and VTA composed of three phases: pre-conditioning , conditioning , and test . Five pulses at 5 Hz and 10–15 pulses at 50 Hz were used for V2L and VTA electrical stimulation , respectively . T1: Timing1; T2: Timing2 . ( B ) Shift in population average of dF/F value . Post hoc tests revealed that dF/F values were significantly different in many of the comparisons between the time points in each conditioning paradigm ( Ryan's test ) . ( C ) Changing dF/F value calculated by subtracting the ‘Before’ value from the ‘2 hr after’ value ( dF/F ( 2 hr-before ) ) to simplify the results of ( B ) . The T1 conditioning paradigm showed a significantly larger temporal change than the others . ( D ) Difference in pattern similarity . Pattern similarity calculated from cosine similarity revealed that only T1 conditioning significantly increased the value of ‘2 hr after’ when compared with the ‘Before’ value . n = 8 animals in each experimental group . Ryan's post hoc test: *p < 0 . 05 , **p < 0 . 01 , *p < 0 . 001 . Error bars represent SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 00810 . 7554/eLife . 02726 . 009Figure 4—figure supplement 1 . Percentage distributions of neurons by dF/F in each conditioning group and reliability of calcium transients occurrence following three repetitive secondary visual cortex lateral area ( V2L ) stimulations . ( A ) Percentage distributions of neurons , pooled from eight animals in each conditioning group , are shown . The number of neurons showing a dF/F value above 0 . 2 increased 2 hr after stimulation ( right panel ) compared with ‘Before’ ( left panel ) , especially in T1 at ‘2 hr after’ . Note , the dF/F probability distribution in T1 conditioning at ‘1 hr after’ ( far right ) is comparable to the ventral tegmental area ( VTA ) only and T2 at ‘2 hr after’ . ( B ) The reliability of calcium transients occurrence across three repetitive V2L stimulations was calculated by Cronbach's alpha . Error bars represent 95% confidence intervals . None of the conditioning group , including T1 , showed a significant shift in the value of Cronbach's alpha , suggesting no reliability shift in either group . T1: Timing1; T2: Timing2 . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 00910 . 7554/eLife . 02726 . 010Figure 4—figure supplement 2 . No correlation between Ca2+ influx evoked by ventral tegmental area ( VTA ) stimulation and neuronal activity increase in response to secondary visual cortex lateral area ( V2L ) stimulation . Scatter plots of the Ca2+ transients evoked by VTA phasic stimulation from the 30 to 50 s time points against the increased value of dF/F , calculated by subtracting the dF/F value of ‘Before’ from ‘2 hr after’ conditioning ( dF/F ( 2 hr−Before ) ) of Timing1 conditioning . Each plot represents one animal and each dot represents a different single neuron of the animal . There is no significant correlation between the amount of Ca2+ transients evoked by VTA stimulation and the increased value of dF/F . This indicates that the higher amount of intracellular Ca2+ is not the main factor modifying the response to input from the sensory cortex . DOI: http://dx . doi . org/10 . 7554/eLife . 02726 . 010 For the conditioning phase , the effect of timing on V2L and VTA stimulation was also tested by using two different time intervals , Timing1 ( T1 ) and Timing2 ( T2 ) ( Figure 4A ) . In T1 , V2L stimulation was applied half a second before VTA phasic stimulation . T1 simultaneously simulates a visual experience and VTA activation . For T2 , V2L stimulation was applied 45 s after VTA phasic stimulation , which is 15 s before the next VTA stimulation , when no Ca2+ transients were observed ( Figure 4A ) . Therefore , T2 is used as a control , at a point when visual and VTA inputs are temporally separated . As additional controls , the V2L and the VTA were stimulated separately during the conditioning phase . Two-way repeated measures ANOVA with temporal change ( before , right after , 1 hr after , and 2 hr after conditioning ) and conditioning paradigm ( V2L only , VTA only , Timing1 , and Timing2 ) as factors , revealed that population averages of dF/F values of Ca2+ transients signal showed significant differences in temporal change ( F ( 3 , 84 ) = 48 . 486 , p < 0 . 0001 ) and the conditioning paradigm × temporal change interaction ( F ( 9 , 84 ) = 3 . 173 , p = 0 . 0024 ) ( Figure 4B ) . T1 conditioning significantly increased the dF/F value more than other conditioning paradigms , indicating that T1 conditioning has the largest modification effect on the PFC response to a sensory input ( one-way ANOVA: F ( 3 , 28 ) = 6 . 504 , p = 0 . 0018; post hoc Ryan's test: p < 0 . 01; Figure 4C ) . This increase in value is due to an increased number of neurons showing high dF/F values ( Figure 4—figure supplement 1A ) . In addition , there was no significant correlation between the Ca2+ level in response to VTA stimulation ( at conditioning ) and the increased Ca2+ level in response to V2L stimulation ‘2 hr after’ T1 conditioning ( test phase ) ( Figure 4—figure supplement 2 ) , suggesting that the increased neuronal response with V2L stimulation is independent of how well the neuron responds to the DA input . This relatively unexpected conclusion may be explained by the fact that the network dynamics of the PFC microcircuit are composed of inhibitory and excitatory neurons , both of which express DA receptors . Finally , the neuronal population dynamics were investigated through analysis of the pattern similarity ( cosine similarity , see the ‘Materials and methods’ section ) of neuronal activity in the PFC across three repetitions of V2L stimuli , conducted at four time points: ‘before’ , ‘right after’ , ‘1 hr after’ , and ‘2 hr after’ after the conditioning phase . Two-way repeated measures ANOVA on the temporal changes of the mean cosine similarities in each conditioning paradigm revealed significant differences in the temporal factor ( F ( 3 , 84 ) = 7 . 676 , p = 0 . 0001 ) and the time × condition paradigm interaction ( F ( 9 , 84 ) = 2 . 483 , p = 0 . 0145; Figure 4D ) . Post hoc tests also showed that the cosine similarity in T1 was significantly increased in ‘2 hr after’ compared with ‘before’ ( Ryan's test , p < 0 . 01; Figure 4D ) . This increase in cosine similarity in T1 conditioning is not simply due to an increased number of high dF/F neurons ( Figure 4—figure supplement 1A ) , because ‘1 hr after’ in T1 conditioning had a significantly increased cosine similarity ( Figure 4D ) but its population probability distribution was not significantly different from that of VTA only ( 2 hr after ) or T2 ( 2 hr after ) ( two-sample Kolmogorov–Smirnov test , p = 0 . 5128 and p = 0 . 1418 , respectively ) ( Figure 4—figure supplement 1A ) . This suggests that the increase in cosine similarity in T1 conditioning was not due to an increased number of high dF/F neurons . In addition , the increased cosine similarity was not due to increased reliability of Ca2+ transients occurrence in response to three repetitive V2L stimulations , because the reliability calculated by Cronbach's alpha did not change between ‘before’ and ‘2 hr after’ in any of the conditioning groups , including T1 ( Figure 4—figure supplement 1B ) . These results indicate that only T1 conditioning improved the pattern similarity in PFC neurons , and the increase in pattern similarity could be the result of circuit ( network ) modification , and not simply the result of increased reliability of Ca2+ transients occurrence or an increased number of high responding dF/F neurons . This study revealed that only coincident visual sensory input with long-lasting Ca2+ transients increased neuronal activity and induced robust repeatable neuronal responses , at the population level , to the same sensory input . By increasing both dF/F and pattern similarity , DA input may enhance PFC activity and establish cortical representation ( Xue et al . , 2010 ) . In fact , it is known that DA release plays a major role in dynamic cortical remodeling in the auditory cortex ( Bao et al . , 2001 ) , suggesting that the phenomenon observed here in the T1 conditioning paradigm may represent the neurophysiological basis for dynamic neural events . Besides network modification in PFC , it has also been demonstrated that PFC neurons show long-lasting Ca2+ transients that depend on the D1 receptor pathway . This newly identified physiological phenomenon might have an important function in Ca2+ regulation of neuronal processes ( Berridge , 1998; Brenowitz et al . , 2006; Roussel et al . , 2006; Bardo et al . , 2006 ) . For example , association learning is only successful if the cue signal is less than 20–30 s before the reward signal . Long-lasting Ca2+ transients may be tightly associated with this time window in association learning . Taken together , the long-lasting Ca2+ transients reported here could be a key phenomenon required to explain the dynamics of the dopaminergic neural network and its role in PFC cognitive functions .
Neuronal retrograde tracer Fluorogold ( 4%; Fluorochrome , Denver , CO ) or FITC-conjugated dextran ( Life Technologies , Carlsbad , CA ) were injected into the M2 or V2L , respectively , and 48 hr later , mice were fixed with 4% formalin in phosphate buffer . Brains were sliced using a vibratome and Fluorogold or FITC was identified by using fluorescence microscopy . All procedures were conducted according to the animal welfare guidelines of the NIH and approved by the NIH Animal Care and Use Committee . C57BL/6 mice ranging in age from 2 to 4 months were used . Throughout all procedures , body temperature was maintained at 37°C using a heating pad . Anesthesia was induced with Avertin ( 2 . 2 . 2-tribromoethanol; Sigma-Aldrich , St . Louis , MO ) , and mice were placed in a stereotaxic device . Stainless bipolar stimulating electrodes were implanted in the V2L ( 2 . 5 mm lateral and 2 . 5 mm posterior from bregma , 0 . 5 mm below the cortical surface ) and VTA ( 0 . 5 mm lateral and 3 . 1 mm posterior from bregma , 4 . 5 mm below the cortical surface ) . A 1 . 5 mm craniotomy ( with the dura carefully removed ) was opened over M2 ( 0 . 5 mm lateral and 1 . 0 mm anterior from bregma ) . Multi-cell bolus loading of neocortical cells with the calcium indicator Oregon-BAPTA Green 1-AM ( OGB-1-AM; Life Technologies ) and astrocyte marker sulforhodamine 101 ( SR101; Life Technologies ) was performed as previously described ( Stosiek et al . , 2003; Nimmerjahn et al . , 2004 ) . This multi-cell bolus loading was performed in superficial layer 2/3 ( L2/3 ) . The craniotomy was then covered with silicone ( Kwik-Sil Adhesive; World Precision Instruments , Sarasota , FL ) and sealed with a glass coverslip . A metal bar was glued directly on the skull with dental acrylic for future attachment to an imaging frame . About ∼2 hr after surgery , when the mouse had completely recovered from the anesthesia , cortical activity , measured by evoked Ca2+ transients , was imaged in the awake , head-fixed mouse . The mouse was placed under the microscope . Cortical activity was imaged using a two-photon microscope ( Olympus Fluoview; Olympus , Japan ) equipped with a 25 × ( 1 . 05 NA ) water-immersion objective ( Olympus ) . Excitation wavelength was 870 nm ( Mai-Tai oscillator; Spectra-Physics , Santa Clara , CA ) . Images ( 256 × 256 pixels ) were acquired at a frame rate of 2 . 3 Hz . Low quality images ( when cells showed unclear boarders ) were not used for analysis . Imaging started at time 0 , however high background signals ( ∼0 . 03 dF/F , lasting 2–3 s ) were present at time 0 due to mechanical noise . Therefore , figures show the Ca2+ transients starting from 10 s to demonstrate the neuronal activity that occurs in response to electrical stimulation ( Figures 1–3 , Figure 2—figure supplement 1 and Figure 3—figure supplement 1 ) . SCH23390 ( 1 mg/kg ) , eticlopride ( 0 . 5 mg /kg ) , CPP ( 3 mg/kg ) , and CNQX ( 10 mg/kg ) ( Sigma–Aldrich ) were administered by i . p . injection . The injected volume was adjusted to 1% of the animal's body weight . For VTA stimulation , a biphasic pulse of 1 ms duration was used in all experiments . To evoke the long-lasting Ca2+ transients , 50–400 μA and 10–15 pulses were applied . The experimental protocol shown in Figures 1–3 used 400 μA and 10 pulses . VTA stimulation with 50–400 μA and 10–15 pulses was used for the conditioning experiment ( Figure 4 ) . For the V2L , a biphasic pulse of 1 ms and 250 μs duration was used in the experiments shown in Figure 2—figure supplement 1 and Figure 4 , respectively . Stimulation with 350 or 400 μA was used in Figure 2—figure supplement 1 . To evoke dF/F values below 0 . 2 , in response to V2L stimulation at ‘before’ , current was adjusted ( 50–300 μA ) in each animal ( experiments shown in Figure 4 ) . Data were analyzed with custom-written programs in MATLAB ( Mathworks , Natick , MA ) and ImageJ ( NIH , Bethesda , MD ) . To remove motion artifacts from in vivo calcium imaging , the ImageJ plugin TurboReg for image alignment ( Thévenaz et al . , 1998 ) was used . Individual cells were semi-automatically detected ( Supplementary file 1 ) . SR101-stained astrocytes were excluded from data analysis . For each cell , fluorescence change was defined as dF/F = ( F1 − F0 ) /F0 , where F1 is fluorescence at any time point , and F0 is the baseline fluorescence , defined as the median of fluorescence values measured within 40 s before and after the time point of F1 ( for responses to VTA stimulation ) , or within 2 . 5 s before and after the time point of F1 ( for responses to V2L stimulation ) . To calculate the decay time constant ( tau ) of Ca2+ transients ( Figure 3B , D ) , the following equation was used for fitting the time course of dF/F during the decay period: dF/F ( t ) = dF/Fmax × e−t/tau , where dF/Fmax is the peak dF/F value and t is elapsed time after dF/F peaks . To calculate the population average of dF/F shown in Figure 4 , the averages across cells from the three V2L stimulations were averaged . Pattern similarity in Figure 4D was measured using cosine similarity ( cosine of the angle between two vectors ) in each pair of vectors , which are composed of the N-dimension of dF/F values of each neuronal response , where N is the total number of neurons . For the three repetitions of V2L stimulation at each time point , three cosine similarities were measured between the first and second , second and third , and first and third V2L stimulations , and then these three cosine similarities were averaged to show the pattern similarity of the time points . In Figure 4B , the result of post hoc tests revealed that dF/F values were significantly different in many of the comparisons between the time points in each conditioning paradigm ( Ryan's test , p < 0 . 05 ) . In Figure 4C , to simplify , the values of dF/F at 2 hr after conditioning were compared , and standardized by subtracting the ‘before’ values from ‘2 hr after’ among the different conditioning groups ( dF/F ( 2 hr−before ) ) . | Around 120 years ago , Ivan Pavlov unintentionally sparked a new field of psychology research . He did so by noting that his dogs had learned to associate the sound of the bell that he rang before feeding them with the food itself , such that they would salivate upon hearing the bell even when there was no food present . This form of learning—now known as associative learning—has since been demonstrated in species from honeybees to humans . For the brain to associate two events , such as the sound of a bell and the delivery of food , it must encode the first event and keep that information available or ‘on-line’ until the occurrence of the second event , at which point the two can be linked together . This process takes place in part of the brain called the prefrontal cortex , but the mechanism by which it occurs is largely unclear . Now , Iwashita has obtained new insights into the molecular basis of associative learning by studying how the activity of the prefrontal cortex is affected by the activity of a second region of the brain . This second region , called the ventral tegmental area , is part of the brain's reward circuit: it becomes active whenever an animal experiences a desirable event , such as receiving food , and supplies a neurotransmitter called dopamine to its target areas , which include the prefrontal cortex . Electrodes were used to mimic the changes in brain activity that occur when a mouse learns to associate a visual stimulus with a reward: this involved repeatedly activating the visual cortex in a conscious mouse , followed by activation of the ventral tegmental area . Short-lived increases in calcium levels were seen in the prefrontal cortex , raising the possibility that these ‘calcium transients’ are the signal that enables the brain to link two events . Moreover , blocking proteins called dopamine D1 receptors in the prefrontal cortex reduced the calcium transients , which is consistent with existing evidence that dopamine from the ventral tegmental area is required for associative learning . Intriguingly , the calcium transients lasted for roughly 30 s , which is also the maximum length of time by which a stimulus and a reward can be separated and still be associated . Given that the calcium transients could not be detected in anesthetized mice , a full understanding of the mechanisms underlying associative learning may require studies of the conscious brain . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"neuroscience"
] | 2014 | Phasic activation of ventral tegmental neurons increases response and pattern similarity in prefrontal cortex neurons |
Adult neural stem cells , located in discrete brain regions , generate new neurons throughout life . These stem cells are specialized astrocytes , but astrocytes in other brain regions do not generate neurons under physiological conditions . After stroke , however , striatal astrocytes undergo neurogenesis in mice , triggered by decreased Notch signaling . We used single-cell RNA sequencing to characterize neurogenesis by Notch-depleted striatal astrocytes in vivo . Striatal astrocytes were located upstream of neural stem cells in the neuronal lineage . As astrocytes initiated neurogenesis , they became transcriptionally very similar to subventricular zone stem cells , progressing through a near-identical neurogenic program . Surprisingly , in the non-neurogenic cortex , Notch-depleted astrocytes also initiated neurogenesis . Yet , these cortical astrocytes , and many striatal ones , stalled before entering transit-amplifying divisions . Infusion of epidermal growth factor enabled stalled striatal astrocytes to resume neurogenesis . We conclude that parenchymal astrocytes are latent neural stem cells and that targeted interventions can guide them through their neuronal differentiation .
Neurogenesis is extremely limited in the adult brain . In most mammals , specialized astrocytes in the subventricular zone ( SVZ ) and hippocampal dentate gyrus ( DG ) are stem cells and generate new neurons continuously , but apart from that , the brain’s ability to replace lost neurons is very limited . However , in response to an experimental stroke or excitotoxic lesion , some astrocytes in the mouse striatum can generate neurons ( Magnusson et al . , 2014; Nato et al . , 2015 ) . This neurogenic response is regulated by Notch signaling and can be activated even in the uninjured mouse striatum by deleting the Notch-mediating transcription factor Rbpj ( Magnusson et al . , 2014 ) . Striatal astrocytes undergo neurogenesis by passing through a transit-amplifying cell stage . But it is not known whether these astrocytes become bona fide neural stem cells . If they do , this could have far-reaching implications for regenerative medicine . Astrocytes make up a large fraction of all brain cells ( 10–20% in mice ) ( Sun et al . , 2017 ) and are distributed throughout the central nervous system . They would thus represent a very abundant source of potential neural stem cells that might be recruited for therapeutic purposes . Although certain injuries and Rbpj deletion can both trigger neurogenesis by astrocytes , it almost exclusively does so in the striatum . And even within the striatum , primarily the astrocytes in the medial striatum readily activate neurogenic properties ( Figure 1a ) . This suggests that neurogenic parenchymal astrocytes either occupy an environmental niche favorable to neurogenesis or that only they have an inherent neurogenic capacity . In order to recruit astrocytes for therapeutic neurogenesis , a first step is to understand the mechanisms underlying this process . If these mechanisms are understood , they could potentially be targeted to induce localized therapeutic neurogenesis throughout the central nervous system . Here , we generated two separate single-cell RNA sequencing datasets to study neurogenesis by parenchymal astrocytes in mice . We found that , at the transcriptional level , Rbpj-deficient striatal astrocytes became highly similar to SVZ neural stem cells and that their neurogenic program was nearly identical to that of stem cells in the SVZ , but not the DG . Interestingly , astrocytes in the non-neurogenic somatosensory cortex also initiated a neurogenic program in response to Rbpj deletion , but all stalled prior to entering transit-amplifying divisions and failed to generate neuroblasts . In the striatum , too , many astrocytes halted their development prior to entering transit-amplifying divisions . We found that stalled striatal astrocytes could be pushed into transit-amplifying divisions and neurogenesis by an injection of epidermal growth factor ( EGF ) , indicating that it is possible to overcome roadblocks in the astrocyte neurogenic program through targeted manipulations . Taken together , we conclude that parenchymal astrocytes are latent neural stem cells . We posit that their intrinsic neurogenic potential is limited by a non-permissive environment . Recruiting these very abundant latent stem cells for localized therapeutic neurogenesis may be possible but is likely to require precise interventions that guide them through their neurogenic program .
To understand the cellular mechanisms underlying neurogenesis by parenchymal astrocytes , we decided to perform single-cell RNA sequencing of striatal astrocytes undergoing neurogenesis in vivo . To this end , we generated two separate single-cell RNA sequencing datasets , each with complementary strengths . For the first dataset , which we call the Cx30-CreER dataset , we used Connexin-30 ( Cx30; official gene symbol Gjb6 ) -CreER mice to delete Rbpj and activate heritable tdTomato or YFP expression in astrocytes throughout the brain . This method targeted up to 100% of astrocytes in the striatum ( tdTomato expression in 92 ± 10% of glutamine synthetase+ , S100+ cells; mean ± S . D . ; n = 6 mice ) . These mice also target neural stem cells in the SVZ , but no other brain cell types ( Magnusson et al . , 2014 ) . Rbpj deletion in healthy mice was chosen as the means by which to activate the neurogenic program of astrocytes . That is because ( 1 ) it constitutes a single , precisely timed trigger throughout the brain , ( 2 ) it mimics the endogenous stimulus of reduced Notch signaling by which stroke induces neurogenesis by astrocytes ( Magnusson et al . , 2014 ) , and ( 3 ) it does not induce potentially confounding cellular processes like reactive gliosis or cell death . Striatal tdTomato+ cells were isolated by flow cytometry from mice homozygous for the Rbpj null allele 2 , 4 and 8 weeks after tamoxifen administration , and from control mice with intact Rbpj , 3 days after tamoxifen administration ( hereafter called ground-state astrocytes ) ( Figure 2—figure supplement 1a–d ) . Sequencing libraries were prepared using Smart-seq2 ( Picelli et al . , 2013 ) . In addition to parenchymal astrocytes , Cx30-mediated recombination also targets SVZ neural stem cells , whose progeny might be sorted along with striatal astrocytes and , thus , confound our results . We therefore generated a second single-cell RNA sequencing dataset , for which labeling of cells was restricted to the striatum . Striatal astrocytes were selectively recombined via injection of an adeno-associated virus ( AAV ) expressing Cre under the control of the astrocyte-specific GFAP promoter into the striatum of Rbpjfl/fl; R26-loxP-Stop-loxP-tdTomato mice ( Figure 1b ) . We isolated tdTomato+ cells 5 weeks after virus injection , a time point at which Ascl1+ astrocytes , transit-amplifying cells and neuroblasts are all present ( Magnusson et al . , 2014 ) . We sequenced these cells using 10X Chromium by 10X Genomics . Despite the GFAP promoter , the AAV-Cre dataset also contained oligodendrocytes and some microglia ( Figure 1—figure supplement 1 ) . These were excluded from the final dataset based on our previous finding that they are not involved in a neurogenic program ( Magnusson et al . , 2014 ) . The AAV-Cre dataset contained a somewhat higher proportion of transit-amplifying cells than the Cx30-CreER dataset ( Figure 2—figure supplement 1f ) , likely because the AAV-Cre dataset only contained cells from 5 weeks after Rbpj-KO , near the peak of transit-amplifying divisions ( Magnusson et al . , 2014 ) . We first asked whether we could reconstruct striatal astrocyte neurogenesis computationally using the AAV-Cre dataset . Using Uniform Manifold Approximation and Projection ( UMAP ) , we found that the sequenced cells indeed included Rbpj-deficient astrocytes ( Cx30+ ) , activated astrocytes and transit-amplifying cells ( Egfr+ , Ascl1+ , Mki67+ , Dcx+ ) and neuroblasts ( Dcx+ ) , up until the migratory neuroblast stage ( Nav3+ ) ( Figure 1c–d ) . We conclude that it is possible to use single-cell RNA sequencing to study the transcriptional mechanisms underlying the astrocyte neurogenic program . We were interested in the transcriptional changes that take place as astrocytes first initiate the neurogenic program . For this question , we primarily used the Cx30-CreER dataset , which included astrocytes both at ground state and at 2 , 4 and 8 weeks after Rbpj deletion ( Figure 2a ) . Furthermore , the Cx30-CreER mice allowed us to study the neurogenic process in the uninjured brain , without the potentially confounding effects of a needle injection . Two weeks after Rbpj deletion , astrocytes had upregulated genes mostly related to transcription , translation and metabolism ( Figure 2b , S1g , Supplementary file 1 ) . To study in detail how gene expression changed over the course of neurogenesis , we next reconstructed the differentiation trajectory of these astrocytes computationally using Monocle ( Trapnell and Cacchiarelli , 2014; Figure 2—figure supplement 2; Methods ) . This analysis captured the progression from astrocyte to neuroblast in pseudotime ( Figure 2c ) . The pseudotemporal axis reflected neurogenesis , as revealed by plotting along pseudotime the expression of genes that are dynamically induced in canonical neurogenesis ( Figure 2d ) . This analysis confirmed the previous finding that neurogenesis by parenchymal astrocytes involves transit-amplifying divisions ( Magnusson et al . , 2014; Niu et al . , 2013 ) , and does not occur through direct transdifferentiation , a process during which intermediate stages of differentiation are skipped ( Briggs et al . , 2017 ) . In order to better understand the transcriptional changes that take place as striatal astrocytes initiate neurogenesis , we used Monocle’s gene clustering algorithm , which clusters genes whose expression levels vary similarly along pseudotime ( Figure 2e ) . These gene clusters were then characterized using the gene ontology ( GO ) tool DAVID ( Huang et al . , 2009 ) . This approach enabled us to see which biological processes changed as differentiation proceeded ( Supplementary file 2 ) . We found that , as astrocytes initiated their neurogenic program , genes associated with metabolism , particularly lipid ( e . g . Fabp5 , Fasn , Elovl5 ) and carbohydrate metabolism ( e . g . Gapdh , Aldoa , Aldoc ) were at first upregulated but dramatically downregulated as cells entered transit-amplifying divisions ( Figure 2e–f , [Cluster 1] ) . This suggested that the neurogenic program requires high metabolic activity to prepare for transit-amplifying divisions . In addition to these metabolic genes , the expression of genes involved in the regulation of transcription ( e . g . mRNA splicing genes Snrpb , Srsf5 , Srsf3 ) and translation ( e . g . ribosomal subunits Rpl17 , Rpl23 , Rps11 ) progressively increased throughout the neurogenic process ( Figure 2e–f , [Cluster 2] ) . Accordingly , we found that the number of genes detected per cell increased in the initial stages of neurogenesis ( Figure 2g ) : pre-division astrocytes expressed 1 . 4 times as many genes as ground-state astrocytes ( p=10−12; 95% C . I . 1 . 3–1 . 5; Materials and methods ) . In theory , such an increase in the number of detected genes could be an artifact caused by a simultaneous increase in cell size; however , no cell size increase was observed during this early phase of neurogenesis , as measured by flow cytometer forward scatter ( Figure 2h ) . The changes to metabolism and gene expression activity described above were the two dominating transcriptional changes that occurred early in the neurogenic process of astrocytes . At the peak of metabolic gene expression , astrocytes initiated transit-amplifying divisions ( Figure 2e; Materials and methods ) . Accordingly , they upregulated genes associated with cell division ( e . g . Ccnd2 , Ccng2 , Cdk4; Figure 2e–f , [Cluster 3] ) . Finally , as they exited transit-amplifying divisions , cells upregulated classical marker genes of neuroblasts ( e . g . Cd24a , Dcx , Dlx1/2 , Tubb3 [βIII tubulin]; Figure 2e–f , [Cluster 3] ) , indicative of early neuronal commitment and differentiation . Interestingly , one group of genes showed a remarkable pattern of initial downregulation followed by upregulation as cells developed into neuroblasts ( Figure 2e–f [Cluster 4] , S4a ) . This gene group consisted mainly of genes encoding proteins localized to synapses ( e . g . Cplx2 , Ctbp2 , Shank1 , Chrm4 , Bdnf ) . This suggested that synapse maintenance was compromised very early as astrocytes initiated neurogenesis , but also that some of the same genes are important for synapse maintenance in both astrocytes and neurons . The gene expression changes described above were confirmed in the AAV-Cre dataset , using a differential expression analysis along pseudotime ( Figure 2i ) . But because the AAV-Cre dataset only contained cells isolated 5 weeks after Rbpj deletion , this dataset did not contain ground-state astrocytes and could not capture changes that happened early in the neurogenic program . Throughout the initial phase of neurogenesis and up until the point where cells entered transit-amplifying divisions , astrocytes maintained normal astrocyte morphology and expression of common marker genes ( e . g . Aqp4 , S100b , Slc1a2 [Glt1] , Glul ) ( Figure 2—figure supplement 3b–c , Video 1 ) . They never showed signs of reactive gliosis ( e . g . morphological changes or upregulation of Gfap ) , indicating that reactive gliosis is not required for neurogenesis by astrocytes in vivo . As the astrocytes entered the neurogenic program , they did eventually lose their typical branched morphology; however , they did so only gradually , and only after entering transit-amplifying divisions . In fact , they retained rudiments of astrocytic processes for several transit-amplifying cell divisions ( Figure 2—figure supplement 3c , Video 1 ) . As an additional sign that they originated from parenchymal astrocytes , many clustered transit-amplifying cells retained trace levels of the astrocyte marker protein S100β , even though they no longer expressed the S100b gene . S100β protein is found in parenchymal astrocytes but not in SVZ stem cells ( Figure 2—figure supplement 3d–g ) . Accordingly , we did not detect any S100β protein in SVZ transit-amplifying cells ( Figure 2—figure supplement 3d–g ) . This suggests that the low levels of S100β protein seen in striatal transit-amplifying cells were lingering remnants of their parenchymal astrocyte origin . It indicates that trace levels of S100β protein can function as a short-term lineage-tracing marker for transit-amplifying cells derived from parenchymal astrocytes . A recent study in Drosophila melanogaster showed that many quiescent neural stem cells are halted in the G2 phase of the cell cycle ( Otsuki and Brand , 2018 ) . However , we analyzed EdU incorporation in dividing striatal astrocytes . We found some single astrocytes that had not yet divided but were EdU+ ( Figure 2—figure supplement 3c ) . This suggested that these astrocytes initiated the cell cycle by synthesizing DNA and thus had not been suspended in the G2 phase . We were interested in a direct comparison between neurogenic parenchymal astrocytes and adult neural stem cells to see how their neurogenic transcriptional programs relate to one another . A number of published single-cell RNA sequencing datasets exist for mouse neural stem cells generated by 10X Chromium or Smart-seq2 . We first analyzed our AAV-Cre dataset together with a previously published 10X Chromium dataset for SVZ neural stem cells and their progeny ( Zywitza et al . , 2018 ) . A UMAP analysis of these cells together with our AAV-Cre dataset revealed that the striatal astrocytes and their progeny clustered closely together with the corresponding stages of SVZ cells ( Figure 3a–b ) . This indicated that neurogenesis in these two brain regions proceeds through the same general developmental stages . Particularly interesting was the observation that our ‘Astrocyte Cluster 2’ grouped closely together with SVZ neural stem cells ( Figure 3b ) , indicating that these cells are transcriptionally similar . Because our AAV-Cre dataset only contained cells isolated 5 weeks after Rbpj deletion , this dataset could not reveal how striatal astrocytes in their ground state relate to SVZ stem cells . We therefore next turned to our Cx30-CreER dataset and analyzed these cells together with SVZ cells from another previously published Smart-seq2 dataset ( Llorens-Bobadilla et al . , 2015; Figure 3c ) . We used Monocle to perform pseudotemporal ordering of our Cx30-CreER dataset together with the SVZ cells , using genes involved in the GO term Neurogenesis ( GO:0022008 ) , and projected the cells onto the pseudotime axis ( Figure 3d ) . Interestingly , on this neurogenic trajectory , ground state astrocytes were located upstream of the SVZ stem cells , as if representing highly quiescent cells that pass through an initial activation phase before joining the SVZ stem cells ( Figure 3d ) . We next compared the gene expression changes that occur in SVZ cells with those in our Cx30-CreER dataset . We used Monocle’s gene clustering analysis of only the SVZ cells , without the striatal cells . We first confirmed the previously published finding that SVZ neurogenesis is accompanied by large-scale changes in genes associated with metabolism ( Figure 3—figure supplement 1a–c; Llorens-Bobadilla et al . , 2015 ) . Intriguingly , however , whereas SVZ neural stem cells appear to be maintained at a high baseline level of metabolic gene expression ( Figure 3—figure supplement 1b–c ) , striatal astrocytes first passed through a phase of metabolic gene upregulation to reach this level , seen by superimposing the striatum-only ( Figure 2e ) and SVZ-only ( Figure 3—figure supplement 1b ) metabolic gene curves ( Figure 3e ) . This suggested that the neurogenic program in striatal astrocytes starts with an activation phase during which these astrocytes begin to resemble SVZ stem cells with regard to metabolism-associated genes . Similarly , evidence for an initial activation phase was found also in gene expression activity: As described above , we detected a 1 . 4-fold difference in the number of genes detected in Rbpj-deficient astrocytes compared to ground-state astrocytes . This difference is similar in magnitude to a previously detected 1 . 7-times difference between adult SVZ stem cells and striatal astrocytes ( Gokce et al . , 2016 ) . This suggested that the gene expression activity of Rbpj-deficient astrocytes began to quantitatively resemble that of neural stem cells . We asked if we could use our RNA sequencing data to address whether the differentiation potency of striatal astrocytes begins to resemble that of SVZ stem cells . ‘Signaling entropy’ is a measure of the interconnectedness of a cell’s proteome . It represents the average number of possible protein-protein interactions by all proteins expressed by a cell . It is a reliable proxy for a cell’s differentiation potency ( Teschendorff and Enver , 2017; Chen and Teschendorff , 2019 ) , the underlying rationale being that a highly interacting proteome enables promiscuous signaling , which in turn facilitates responding to a variety of differentiation signals . As a validation of the signaling entropy concept , we plotted signaling entropy versus pseudotime in the SVZ dataset and saw that SVZ cells exhibit their maximum signaling entropy as activated stem cells or transit-amplifying cells ( Figure 3—figure supplement 1d ) . We found that the signaling entropy of striatal astrocytes increased after Rbpj deletion ( Figure 3f ) , to the levels seen in SVZ stem cells ( Figure 3g ) ( One-way ANOVA: F = 35 . 62 , p<0 . 0001; Ground-state [0 . 8467 ± 0 . 0214] vs . pre-division astrocytes [0 . 8771 ± 0 . 0171 , mean ±S . D] , p<0 . 0001 [unpaired t-test]; pre-division astrocytes vs . pre-division SVZ [0 . 8848 ± 0 . 0232] , p=0 . 0721 ) . This supported the conclusion that striatal astrocytes achieve the same level of differentiation potency as SVZ neural stem cells . Right before entering transit-amplifying divisions , astrocytes were in their most neural stem cell-like state . In addition to the similarities described above , striatal astrocytes in this pre-division state had upregulated some marker genes of neural stem cells ( e . g . Fabp7 [Blbp] , Egfr , Ascl1 , Cd9 , Cd81 , Slc1a3 [Glast]; Figure 3h and Figure 6 ) . Other such markers were already expressed by ground-state astrocytes and their RNA levels did not change after Rbpj deletion ( e . g . Tnfrsf19 ( Troy ) , Sox2 , Nr2e1 ( Tlx ) , Msi1 [Musashi-1] , Rarres2; Figure 3—figure supplement 1e ) . Interestingly , some classical neural stem cell markers were never upregulated by Rbpj-deficient astrocytes . For example , Gfap , Prom1 , Thbs4 and Vim remained at low levels throughout the neurogenic process ( Figure 3i ) , showing that neurogenic astrocytes do not become transcriptionally identical to SVZ stem cells . Supplementary file 1 includes a full list of genes differentially expressed between ground-state astrocytes and those in their stem cell-like pre-division state . In the adult brain , active neural stem cells exist in two regions , the SVZ and the DG , and produce distinct neuronal subtypes . We asked whether the neurogenic program of striatal astrocytes is more similar to that of stem cells in the SVZ or the DG . For this , we integrated our AAV-Cre dataset with the aforementioned SVZ dataset from Zywitza et al . , 2018 . and another previously published 10X Chromium dataset of DG neurogenesis ( Hochgerner et al . , 2018; Figure 4a ) . We used our AAV-Cre dataset rather than the Cx30-CreER dataset for this comparison because previously published 10X Chromium datasets exist for the SVZ and DG that contain both stem cells , transit-amplifying cells and neuroblasts . A clustering analysis revealed that cells clustered together with other cells of the same maturational stage , regardless of which brain region they came from ( Figure 4a ) . This indicates that neurogenesis proceeds along the same general steps in all three regions . To perform a more focused comparison of the neurogenic transcriptional programs in the striatum , SVZ and DG , we next performed pairwise correlation between striatal cells and the corresponding maturational stages in the SVZ and DG ( Figure 4b–c ) . We found that the transcriptomes of the striatal cells were more similar to those of SVZ stem cells than to those of DG stem cells in every maturational stage , in particular at the transit-amplifying cell and neuroblast stages ( Figure 4b–c ) . We next took a more detailed look at neuroblasts , and only included transcription factors in our analysis ( Hu et al . , 2019 ) , reasoning that the neuroblast transcription factor profile is indicative of commitment to neuronal subtypes . We performed three pairwise differential expression analyses between the neuroblasts in the striatum , SVZ and DG and plotted the most differentially expressed transcription factors in a heatmap ( Figure 4d ) . This revealed the close transcriptional similarity between striatal and SVZ neuroblasts , suggesting that striatal astrocytes and SVZ stem cells are predisposed to generate very similar neuronal subtypes ( GABAergic interneurons ) , different from those in the DG ( excitatory granule neurons ) . In summary , the transcriptional program governing neurogenesis by striatal astrocytes is highly similar , but not identical , to that of SVZ cells , and less similar to that of DG stem cells . We speculate that this similarity likely reflects a shared developmental history . It has , for example , been shown that the region-specific transcriptional signature among astrocytes is conserved when astrocytes are reprogrammed directly into neurons ( Herrero-navarro et al . , 2020 ) . We previously found that , in addition to the striatum , Rbpj deletion triggers astrocyte neurogenesis in layer 1 of the medial cortex ( Magnusson et al . , 2014 ) . Now , we asked whether astrocytes in other parts of the brain also show signs of activating a neurogenic program , even though they do not complete neurogenesis . To detect faint signs of an initiated neurogenic program , we first performed an immunohistochemical staining against the early neurogenesis-associated transcription factor Ascl1 in Cx30-CreER; Rbpjfl/fl mice , 4 weeks after tamoxifen administration . Importantly , we used tyramide-based signal amplification to boost the fluorescent signal in order to be able to detect low levels of Ascl1 protein . Intriguingly , this showed that many astrocytes throughout the brain had upregulated low levels of Ascl1 in response to Rbpj deletion ( Figure 5a–c ) . This suggested that astrocytes in all brain regions initiate at least one early step in a neurogenic program in response to Rbpj deletion . To learn more , we next performed single-cell RNA sequencing on astrocytes isolated from the non-neurogenic somatosensory cortex of the same Cx30-CreER mice that we had used for the striatum analysis ( Figure 5d ) . We chose the Cx30-CreER model for this analysis because it induces Rbpj deletion simultaneously throughout the brain , does not lead to any potentially confounding tissue damage , and does not suffer from viral tropism-induced differences in recombination efficiency between the striatum and cortex . We used Monocle to arrange the cortical cells in pseudotime together with the striatal cells and found that , regardless of brain region , cells from the different time points scored similarly along the initial stages of this pseudotime axis . However , all the cortical astrocytes halted their development prior to entering transit-amplifying divisions ( Figure 5e ) . A density plot of cortical astrocytes from the different time points helped better illustrate the gradual progression along this neurogenic axis ( Figure 5f ) . Two weeks after Rbpj deletion , cortical astrocytes had upregulated genes mainly associated with metabolism and gene expression , just like striatal astrocytes ( Figure 5g , Supplementary file 1 ) . They also steadily increased gene expression activity ( Figure 5h ) , again similar to striatal cells . In addition , signaling entropy continuously increased after Rbpj deletion ( 0 . 85 ± 0 . 018 vs . 0 . 87 ± 0 . 018 [mean ± S . D . ] in ground-state vs . pre-division astrocytes , respectively [p<0 . 0001 , unpaired t test]; Figure 5i–j ) to almost the same level as in the striatal Rbpj-deficient astrocytes ( Figure 5j; pre-division striatal vs . cortical astrocytes: p=0 . 0091 ) , suggesting that , like the striatal astrocytes , the cortical ones were shifting toward a more stem cell-like state . At their highest pseudotime score , cortical astrocytes were at their most stem cell-like state , characterized by upregulation of neural stem cell genes such as Ascl1 , Egfr , Fabp7 , Pou3f4 , Msi1 , Slc1a3 , Cd81 to the same RNA levels as in striatal astrocytes or SVZ stem cells ( Figure 5k and Figure 6 ) . One neural stem cell marker , Cd9 ( Llorens-Bobadilla et al . , 2015 ) , was conspicuously higher in striatal and SVZ cells than cortical cells ( Figure 5—figure supplement 1 ) . Another , Neurog1 , was upregulated by both striatal and cortical astrocytes but was not expressed in the SVZ ( Figure 5—figure supplement 1 ) . Taken together , it seemed as if cortical astrocytes were preparing to activate the full neurogenic program but that they were unable to enter transit-amplifying divisions . Cortical astrocytes can be isolated from an injury site and generate neurospheres in vitro ( Shimada et al . , 2012; Buffo et al . , 2008; Sirko et al . , 2013 ) . Given this intrinsic capacity to activate neurogenic properties , we therefore wondered whether it might instead be the environment of the cortex that prevents these astrocytes from completing their neurogenic program in vivo . To gauge the neurogenic permissiveness of the striatal and cortical environments , we isolated neural progenitor cells by sorting tdTomato+ cells from the SVZ and DG of Cx30-CreER; tdTomato mice ( with or without Rbpj deletion ) 3–6 days after tamoxifen administration ( Supplementary file 3 ) . These cells were then immediately grafted into the striatum or cortex of littermate mice ( Figure 5l ) , to see how many of the grafted cells would form neuroblasts and neurons . When the recipient mice were sacrificed and analyzed 4 weeks later , we found that more of the transplanted cells had developed into Dcx+ neuroblasts ( Figure 5m ) and NeuN+ neurons ( Figure 5n ) in the striatum than in the cortex ( Dcx: striatum 52 ± 21% , cortex 20 ± 22% , mean ± S . D . ; NeuN: striatum 10 ± 9% , cortex 2 ± 3% ) . This suggested that the cortex has an environment less permissive to neurogenesis than the striatum . Such a hostile environment may help explain why Rbpj-deficient astrocytes in the cortex fail to complete the neurogenic program that they initiate . Although we focused our analysis on the somatosensory cortex , our immunohistochemical staining of Ascl1 ( Figure 5b ) suggests that an aborted neurogenic program may occur throughout the entire brain in response to Rbpj deletion . Although Rbpj deletion does trigger neurogenesis by astrocytes in the striatum , this process is most active in the medial striatum ( Figure 1a ) . This uneven distribution suggests that in addition to Notch signaling , other mechanisms regulate astrocyte neurogenesis . We wanted to know if our RNA sequencing datasets could help us learn why many striatal astrocytes fail to complete neurogenesis . We noted that , after Rbpj deletion , it takes approximately 3–4 weeks until the first striatal astrocytes enter transit-amplifying divisions ( Figure 2c ) . Yet , many cells isolated even 8 weeks after Rbpj deletion remained as astrocytes that seemed to have halted their development immediately prior to the transit-amplifying cell stage ( Figure 2c ) . We asked whether stalled striatal astrocytes could resume their neurogenic program if they were stimulated to enter transit-amplifying divisions . Striatal astrocytes upregulate Egfr in response to Rbpj deletion ( Figure 3h ) . We performed a single injection of the mitogen EGF directly into the lateral striatum of Cx30-CreER; Rbpjfl/fl mice 7 weeks after tamoxifen administration ( Figure 6a ) . In Rbpjwt/fl control mice , EGF injection on its own was not enough to induce striatal neurogenesis; neither did it induce any neuroblast migration from the SVZ ( Figure 6b , d ) or any detectable increase in the amount of SVZ neuroblasts ( Figure 6—figure supplement 1a–c ) . In wild-type mice , EGF did not trigger cell divisions in striatal astrocytes or non-astrocytes ( Figure 6—figure supplement 1d ) . We concluded that EGF injection into the lateral striatum did not perturb SVZ neurogenesis or trigger confounding striatal proliferation . Interestingly , however , EGF injection into the striatum of Rbpjfl/fl mice did lead to a doubling in the number of reporter-positive neuroblasts compared to the vehicle-injected contralateral striatum ( 216 ± 90 [EGF hemisphere , mean ± S . D . ] , 110 ± 61 [vehicle hemisphere]; p=0 . 023 [paired t-test]; Figure 6c–d ) . This increase was the result of an increased number of astrocytes entering transit-amplifying divisions rather than increased proliferation by the same number of astrocytes: EGF-injected striata contained on average twice as many neuroblast clusters as vehicle-injected striata ( 21 ± 10 [EGF hemisphere , mean ±S . D . ] , 8 ± 5 [vehicle hemisphere]; p=0 . 049 [paired t-test]; Figure 6e ) , but cluster size did not differ ( p=0 . 59 [Mann-Whitney test]; Figure 6f ) . Some animals also showed a dramatic increase in the number of Ascl1+ transit-amplifying cell clusters in the EGF-injected striatum ( Figure 6g–h ) . This increase , however , was less pronounced ( p=0 . 132 [paired t-test] ) than that of Dcx+ cells , possibly because most of the EGF-stimulated astrocytes had already passed the transit-amplifying stage and developed into neuroblasts 2 weeks after the EGF injection . To investigate the local astrocyte origin of these Ascl1+ clusters , we used residual S100β protein as a short-term lineage-tracing marker of striatum-generated transit-amplifying cells ( see Figure 2—figure supplement 3d–g ) . We found that S100β protein levels were the same in Ascl1+ clusters in the EGF- and vehicle-injected striata ( Figure 6i; p=0 . 159 [unpaired t-test] , with a trend toward more S100βhigh clusters in the EGF-injected striatum ) . This showed that EGF had caused no influx of S100β-negative cells from the SVZ into the striatum . Our results demonstrate that EGF administration can enable stalled neurogenic astrocytes in the striatum to initiate transit-amplifying divisions and resume neurogenesis . This effect was easily measurable even after a single , localized EGF injection . Yet , striatal neurogenesis was still primarily localized to the medial striatum , suggesting that many stalled astrocytes may require a stronger stimulus than a single injection to re-initiate neurogenesis . Further studies would be needed to tell whether longer and more broadly localized EGF exposure could recruit an even larger proportion of stalled striatal astrocytes . EGF exposure in the somatosensory cortex was not sufficient to make stalled Rbpj-deficient astrocytes enter transit-amplifying divisions there ( Figure 6—figure supplement 1e ) . Interestingly , however , the mere act of inserting a thin syringe needle was enough of a stimulus to do so in some stalled cortical astrocytes ( Figure 6—figure supplement 1e ) . We have thoroughly explored the effect of injuries on recruiting halted astrocytes in two separate studies ( stab wound in the cortex Zamboni et al . , 2016 and stroke in the striatum Santopolo et al . , 2020 ) . Taken together , these results show that it is possible to enhance and modulate the astrocyte neurogenic program through targeted interventions , an observation that should be useful when attempting to recruit astrocytes as a reservoir for new neurons throughout the central nervous system .
The brain has almost no capacity to replace lost neurons on its own . Although the SVZ and DG do contain neural stem cells , it is not feasible to use these regions as sources of replacement neurons for the rest of the brain . That is mainly because neurons generated in the SVZ or DG would need to migrate unfeasibly far to reach distant injury sites in the human brain . However , a therapeutic workaround could be to artificially stimulate neurogenesis by cells close to an injury . Astrocytes could be such a reservoir for new neurons because they have an intrinsic neurogenic potential and are very abundant . However , it will be necessary to understand the mechanisms governing astrocyte neurogenesis before the process can be harnessed and tailored to generate specific neuronal subtypes of significant amounts . Here , we used single-cell RNA sequencing to study the transcriptional program of parenchymal astrocytes as they initiate neurogenesis in response to deletion of the Notch-mediating transcription factor Rbpj . We find that the neurogenic program of parenchymal astrocytes is almost identical to that of SVZ neural stem cells . Based on the analyses in this study , we propose a model where parenchymal astrocytes are regarded as latent neural stem cells ( Figure 7 ) . Specifically , we find that initiation of the neurogenic program in astrocytes is dominated by a steady upregulation of genes associated with transcription and translation , as well as a transient wave of metabolism-associated genes . These findings fit well with what is known about neural stem cell activation in the DG and SVZ , where expression levels of genes associated with transcription and translation steadily increase throughout the neurogenic program ( Llorens-Bobadilla et al . , 2015; Shin et al . , 2015 ) . Our data suggest that parenchymal astrocytes start off at an even lower baseline level of gene expression than neural stem cells , and that they have increased the number of expressed genes ~ 1 . 4 fold by the time they reach their most stem cell-like state , prior to transit-amplifying divisions . Metabolic gene expression peaked right before striatal astrocytes entered transit-amplifying divisions , suggesting that the rapid burst of cell divisions is the most metabolically demanding step of the neurogenic program . In the DG , exit from quiescence requires increased lipid metabolism ( Knobloch et al . , 2013 ) , and once activated stem cells enter transit-amplifying divisions , metabolic genes are downregulated , both in the DG and SVZ ( Llorens-Bobadilla et al . , 2015; Shin et al . , 2015 ) . This similarity suggests that , like parenchymal astrocytes , neural stem cells in the DG and SVZ experience a metabolic peak prior to entering transit-amplifying divisions . We find that it takes 2–4 weeks for astrocytes to reach this pre-division state . One possible interpretation of this is that the most time-consuming portion of the astrocyte neurogenic program is the initial exit from deep quiescence , as is the case in quiescent muscle stem cells ( Siegel et al . , 2011 ) . Another ( and compatible ) interpretation is that Rbpj protein may remain bound at its target promoters long after the Rbpj gene has been deleted , and that the neurogenic program is initiated only when the protein level has fallen below a certain threshold . Such stable transcription factor-DNA interaction is a mechanism by which some transcription factors buffer against fluctuating transcription levels ( Traets et al . , 2020 ) . One of our main findings is that all astrocytes in the striatum and somatosensory cortex initiate a neurogenic program in response to Rbpj deletion , even though only some astrocytes in the striatum and medial cortex manage to enter transit-amplifying divisions . This suggests that all astrocytes in the striatum and somatosensory cortex , and perhaps in other regions as well , are latent neural stem cells . We hypothesize that the inability of most astrocytes to enter transit-amplifying divisions is due to a lack of proper environmental signals rather than an intrinsic inability . This hypothesis is based on four observations: 1 ) A large body of literature has demonstrated that parenchymal astrocytes isolated from the injured mouse cortex can generate neurospheres in vitro ( Shimada et al . , 2012; Buffo et al . , 2008; Sirko et al . , 2013 ) , proving that astrocytes have intrinsic neural stem cell potential but lack the proper in vivo environment to realize this potential . 2 ) We here transplanted neural stem cells to the cortex and striatum and found that fewer developed into neuroblasts and neurons in the cortex ( Figure 5l–n ) , showing that the cortex has a non-permissive environment . 3 ) A single injection of EGF into the striatum was enough to make some astrocytes , which had stalled in a pre-division state , enter transit-amplifying divisions ( Figure 6 ) , showing that extracellular signals can bolster and improve the astrocyte neurogenic program . 4 ) In Cx30-CreER; Rbpjfl/fl mice , most astrocyte-derived transit-amplifying cells and neuroblasts appear in the medial striatum , close to the SVZ . It is possible that the nearby SVZ provides a gradient of neurogenic factors into the medial striatum and establishes a more permissive environment there . Such a gradient has indeed been shown to appear if growth factors are infused into the lateral ventricle ( Yan et al . , 1994; Anderson et al . , 1995 ) . In this study , we use Rbpj deletion as our trigger to induce neurogenesis by astrocytes . We propose that our results apply also to injury-induced astrocyte neurogenesis even in the absence of Rbpj deletion , such as that observed in the stroke-injured striatum . Decreased Notch signaling is the mechanism that triggers neurogenesis by striatal astrocytes in stroke-injured mice ( Magnusson et al . , 2014 ) . Therefore , the Rbpj deletion model is a biologically relevant way to activate the astrocyte neurogenic program without triggering confounding tissue processes that occur in response to injury , such as inflammation and reactive astrogliosis . Indeed , the precision afforded by our Rbpj deletion model allowed us to observe here that the astrocyte neurogenic program is not dependent on simultaneous reactive gliosis . Instead , neurogenesis and reactive astrogliosis are likely controlled by two independent , but compatible , transcriptional programs . Is there practical usefulness in designating parenchymal astrocytes as latent neural stem cells ? We argue that there is . This nomenclature highlights an aspect of astrocyte biology that exists in parallel to the normal important functions of astrocytes in the healthy brain . Labeling parenchymal astrocytes as latent neural stem cells can help focus research efforts toward activating and fine-tuning the inherent neurogenic capacity of these cells to generate therapeutically useful neuronal subtypes . It is likely that astrocytes in different brain regions use different mechanisms to lock their astrocyte identity in place . Our study demonstrates that it is feasible to identify such mechanisms and seek to target them for enhancing neurogenesis by astrocytes throughout the central nervous system .
Cx30-CreER ( Slezak et al . , 2007 ) , R26-tdTomato ( Ai14 ) ( Madisen et al . , 2015 ) , R26-YFP ( Srinivas et al . , 2001 ) , Rbpjfl/flTanigaki et al . , 2002 and wild-type mice were of the C57BL/6 strain and were >2 months old . Genetic recombination was induced by intraperitoneal tamoxifen injections for five consecutive days ( Sigma-Aldrich , St . Louis , MO; 20 mg/ml in 1:9 ethanol:corn oil; 2 mg/injection ) . Time points were counted from the last day of injection . Mice with no Rbpj deletion ( for isolation of ground state astrocytes ) were sacrificed one day after a three-day tamoxifen regimen . For EdU administration ( Figure 2—figure supplement 3c ) , EdU ( Thermo Fisher , Waltham , MA ) was administered through the drinking water ( 0 . 75 g/l EdU in water containing 1% sucrose ) for two weeks leading up to the animals’ death . Mice were housed in standard conditions with 12/12 hr light/dark cycles and free access to food and water . Experimental procedures were approved by the Stockholms Norra Djurförsöksetiska Nämnd ( Permit reference numbers N571-11 and N155-16 ) . Immunohistochemical staining was performed as described previously ( Magnusson et al . , 2014 ) . We used antibodies against Ascl1 ( Cosmo Bio , Tokyo , Japan; rabbit , 1:2000 ) , detected with or without ABC kit [Vector Laboratories , Burlingame , CA] and TSA [PerkinElmer , Waltham , MA] ) , Dcx ( Santa Cruz Biotechnology , Dallas , TX; goat , 1:500 ) , GFP/YFP ( Aves Labs , Davis , CA , USA; chicken , 1:2000 ) , Ki67 ( eBioscience , San Diego , CA , USA; rat , 1:2000 ) , S100 ( DAKO , Glostrup , Denmark; rabbit , 1:200 , directly conjugated to A647 using Alexa Fluor Antibody Labeling Kit [Thermo Fisher] ) . EdU was detected using Click-iT EdU Alexa Fluor 647 Imaging Kit ( Thermo Fisher ) . One mouse brain was used for each time point , for a total of 4 brains . From each brain , a ~ 1 mm thick coronal slice was isolated , spanning rostrocaudal levels Bregma −0 . 2 to +1 . 0 ( Figure 2—figure supplement 1a–c ) . From this section , the SVZ was isolated and discarded . The striatum and cortex were then isolated from both hemispheres . Tissue was cut into fine pieces and incubated with papain solution ( 20 units/ml ) containing L-cysteine ( 1 mM ) , EDTA ( 0 . 5 mM ) and DNase ( 2000 units/ml ) ( Papain Dissociation System [LK003160] , Worthington , Lakewood , NJ ) , dissolved in EBSS ( Thermo Fisher ) , at 37°C for 30 min , with occasional agitation . Then , the tissue was triturated with smaller and smaller pipet tips and filtered through a 70 μm cell strainer ( BD , Franklin Lakes , NJ ) . After a 10 min centrifugation at 200*g in a swing-bucket rotor at 4°C , supernatant was removed and cells were resuspended in 3 ml sterile-filtered washing buffer ( Dulbecco’s PBS containing sodium pyruvate , streptomycin sulfate , kanamycin monosulfate , glucose and calcium chloride; Sigma-Aldrich , D4031 ) containing 0 . 5% bovine serum albumin ( BSA; Sigma-Aldrich , A9418 ) . To remove debris , we then used gradient centrifugation and myelin removal beads in the following way . The cell suspension was carefully layered on top of 4 ml of 90% Percoll ( GE Healthcare , Chicago , IL; 17-0891-01 , diluted in 1 . 5 M sterile-filtered NaCl ) in a 15 ml Falcon tube and centrifuged at 200*g for 15 min in a swing-bucket rotor at 4°C . After removing the small top phase , the large bottom phase ( including the white interphase ) was resuspended in 17 . 5 ml washing buffer in a 50 ml Falcon tube and repeatedly aspirated with a Pipette Boy . This cell suspension was then centrifuged at 200*g for 10 min at 4°C . The pellet was set aside , and to maximize cell yield the supernatant was again repeatedly aspirated and centrifuged . Then , the two pooled cell pellets were resuspended in 1 ml sterile-filtered MACS buffer ( 5% BSA and 2 mM EDTA in PBS ) . The cell suspension was incubated with 80 μl myelin removal beads II ( Miltenyi Biotec , 130-096-731 ) for 15 min . Meanwhile , a MACS Large Cell Column ( Miltenyi Biotec , Bergisch Gladbach , Germany; 130-042-202 ) was placed in a Magnetic MACS Multistand ( Miltenyi Biotec ) and activated by allowing 500 μl of MACS buffer to drop through . The cell suspension was passed through the MACS column and the flow-through was collected in a FACS tube pre-coated with 1% sterile-filtered BSA . MACS buffer ( 500 μl ) was then passed through the column twice to increase cell yield . The cell suspension was then incubated with LIVE/DEAD Blue Dead Cell Stain Kit ( Thermo Fisher , L23105 ) for 30 min . Immediately afterwards , an a BD Influx FACS machine ( 86 μm nozzle ) was used to sort live tdTomato+ single cells into 96-well skirted PCR plates ( Thermo Fisher ) containing lysis buffer ( containing dNTP [2 . 5 mM] , oligo-dT [2 . 5 mM] , RNAse Inhibitor [Takara Bio , Kusatsu , Japan; 1:40] and ERCC [Ambion , TX; 1:4 , 000 , 000] ) with the machine set on 'Index Sort' mode to record cell size information and fluorescence parameters used . Plates were immediately frozen on dry ice and stored at −80°C until processing . For preparation of single-cell cDNA libraries , the plates were thawed at 72°C for 4 min and immediately placed on ice . A reverse transcription of mRNA was performed and the resulting cDNA libraries were immediately subject to amplification for 19 cycles according to the Smart-Seq2 protocol ( Picelli et al . , 2014a ) . The final libraries were subjected to bead purification ( AMPure , Beckman Coulter , Brea , CA; 0 . 6:1 bead-to-sample ratio ) to remove excess primers and primer dimers . Library quality was checked using an Agilent Bioanalyzer and DNA High Sensitivity microfluidics chips . After determining that libraries were of sufficiently high quality for downstream sequencing , 0 . 5–1 . 5 ng of cDNA from each single-cell library was used for tagmentation reactions with a custom Tn5 enzyme/oligo ( comparable to the Nextera XT protocol ) ( Picelli et al . , 2014b ) . Nextera XT dual indexes were used ( 96 index kit ) and each library was subjected to an additional 10 cycles of amplification . The final products were again subjected to bead cleaning to remove excess primers and monitored for successful amplification by random sampling of different wells and visualization of library size and quality on the Bioanalyzer . Successful tagmentation reactions were pooled as individual 96-well plates and sequenced using an Illumina HiSeq 2500 machine with 2*125 bp sequencing on High Output mode . For each sorted cell , a ratio was made between its forward-scatter value and the average forward-scatter value of that sort’s tdTomato-negative cell population . Thirteen Rbpjfl/fl mice were subjected to intrastriatal injection of AAV8-GFAP ( 0 . 7 ) -iCre-WPRE virus ( VB4887; Vector Biolabs , Malvern , PA , USA ) using coordinates in relation to bregma: 1 . 2 mm anterior , 2 . 5 mm lateral , 3 mm deep ) . Five weeks later , mice were sacrificed and the brain collected for isolation of single cells . We used a brain slicer to produce 1 . 5 mm thick sections spanning the rostrocaudal region transduced by the virus . Prior to isolating striata for tissue dissociation , we inspected fluorescent labeling at the microscope to ensure that only striatal cells ( and not SVZ cells ) had been transduced . At this stage , two mice were excluded because they had recombined cells in the SVZ . For the remaining 11 mice , striata were microdissected from each slice following the same approach outlined above ( sample preparation for the Cx30-CreER experiment ) , pooled as one sample , and digested using Miltenyi’s Adult Brain Dissociation kit . Single cells were finally FAC-sorted based on tdTomato signal as described above ( i . e . for the Cx30-CreER dataset ) . Library preparation was performed using Chromium Single Cell 3’ Library and Gel Bead Kit ( v3 , 10X Genomics , Pleasanton , CA ) according to the manufacturer’s instructions . Libraries were sequenced on a NovaSeq6000 S1 flowcell using a custom read setup: 28 nt ( read 1 ) , eight nt ( i7 index ) , 166 nt ( read 2 ) . Raw reads were obtained from the sequencer and aligned to the mouse genome ( mm10 ) supplemented with tdTomato DNA sequence . We used cellranger v2 . 0 . 1 for pre-processing of raw reads followed by Seurat R package ( v2 . 3 . 4 ) ( Stuart et al . , 2019 ) for downstream analysis of mapped data . Aligned transcripts were normalized for sequencing depth , log-transformed , and scaled using Seurat’s default parameters . We excluded genes expressed in less than 3 cells , and cells with mitochondrial content exceeding 25% ( n = 418 ) . Next , we ran principal component analysis ( PCA ) on 2359 highly variable genes and selected the first 10 dimensions for non-linear dimensionality reduction ( UMAP ) and clustering . The analysis allowed us to identify distinct cell types composing the dataset ( Figure 1—figure supplement 1 ) . We removed contaminating cells lacking expression of tdTomato ( e . g . microglia , n = 87 ) , and cells expressing tdTomato transcripts but not involved in the astroglial neurogenic program ( i . e . cells belonging to the oligodendrocyte lineage , n = 1 , 438 ) . After filtering , we retained 1380 cells for further processing . We included the top 3249 highly variable genes for running PCA and UMAP ( using the first 20 PCs ) on the filtered dataset . We next performed clustering based on the shared nearest neighbour algorithm ( Seurat , v2 . 3 . 4 ) to identify distinct subpopulations within the sample ( Figure 1 ) . We next , ran pairwise comparisons using Wilcoxon’s test to determine cluster-specific transcriptional profiles . Clustering and differential expression analyses allowed us to record the presence of two astrocyte clusters characterized by expression of Gjb6 ( Astrocyte Cluster 1 ) and upregulation of Ascl1 ( Astrocyte Cluster 2 ) ; a cluster encompassing transit-amplifying cells enriched in Mki67 transcripts; and two clusters of neuroblasts expressing Dcx and Robo2 , as well as Foxp2 and Nav3 . For analyzing differential expression along pseudotime of the AAV-Cre dataset ( Figure 2i ) , Monocle ( v . 2 . 12 . 0 ) was used . We aimed at comparing the neurogenic program initiated by striatal astrocytes with that recorded in the neurogenic niches of the SVZ and DG . For this purpose , we compared the transcriptional profiles of cells contained in the current AAV-Cre dataset to that of SVZ ( Zywitza et al . , 2018 ) ( accession code: GSE111527 ) and DG ( Hochgerner et al . , 2018 ) ( accession code: GSE95315 ) neurogenesis . From the complete SVZ dataset , we only considered samples collected from wild-type female animals ( an002 , an003_F , and an003_L ) . Both SVZ and DG datasets were processed separately , in order to identify distinct cell types and retain only clusters implicated in the neurogenic program . Specifically , we ran PCA on highly variable genes ( n = 6497 for SVZ and n = 2224 for DG ) , followed by tSNE and clustering performed on the top 15 PCs , and selected clusters of cells based on expression of classic neurogenesis-related markers ( e . g . Aqp4 , Thbs4 , Sox4 , Top2a , Dcx ) for downstream analysis . Filtered datasets were integrated with striatal data by means of canonical correlation analysis ( CCA , Seurat ) considering the top 2000 HVGs merged across samples . Datasets were aligned using the first 13 dimensions of CCA and visualized in UMAP space . To investigate the relative distance of cell types across samples we ran hierarchical clustering based on a distance matrix constructed in the aligned space ( i . e . using CCA dimensions ) , followed by differential expression analysis that allowed insight into shared and unique transcriptional programs active in the different neurogenic niches . Finally , we ran correlation analysis on the average expression of each gene to determine similarities in cell-specific transcriptional profiles across brain regions ( data were visualized as scatterplots using ggplot 3 . 3 . 0 ) . Donor mice ( Cx30-CreER; R26-tdTomato , either with or without inducible Rbpj deletion , see Supplementary file 3 ) , were sacrificed 3–6 days after intraperitoneal tamoxifen administration . The SVZ and DG were dissected and cells isolated using the same protocol used to isolate cells for the Cx30-CreER dataset . Live tdTomato+ cells were sorted using an 86 µm nozzle into a FACS tube pre-coated with 1% BSA , containing 250 µl wash buffer ( see cell isolation protocol for the Cx30-CreER dataset ) . Cell suspensions were kept on ice at all times and chilled during cell sorting . After sorting , cell suspensions were transferred to pre-coated microcentrifuge tubes and centrifuged at 200*g for 10 min in a swing-bucket rotor at 4°C . The swing-bucket rotor was important as it enabled the small cell pellet to collect at the bottom of the tube . Supernatant was removed carefully . Tubes were then briefly centrifuged in a tabletop centrifuge to bring down a few extra microliters of fluid from the walls of the tube . This amount of liquid was enough to resuspend the cell solutions , which was done by carefully flicking the tubes . If more volume needed to be added , wash buffer was used . Cell suspension was transferred to 0 . 2 ml tubes , placed on ice and immediately transplanted to the brains of littermate mice . One microliter of this solution was injected into each of the following coordinates ( relation to bregma ) : Striatum: 0 . 5 mm anterior , 2 . 0 mm lateral , 3 . 0 mm deep . Cortex: 0 . 5 mm anterior , 2 . 5 mm lateral , 1 . 0 mm deep . The entire protocol took ~16–18 hr to perform from beginning to end . During the analysis of transplantation-recipient mice , mice were excluded if <20 transplanted cells were found . In Cx30-CreER; Rbpjfl/fl and Rbpj+/fl mice , intrastriatal EGF injection was performed seven weeks after tamoxifen administration . EGF ( Sigma-Aldrich E4127 , 200 ng/μl ) was dissolved in PBS containing 0 . 1% bovine serum albumin ( BSA ) . Four Cx30-CreER; Rbpjfl/fl mice and four Cx30-CreER; Rbpjfl/wt control mice were anesthetized with isoflurane ( 4% at 400 ml/min flow rate to induce anesthesia , followed by ~2% at 200 ml/min ) . Then , 2 μl of EGF solution was delivered into the lateral striatum through a single injection ( 0 . 7 mm anterior , 2 . 4 mm lateral of bregma; 3 mm deep of dura mater ) using a 36-gauge beveled needle ( World Precision Instruments , Sarasota , FL , USA ) . A vehicle solution consisting of 0 . 1% BSA in PBS was injected into the contralateral striatum of the same mice . Mice were sacrificed and analyzed 2 weeks after EGF injection . For estimating whether EGF induces proliferation in wild-type mice ( Figure 6—figure supplement 1d ) , mice were sacrificed 48 hr after EGF injection . Evaluation of EGF’s effect on cortical astrocytes was evaluated as follows . Animals were first injected with 0 . 3 μl of AAV-GFAP-Cre into the somatosensory cortex to delete Rbpj locally ( 1 . 0 mm anterior , 1 . 5 mm lateral of bregma . Liquid was injected at both 1 . 0 and 0 . 5 mm deep of dura mater ) . Both left and right hemispheres were injected with the virus . Four weeks later , 2*0 . 3 μl of EGF solution was delivered into one hemisphere using the same injection coordinates . A vehicle solution consisting of 0 . 1% BSA in PBS was injected into the other hemisphere of the same mouse . Mice were sacrificed and analyzed 2 , 3 and 4 weeks after EGF injection . To estimate the number of neuroblasts in the SVZ ( Figure 6—figure supplement 1 ) , we performed an immunohistochemical staining against Dcx and used Dcx signal intensity as a proxy for cell numbers . We took tiled images of the SVZ in 3–5 brain sections per mouse using a Zeiss LSM 700 confocal microscope , using the same microscope settings for the left and right SVZ . CellProfiler v . 3 . 0 . 0 ( Carpenter et al . , 2006 ) was used to identify Dcx+ cells or cell clusters and calculate total signal intensity ( cell area*mean intensity ) on the EGF- and vehicle-injected hemispheres in all images . To detect residual S100β protein levels in striatal Ascl1+ cells , we took images of all Ascl1+ cells found in the striatum and SVZ of five brain sections of one mouse , adjusting imaging depth for each cell and using the same microscope settings for every cell . Then , we measured Ascl1 and S100β signal intensity in each cell using ImageJ’s Measure function ( Schneider et al . , 2012 ) . We used LandSCENT v . 0 . 99 . 3 ( Chen and Teschendorff , 2019 ) to calculate signaling entropy on the single-cell level . Because the LandSCENT package only contains a protein-protein interaction network for human genes , we first converted our dataset’s mouse gene names into their human orthologous counterparts using biomaRt . Only genes that were expressed ( TPM ≥10 ) in ≥10% of cells were included in the analysis . | Regenerative medicine aims to help the body replace damaged or worn-out tissues , often by kick-starting its own intrinsic repair mechanisms . However , the brain cannot easily repair itself , and therefore poses a much greater challenge . This is because nerve cells or neurons , which underpin learning , memory , and many other abilities , are also the brain’s greatest weakness when it comes to tissue repair . In most parts of the adult brain , neurons are never replaced after they die . This means that damage to brain tissue – for example , after a stroke – can have severe and long-lasting consequences . Neural stem cells are one type of brain cell that can turn into new neurons if needed , but they are only found in a few parts of the brain and cannot fix damage elsewhere . More recent work in mice has shown that astrocytes , a common type of support cell in the brain that help keep neurons healthy , could also generate new neurons following a stroke . However , the ability was restricted to small numbers of astrocytes in a specific part of the brain . Here , Magnusson et al . set out to determine the molecular mechanisms behind this regenerative process and why it is unique to certain astrocytes . The researchers used a technique called single-cell RNA sequencing to analyze the genetic activity within individual mouse astrocytes that had been exposed to conditions mimicking a stroke . This method revealed which genes are switched on or off , thus generating a profile of gene activity for each astrocyte analyzed . This experiment showed that the profiles of astrocytes that had started to produce neurons were in fact nearly identical to neural stem cells . Even the astrocytes that could not generate neurons took the first few steps towards this genetic state; however , they stalled early in the process . Treating the brains of mice withepidermal growth factor , a powerful molecular signal that stimulates cell growth , kick-started nerve cell production in a subset of these cells – showing that at least some of the non-regenerative astrocytes could be stimulated to make neurons if given the right treatment . The results of this study shed new light on how some astrocytes in the brain gain the ability to form new neurons . In the future , this knowledge could help identify a source of replacement cells to regenerate the injured brain . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"neuroscience"
] | 2020 | Activation of a neural stem cell transcriptional program in parenchymal astrocytes |
Ubiquitin is essential for eukaryotic life and varies in only 3 amino acid positions between yeast and humans . However , recent deep sequencing studies indicate that ubiquitin is highly tolerant to single mutations . We hypothesized that this tolerance would be reduced by chemically induced physiologic perturbations . To test this hypothesis , a class of first year UCSF graduate students employed deep mutational scanning to determine the fitness landscape of all possible single residue mutations in the presence of five different small molecule perturbations . These perturbations uncover 'shared sensitized positions' localized to areas around the hydrophobic patch and the C-terminus . In addition , we identified perturbation specific effects such as a sensitization of His68 in HU and a tolerance to mutation at Lys63 in DTT . Our data show how chemical stresses can reduce buffering effects in the ubiquitin proteasome system . Finally , this study demonstrates the potential of lab-based interdisciplinary graduate curriculum .
Protein homeostasis enables cells to engage in dynamic processes and respond to fluctuating environmental conditions ( Powers et al . , 2009 ) . Misregulation of proteostasis leads to disease , including many cancers and neurodegenerative diseases ( Balch et al . , 2008; Lindquist and Kelly , 2011 ) . Protein degradation is an important aspect of this regulation . In eukaryotes ~80% of the proteome is degraded by the highly conserved ubiquitin proteasome system ( UPS ) ( Zolk et al . , 2006 ) . The high conservation of the UPS is epitomized by ubiquitin ( Ub ) , a 76 amino acid protein post-translational modification that is ligated to substrate amine groups , including on Ub itself in poly-Ub linkages , via a three enzyme cascade ( Finley et al . , 2012 ) . Perhaps due to its central role in regulation , the sequence of ubiquitin has been extremely stable throughout evolution . Only three residues vary between yeast and human ( 96% sequence identity ) . This remarkable conservation implies that the UPS does not acquire new functions through mutations in the central player , Ub . Instead the evolution of proteins that add Ub to substrate proteins ( E2/E3 enzymes ) , remove Ub ( deubiquitinating enzymes , DUBs ) , or recognize Ub ( adaptor proteins ) combine to create new functions , many of which rely on various poly-Ub topologies ( Sharp and Li , 1987; Zuin et al . , 2014 ) . The role of Lys48 linked poly-Ub in protein degradation ( Thrower et al . , 2000 ) appears to be universally conserved , but the functions of other linkages are more plastic . Although mass spectroscopy of cell lysates has shown that every possible poly-Ub lysine linkage exists within yeast cells ( Peng et al . , 2003 ) , only the roles of Lys11 linked poly-Ub in ERAD ( Xu et al . , 2009 ) and Lys63 linked poly-Ub in DNA damage ( Zhang et al . , 2011 ) and endocytosis ( Erpapazoglou et al . , 2014 ) are well characterized in yeast . Both of these linkages are central to stress responses , mirroring some of the established roles for non-Lys48 linkages in other organisms ( Komander and Rape , 2012 ) . Given this central role in coordinating a diverse set of stress responses , perhaps the high sequence conservation of ubiquitin is not surprising . However , classic Alanine-scanning studies showed that ubiquitin is quite tolerant of mutation under normal growth conditions ( Sloper-Mould et al . , 2001 ) . The high mutational tolerance of Ub was further confirmed using EMPIRIC ( 'extremely methodical and parallel investigation of randomized individual codons' ) , where growth rates of yeast strains harboring a nearly comprehensive library of all ubiquitin point mutations were assessed in bulk by deep sequencing ( Roscoe et al . , 2013 ) . Subsequent studies revealed that many of the constraints on the Ub sequence are enforced directly by the E1-Ub interaction ( Roscoe and Bolon , 2014 ) ; however , the surprisingly high number of tolerant positions remained unexplained . To address the paradox of the high sequence conservation and mutational tolerance of ubiquitin , we posed the problem to the first year students in UCSF’s iPQB ( Integrative Program in Quantitative Biology ) and CCB ( Chemistry & Chemical Biology ) graduate programs . Previous EMPIRIC experiments on HSP90 suggested that reducing protein expression could reveal fitness defects that are otherwise buffered ( Jiang et al . , 2013 ) . Similarly , the high expression level of Ub might buffer some fitness defects . As an alternative approach to reducing expression , the buffering might also be exposed by chemical perturbations . Chemical-genetic approaches have been successful in elucidating protein function by assessing the context dependence of mutations on organismal fitness ( Hietpas et al . , 2013; Stockwell , 2000 ) . There is substantial evidence to link the pool of free Ub to the eukaryotic stress responses to many chemicals . For example , Ub overexpression is protective against the general translational inhibitor cycloheximide ( Hanna et al . , 2003 ) and the deubiquitinating enzyme Ubp6 is upregulated upon stress to increase the pool of free Ub in the cell ( Hanna et al . , 2007 ) . Collectively , these results suggest that the sequence of Ub is subject to many constraints arising from interacting with diverse proteins while mediating the stress responses to distinct chemical perturbations . The students performed the bulk competition experiments , deep sequencing and data analysis as part of an 8-week long research class held in a purpose-built Teaching Lab . In small teams of 4–5 students working together for 3 afternoons each week , they each examined a chemical stressor: Caffeine , which inhibits TOR and consequently the cell cycle ( Reinke et al . , 2006; Wanke et al . , 2008 ) ; Dithiothreitol ( DTT ) , which reduces disulfides and induces the unfolded protein response ( Frand and Kaiser , 1998 ) and the ER associated decay ( ERAD ) pathway ( Friedlander et al . , 2000 ) ; Hydroxyurea ( HU ) , which causes pausing during DNA replication and induces DNA damage ( Koç et al . , 2004; Petermann et al . , 2010 ) ; or MG132 , which inhibits the protease activity of the proteasome ( Jensen et al . , 1995; Rock et al . , 1994 ) . We expected MG132 to desensitize the yeast to deleterious mutations in Ub , as the inhibition acts on the final degradation of UPS substrates . For the other three chemicals we expected that specific sites on Ub would become sensitized to mutation . These sites could represent important Ub/protein binding interfaces that are required for Ub to bind to adaptor proteins and ligation machinery required to respond to a specific stress . Furthermore , we expected that Caffeine induced stress would be mediated through Lys48 linked poly-Ub ( cell cycle ) , DTT induced stress would be mediated through Lys11 linked poly-Ub ( ERAD ) , and HU induced stress would be mediated through Lys63 linked poly-Ub ( DNA damage response ) . Our data collectively show that stress reduces a general buffering effect and unmasks a shared set of residues that become less tolerant to mutation . Additionally , we have identified a small set of mutations that are specifically aggravated or alleviated by each chemical . Small fitness defects that are undetectable by EMPIRIC may still be significant over evolutionary timescales and are likely to similarly be enriched for certain chemicals ( Boucher et al . , 2014 ) . We suggest that expanding the set of environmental stresses , and improved measurement of smaller fitness defects , might be able to explain the high sequence conservation of ubiquitin , as different positions in the protein are important for interactions mediating the specific responses to a wide variety of perturbations .
Previously , the fitness landscape of Ub in yeast was determined using eight competition experiments using the EMPIRIC strategy of deep sequencing short regions of all possible single amino acid substitutions during a growth competition experiment in rich media ( Roscoe et al . , 2013 ) . These experiments measured all point mutants contained in short 30 base pair ( bp ) /10 amino acid residue stretches of the Ub open reading frame ( ORF ) , which necessitated 8 separate competition experiments . To increase the throughput and reduce the cost of the experiment , we designed a barcoding strategy ( Fowler et al . , 2014 ) , that allowed us to determine allele fitness in a single experiment using EMPIRIC with barcodes ( EMPRIC-BC ) . We synthesized eighteen bp random barcodes ( N18 BCs ) , which were ligated upstream of the Illumina sequencing primer binding site . The specific association of each unique N18 BC with a given mutant Ub allele was then established through paired end sequencing of the Ub ORF and the N18 BC ( Figure 1A ) The resulting lookup table of BCs and alleles was then employed in our competition experiments to count alleles by directly sequencing the N18 BCs . In addition to simplifying the experiment , this strategy enabled us to count the alleles with a short , single end sequencing run , substantially reducing cost . The library is ~97 . 5% complete at the amino acid level . We observed a slight GC bias in the codon coverage ( Figure 1B–C ) , which is likely due to the cloning method that initially generated the Ub mutants ( Hietpas et al . , 2012 ) . Most substitutions are associated with many N18 BCs , with a median of fifteen unique barcodes representing a specific amino acid substitution ( Figure 1D ) . 10 . 7554/eLife . 15802 . 003Figure 1 . Barcoding enables a bulk competition experiment of ~1500 Ubiquitin variants . ( A ) Prior to the competition experiment , ubiquitin alleles were specifically associated with unique barcodes through a paired end sequencing . To monitor the frequency of different alleles during the competition experiments , we directly sequenced the barcodes in a short single end read . ( B ) The library contains most codon substitutions and almost all are associated with multiple barcodes . A slight GC bias is seen in the cloning . WT codons are shown in green and missing alleles are shown in grey . ( C ) The amino acid coverage of the library is almost complete . WT residues are shown in green and missing alleles are shown in grey . ( D ) Examining the number of barcodes per amino acid substitution shows that 2 . 5% of the library is missing and the median number of barcodes per substitution is 15 . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 003 To determine the differential fitness landscape of Ub under different chemical stresses , we first conducted an EMPIRIC-BC experiment under 0 . 5% DMSO to serve as a control ( Figure 2 ) . The resulting fitness landscape is quite similar to the previously published dataset , which was collected under no chemical stress ( Roscoe et al . , 2013 ) ( Figure 3A ) . The lowest fitness scores occurred at premature stop codons and residues that are critical to build Lys48 poly-Ub linkages ( Lys48 , Ile44 , Gly75 , Gly76 ) . As previously observed , much of the protein surface is tolerant to mutation . Based on the average value of the stop codon substitutions , we set a minimum fitness score of -0 . 5 ( Figure 3B ) . Comparisons of biological replicates indicated that the data were reproducible and well fit by a Lorentzian function centered at 0 , as expected for the ratio of two Gaussian distributions ( Figure 3C , D ) . 10 . 7554/eLife . 15802 . 004Figure 2 . Competition experiment based on a galactose inducible Ub . The fitness of all ubiquitin mutants was measured in a single culture by shutting off the galactose-driven wild type copy . This allows a constitutively expressed mutant to be the sole source of ubiquitin for the cell . The library was grown for 48 hr in galactose to remove dominant negative alleles and then expression of the wild type copy repressed by the addition of glucose . Upon repression of the wild type copy , chemical perturbations were added and the yeast were grown for multiple generations . Fitness scores were calculated for each mutant based on the relative frequencies of mutant and wild type alleles over multiple generations . The ratio of ( mutant counts ) : ( wild type counts ) was computed for each time point and a line fit to these ratios vs . generation time . The fitness score is the slope of the linear fit . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 00410 . 7554/eLife . 15802 . 005Figure 3 . Ubiquitin fitness scores determined in DMSO are replicable and define the 'unperturbed' Ub fitness landscape . ( A ) Heatmap showing the fitness of observed ubiquitin alleles . Scores presented are the average of three biological replicates . Wild type amino acids are shown in green and mutations without fitness values ( due to lack of barcode or competition sequencing reads ) are shown in grey . The average fitness score of each position and the averages of substitutions binned by amino acid characteristics are shown below . The single column on the far right shows the average of each amino acid substitution across all positions . ( B ) The distribution of fitness values is shown and colored based on fitness score . Grey bins reflect fitness scores that were reset to -0 . 5 . ( C ) Biological replicates of the competition experiment in DMSO are well correlated ( R2 = 0 . 79 ) . Each point represents the fitness score of a mutant in two biological replicates . Points are colored based on the average standard deviation of the barcodes contributing to each fitness score . ( D ) The distribution of the residuals to the identity line between two DMSO replicates is symmetric and well modeled by a Lorentzian ( X0 = 0 , Γ = 0 . 0035 , scaled by 1600 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 00510 . 7554/eLife . 15802 . 006Figure 3—figure supplement 1 . Error estimates for fitness scores determined in DMSO . We calculated the standard deviation of the distributions of barcode fitness scores that contribute to each amino acid mutant fitness score . Large errors of the stop codon substitutions are due to variation of fitness scores below the -0 . 5 floor . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 00610 . 7554/eLife . 15802 . 007Figure 3—figure supplement 2 . Fitnesses determined in DMSO are well correlated to the previously determined unperturbed fitnesses . A linear regression ( R2 = 0 . 785 ) and Pearson’s correlation coefficient ( CC = 0 . 886 ) were calculated between the fitness scores determined in DMSO and the previously published unperturbed dataset ( Roscoe et al . , 2013 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 007 Because our competition experiments require cells growing for multiple generations in log phase , we conducted our experiments at chemical concentrations that inhibit yeast growth by 25% . These chemical concentrations are not as high as used in previous transcriptional studies of yeast chemical stress responses ( Gasch et al . , 2000 ) . For Caffeine ( 7 . 5 mM ) , and DTT ( 1 mM ) we determined the IC25 for each drug by following growth via optical density ( Figure 4—figure supplement 1 ) . Since HU ( 25 mM ) induced a lag phase followed by near wild type like growth , we determined the IC25 concentration by monitoring yeast growth from two to five hr post treatment . DMSO ( 0 . 5% ) and MG132 ( 50 uM ) did not inhibit growth . Next , we performed the EMPIRIC-BC experiment with each chemical perturbation ( Figure 4B ) . In Caffeine , DTT , and HU ( Figure 4A–D , Figure 4—figure supplements 1–3 ) many mutations are sensitized , and become less fit than in DMSO . Generally this increased sensitivity is localized around the C-terminus , which is essential for E1 activation , and the hydrophobic patch , which is the dominant interface for protein-protein interactions . 10 . 7554/eLife . 15802 . 008Figure 4 . Perturbations sensitize ubiquitin to mutations . The difference in fitness between DMSO and a perturbation for each Ub allele: ( A ) Caffeine , ( B ) DTT ( C ) Hydroxyurea ( D ) MG132 . Wild type amino acids are shown in red and mutations without fitness values ( due to lack of barcode or competition sequencing reads ) are shown in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 00810 . 7554/eLife . 15802 . 009Figure 4—figure supplement 1 . Perturbations sensitize ubiquitin to mutations . Heatmaps showing the fitness of observed Ub alleles under: ( A ) Caffeine , ( B ) DTT ( C ) Hydroxyurea ( D ) MG132 . Wild type amino acids are shown in green and mutations without fitness values ( due to lack of barcode or competition sequencing reads ) are shown in grey . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 00910 . 7554/eLife . 15802 . 010Figure 4—figure supplement 2 . Perturbations sensitize ubiquitin to mutations . Heatmaps showing the error estimates for each observed Ub alleles under; ( A ) Caffeine , ( B ) DTT ( C ) Hydroxyurea ( D ) MG132 . Wild type amino acids are shown in green and mutations without fitness values ( due to lack of barcode or competition sequencing reads ) are shown in grey . Growth rates were calculated by monitoring OD600 over 8 hr and normalized to the unperturbed SUB328 growth rate . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01010 . 7554/eLife . 15802 . 011Figure 4—figure supplement 3 . Growth curves . We determined the concentration to inhibit SUB328 growth by 25% by monitoring optical density . Error bars represent standard deviation of multiple measurements . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 011 To compare the responses to each perturbation , for each pairwise comparison we plotted the fitness scores for each mutant as a scatter plot and calculated the residual to the identity line . We compared the distribution of these residuals to the distribution of residuals calculated by the DMSO self comparison ( Figure 5 ) . Caffeine , DTT and HU generally sensitize the protein to mutation , which is evident in the enrichment of mutations with reduced fitness compared to the DMSO self distribution . These newly sensitized mutations are largely shared between these different chemical perturbations . 10 . 7554/eLife . 15802 . 012Figure 5 . Residual distributions highlight a shared mutational response between Caffeine , DTT and HU . The residuals between datasets shows are shown with the Lorentzian representing the biological replicates of DMSO in red . When compared to DMSO , three perturbations ( Caffeine , DTT and HU ) shift the distributions to the left , which highlights the increased sensitivity to mutation . In contrast , MG132 slightly shifts the distribution to the right , which highlights the alleviating interaction between MG132 and deleterious ubiquitin alleles . Comparisons between Caffeine , Hydroxyurea and DTT are symmetric but with longer tails than the control experiments . This result suggests a shared response comprised of many sensitized residues and a smaller number of perturbation-specific signals . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 012 In contrast to the sensitizing effects of DTT , Caffeine , and HU , the proteasome inhibitor MG132 increases mutational tolerance throughout the protein . This effect can be seen in the slight shift of the residuals distribution to the right when compared to the DMSO self distribution ( Figure 5D ) . The effect is small at the MG132 concentration we assayed , which is likely due to the poor penetrance of MG132 in yeast cells containing a wild type allele of ERG6 ( Lee and Goldberg , 1996 ) . This alleviating interaction is likely because MG132 directly perturbs proteasome , reducing the impact of defects related to Lys48 linked poly-Ub chains and leaving functions related to other , non-degradative poly-Ub topologies unperturbed . Also , reducing proteasome activity may increase the free pool of Ub in the cell by reducing the number of Ub proteins degraded by the proteasome . This increased pool of Ub could buffer the effects of deleterious Ub mutants participating in non-proteasomal functions . The consequence of an increased pool of free Ub might therefore lead to the general alleviating interactions observed between Ub mutants and MG132 treatment . Additionally , the difference distributions are wider than the distribution between DMSO replicates . This result shows that the perturbations unmask previously buffered fitness defects that are phenotypically important for each perturbation . If the biological role of each residue were independent of perturbation then the distributions would be shifted without affecting the shape . Instead these data show that each perturbation uncovers unique roles of Ub residues in responding to a specific perturbation . One potential explanation of the buffer unmasked by the chemical perturbations is the stability of the Ub protein itself . Although Ub is highly stable ( Ibarra-Molero et al . , 1999; Wintrode et al . , 1994 ) , mutations that destabilize it may lead to misfolding or perturb Ub/protein interactions important for UPS function . To assess the degree to which mutational destabilization of ubiquitin itself is predictive of a decrease in mean fitness for each perturbation , we used the macromolecular modeling software Rosetta to estimate changes in protein stability ( Kellogg et al . , 2011; Kortemme and Baker , 2002 ) for every mutation in our library . With the resulting predictions , we classified each ubiquitin mutation as either destabilizing ( change in Rosetta Energy Units ( REU ) > = 1 . 0 ) or neutral/stabilizing ( change in REU < 1 . 0 ) . We observed a significant difference in experimental fitness between the two predicted classes for all conditions ( Figure 6 ) . This result holds independently of the absolute mean experimental fitness score of each perturbation , meaning that the difference in mean experimental fitness between predicted destabilizing and neutral mutations is not simply the result of lower mean destabilizing fitness scores . These results suggest that ubiquitin stability is more important for fitness in each of the perturbed conditions than in unperturbed yeast . Under stress , subtle changes in Ub stability could induce fitness defects that are otherwise buffered under control ( DMSO ) conditions . Furthermore , even small changes to ubiquitin stability could induce considerable changes to the Ub conformational ensemble that could destabilize Ub/protein complexes ( Lange et al . , 2008; Phillips et al . , 2013 ) . Adaptability within the UPS could buffer these defects in DMSO , but they can be revealed upon chemical stress . 10 . 7554/eLife . 15802 . 013Figure 6 . Fitness score data binned by Rosetta stability predictions . Fitness scores for each of the 5 sets of experimental conditions are shown along the y-axis as boxplots . Scores are grouped first by their respective experimental condition , and then by the change in stability of the ubiquitin monomer of the mutation estimated by Rosetta . Mutations that Rosetta predicts to be neutral or stabilizing ( REU ( Rosetta Energy Units ) < 1 . 0 ) are shown in blue boxes; mutations predicted to be destabilizing ( REU >= 1 . 0 ) are shown in green boxes . The mean of each fitness score distribution is shown as a white dot . The p-value of the two-sided T test between the fitness mean of mutations predicted to be stabilizing and those predicted to be neutral/stabilizing is shown at the bottom of the plot . Experimental conditions are arranged from left to right along the x-axis in order of decreasing p-value . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01310 . 7554/eLife . 15802 . 014Figure 7 . The structure of Ub highlighting important residues . Cartoon model of Ub ( PDB 1UBQ ) with important residues colored as follows: Lys48 - orange , Lys63 - light blue , Lys11 - green , other Lys residues - yellow , hydrophobic patch ( Leu8 , Val40 , Ile 44 ) - red , C-terminal diGly motif ( Gly75 and 76 ) - purple , Arg42 - pink , His68 - olive . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01410 . 7554/eLife . 15802 . 015Figure 8 . Average fitness values show sensitization by the perturbations at each position in ubiquitin . ( A ) Based on the average fitness score , positions were binned into tolerant ( >=-0 . 075 - Blue ) , intermediate ( <-0 . 075 to > -0 . 35 - Pink ) and sensitive ( <= -0 . 35 - Red ) . ( i ) DMSO ( ii ) Caffeine ( iii ) DTT ( iv ) Hydroxyurea show a shift from tolerant to intermediate and sensitive positions . ( B ) Positions binned by average fitness score mapped onto the ubiquitin structure . C-alpha atoms are shown in spheres and the residues are colored as in A . Met1 is colored grey . ( C ) New sensitive positions induced by the perturbation describe a shared response to perturbation with 8 of 13 positions shared between Caffeine , DTT and HU . ( D ) New intermediate positions highlight the similarity between HU and Caffeine , with DTT sensitizing a unique set of residues . ( E ) New tolerant positions are unique to each perturbation . ( F ) Average position fitness scores mapped onto ubiquitin . ( i ) DMSO ( ii ) Minimum average fitness score in all perturbations . C-alpha atoms are shown in spheres and the residues are colored according to average fitness . Met1 is colored grey . ( G ) Minimum average fitness scores – DMSO average fitness scores mapped onto ubiquitin . C-alpha atoms are shown in spheres and the residues are colored according to the difference in fitness . Met1 is colored grey . With this small set of perturbations most positions are sensitized . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 015 To assess the role of specific positions in Ub we averaged the fitness score of each amino acid mutation at a given position . We then binned each position into sensitive ( ≤-0 . 35 ) , intermediate ( -0 . 35 to -0 . 075 ) and tolerant ( ≥-0 . 075 ) and examined the distribution of average fitness in each condition ( Figure 8A ) . These distributions again show that most positions in Ub are tolerant to mutation in DMSO , but many positions are sensitized upon chemical perturbation . In DMSO only residues with well-established roles are sensitive: Arg42 ( E1 activation ) , Ile44 ( hydrophobic patch hotspot ) , Lys48 ( essential Lys48 linked poly-Ub ) and Gly75-Gly76 of the C-terminus ( E1 activation ) ( Figure 7 ) . The face opposite the hydrophobic patch is mostly tolerant and the protein core and residues adjacent to the sensitive residues are mostly intermediate ( Figure 8B - i ) . When treated with Caffeine , DTT or HU , a shared set of residues become sensitive ( Figure 8B ii– iv , Figure 8C ) . These residues are either: located adjacent to DMSO sensitive residues ( e . g . Leu73 , which is in the C-terminal tail ) ; residues with important biological functions that of intermediate sensitivity in DMSO ( e . g . Leu8 , Val70 , which are important hydrophobic patch residues ) ; or core residues ( e . g . Ile36 , Leu71 ) . These positions tolerated a small set of substitutions in DMSO but upon perturbation became only tolerant of mutations that share physical chemistry with the wild type residue . Examining the positions made intermediate by the perturbations highlights the similarities and differences between the DTT and Caffeine/HU datasets ( Figure 8D ) . All three perturbations shift a shared set of residues to the intermediate bin . These residues are mostly surface residues on the tolerant face of Ub . In DMSO they tolerate a wide range of amino acids . Upon perturbation the mutational tolerance is reduced to amino acids generally compatible with surface residues . For example in DMSO , Asp32 is tolerant to any substitution except Proline . Upon perturbation , this position is restricted to polar and negative substitutions . Additionally , DTT uniquely shifts five positions into the intermediate bin . This is due to subtle changes in the tolerance of positions that are otherwise highly tolerant . For example , mutations at Arg54 are well tolerated in all other conditions . However , in DTT mutations to negative residues become deleterious while all other substitution remain tolerated . This suggests that Arg54 may participate in a salt bridge during a protein-protein interaction that is involved in mediating the cellular response to DTT . We also uncovered newly tolerant positions , which are uniquely tolerant to each of the perturbations ( Figure 8E ) . These positions tend to be mildly sensitive to most mutations in DMSO , suggesting that these residues are involved in biological pathways that are important for cellular function , but not essential . When perturbed , these positions are mildly desensitized to mutation , with little regard for mutant amino acid identity . The most striking example is at Lys63 in DTT . In all other conditions any mutation of this residue is mildly deleterious . Because Lys63 linked poly-Ub chains are important for efficient cargo sorting in the endosome , this sensitivity is likely due to an endocytic defect . However , upon DTT treatment most mutants of Lys63 show WT-like fitness . This suggests an epistatic interaction between DTT treatment and mutations at Lys63 . The endocytic defect caused by mutation Lys63 is likely masked by a similar endocytic defect induced by DTT treatment . Average difference maps showing the ( DMSO - perturbation ) fitness score highlight the features that underlie the sensitized and desensitized positions ( Figure 4 ) . In a final effort to resolve the dichotomy between the Ub fitness landscape and the evolutionary record , we visualized the average fitness of each position in DMSO and compared it to the minimum of the average fitness of each position for all perturbations ( Figure 8F , G ) . The data in DMSO again shows that biologically relevant positions are sensitive , the face opposite the hydrophobic patch is extremely tolerant to mutation , and that core residues are intermediately tolerant . Perturbations dramatically increase mutational sensitivity at the C-terminus , around the hydrophobic patch and at some core positions . However , much of the tolerant face of the protein remains tolerant to mutation in all of the perturbations . By exploring a wider array of perturbations we should be able determine the environmental pressures that constrain these tolerant positions and explain the extreme conservation of Ub . To determine the elements of the shared response to HU , Caffeine and DTT , we defined 'shared sensitizing mutations' as those that were both sensitizing ( delta fitness ≤ -0 . 2 for all perturbations ) and consistent between perturbations ( within 0 . 1 of the regression line ) ( Figure 9A and Figure 9—source data 1 ) . Most of these mutations change from being mildly deleterious to being nearly null upon chemical stress . For example , in DMSO Ub tolerates mutation to small hydrophobics and other polar residues at Thr7 . However , chemical stresses causes mutations of small hydrophobic or charged residues at this position to be deleterious . As Thr7 is adjacent to the hydrophobic patch residue Leu8 , this sensitization is likely due to non-polar substitutions disrupting Ub adaptor protein binding and poly-Ub packing ( Komander and Rape , 2012 ) . Additionally , typically destabilizing substitutions such as Proline or Tryptophan generally become more deleterious under perturbation . We also investigated specific signals outside of the shared sensitizing response ( Figure 9B ) . We identified perturbation specific mutations by comparing the change in fitness scores of each of the sensitizing perturbations ( Figure 9—source data 2 , 3 ) . Because these mutants are not sensitized by all of the perturbations they likely alter binding to specific adaptors , conjugation machinery , or substrates . Most of these perturbation specific mutants are tolerated in Caffeine and HU , but sensitized by DTT treatment . However , the H68Y mutation differs as DTT and HU treatments sensitize this mutation whereas Caffeine treatment does not . His68 is an important position found at the interface between Ub and adaptor domains such as UIM and UBA domains . These domains are important for the trafficking of ubiquitinated proteins . His68 lies adjacent to the hydrophobic patch and binding to UIM domains is reduced when it is protonated ( Fujiwara et al . , 2004 ) . In contrast , when His68 is mutated to Val in Ub , the binding to UIM domains is increased , likely mimicking the deprotonated state that forms a hydrophobic surface ( Fujiwara et al . , 2004 ) . 10 . 7554/eLife . 15802 . 016Figure 9 . A shared response to different chemical perturbations . ( A ) DMSO fitness - Caffeine fitness vs . DMSO fitness - DTT fitness . The markers are colored based on DMSO fitness - Hydroxyurea fitness . ( B ) DMSO fitness - Caffeine fitness vs . DMSO fitness - Hydroxyurea fitness . The markers are colored based on DMSO fitness - DTT fitness . ( C ) DMSO fitness - DTT fitness vs . DMSO fitness - Hydroxyurea fitness . The markers are colored based on DMSO fitness - Caffeine fitness . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01610 . 7554/eLife . 15802 . 017Figure 9—source data 1 . Shared response mutants representing mutations that are equally perturbed by all three sensitizing perturbations . Mutants in the shared response were determined by fitting a line to the fitness scores . The distance from each point to that line was calculated . If the distance was less than 0 . 1 and the average Δ ( DMSO - Perturbation ) fitness was less than -0 . 2 the mutant was considered part of the shared response . E1 activity relative to WT Ub ( Roscoe and Bolon , 2014 ) is listed and may explain the sensitization of some of the shared response mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01710 . 7554/eLife . 15802 . 018Figure 9—source data 2 . Perturbation specific mutations represent alleles that are differentially affected by Caffeine , DTT and Hydroxyurea . Perturbation specific mutations were determined by fitting a line to the delta ( DMSO - perturbation ) fitness scores . The distance from each point to that line was calculated . If the distance was greater than 0 . 35 the mutant was classified as perturbation specific . Mutants with high experimental errors were deemed outliers and removed from this list . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 01810 . 7554/eLife . 15802 . 019Figure 9—source data 3 . Specific information regarding highlighted mutants . DOI: http://dx . doi . org/10 . 7554/eLife . 15802 . 019 Lys11 is similarly important for Ub biology and shows a specific sensitization to DTT . Lys11 linked poly-Ub chains are the second most abundant linkage in yeast . These chains likely signal for degradation by the proteasome , like Lys48 linked chains , and have been implicated in the response to ER stress ( Xu et al . , 2009 ) . In DMSO all substitutions , except to negative and aromatic residues , are tolerated . However , substitutions to Leu , Ile , His and Asn are sensitized uniquely in DTT . These data suggest that Lys11 is mediating an interaction to DTT induced stress . Although previous studies have indicated a synthetic lethal interaction between Lys11Arg and DTT ( Xu et al . , 2009 ) , in our experiments , at lower DTT concentrations , the relatively high fitness of Lys11Arg suggests that the structural role of the positively charged residues and not poly-Lys11 Ub linkages may dominate the physiological response . In addition to fitness defects that are likely due to perturbing Ub/protein interfaces , we also observed defects due to perturbing ploy-Ub chain structure and dynamics . Lys63 linked poly-Ub chains exist in three distinct conformations in solution ( Liu et al . , 2015 ) . The populations of these conformational states help determine binding partner selection between Lys63 linked chains and adaptor proteins . Mutating Glu64 to Arg biases the chains towards the open conformation ( Liu et al . , 2015 ) . In DMSO , the mutation of Glu64 to a positive residue caused a fitness defect . In Caffeine and HU these mutants are sensitized and the fitness defect is further increased . However , DTT treatment increased the tolerance to positive mutations at this position , again suggesting an interaction between Lys63 linked poly-Ub and DTT treatment .
We have determined the fitness landscape of Ub in yeast grown in the presence of five chemical perturbations . We identified newly sensitized positions in the protein , which supports the hypothesis that the Ub sequence is highly constrained by its role in a wide array of environmental stress responses . Although each perturbation had some unique features , we observed a general buffering effect that may have obscured mutational sensitivity in the previously determined Ub fitness landscape . Perhaps the most surprising result in our study was the failure to recapitulate the synthetic lethal interaction between Lys11Arg and DTT ( Xu et al . , 2009 ) . This interaction was observed using the same strain ( SUB328 ) , however fitness was determined through a dilution spot assay on an agar plate containing 30 mM DTT . Our experiments were conducted in liquid culture with 1 mM DTT refreshed every sampling period . It is likely that we did not achieve a stress regime where Lys11 poly-Ub is essential for DTT tolerance . The Lys11Arg mutation induces the upregulation of proteins involved in ERAD including Ubc6 , the ERAD E2 . Also , the turnover of known ERAD substrates is unaffected by the Lys11Arg mutation , suggesting that Lys48 linked chains can be substituted for Lys11 linked chains ( Xu et al . , 2009 ) . These adaptations could be sufficient to counteract the loss of Lys11 poly-Ub in our experiments , but are insufficient at higher concentrations of DTT . It would be interesting to explore these two regimes and determine the concentration of DTT that induces the lethality of the Lys11Arg mutant . Taken together , these data represent a step towards understanding the apparent dichotomy between the Ub conservation and the previously determined Ub fitness landscape . While much of the protein is tolerant to mutation when cells are grown with traditional laboratory conditions , new stress conditions reveal hidden mutational sensitivity . We show that thirteen new positions are extremely sensitized in at least one stress condition with an additional thirteen new positions intermediately sensitized . While the incorporation of these new stresses provides a rationale for an additional 1/3 of the protein , we cannot currently explain the conservation of some positions in the 'tolerant' face of the protein . Expanding the set of chemical perturbations assayed may begin to address this dichotomy further . It is also possible that mutations at tolerant positions create fitness defects that are too subtle to be determined by our current methods . These subtle defects can lead to the sequence conservation observed in Ub when a large population undergoes selection over a longer evolutionary time ( Boucher et al . , 2014 ) . Future experiments may be able to identify these effects by: i ) increasing the dose of the perturbation; ii ) reducing the expression of the mutant Ub; or iii ) performing the selection over more generations ( Rockah-Shmuel et al . , 2015 ) . The observed fitness defects can be due either to the functional properties of the mutant or the concentration of free Ub in the cell . The mutants in the library are all expressed on the same plasmid and the same promoter , which gave us confidence to interpret the effect of the mutants on Ub thermostablitly and biological function . However , it is important to note that the fitness defects may be due to decreased Ub mutant expression or increased Ub mutant degradation . Historically , many temperature sensitive mutants are not deficient in protein activity , but rather have increased protein turnover by the proteasome due to destabilizing mutations ( Gardner et al . , 2005 ) . Ub turnover is unique in that free Ub is not degraded ( Shabek et al . , 2007 ) and conjugated Ub is degraded by the proteasome when the DUBs Upb6 and Rpn11 do not remove Ub from the substrate , causing Ub to be pulled into the proteasome along with the substrate protein ( Hanna et al . , 2007 ) . Additionally , the free Ub pool can be increased by the activity of the DUB Doa4 ( Kimura et al . , 2009 ) , which cleaves Ub conjugates into free Ub . Therefore some of the mutants may increase Ub turnover either by interfering with DUB recognition or by destabilizing Ub and easing Ub unraveling by the proteasome ( Lee et al . , 2001; Prakash et al . , 2004 ) . Selection experiments that probe protein-protein interactions more directly , similar to those performed between the yeast E1 and Ub ( Roscoe and Bolon , 2014 ) , may be able to directly determine whether some mutants have abnormal proteasomal engagement and turnover . The inability of SUB328 to regulate the free Ub pool by increasing expression may also limit the interpretation of our results . Physiologically Ub protein levels are maintained by strong expression of the UBI4 locus in response to stress conditions ( Finley et al . , 1987 ) . In the strain used in these experiments ( SUB328 ( Finley et al . , 1994 ) , the native Ub loci have been deleted and complemented with both a plasmid containing WT Ub and a plasmid containing the mutant Ub . Therefore , SUB328 can no longer respond to stress by increasing the expression of UBI4 and is entirely dependent on DUB based mechanisms to maintain the free Ub pool , increasing the fitness defects of mutants that interfere with Doa4 , Upb6 , and Rpn11 activity . Alternatively , SUB328 expressing a deficient Ub mutant might increase Ub expression by increasing the copy number of the plasmid . While integrating these Ub alleles into the genome would remove copy number variation , it would also dramatically decrease number of Ub variants that could be assessed due to the relative inefficiency of integration when compared to transformation . While these caveats are a concern , the SUB328 experimental system has been successfully used for many years to assess the effect of Ub mutations in yeast ( Roscoe et al . , 2013; Sloper-Mould et al . , 2001; Lee et al . , 2014 ) . Furthermore , we have only interpreted mutants that cause large defects , such as the biologically sensible fitness defects of mutations at known important residues such as Leu8 or Lys48 and residues with fitness values that vary between conditions such as His68 . The fitness effects at these variable positions are not simply due to expression or turnover defects , which should decrease fitness uniformly across the chemical stresses . Therefore , mutants that affect protein turnover or expression are likely members of the 'shared response' mutants described above . In the case where a perturbation increased Ub turnover we would expect that most positions would be uniformly sensitized . Instead we observe a stereotyped pattern of sensation for the three sensitizing perturbations , suggesting that Ub turnover is similar for all three perturbations and the 'perturbation specific mutations' are independent of protein turnover and/or expression defects . These experiments also demonstrate the success of graduate-level project based courses ( Vale et al . , 2012 ) as key components of a first-year curriculum . Our students were able to generate high quality data and useable computational pipelines during the 8 weeks of class time . These successes are notable because few students began the class with a background in both areas . By creating a project lab environment that encouraged team based learning and teaching , we enabled students to quickly acquire relevant skills within the context of an active research project . The wide variety of stress responses that Ub mediates and the vast chemical space that can be safely and economically addressed in a classroom make yeast and Ub ideal systems for continuing these studies . It is our hope that other graduate programs can similarly offer project based classes in their curriculums and we will make our reagents available for use to further that goal .
Yeast strain SUB328 ( MATa lys2-801 leu2-3 , 2–112 ura3-52 his3-Δ200 trp1-1 ubi1-Δ1::TRP1 ubi2-Δ2::ura3 ubi3-Δub-2 ubi4-Δ2::LEU2 ( pUB146 ) ( pUB100 ) ) ( Finley et al . , 1994 ) was used , which expresses ubiquitin from a galactose-inducible promoter in pUB146 . pUB100 expresses the Ubi1 tail . A library of ubiquitin genes was saturated with point mutations ( Roscoe et al . , 2013 ) . Barcodes were added by ligating N18 oligos flanked by EagI and AscI sites into each of the eight previously create Ub libraries . These libraries were bottlenecked by transformation into E . coli and then pooled to create the single N18BC-UbLib . This pooled library was transformed into E . coli to create the final N18BC-UbLib . To associate the N18BCs to a given Ub allele , we performed a paired end read on the Illumina MiSeq . Because Ub is a small gene , we were able to read the entire ORF with a 260 bp read and the associated N18BC with a 30 bp read . To prepare the library for sequencing , plasmid DNA was extracted from E . coli using the Omega Bio-Tek mini-prep kit . A ~700 bp product was amplified with primers containing the Illumina PE1 and PE2 primer sequences for 9 cycles to minimize PCR recombination . These products were separated on agarose gel , and excised products were purified by silica column . This library was prepared for sequencing on the Illumina MiSeq . The concentration to reduce the growth rate of SUB328 ( WT Ub ) by 25% was determined by monitoring the growth of cells by optical density measurements at 600nm over 8 hr . MG132 and DMSO did not affect SUB328 ( WT Ub ) growth rate at any tested concentration . Hydroxyurea treatment induces a lag-phase followed by WT like growth . | The ability of an organism to grow and reproduce , that is , it’s “fitness” , is determined by how its genes interact with the environment . Yeast is a model organism in which researchers can control the exact mutations present in the yeast’s genes ( its genotype ) and the conditions in which the yeast cells live ( their environment ) . This allows researchers to measure how a yeast cell’s genotype and environment affect its fitness . Ubiquitin is a protein that many organisms depend on to manage cell stress by acting as a tag that targets other proteins for degradation . Essential proteins such as ubiquitin often remain unchanged by mutation over long periods of time . As a result , these proteins evolve very slowly . Like all proteins , ubiquitin is built from a chain of amino acid molecules linked together , and the ubiquitin proteins of yeast and humans are made of almost identical sequences of amino acids . Although ubiquitin has barely changed its sequence over evolution , previous studies have shown that – under normal growth conditions in the laboratory – most amino acids in ubiquitin can be mutated without any loss of cell fitness . This led Mavor et al . to hypothesize that treating the yeast cells with chemicals that cause cell stress might lead to amino acids in ubiquitin becoming more sensitive to mutation . To test this idea , a class of graduate students at the University of California , San Francisco grew yeast cells with different ubiquitin mutations together , and with different chemicals that induce cell stress , and measured their growth rates . Sequencing the ubiquitin gene in the thousands of tested yeast cells revealed that three of the chemicals cause a shared set of amino acids in ubiquitin to become more sensitive to mutation . This result suggests that these amino acids are important for the stress response , possibly by altering the ability of yeast cells to target certain proteins for degradation . Conversely , another chemical causes yeast to become more tolerant to changes in the ubiquitin sequence . The experiments also link changes in particular amino acids in ubiquitin to specific stress responses . Mavor et al . show that many of ubquitin’s amino acids are sensitive to mutation under different stress conditions , while others can be mutated to form different amino acids without effecting fitness . By testing the effects of other chemicals , future experiments could further characterize how the yeast’s genotype and environment interact . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology"
] | 2016 | Determination of ubiquitin fitness landscapes under different chemical stresses in a classroom setting |
Endothelial integrity is vital for homeostasis and adjusted to tissue demands . Although fluid uptake by lymphatic capillaries is a critical attribute of the lymphatic vasculature , the barrier function of collecting lymphatic vessels is also important by ensuring efficient fluid drainage as well as lymph node delivery of antigens and immune cells . Here , we identified the transmembrane ligand EphrinB2 and its receptor EphB4 as critical homeostatic regulators of collecting lymphatic vessel integrity . Conditional gene deletion in mice revealed that EphrinB2/EphB4 signalling is dispensable for blood endothelial barrier function , but required for stabilization of lymphatic endothelial cell ( LEC ) junctions in different organs of juvenile and adult mice . Studies in primary human LECs further showed that basal EphrinB2/EphB4 signalling controls junctional localisation of the tight junction protein CLDN5 and junction stability via Rac1/Rho-mediated regulation of cytoskeletal contractility . EphrinB2/EphB4 signalling therefore provides a potential therapeutic target to selectively modulate lymphatic vessel permeability and function .
The endothelium of blood and lymphatic vessels forms a barrier that controls the movement of fluid , molecules , ions and cells between the blood/lymph and the tissue . The barrier function of endothelial cells ( ECs ) varies among different organs and vessel types . The blood brain barrier ( BBB ) , for example , is formed of a continuous layer of ECs connected by specialized tight junctions and adherens junctions ( Zhao et al . , 2015 ) . In contrast , blood vessels of the kidney and small intestine are lined by fenestrated ECs to facilitate rapid exchange , uptake and secretion of fluids , solutes and molecules ( Aird , 2012 ) . The architecture of lymphatic endothelial cell ( LEC ) junctions also differs between the different vessel types ( Baluk et al . , 2007 ) . Highly permeable button-like junctions of lymphatic capillaries allow uptake of fluid from the interstitium . The lymph is then drained and transported via collecting lymphatic vessels that are equipped with continuous zipper-like junctions preventing excessive leakage ( Baluk et al . , 2007; Potente and Mäkinen , 2017 ) . The molecular mechanisms that establish and maintain such functionally specialised junctional features of different vessel types are poorly understood . Vascular integrity is regulated by junctional adhesion molecules , of which the adherens junction molecule vascular endothelial ( VE ) -cadherin and the tight junction molecule Claudin 5 ( CLDN5 ) have received particular attention . Dysregulation of VE-cadherin can result in junctional disruption , increased vessel permeability and severe ( lymph ) edema formation ( Frye et al . , 2015; Hägerling et al . , 2018; Yang et al . , 2019 ) . Interestingly , different effects of gene disruption in mice have been observed depending on the organ and the vessel type . For example , endothelial loss of VE-cadherin in adult mice results in junctional disruption and increased vascular permeability in the heart and lung , but not the skin and brain ( Frye et al . , 2015 ) . LEC-specific deletion of VE-cadherin in adult mice similarly leads to organ-specific disruption of endothelial junctions affecting the mesenteric collecting vessels but not dermal lymphatic vessels ( Hägerling et al . , 2018; Yang et al . , 2019 ) . Unlike VE-cadherin expressed in all ECs , CLDN5 is absent from certain phenotypically ‘leaky’ vessels ( Benz et al . , 2019 ) . For instance , in the skin , CLDN5 is not expressed in blood vessels that are permissive to VEGF-induced vascular leakage ( Honkura et al . , 2018 ) . In line with a function in junctional stabilization , CLDN5 is critical for the establishment and maintenance of the BBB ( Greene et al . , 2018; Nitta et al . , 2003 ) . CLDN5 is however highly expressed in the phenotypically ‘permeable’ lymphatic endothelium including lymphatic capillaries ( Baluk et al . , 2007 ) , where its function is not known . The junctional adhesion molecules are intracellularly associated to the actin cytoskeleton ( Dejana et al . , 2009 ) . The molecular control of endothelial cytoskeletal dynamics presents therefore a second important leverage point that regulates vascular integrity . Quiescent endothelia are characterized by a balance of actin stabilization and myosin-based actin pulling forces that are constantly applied to endothelial junctions . Elevation of myosin-based actin contractility activates endothelial junctions . Rho GTPases are key regulators of such cytoskeletal dynamics ( Dorland and Huveneers , 2017 ) . Differential expression and function of Rho GTPase regulators , such as the guanosine nucleotide exchange factor Vav3 , has been proposed to contribute to barrier diversity across different blood vessels ( Hilfenhaus et al . , 2018 ) . Upstream regulators of Rho GTPases may provide another level for organ- and vessel-type specific barrier regulation . Here , we identified EphrinB2/EphB4 signalling as a critical and selective regulator of collecting lymphatic vessel integrity . Inducible EC-specific deletion of Efnb2 or Ephb4 disrupted lymphatic endothelial junctions in several organs while dermal and pulmonary blood vessel barrier function was not compromised . Studies in primary human LECs further showed that inhibition of EphrinB2/EphB4 signalling led to reduction in junctional CLDN5 while VE-cadherin was not affected . LEC-specific deletion of Cldn5 in postnatal mice did not however fully recapitulate the junctional phenotype observed in Efnb2/Ephb4-deficient vessels . Interestingly , we found that basal EphrinB2/EphB4 signalling regulates junctional localisation of CLDN5 and junction stability via Rac1/Rho-mediated control of cytoskeletal contractility in primary LECs . Our results suggest that EphrinB2/EphB4 signalling provides a potential therapeutic target for intervention of diseases associated with abnormal lymphatic vessel permeability .
EphrinB2 and EphB4 play important roles in the embryonicand early postnatal development of blood and lymphatic vessels ( Adams et al . , 1999; Gerety et al . , 1999; Mäkinen et al . , 2005; Zhang et al . , 2015 ) . To study the role of EphrinB2/EphB4 signalling in the remodeling and quiescent vasculature , we conditionally deleted Ephb4 or Efnb2 in postnatal mice using the tamoxifen-inducible Pdgfb-iCreERT2iresGFP line ( Claxton et al . , 2008 ) . We studied the effect on the dermal vasculature of the ear where Pdgfb-CreERT2 targets the endothelium of all blood vessels and collecting lymphatic vessels , but not lymphatic capillaries ( Wang et al . , 2017 ) . Gene deletion was induced by 4-Hydroxytamoxifen ( 4-OHT ) administration at 3 weeks of age , when dermal endothelial cell proliferation has stopped ( Figure 1—figure supplement 1A ) . Specificity of the Pdgfb-CreERT2-mediated recombination was confirmed using the R26-mTmG reporter in whole-mount stained ears ( Figure 1A left panel ) . Immunostaining of the ear skin of a 7-week-old 4-OHT-treated Ephb4flox/flox;Pdgfb-CreERT2 mouse revealed abnormal collecting vessels with ectopic expression of the lymphatic capillary marker LYVE1 ( Figure 1A ) and disruption of VE-cadherin+ cell-cell junctions ( Figure 1B ) . A similar , albeit more severe phenotype was observed in the Efnb2flox/flox;Pdgfb-CreERT2 mice already 2 weeks after 4-OHT administration ( Figure 1C , D ) . Unexpectedly , the cell-cell junctions of lymphatic capillaries not targeted by the Pdgfb-CreERT2 transgene also appeared disorganized in the Efnb2 mutant mice ( Figure 1—figure supplement 1B ) , suggesting secondary effects caused by disruption of collecting vessels . Gross morphology of the blood vessels and the architecture of blood endothelial junctions were however unaltered in both mutants ( Figure 1A–D ) . We next performed a modified Miles assay , to assess whether barrier function of the blood endothelium was affected by endothelial deletion of Efnb2 . In agreement with a lack of morphological alterations of the junctions , permeability of the skin or lung vasculature was not changed in Efnb2 mutants compared to control littermates under homeostasis ( Figure 1E ) . To confirm these findings , we deleted endothelial Efnb2 using another tamoxifen inducible Cre line , the Cdh5-CreERT2 ( Wang et al . , 2010 ) . Efficient depletion of EphrinB2 was shown exemplarily by Western blot from total lung lysates 8 days after the first tamoxifen administration ( Figure 1F ) . However , no significant changes were observed either under homeostasis , or after VEGF- or histamine-induced vascular leakage in the lungs of Efnb2flox/flox;Cdh5-CreERT2 mutant compared to Cre-negative littermate controls ( Figure 1F ) . In summary , these results demonstrate that loss of EphrinB2/EphB4 signalling in the postnatal vasculature leads to disruption of dermal lymphatic endothelial cell-cell junctions , while dermal and pulmonary blood endothelial cell junctions and barrier integrity are not altered . To further investigate the role of EphrinB2/EphB4 signalling in the regulation of LEC junctions , we deleted Efnb2 or Ephb4 selectively in lymphatic endothelia using tamoxifen-inducible Prox1-CreERT2 mice ( Bazigou et al . , 2011 ) . Gene deletion was induced by 4-OHT administration at postnatal day ( P ) 12 , when remodeling of the ear vasculature into lymphatic capillaries and collecting vessels is initiated ( Lutter et al . , 2012 ) and increased mechanical tension at lymphatic junctions is expected ( Dorland and Huveneers , 2017 ) . GFP reporter expression demonstrated efficient and specific Cre-mediated targeting of the lymphatic vasculature ( Figure 2A ) . Efnb2 gene deletion was confirmed by qRT-PCR analysis of dermal LECs sorted from mutant and control ears ( Figure 2—figure supplement 1A ) . Loss of Efnb2 was also evidenced by defective lymphatic valve formation ( Figure 2A ) , consistent with the previously described role of EphrinB2-EphB4 signalling in this process ( Mäkinen et al . , 2005; Zhang et al . , 2015 ) . Whole-mount staining for CLDN5 revealed disrupted cell-cell junctions and disintegration of the endothelial layer in Efnb2-deficient collecting lymphatic vessels ( Figure 2A ) . Similar defects were observed after LEC-specific deletion of Ephb4 ( Figure 2A ) . Ultrastructural analysis of the Efnb2 mutant ear vasculature using transmission electron microscopy confirmed disruption of cell-cell junctions , characterized by large intercellular gaps ( Figure 2B , C ) and convoluted junctions with increased junction overlap compared to controls ( Figure 2B , D ) . We also observed signs of cell degeneration ( Figure 2B ) . Deletion of Efnb2 in the quiescent dermal lymphatic vasculature at 8 weeks and analysis at 18 weeks of age also showed junctional alterations , albeit less severe than in the juvenile mice ( Figure 2—figure supplement 1B ) . Next , we asked whether EphrinB2/EphB4 function in regulating LEC junctions is conserved in different vascular beds . Analysis of the outer lymphatic endothelial layer of the subcapsular sinus of inguinal lymph nodes that forms during embryonic development ( Bovay et al . , 2018 ) showed disorganized junctions in 4-week-old Efnb2 and Ephb4 mutant mice after deletion at P12 ( Figure 2E ) . In addition , analysis of the mesenteric vasculature of P11 Efnb2 and Ephb4 mice neonatally ( P4 ) treated with 4-OHT showed disorganisation of lymphatic endothelial cell-cell junctions in the mutant vessels compared to Efnb2 heterozygous or Cre-negative controls ( Figure 2F ) . As expected , defective lymphatic valves and chylothorax , indicative of valve dysfunction ( Nitschké et al . , 2017 ) , were also observed ( Figure 2—figure supplement 2A , B ) . Taken together , these data demonstrate that EphrinB2/EphB4 signalling regulates junctional integrity of lymphatic vessels of different vascular beds in juvenile and adult mice . Next , we asked if junctional disruption in the Efnb2/Ephb4-deficient lymphatic vessels is a cell-autonomous defect or caused secondarily due to valve disruption and consequent alterations in the transvalvular flow pattern affecting the vessel wall downstream of the valve ( Wilson et al . , 2018 ) . Analysis of the P11 neonatal mesenteric vasculature of the Efnb2GFP reporter mice showed that Efnb2 expression was not restricted to valves but was present in all LECs of collecting vessels ( Figure 3A ) . In addition , defects in cell-cell junctions were observed in Efnb2-deficient lymphatic vessel regions both upstream and downstream of the valve ( Figure 3B ) . To quantify the state of the LEC junctions , we described four junctional categories based on previous definition of blood endothelial junctions ( Neto et al . , 2018 ) : linear junctions , thick/reticular junctions , jagged junctions and discontinuous junctions as exemplified in Figure 3C . Linear morphology was previously described to represent mature stable junctions while reticular and jagged morphologies indicate active remodeling ( Dorland and Huveneers , 2017 ) . Finger-like structures were found in all lymphatic junctional categories , and were thus not included in the definition . In wild type vessels , linear junctions were most frequent in both the upstream and downstream regions of the valve ( Figure 3C ) . Discontinuous junctions were not observed ( Figure 3C ) , and were thus considered pathological . Efnb2-deficient vessels showed an almost complete loss of linear junctions at the expense of an increase in discontinuous junctions in collecting vessels , both upstream and downstream of the defective valves ( Figure 3C ) . Quantification of the morphology of LEC junctions in the lymph node capsule ( Figure 2E ) confirmed the abnormal high prevalence of discontinuous junctions in the Efnb2-deficient mice ( Figure 2—figure supplement 3 ) . To directly test if valve dysfunction caused by loss of Efnb2 contributes to the disruption of LEC junctions in collecting vessels , we generated a tamoxifen-inducible Cre line that specifically targets lymphatic valves . To this end , a BAC-transgenic mouse line with expression of CreERT2 controlled by the regulatory elements of the lymphatic valve-specific Cldn11 gene ( Takeda et al . , 2019 ) was generated ( Ortsäter and Mäkinen , unpublished data ) . Specificity and efficiency of Cre-mediated recombination was validated using the R26-mTmG reporter ( Figure 3D and Figure 3—figure supplement 1A ) . As expected , neonatal deletion of Efnb2 using the Cldn11-CreERT2 mice resulted in malformed lymphatic valves , characterized by disorganisation of PROX1high cells , in mesenteric lymphatic vessels ( Figure 3D and Figure 3—figure supplement 1B ) . However , quantification of the state of the cell junctions in regions of collecting vessels located upstream or downstream of morphologically abnormal Cre-targeted ( GFP+ ) valve showed no alterations compared to the control ( Figure 3E ) . These results show that the disruption of lymphatic endothelial cell-cell junctions in Efnb2-deficient collecting vessels is not caused secondarily by valve dysfunction . Instead they suggest a collecting vessel LEC-autonomous function of EphrinB2/EphB4 signalling in regulating junctional integrity of lymphatic vessels . Next , we studied the mechanism by which EphrinB2/EphB4 signalling regulates lymphatic endothelial junctions in primary human dermal lymphatic endothelial cells ( HDLECs ) . To first test if the signalling pathway is activated in a basal state , EphB4 was immunoprecipitated using EphrinB2-Fc from HDLECs treated with control or EFNB2 siRNA . Western blot analysis of the immunoprecipitations for phospho-tyrosine ( 4G10 ) showed basal phosphorylation of EphB4 , which was abolished upon EFNB2 silencing ( Figure 4A ) . EphB4 protein levels on the surface of EFNB2 silenced LECs were not however compromised as stimulation with crosslinked EphrinB2-Fc could potently activate EphB4 protein ( Figure 4A ) . To induce an acute loss of basal EphB4 signalling , we incubated HDLECs with an EphrinB2 function blocking antibody ( B11 ) , which binds to EphrinB2 , thereby preventing ligation to and activation of the EphB4 receptor ( Abéngozar et al . , 2012 ) . Staining for VE-cadherin revealed formation of thick/reticular and jagged HDLEC junctions , characterized by an increase in perpendicularly oriented VE-cadherin and junctional overlaps ( Figure 4B ) . These morphological changes in junctions were observed already 3 hr after EphrinB2 inhibition ( Figure 4B ) and became more pronounced over time ( Figure 4—figure supplement 1 ) . To assess the functional consequence of EphrinB2 inhibition on monolayer permeability , confluent HDLECs grown on transwell filters were pre-incubated for 3 hr with the B11 antibody . Permeability to 40 kDa FITC-dextran was substantially increased in EphrinB2 inhibited compared to control HDLECs ( Figure 4C ) . Notably , while the junctional localisation of VE-cadherin was minimally affected by EphrinB2 blockade , we observed reduction of junctional CLDN5 from VE-cadherin+ junctions ( Figure 4D ) . A similar phenotype was observed after siRNA-mediated silencing of EFNB2 ( Figure 4—figure supplement 2A , B ) . Surprisingly , depletion of VE-cadherin in HDLECs did not result in disrupted CLDN5 junctions ( Figure 4E and Figure 4—figure supplement 3 ) , suggesting that the localisation of CLDN5 at LEC junctions is independent of VE-cadherin . This is in contrast to BECs where junctional expression of CLDN5 is dependent on VE-cadherin ( Taddei et al . , 2008 ) . To test if depletion of junctional CLDN5 explained the increased monolayer permeability in EphrinB2-inhibited HDLECs , we first studied the effect of CLDN5 silencing on the junctional organisation and monolayer permeability in HDLECs . Although CDLN5 siRNA-treated LECs showed an increase in thick and reticular VE-cadherin+ junctions , monolayer permeability to 40 kDa FITC-dextran was only moderately increased ( Figure 4E , F ) . Depletion of VE-cadherin by CDH5 silencing also resulted in a minor increase in HDLEC monolayer permeability ( Figure 4F ) . To investigate the in vivo consequence of CLDN5 depletion on LEC junctions , we generated a floxed Cldn5 allele ( Figure 4—figure supplement 4A , B ) . Validation of the allele showed the expected loss of CLDN5 protein in Cldn5flox/flox;PGK-Cre ( i . e . germline Cldn5 homozygous ) embryos ( Figure 4—figure supplement 4C ) . Next , we conditionally deleted Cldn5 in lymphatic endothelia at P4 using the tamoxifen-inducible Prox1-CreERT2 mice ( Figure 4G ) . Whole-mount immunostaining of mesenteries of P11 Cldn5 mutant mice revealed efficient depletion of CLDN5 from lymphatic junctions , with the exception of CLDN5 hot spots remaining in particular in lymphatic valves ( Figure 4G ) . Cldn5 deficient vessels showed an increase in thick and reticular VE-cadherin junctions compared to control vessels ( Figure 4G ) . However , we did not observe discontinuous junctions , characteristic of the lymphatic endothelium of EphrinB2/EphB4 mutant vessels . Loss of CLDN5 also did not result in obvious valve deformation or chylothorax formation at P11 ( Figure 4G and data not shown ) . Together , these results demonstrate that depletion of junctional adhesion molecule CLDN5 is not sufficient to explain the breakdown of lymphatic junctions observed upon EphrinB2 blockade in vitro or Efnb2 deletion in vivo . Endothelial junctions are tightly controlled via the actin cytoskeleton ( Dorland and Huveneers , 2017 ) . In human umbilical vein endothelial cells ( HUVECs ) , blockade of EphrinB2/EphB4 signalling leads to increased F-actin stress fibre formation ( Abéngozar et al . , 2012 ) . We thus sought to study if changes in the actin cytoskeleton upon loss of basal EphrinB2/EphB4 signalling contribute to disruption of HDLECs junctions . Like in HUVECs ( Dorland and Huveneers , 2017 ) , stable and quiescent HDLEC junctions are aligned by thick parallel cortical actin bundles . Remodeling HDLEC junctions are instead marked by a loss of cortical actin with concomitant increase in actin stress fibres and radial actin associated with perpendicular-oriented VE-cadherin+ junctions ( Sabine et al . , 2015 ) . EphrinB2 inhibition with the B11 antibody induced disruption of cortical actin as shown by phalloidin staining of HDLECs ( Figure 5A ) . Central actin ( radial actin and actin stress fibres ) was instead increased by 61% in B11-treated compared to untreated HDLECs ( Figure 5B ) . In the blood vasculature , the tyrosine receptor kinase Tie2 ( TEK ) signals through the Rho family GTPases Rac1 and RhoA to organize the cytoskeleton into a junction-fortifying arrangement that enhances barrier function ( Mammoto et al . , 2007 ) . To test if Rac1 is similarly involved in EphrinB2/EphB4-mediated regulation of LEC junctions , we first assessed Rac1 activity in HDLECs after EphB4 activation or inactivation using a pull-down assay . Activation of EphB4 by crosslinked EphrinB2-Fc led to a 2-fold increase in Rac1 activity ( Figure 5C ) . Conversely , inhibition of EphB4 activity by treatment with EFNB2 siRNA led to a 44% decrease in Rac1 activity ( Figure 5D ) . Inhibition of Rac1 activity can lead to an increase in RhoA activity ( Wu et al . , 2009b ) . In blood ECs , spatio-temporal and antagonistic signalling of these Rho GTPases has been shown to regulate junctional integrity ( van Buul et al . , 2014 ) . We thus tested whether interference with the Rho signalling pathway can inhibit junctional disruption caused by EphrinB2 blockade . To that end , we pre-incubated HDLECs with the well-established Rho kinase ( ROCK ) inhibitor Y-27632 for 45 min before treating the cells with the B11 antibody for 3 hr . Immunostaining revealed that junctional reduction of CLDN5 upon EphrinB2 blockade was inhibited in HDLECs pre-treated with the ROCK inhibitor Y-27632 ( Figure 5E ) . Notably , cortical actin was maintained ( Figure 5E ) and the increase in central actin was markedly reduced in B11-treated HDLECs when ROCK activity was also inhibited ( Figure 5E , F ) . ROCK inhibition alone did not affect the actin cytoskeleton ( Figure 5F ) or monolayer permeability of control HDLECs to 40 kDa FITC-dextran ( Figure 5G ) . However , it efficiently inhibited EphrinB2 blockade-induced increase in monolayer permeability ( Figure 5G ) . We conclude that continuous basal EphrinB2/EphB4 signalling has a critical homeostatic function in regulating Rac1/Rho-mediated cytoskeletal organisation and thereby CLDN5-mediated lymphatic junction barrier integrity .
Endothelial barrier function is maintained via cell-cell junctions that need to be tightly controlled to allow selective permeability without compromising the overall junctional integrity . Here we identify the transmembrane ligand EphrinB2 and its tyrosine kinase receptor EphB4 as critical regulators of collecting lymphatic vessel integrity . We found that continuous basal EphrinB2/EphB4 signalling maintains cell junction stability selectively in LECs by controlling Rac1/Rho-mediated regulation of the actin cytoskeleton and CLDN5 localisation . Although endothelial cell-cell junctions of all blood and lymphatic vessels share common basic structural features and molecular composition , there are important differences that reflect different vessel functions ( Potente and Mäkinen , 2017 ) . Fluid uptake function of lymphatic capillaries is known as a critical attribute of the lymphatic vasculature and mediated via specialized highly permeable button-like junctions ( Baluk et al . , 2007 ) . The collecting lymphatic vessels are instead equipped with continuous zipper-like junctions , similar to those found in most blood vessels , that prevent excessive leakage and thus ensure efficient lymph transport ( Baluk et al . , 2007 ) . Altered collecting lymphatic vessel permeability has been implicated in a variety of pathological conditions , highlighting the need to understand the underlying mechanisms . For example , collecting vessel permeability is increased upon infection with Yersinia pseudotuberculosis ( Fonseca et al . , 2015 ) , which results in development of fibrosis ( Ivanov et al . , 2016 ) , increased inflammation and compromised immunity due to reduced flow of lymph and immune cell transport to the lymph nodes ( Kuan et al . , 2015 ) . Excessive leakage of lymph , caused by dysfunctional lymphatic endothelial junctions in the regions of collecting vessel valves , may also contribute to primary ( inherited ) lymphedema ( Mahamud et al . , 2019; Sabine et al . , 2015 ) . EphrinB2 and EphB4 are critical regulators of both blood and lymphatic vessel development ( Adams et al . , 1999: Gerety et al . , 1999; Mäkinen et al . , 2005; Zhang et al . , 2015 ) , and they show continuous expression in endothelial cells of both vessel types in adult tissues ( Luxán et al . , 2019; Mäkinen et al . , 2005 ) . Unexpectedly , we found that postnatal deletion of Efnb2 or Ephb4 in endothelial cells using the Pdgfb-CreERT2 or Cdh5-CreERT2 selectively disrupted the integrity of collecting lymphatic vessels in multiple organs , but did not compromise the barrier function of dermal and pulmonary blood vessels . The different requirements of BECs and LECs for junctional stability may be related to different dependencies on cell-cell vs . cell-matrix adhesions and substrate stiffness . In contrast to arteries and veins , collecting lymphatic vessels lack an elastic lamina , have a thinner basement membrane and less mural cell support ( Potente and Mäkinen , 2017 ) . Collecting vessel LECs may thus experience a less rigid extracellular matrix ( ECM ) , which has been associated with an increase in cell-cell coupling in favour of cell-ECM adhesions that are strengthened on stiffer matrix ( Polio et al . , 2019 ) . Although this hypothesis would need to be tested experimentally , it is interesting to note that Efnb2 is upregulated in LECs grown on soft in comparison to stiff matrices ( Frye et al . , 2018 ) . However , blood endothelial progenitors showed an increase in expression of the arterial marker EphrinB2 cultured on stiff ( arterial ) matrices ( Xue et al . , 2017 ) , suggesting that upregulation of Efnb2 on soft matrices is an LEC-specific phenomenon . Although disruption of LEC junctions was observed in both Efnb2 and Ephb4 deficient vessels , the phenotype appeared stronger and developed more rapidly upon Efnb2 deletion . It is possible that EphrinB2 reverse signalling plays a role and accounts for this difference . Our previous work showed that genetic deletion of the PDZ binding domain of EphrinB2 led to defects in developmental lymphatic vessel remodeling and collecting vessel maturation ( Mäkinen et al . , 2005 ) . Although this suggested a role for EphrinB2 reverse signalling , it was later shown that the PDZ mutation affects not only the ligand-mediated reverse but also dampens receptor-mediated forward signalling ( Zhang et al . , 2015 ) . Genetic deletion and antibody-mediated inhibition of EphB4 further established the forward signalling through the receptor as being the main Ephrin signalling pathway for collecting vessel and valve development ( Zhang et al . , 2015 ) . The basal EphrinB2-dependent phosphorylation of EphB4 we observed in LECs supports , although does not prove this notion . It is also possible that the turnover of EphrinB2 and EphB4 proteins is different , such that faster depletion of EphrinB2 upon genetic deletion could explain the rapid development of a severe phenotype . Although we did not observe blood vessel defects in the skin and lung of Efnb2 or Ephb4 deleted mice , a role for EphB4 in maintaining vascular integrity in the heart was recently demonstrated ( Luxán et al . , 2019 ) . Endothelial-specific deletion of Ephb4 in blood vessels in adult mice resulted in a hypertrophic cardiac phenotype that was attributed to a dysfunctional caveolae-mediated recycling of focal adhesion and junctional molecules specifically in coronary capillaries but not in skeletal muscle capillaries ( Luxán et al . , 2019 ) . It has been suggested that organ-specific differences exist in the susceptibility to disruption of vascular integrity due to differences in the magnitude of mechanical stress they constantly have to resist ( Frye et al . , 2015; Hägerling et al . , 2018; Luxán et al . , 2019 ) . It is therefore possible that junctional alterations occur at a later time-point in Efnb2/Ephb4-deficient dermal or pulmonary blood vessels and/or in context of specific pathological conditions . However , it is also interesting to note that LEC-specific deletion of Efnb2 or Ephb4 using the Prox1-CreERT2 mice led to global disruption of junctional integrity of collecting lymphatic vessels in all tissues analysed ( skin , mesentery and lymph node ) , suggesting a general role of this pathway in LEC junction maintenance . It would be of interest to assess the effect of disruption of collecting vessel junctions on the ability of the vessels to take up and transport fluid , but also to study the long-term functional consequences in the context of pathology such as adult-onset obesity or inflammation . ECs are subjected to fluid shear stress ( FSS ) due to the friction between the ECs and the blood or lymph . FSS has been extensively described to regulate endothelial junctions and lymphatic valve morphogenesis Gordon et al . , 2020 . In agreement with previous studies ( Mäkinen et al . , 2005; Zhang et al . , 2015 ) , we found that loss of EphrinB2-EphB4 signalling resulted in lymphatic valve defects . This raised the possibility that altered flow patterns contribute to disruption of EC junctions on the collecting vessel wall . Interestingly , genetic inactivation of the mechanosensitive transcription factor Foxc2 also results in degeneration of lymphatic valves and disruption of LEC junctions ( Sabine et al . , 2015 ) . The junctional phenotype in the Foxc2-deficient vessels is however restricted to the valve sinus , where flow recirculation maintains high FOXC2 expression ( Sabine et al . , 2015 ) . Expression of Efnb2 in all LECs of collecting vessels instead suggested a role beyond valve areas . In agreement with this , LEC-specific deletion of Efnb2 using the Prox1-CreERT2 resulted in junctional disruption in collecting vessels both upstream and downstream of the valve . In further support of a cell-autonomous function of EphrinB2 in all collecting vessel LECs , valve-specific deletion of Efnb2 using the Cldn11-CreERT2 resulted in valve disruption but cell-cell junctions of the lymphatic vessel wall were not affected . Lymphatic capillaries also express Efnb2 ( Bazigou et al . , 2011 ) . However , we were not able to assess the potential cell-autonomous role of EphrinB2 in the formation and maintenance of button junctions in lymphatic capillaries because they were affected secondary to deletion of Efnb2 in collecting vessels . Interestingly , LEC-specific deletion of VE-cadherin was previously shown to result in junctional defects in and fragmentation of lacteal lymphatic capillaries in the intestinal villi secondary to defects in the mesenteric collecting lymphatic vessels ( Hägerling et al . , 2018 ) . These observations suggests that normal function of collecting vessels , ensuring efficient lymph flow , is critical for the maintenance of LEC junctions in lymphatic capillaries . ECs are interconnected via VE-cadherin-mediated adherens junctions that have important roles in the development and maintenance of lymphatic vessels ( Dartsch et al . , 2014; Hägerling et al . , 2018; Yang et al . , 2019 ) . Interestingly , however , organ-specific differences exist in the role of VE-cadherin in vessel maintenance . LEC-specific deletion of Cdh5 ( encoding VE-cadherin ) in adult mice resulted in deterioration of mesenteric collecting vessels , while dermal lymphatic vessels were not affected ( Hägerling et al . , 2018 ) . This is different from the phenotype in both vascular beds observed upon loss of EphrinB2/EphB4 signalling , which suggests the involvement of VE-cadherin-independent mechanisms of junction disruption . The role of other molecular players was further supported by lack of effect of EphrinB2/EphB4 inhibition on VE-cadherin localisation or levels in HDLECs in vitro . Instead , we found that EphrinB2/EphB4 blockade resulted in reduction in the junctional localisation of CLDN5 . Genetic deletion of Cldn5 in LECs in vivo did not however fully recapitulate the phenotype observed in Efnb2/Ephb4-deficient lymphatic vessels . Although the morphology of cell junctions was altered , we could not observe a complete breakdown of the junctions as in the Efnb2/Ephb4 mutants . Although our data suggest a more prominent role of CLDN5 in controlling LEC junction integrity compared to VE-cadherin , cross-talk and compensation between CLDN5 and VE-cadherin is likely to occur . Quiescent endothelia are characterized by a balance of actin stabilization and myosin-based actin pulling forces that are constantly applied to endothelial junctions . Elevation of myosin-based actin contractility activates endothelial junctions , but abnormally increased contractility can ultimately lead to adherens junction disruption and loss of vascular integrity ( Faurobert et al . , 2013 ) . In agreement with previous findings in HUVECs ( Abéngozar et al . , 2012 ) , we found that loss of EphrinB2/EphB4 signalling in primary LECs led to an increase in actin stress fibres that was attributed to increased Rho activity . Although in vivo validation remains to be done , our findings support a concept that continuous basal activation of EphrinB2/EphB4 signalling provides a homeostatic control of Rac1/Rho-mediated cytoskeletal contractility to regulate LEC junction integrity . Interestingly , loss of ( valve-specific ) FOXC2 and FOXC1 in LECs was recently shown to result in increased actin stress fibre formation and Rho activity , suggesting similar mechanisms of junctional regulation in valve LECs ( Norden et al . , 2020 ) . It may therefore not be surprising that ablation of the adhesion molecule CLDN5 was not sufficient to induce a breakdown of LEC junctions in the absence of activation of the cytoskeleton . A similar concept has been suggested in established blood vessels , whereby basal Tie2 receptor signalling mediates stabilization of pulmonary endothelial junctions via the actin cytoskeleton even in the absence of VE-cadherin ( Frye et al . , 2015 ) . In conclusion , we show that continuous basal activation of EphrinB2/EphB4 signalling provides a critical homeostatic mechanism regulating Rac1/Rho-mediated cytoskeletal contractility in LECs and integrity of cell junctions in collecting lymphatic vessels in vivo . Activation of the EphB4 receptor or downstream signalling components may thus provide an opportunity to overcome pathological hyperpermeability and restore basal permeability of collecting lymphatic vessels , thereby presenting a potential strategy for selective modulation of lymphatic vessel function .
Efnb2flox ( Grunwald et al . , 2004 ) , Ephb4flox ( Martin-Almedina et al . , 2016 ) , Pdgfb-iCreERT2iresGFP ( Claxton et al . , 2008 ) , R26-mTmG ( Muzumdar et al . , 2007 ) , Prox1-CreERT2 ( Bazigou et al . , 2011 ) and Efnb2GFP ( Davy and Soriano , 2007 ) mice were described previously . Ephb4 and Efnb2 mutant mice were analysed and showed the same phenotype both in the presence and absence of the R26-mTmG reporter allele . Cldn11-CreERT2 were generated using BAC transgenesis , by inserting a cDNA encoding CreERT2 under the control of the regulatory elements of the lymphatic valve-specific Cldn11 gene ( H Ortsäter and T Mäkinen , manuscript in preparation ) . ES cell line containing ‘knockout-first’ Cldn5 allele ( Cldn5tm1a ( EUCOMM ) Wtsi ) was obtained from The European Conditional Mouse Mutagenesis Program ( EUCOMM ) . After obtaining germ line transmission , LacZ-neo cassette was removed by crossing with a FlpO deleter strain ( Wu et al . , 2009a ) followed by mating to C57BL/6J for at least three generations . For induction of Cre-mediated recombination , 4-hydroxytamoxifen ( 4-OHT , H7904 , Sigma Aldrich ) was administered by intraperitoneal injection at P4 ( 1 × 50 µg , dissolved in Ethanol ) , or at P12 ( 1 × 1 mg ) or P21-22 ( 3 × 1 mg , dissolved in peanut oil ) . For induction of Cre-mediated recombination at 4 and 8 weeks , tamoxifen ( 1–2 mg , T5648 , Sigma Aldrich ) was dissolved in peanut oil and administered by intraperitoneal injections on 3–5 consecutive days . All strains were maintained and analysed on a C57BL/6J background ( backcrossed minimum six times ) . Experimental procedures were approved by the Uppsala Animal Experiment Ethics Board , Sweden ( permit numbers: C416/12 and C130/15 ) and by the Landesamt für Natur , Umwelt und Verbraucherschutz ( LANUV ) , Nordrhein-Westfalen , Germany ( permit number: 8 . 87–50 . 10 . 36 . 09 . 063 ) . Animals were kept in a barrier facility under pathogen–free conditions . Human primary dermal lymphatic endothelial cells ( HDLEC from juvenile foreskin , cat . C12216 ) were obtained from PromoCell and tested negative for mycoplasma contamination . Cells were cultured on bovine Fibronectin-coated dishes in complete ECGMV2 medium ( PromoCell ) at 37° with 5% CO2 . Dharmacon siGENOME siRNA was used to knock-down human CLDN5 ( Cat# D-011409-03-0010 ) and CDH5 ( Cat# D-003641-03-0010 ) . Dharmacon ON-TARGETplus SMARTpool siRNA was used to knock-down human EFNB2 ( Cat# L-003659-00-0005 ) . As a control AllStars Negative Ctrl siRNA was used ( Cat# 1027281 , Qiagen ) . Cells were subjected to transfection 24 hr after plating , using Lipofectamine 2000 ( Invitrogen ) according to the manufacturer’s instructions . To determine paracellular permeability , 4 × 104 HDLECs were seeded on fibronectin–coated Transwell filters ( Costar 3413 , 0 . 4 μm pore size; Corning ) and grown to confluence . For EphrinbB2 blockade , HDLECs were serum-starved ( EGMV2 basal medium + 0 . 5% FBS ) and then incubated with EphrinB2 blocking antibody B11 ( 80 μg/ml ) or isotype ctrl or PBS for 2 hr . To study the impact of Rho inhibition , HDLECs were pre-incubated with the ROCK inhibitor ( Y-27632 ) for 30 min before the EphrinB2 blocking antibody B11 was added . After 2 hr , 0 . 25 mg/ml FITC-dextran ( 40 kDa; Sigma-Aldrich ) was added to the upper chamber of the transwells and transcellular diffusion was allowed for 1 hr . Fluorescence in the lower chamber was measured with a microplate reader ( Synergy HTX Multi-Mode Microplate Reader ) , and monolayer integrity was confirmed by immunofluorescence staining for VE-cadherin after each assay . 30 min prior to activation of EphB4 by its ligand EphrinB2 , recombinant human EphrinB2-Fc ( 0 . 5 μg/ml , 7397-EB , R and D Systems ) was clustered using Goat Anti-Human IgG ( 5 μg/ml , 109-005-098 , Jackson ImmunoResearch ) . For cell lysis and detection of phospho-tyrosine after immunoprecipitation , HDLECs were lysed in lysis buffer containing 20 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 2 mM CaCl2 , 1 mM Na3VO4 , 1% Triton X-100 , 0 . 04% NaN3 , and 1 × complete EDTA-free protease inhibitors ( Roche ) . EphB4 was immunoprecipitated from cell lysates by incubation with 3 μg human EphrinB2-Fc ( 7397-EB , R and D Systems ) and 30 µl Protein G-Sepharose for 3 hr at 4°C . Immunocomplexes were washed five times with lysis buffer and analysed by SDS-PAGE . Mouse lungs were homogenized with an Ultra Turrax ( IKA-Werke ) in RIPA buffer containing 1% NP-40 , 1% sodium deoxycholate , 0 . 01 M NaPi , 150 mM NaCl , 2 mM EDTA , 1 mM Na3VO4 , and 2 × Complete EDTA-free protease inhibitors , followed by incubation for 4 hr at 4°C . Lysates were centrifuged at 4°C for 1 hr at 20 , 000 g and supernatant was used for direct blot analysis . Total cell or organ lysates or immunoprecipitated material was separated by SDS-PAGE and transferred to PVDF membranes ( 88520 , Thermo Fisher Scientific ) by wet blotting . For detection of phosphorylated tyrosine , milk powder in the blocking buffer was replaced by 2% BSA , and 200 μM Na3VO4 was added . The following antibodies were used: Goat anti-EphrinB2 ( AF496 , R and D Systems , 0 . 5 μg/ml ) , goat anti-EphB4 ( AF3038 , R and D Systems , 0 . 5 μg/ml ) , goat anti-VE-cadherin ( sc-6458 , Santa Cruz Biotechnology , 0 . 2 μg/ml ) , rabbit anti-CLDN5 ( 34–1600 , Invitrogen , 0 . 5 μg/ml ) , mouse anti-phosphotyrosine 4G10 ( 05–321 , Merck , 0 . 5 μg/ml ) , rabbit anti-β-actin ( 4967 , Cell Signalling Technologies , 0 . 1 μg/ml ) . Serum-starved ( EGMV2 basal medium + 0 . 5% FBS ) confluent HDLEC cultures remained resting or were stimulated for 30 min with 0 . 5 µg ml−1 clustered recombinant human EphrinB2-Fc ( 7397-EB , R and D Systems ) and subsequently subjected to Rac1-GTP pulldown assay according the manufacturer’s instructions ( BK030 , Cytoskeleton Inc ) . Briefly , equal amounts of protein lysates were incubated for 1 hr at 4°C under gentle agitation with 10 µg PAK-RBD beads . After extensive washes PAK-RBD-bound proteins were denatured by the addition of 2 × Laemmli buffer and incubation at 97°C for 3 min . Samples were immunoblotted onto PVDF membranes ( 88520 , Thermo Fisher Scientific ) , probed with mouse anti-Rac1 antibodies ( ARC03 , Cytoskeleton Inc ) and corresponding donkey anti-mouse-HRP-conjugated secondary antibodies ( 715-035-151 , Jackson ImmunoResearch ) . Membranes were visualized using a ChemiDoc MP imaging system ( Biorad ) . Sorting of blood and lymphatic endothelial cells was done using ear skin of 4-weeks-old Efnb2flox;R26-mTmG;Prox1-CreERT2 mice . Tissue were collected and immediately digested at 37°C in PBS supplemented with Collagenase IV ( Life technologies ) 10 mg/ml , DNase1 ( Roche ) 0 . 1 mg/ml and FBS 0 . 5% ( Life technologies ) for 25 min with vigorous shaking every 5 min . Collagenase activity was quenched by dilution with FACS buffer ( PBS , 0 . 5% FBS , 2 mM EDTA ) and digestion products were filtered twice through 70 µm nylon filters ( BD Biosciences ) . Cells were again washed with FACS buffer and then processed for enrichment of CD31/PECAM1 positive cells using magnetic beads according to the manufactures instructions ( Miltenyi ) . After enrichment , Fc receptor binding was blocked with rat anti-mouse CD16/CD32 , ( eBioscience ) . Samples were thereafter stained with anti-podoplanin ( clone eBio8 . 1 . 1 , conjugation eFluor660 , 2 µg/ml ) , anti-CD31/PECAM1 ( 390 , PE-Cy7 , 0 . 7 µg/ml ) , anti-CD45 ( 30-F11 , eFluor450 , 4 µg/ml ) , anti-CD11b ( M1/70 , eFluor450 , 4 µg/ml ) and anti-TER-119 ( TER-119 , eFluor450 , 4 µg/ml ) all obtained from eBioscience . Prior to sorting cells were incubated with Sytox blue ( Life technologies ) to label dead cells . Cells were sorted directly to tubes containing RNA extraction buffer ( Qiagen ) on a BD FACSAria III cell sorter configured with four lasers ( 405 , 488 , 561 and 643 nm ) and equipped with a 85 µm nozzle . Single cells were gated from FSC-A/SSC-A , FSC-H/FSC-W and SSC-H/SSC-W plots followed by exclusion of all cells with a eFluor450/Sytox blue positive signal . For compensation , the AbC anti-rat/hamster compensation bead kit ( Life Technologies ) was used . ECs ( CD31/PECAM1+ ) were then sorted as either LECs ( PDPN+GFP+ ) or BECs ( PDPN-GFP-Tomato+ ) . Tissue samples for FACS analysis of proliferating cells were taken from wild-type P3 pups ( developing ear buds ) and from P10 , P21 or 5 weeks old mice ( ear skin ) . Tissue samples were dissected , cut into small pieces and digested in Collagenase IV ( Life technologies ) 4–10 mg/ml , DNase1 ( Roche ) 0 . 1–0 . 2 mg/ml and FBS 1–10% ( Life technologies ) in PBS at 37°C for 10 to 30 min ( depending on stage ) with vigorous shaking every 5 min . Samples were processed as described above except that no enrichment for CD31/PECAM1 positive cells was performed . Staining of cell surface markers was performed by incubation with antibodies targeting podoplanin ( eBio8 . 1 . 1 , PE ) , CD31/PECAM1 ( 390 , PE-Cy7 ) , CD45 ( 30-F11 , PerCP-Cyanine5 . 5 ) , CD11b ( M1/70 , PerCP-Cyanine5 . 5 ) all obtained from eBioscience . After staining cells were thoroughly washed with PBS and stained for dead cells using the blue LIVE/DEAD fixable dead cell stain kit from Life technologies . Thereafter cells were fixed and permeabilized using the Foxp3/Transcription factor staining kit ( eBioscience ) according to the manufacturer’s instructions . Finally cells were incubated with rat serum and then a Ki-67 antibody ( SolA15 , eFluor 660 , eBioscience , 1:100 ) . Cells were analysed on a BD LSR Fortessa cell analyzer configured with five lasers ( 355 , 405 , 488 , 561 and 643 nm ) . Compensation was performed using the anti-rat/hamster compensation bead kit and the ArC amine reactive compensation bead kit ( Life technologies ) . Single viable cells were gated as described above and in Martinez-Corral et al . , 2020 and dead cells were excluded in the 355-UV laser dump channel . FMO controls were used to set up the subsequent gating scheme to obtain cell populations and quantification of proliferating cells . Flow data were processed using FlowJo software version 10 . 5 . 0 ( TreeStar ) . Ear skin was dissected and minced in 5 mg/ml Collagenase II ( Roche ) and 0 . 2 mg/ml DNaseI ( Roche ) digestion mix . From FACS-sorted LECs ( CD31+/PDPN+ ) total RNA was extracted by RNeasy Micro kit ( QIAGEN ) and all obtained RNA was reverse transcribed using oligo dT ( SuperScript III First-Strand Synthesis System , Invitrogen ) . cDNA was pre-amplified using the TaqMan PreAmp Master Mix Kit . Gene expression levels were analysed using TaqMan Gene Expression Assay ( AppliedBiosystems ) and the StepOne Plus Real-Time PCR system ( Applied Biosystems ) following manufacturer’s instructions . Relative gene expression levels were normalized to GAPDH . The following probes were used: Gapdh Mm_03302249_g1 , Efnb2 Mm00438670_m1 ( ThermoFisher Scientific ) . A modified Miles assay for the induction of vascular permeability in the skin was performed as described previously ( Frye et al . , 2015 ) . Evans blue dye ( Sigma-Aldrich ) was injected into the tail vein ( 100 μl of a 1% solution in PBS ) of Efnb2 mutant and control mice ( n = 2–3 animals/group from three independent experiments ) . After 15 min , 50 μl PBS , 100 ng murine VEGF165 in 50 μl PBS , or 225 ng histamine in 50 μl PBS was injected intradermally into the shaved back skin . 30 min later , skin areas were excised and extracted with formamide for 5 d at RT , and the concentration of the dye was measured at 620 nm with a spectrophotometer ( Shimadzu ) . Efnb2 mutant and control mice ( n = 3–4 animals/group from two independent experiments ) were intravenously injected with Evans blue dye ( Sigma Aldrich , 100 μl of a 1% solution in PBS ) , 30 min later sacrificed , and the body circulation was perfused with PBS . Weight-adjusted skin patches and lungs were removed and extracted with formamide for 5 d at RT . The concentration of the dye was measured at 620 nm with a spectrophotometer ( Shimadzu ) . Specimens were fixed in 1% glutaraldehyde and 4% formaldehyde in 0 . 1 M phosphate buffer , pH 7 . 4 , followed by post-fixation in 1% osmium tetroxide , dehydration in acetone and embedding in Epon LX112 ( Ladd Research Industries ) . 150 nm sections were stained with toluidine blue to select regions of interest . 80 nm sections were cut with a Leica Ultracut UCT microtome and imaged using Tecnai Spirit transmission electron microscope ( Fei Europe ) and Quemesa CCD camera ( Olympus Soft Imaging Solutions GMBH ) . Measurements were performed from TEM micrographs at a 11000 × magnification using ImageJ . For whole-mount immunostaining , tissues were fixed in 4% paraformaldehyde ( PFA , all expect CLDN5 ) or ice-cold methanol ( CLDN5 immunostaining ) overnight at 4°C , permeabilised in 0 . 3% Triton-X100 in PBS ( PBSTx ) and blocked in PBSTx plus 3% BSA ( blocking buffer ) . Primary antibodies were incubated at 4°C overnight in blocking buffer . After washing in PBSTx , the samples were incubated with Alexa Fluor-conjugated secondary antibodies in blocking buffer , before further washing and mounting in Dako Fluorescence Mounting Medium . For staining of HDLECs , cells were fixed with 4% PFA in PBS for 20 min at RT or with ice-cold methanol ( CLDN5 immunostaining ) for 20 min at 4° and permeabilized using 0 . 5% Triton-X100 in PBS for 5 min at RT followed by blocking with 3% BSA in PBSTx for 1 hr . Primary antibodies were incubated for 1 hr at RT , washed twice with PBSTx and subsequently incubated with secondary antibodies for 45 min at RT before further washing and mounting in Dako Fluorescence Mounting Medium . The following antibodies were used: goat anti-VE-cadherin ( C19 , cat . sc-6458 , Santa Cruz Biotechnology , 2 μg/ml ) , mouse anti-VE-cadherin ( F8 , cat . sc-9989 , Santa Cruz Biotechnology , 2 μg/ml ) , rat anti-LYVE1 ( cat . MAB2125 , R and D Systems , 1 μg/ml ) , rabbit anti-GFP ( ab290 , Abcam , 1 μg/ml ) , rabbit anti-CLDN5 ( 34–1600 , Invitrogen , 0 . 5 μg/ml ) , sheep anti-mouse FoxC2 ( AF6989 , R and D Systems , 2 μg/ml ) . Secondary antibodies conjugated to AF488 , AF647 , Cy2 , Cy3 or Cy5 were obtained from Jackson ImmunoResearch ( all used 1:200 ) . Additionally , Alexa Fluor 568 Phalloidin ( A12380 , ThermoFisher Scientific ) and Alexa Fluor 594 Phalloidin ( A12381 , ThermoFisher Scientific ) were used . Confocal images of tissues ( whole-mount immunostaining ) and cells represent maximum intensity projections of Z-stacks that were acquired using Leica SP8 inverted microscope with HCX PL APO CS 10 ×/0 . 40 DRY or HC PL APO CS2 63 ×/1 . 30 GLYC objectives and Leica LAS-X software . To quantify the state of in vivo LEC junctions ( Figure 3C , E ) , four junctional categories were defined: straight , thick/reticular , jagged and discontinuous junctions . 2–3 images ( 66 µm2 ) per mouse and region ( upstream and downstream of the valve ) were acquired from 2 ( Cldn11-CreERT2 ) - 3 ( Prox1-CreERT2 ) independent experiments and divided each image in 16 patches . Classification of each patch into the four categories was done manually and blinded . Wild type LECs of the inguinal lymph node capsule showed a more irregular junctional morphology ( Figure 2E and Figure 2—figure supplement 3 ) , therefore we defined three junctional categories: linear/thick/reticular , jagged and discontinuous junctions . Six images ( 80 µm2 ) per genotype were acquired from two independent experiments and each image was divided into 25 patches . Classification of each patch into the three categories was done manually . To quantify valve morphology in the Efnb2flox;Cldn11-CreERT2 mice , we analysed the organization and alignment of PROX1high nuclei within the valve region of Cre-targeted ( GFP positive ) P11 Efnb2flox;R26-mTmG;Cldn11-CreERT2 mesenteries . We defined two categories: normal alignment of PROX1high nuclei and disorganized alignment of PROX1high nuclei along valve leaflets as illustrated in Figure 3—figure supplement 1 . Classification of valves was done manually with 10–11 valves per mesentery from three littermate mice . All other quantifications were done using Fiji ImageJ . For quantification of intercellular gaps and junctional overlap from TEM micrographs , 12–16 lymphatic vessels with 6 to 28 junctions per vessel were measured . The total number of junctions was 194 from three heterozygous control mice and 250 junctions from two Efnb2 mutant mice . For quantification of central actin ( Figure 5B , F ) , we applied a threshold for junctional VE-cadherin staining , created a threshold-based mask and pixel intensities of the junctional actin fraction and overall actin pixel intensity was subtracted . 6–11 images ( 250 µm2 , maximum intensity projection images with 12 z-stacks ) were acquired from 2 to 3 independent experiments . For junctional CLDN5 immunostaining pixel intensity measurements , we applied a defined threshold for junctional VE-cadherin staining , created a threshold-based mask and analysed pixel intensities of CLDN5 immunostaining in the junctions . CLDN5 pixel intensities ( integrated density ) were detected from 6 to 7 images ( 250 µm2 , maximum intensity projection images with 12 z-stacks ) from three independent experiments . Detailed number of experiments are indicated in figure legends . Western blot signal quantifications were done using BioRad Image Lab Software . GraphPad Prism was used for graphic representation and statistical analysis of the data . We used 2-tailed unpaired Student’s t-test to compare between two means , assuming equal variance , multiple t-tests ( with correction for multiple comparison using the Holm-Sidak method ) to compare between multiple conditions and One-sample t-test to compare sample mean with a normalized control value = 1 or 100 . Differences were considered statistically significant when p or adjusted p<0 . 05 . | Lymph vessels are thin walled tubes that , similar to blood vessels , carry white blood cells , fluids and waste . Unlike veins and arteries , however , lymph vessels do not carry red blood cells and their main function is to remove excess fluid from tissues . The cells that line vessels in the body are called endothelial cells , and they are tightly linked together by proteins to control what goes into and comes out of the vessels . The chemical , physical and mechanical signals that control the junctions between endothelial cells are often the same in different vessel types , but their effects can vary . The endothelial cells of both blood and lymph vessels have two interacting proteins on their membrane known as EphrinB2 and its receptor , EphB4 . When these two proteins interact , the EphB4 receptor becomes activated , which leads to changes in the junctions that link endothelial cells together . Frye et al . examined the role of EphrinB2 and EphB4 in the lymphatic system of mice . When either EphrinB2 or EphB4 are genetically removed in newborn or adult mice , lymph vessels become disrupted , but no significant effect is observed on blood vessels . The reason for the different responses in blood and lymph vessels is unknown . The results further showed that lymphatic endothelial cells need EphB4 and EphrinB2 to be constantly interacting to maintain the integrity of the lymph vessels . Further examination of human endothelial cells grown in the laboratory revealed that this constant signalling controls the internal protein scaffold that determines a cell’s shape and integrity . Changes in the internal scaffold affect the organization of the junctions that link neighboring lymphatic endothelial cells together . The loss of signalling between EphrinB2 and EphB4 in lymph vessels reflects the increase in vessel leakage seen in response to bacterial infections and in some genetic conditions such as lymphoedema . Finding ways to control the signalling between these two proteins could help treat these conditions by developing drugs that improve endothelial cell integrity in lymph vessels . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"cell",
"biology"
] | 2020 | EphrinB2-EphB4 signalling provides Rho-mediated homeostatic control of lymphatic endothelial cell junction integrity |
Two structure determination methods , based on the molecular dynamics flexible fitting ( MDFF ) paradigm , are presented that resolve sub-5 Å cryo-electron microscopy ( EM ) maps with either single structures or ensembles of such structures . The methods , denoted cascade MDFF and resolution exchange MDFF , sequentially re-refine a search model against a series of maps of progressively higher resolutions , which ends with the original experimental resolution . Application of sequential re-refinement enables MDFF to achieve a radius of convergence of ~25 Å demonstrated with the accurate modeling of β-galactosidase and TRPV1 proteins at 3 . 2 Å and 3 . 4 Å resolution , respectively . The MDFF refinements uniquely offer map-model validation and B-factor determination criteria based on the inherent dynamics of the macromolecules studied , captured by means of local root mean square fluctuations . The MDFF tools described are available to researchers through an easy-to-use and cost-effective cloud computing resource on Amazon Web Services .
Cryo-electron microscopy ( cryo-EM ) has evolved into one of the most effective structure determination tools in modern day structural biology , achieving in recent years resolutions rivalling those of X-ray crystallography or NMR spectroscopy ( Cheng , 2015 ) . Furthermore , cryo-EM based structure determination overcomes two major bottlenecks faced in traditional X-ray crystallography , namely , the arduous task of preparing well-ordered crystals of macromolecules ( Unger , 2002 ) , and the more fundamental problem with capturing these molecules in unphysiological states as a result of crystal contacts ( Neutze et al . , 2015 ) . Consequently , cryo-EM provides a natural way of resolving the structures of large macromolecular complexes . Historically , computational methods were required to bridge the resolution gap between crystallography and cryo-EM to produce atomic-resolution models of biomolecular complexes . Various real-space refinement methods that combine crystallographic structures and cryo-EM densities for structure determination have been developed , including DireX ( Schröder et al . , 2007 ) , Flex-EM ( Topf et al . , 2008 ) , Rosetta ( DiMaio et al . , 2015 ) , FRODA ( Jolley et al . , 2008 ) , Phenix real space refinement ( Afonine et al . , 2013 ) , and Molecular Dynamics Flexible Fitting ( MDFF ) ( Trabuco et al . , 2008 , 2009; McGreevy et al . , 2016 ) . MDFF , in particular , has proven to be an extremely successful refinement method as evidenced by its numerous applications ( Goh et al . , 2015; McGreevy et al . , 2016 ) ranging from the intricate ribosomal machinery ( Villa et al . , 2009; Trabuco et al . , 2011; Frauenfeld et al . , 2011; Wickles et al . , 2014 ) to a host of non-enveloped viruses ( Zhao et al . , 2013 ) . So far this success has been limited to structure determination from typically low-resolution cryo-EM maps in the 7–25 Å range which , indeed , represented the state-of-the-art at the time of MDFF’s inception ( Trabuco et al . , 2008 ) . However , seminal advances in detection hardware and programs over the past three years ( Li et al . , 2013; Milazzo et al . , 2011 ) have enabled now the routine availability of high-resolution ( <5 Å ) EM maps for a range of biological systems including ion channels ( Liao et al . , 2013 ) , enzymes ( Bartesaghi et al . , 2014 , 2015 ) , membrane fusion machinery ( Zhao et al . , 2015 ) , and key functional components of the ribosome ( Fischer et al . , 2015; Brown et al . , 2015 ) . High-resolution maps pose an imminent challenge to the traditional map-guided structure determination methods as the maps now characterize near-atomic scale features , the interpretation of which requires extremely precise structure building and validation protocols ( DiMaio et al . , 2015 ) . For example , conformation of the protein sidechains , which are more flexible than the backbone , are now discernible within the maps and , thus , require precise modeling of the dihedral angles up to Cβ atoms while also respecting the map boundaries ( Barad et al . , 2015 ) . In order to produce atomic models with correct backbone and sidechain geometries , as well as minimal potential energy , structure determination tools must be augmented with chemically accurate force fields and exhaustive search algorithms respecting density constraints . Inspired by crystallographic modeling techniques , where such structure-building requirements have already been addressed for the resolution of 3–5 Å diffraction data ( DiMaio et al . , 2013; Murshudov et al . , 2011; McGreevy et al . , 2014 ) , tools such as Rosetta have introduced Monte Carlo simulation-based segment building and refinement protocols with heuristic force fields ( DiMaio et al . , 2015 ) , to handle high-resolution EM maps . Other notable automated model-building tools that can be used for the refinement of high-resolution EM maps include Buccaneer ( Cowtan , 2006 ) , ARP/wARP ( Langer et al . , 2008 ) , and Moulder ( Topf et al . , 2006 ) . Driven by a vision to extend the capabilities of flexible fitting approaches ( Topf et al . , 2008; Trabuco et al . , 2008; Tama et al . , 2004; Suhre et al . , 2006; Kovacs et al . , 2008; Wu et al . , 2013 ) for addressing high-resolution maps , two new MDFF methods are introduced here . These methods , denoted cascade MDFF ( cMDFF ) and resolution exchange MDFF ( ReMDFF ) , augment the traditional MDFF method ( Trabuco et al . , 2008 , 2009; McGreevy et al . , 2016 ) ( called direct MDFF henceforth ) with enhanced conformational sampling techniques , namely simulated annealing ( Brünger , 1988 ) and replica exchange molecular dynamics ( Sugita and Okamoto , 1999 ) . The central idea behind the techniques introduced is to fit a search model sequentially to a series of maps of progressively higher resolutions , ending with the original experimental resolution; all but the last in the series are computationally blurred lower-resolution derivatives of the original map , so that larger-scale features of the structure are determined first by fitting to the blurred densities , and smaller-scale refinements are performed subsequently during the fitting to higher-resolution densities . Altogether , this treatment enables a richer conformational sampling of the model within the map than direct MDFF , thereby allowing accurate modeling of the global and local structural features from the map; a similar treatment has previously been employed to increase the radius of convergence of MDFF protocols , but with crystallographic data ( Singharoy et al . , 2015 ) . The cMDFF and ReMDFF methods are demonstrated for structure analysis based on 3 . 2-Å and 3 . 4-Å resolution maps of β-galactosidase ( Bartesaghi et al . , 2014 ) and the TRPV1 channel ( Liao et al . , 2013 ) , respectively . The two methods were found to resolve atomic structures with accuracy greater than that of direct MDFF and comparable to that of Rosetta , even with poor choices of search models . The accuracy is evaluated in terms of the quality of fit measured through global and local cross-correlations ( GCC and LCC ) , integrated Fourier shell coefficients ( iFSC ) , and EMRinger scores ( Barad et al . , 2015 ) , as well as in terms of the quality of structural integrity measures like MolProbity ( Chen et al . , 2010 ) . In the second part of the present study we establish that structural flexibility , as measured by root mean square fluctuations or RMSF within the MDFF simulation , provides an ensemble-based indicator of local and overall resolution of a map offering , thus , a quality measure of an EM map based on the inherent dynamics of the imaged macromolecule . In line with this new finding , RMSF values are shown to provide a physical basis for the determination of optimal sharpening B-factors that maximize the signal-to-noise ratio within a map . These B-factors are determined at three different levels of model description: whole-system , per-domain , and per-residue . Finally , use of the ReMDFF method on cloud computing platforms is discussed . Cloud computing is now a highly suitable approach for computational biology and can be employed for large-scale scientific computing , data analysis , and visualization tasks . For example , Amazon Web Services has been previously demonstrated to be a low-cost cloud computing platform for processing cryo-EM data ( Cianfrocco et al . , 2015 ) . We demonstrate now the usage of Amazon Web Services , highlighting the platform’s capability for rapidly fitting structures to EM density with ReMDFF . The web-interface makes it readily possible for experimental groups around the world to deploy MDFF in an easy and economical way , bypassing the need for their own staff , software , and hardware resources .
In direct MDFF , an initial atomic structure is subjected to an MD simulation with an additional potential energy term VEM that is proportional to the sign-inverse of the EM map . Through VEM , steering forces locally guide atoms towards high-density regions , thereby fitting the structure to the map ( see Materials and methods ) . The equilibrium structure obtained in the simulation represents a global minimum in VEM . For maps in the low resolution range ( 6–15 Å ) , this global minimum is broad , accomodating an ensemble of conformations defined by the overall shape of the macromolecule ( Trabuco et al . , 2008 , 2011 ) . In contrast , at the mid-resolution range of 4–6 Å , densities corresponding to the backbones become discernible , and at sub-4 Å resolutions , even sidechains can be resolved . At such high resolutions , VEM now features multiple proximal local minima which correspond to recurring spatial patterns within a macromolecule , such as helices aligned in parallel or strands in a β-sheet . As shown in Figure 1 , the energy barriers separating these local minima are typically twice as high as those in the case of low-resolution maps . The existence of such potential minima in high-resolution maps exposes MDFF to a long-known weakness of traditional MD-based algorithms , namely entrapment of the fitted structure within undesired local minima instead of reaching the global minimum of VEM . Not unexpectedly , therefore , direct MDFF yields structurally poor or functionally irrelevant models with high-resolution EM maps ( Figure 2—figure supplement 1 ) ( DiMaio et al . , 2015 ) . 10 . 7554/eLife . 16105 . 003Figure 1 . Visual summary of advanced MDFF methodology . A graphic table illustrating MDFF refinement of a model of carbon monoxide dehydrogenase using a high-resolution map . The map represents an open conformation while the initial search model was obtained through crystallography of a closed conformation . This search model was independently fitted , using direct MDFF , to individual members of a set of maps obtained by applying Gaussian blurs of various half-widths ( σ , first column ) to the experimental density . These maps are visualized as a 3D surface in the second column , while the resulting MDFF potentials VEM are represented in cross-section in the third column . Notice the increase in number of contiguous density regions as σ increases . This increase in contiguity is manifested in the lowering of high VEM barriers ( red ) for small σ values to low or flat energy profiles ( blue ) for larger σ values , as observed in the VEM potential cross-sections . Reduced barrier heights allow the structure to explore the conformational space freely during fitting . The structure after 500 ps of fitting , shown in red , is superimposed on the known target structure , shown in blue , in the fourth column . The time evolution of RMSD with respect to the target during fitting is shown in the fifth column . The RMSD plots show that direct fitting to lower resolution maps requires fewer time steps to reach convergence . In fact , the structure never becomes less deviated than the initial 7-Å RMSD from the target in the direct MDFF of the highest-resolution map ( i . e . in the absence of Gaussian blurring ) . The inset shows refinements of the same structure through cMDFF and ReMDFF employing the same set of maps . A clear improvement over direct MDFF is apparent , with convergence to within 1 . 7 Å and 1 . 0 Å of the target achieved within 1000 and 100 ps for cMDFF and ReMDFF respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 16105 . 003 To mitigate this weakness in direct MDFF , we introduce cMDFF and ReMDFF . In cMDFF , the structure is fitted , in a series of MDFF simulations , to maps of gradually increasing resolution . First , the experimental map is smoothened by applying a series of Gaussian blurs with increasing half-widths , σ , to obtain a set of theoretical maps with gradually decreasing resolution; σ = 0 Å corresponds to the experimental map , and σ > 0 Å corresponds to a smoothened one ( see Materials and methods ) . Illustrated in Figure 1 , the density-dependent potential derived from the smoothest map ( i . e . the one with the largest σ value ) features a clear global minimum representing the large-scale structural features of the protein . Second , a search model is fitted to this map , allowing resolution of these large-scale features . Third , the resulting structure is employed as the search model for fitting to the next higher-resolution map in the series . These fitting and search model refreshment steps are repeated through the series of maps in order of decreasing σ , until the structure is finally fitted to the experimental map ( see Video 1 for TRPV1 refinement . An additional virtual reality version can be found at https://www . youtube . com/watch ? v=UwwVC6C9tw0 ) . 10 . 7554/eLife . 16105 . 004Video 1 . cMDFF Refinement of TRPV1 . DOI: http://dx . doi . org/10 . 7554/eLife . 16105 . 004 The gradual increase in map resolution over the course of the simulations allows the structure to explore a greater conformational space than in direct MDFF . The structure thus avoids entrapment within local minima of the density-dependent potential and is accurately fitted to the near-atomic density features of the experimental map while also resolving the larger scale features . In ReMDFF , the cascade scheme is infused with a greater degree of automation . Multiple MDFF simulation replica are run in parallel with each replica fitting a model to a map of a specific resolution . Based on a Metropolis formula analogous to that of conventional replica exchange molecular dynamics simulations ( Sugita and Okamoto , 1999 ) , but now derived in terms of density and not temperature ( see Materials and methods ) , the models are exchanged at regular time intervals between maps from neighboring pairs of replica . Stepwise improvements in fit occur during exchanges between a poorly fitted model at a high resolution with a well-fitted model at a lower resolution . This well-fitted model is further refined against the high-resolution map until convergence is reached , and exchange between the chosen resolutions ceases . Further details are described in Materials and methods . ReMDFF has advantages over cMDFF both in terms of efficiency and automation as it can take advantage of modern parallel computing hardware and the powerful and adaptive replica exchange interface of NAMD ( Jiang et al . , 2014 ) . Nonetheless , as presented in the following , both cMDFF and ReMDFF outperform direct MDFF in quality and speed across a range of high-resolution examples . In an initial proof-of-principle computation , cMDFF and ReMDFF were applied to fit a structure of carbon monoxide dehydrogenase to a 3-Å synthetic density map . The same techniques were subsequently applied to obtain refined structures of two more protein systems , namely TRPV1 ( Liao et al . , 2013 ) and β-galactosidase ( Bartesaghi et al . , 2014 ) , for which experimental densities of 3 . 4 Å and 3 . 2 Å resolution respectively are available . In each case , a direct MDFF simulation was also performed for the purpose of comparison . The MDFF-derived structures were then subjected to a model validation analysis , to evaluate the quality of the models with established protocols in the cryo-EM field . Additional examples chosen for this analysis include γ-secretase at 4 . 5 Å ( Lu et al . , 2014 ) and 3 . 4-Å ( Bai et al . , 2015 ) , β-galactosidase at 2 . 2-Å ( Bartesaghi et al . , 2015 ) and the proteasome at 3 . 3 Å ( Li et al . , 2013 ) resolution . Comparisons between the direct and advanced MDFF protocols , and wherever possible , with other available fitting techniques , such as Rosetta ( DiMaio et al . , 2015 ) , elucidate the general pros and cons of the flexible fitting strategy . An EM density map represents a thermodynamic ensemble of atomic conformations ( Schröder et al . , 2007; Brunger et al . , 2012; Schröder et al . , 2010 ) . Conventionally , however , only a single model representing a best fit to the map is reported . One may ask how statistically representative a single model can be . To quantify the deviation of a fitted model from the rest of an ensemble of simulated molecules , root mean square fluctuation ( RMSF ) of the model relative to the ensemble-averaged structure was computed during an MDFF refinement simulation employing the protocol described in Methods . In the following , the RMSF of a fitted model is first shown to be indicative of the quality of fit of the model , as well as to represent the degree of natural conformational variation exhibited within the thermodynamic ensemble underlying the map . Second , the RMSF values are found to correlate both locally and globally with the resolution of an EM map , providing an interpretation of map quality based on the inherent ( i . e . , natural ) dynamics of the macromolecule under observation . Finally , RMSF values are also employed to identify optimal B-factor values for the sharpening of a map . Altogether , the results of the present study demonstrate that RMSF of a fitted model during an MDFF refinement provides valuable information on the model . ReMDFF involves many independent , though sporadically communicating , MDFF simulations that can be run well on a parallel computer as exploited by the NAMD software for the case of replica exchange simulations ( Jiang et al . , 2014 ) . As a result , ReMDFF provides an efficient and automated method which can converge on a final fitted structure more quickly than direct or cMDFF ( Figure 2—figure supplement 3 ) . However , a potential bottleneck may exist with respect to the computing hardware that a researcher has access to . Cloud computing offers a potential solution , allowing a researcher to focus on the scientific challenges of their project without having to worry about local availability and administration of suitable computer hardware . To prove the feasibility of performing ReMDFF simulations on a cloud platform , we performed ReMDFF for a test system , carbon monoxide dehydrogenase ( PDB 1OAO:chain C ) , on the Amazon Web Services ( AWS ) Elastic Compute Cloud ( EC2 ) platform . The test system converges to the known target structure in approximately 0 . 1 ns simulation time . The time to convergence requires very little wall clock time , and , therefore , incurs a small monetary cost to a user ( Table 3 ) . However , it should be noted that human error and varying experience level can easily add to the incurred cost of cloud usage . Some systems may require multiple simulations to achieve a high quality structure and , therefore , additional time beyond the example discussed here . Furthermore , preparing a structure for simulation may require additional time and resources over the purely simulation-oriented results presented here . The files and information necessary to run ReMDFF on the test system using EC2 cloud computing resources are available at ( http://www . ks . uiuc . edu/Research/cloud/ ) . The Implementation section of the Methods contains further details for setting up and running ReMDFF simulations . 10 . 7554/eLife . 16105 . 027Table 3 . Performance and cost results for ReMDFF of carbon monoxide dehydrogenase on Amazon Web Services ( AWS ) Elastic Compute Cloud ( EC2 ) platform . Costs are incurred on a per-hour basis , with a 1 hr minimum . DOI: http://dx . doi . org/10 . 7554/eLife . 16105 . 027Instance typeCPUPerformance ( ns/day ) Time ( hours ) Simulation cost ( $ ) c3 . 8xlarge305 . 880 . 411 . 68c3 . 4xlarge123 . 330 . 720 . 84c3 . 2xlarge61 . 351 . 780 . 84
Flexible fitting methods have facilitated structure determination from low-resolution EM maps for more than a decade ( Tama et al . , 2004; Suhre et al . , 2006; Velazquez-Muriel et al . , 2006; Orzechowski and Tama , 2008; Topf et al . , 2008; Kovacs et al . , 2008; Lopéz-Blanco and Chacón , 2013; Wu et al . , 2013 ) and continue to be the methods of choice for resolving molecular systems with atomic resolution . MDFF , in particular , has been a front-runner among methods that have facilitated the discovery of some of the most complicated structures in modern day structural biology ( Hsin et al . , 2009; Sener et al . , 2009; Gumbart et al . , 2011; Frauenfeld et al . , 2011; Zhao et al . , 2013; Wickles et al . , 2014 ) . cMDFF and ReMDFF , the new variants of MDFF introduced above , offer now accurate fitting of atomic-level structures within sub-5 Å EM maps , a feat thus far inaccessible to direct MDFF . These new methods extend the radius of convergence of MDFF to at least 25 Å , fitting models to maps of resolutions as high as 3 . 2 Å . This radius of convergence is at least twice that reported for Rosetta refinements of the 20S proteasome ( DiMaio et al . , 2015 ) . Such a broad radius of convergence will allow the refinement of extremely poorly guessed initial models with MDFF , as demonstrated in the cases of β-galactosidase and TRPV1 reported here . ReMDFF simulations involving the so-called replica-exchange molecular dynamics converge quickly using a small number of replicas and are thus amenable to cloud computing applications . Running ReMDFF on the cloud lowers greatly the barrier to usage of the method , providing a cost-effective and practical solution to fitting structures to high-resolution cryo-EM densities for researchers who neither own nor can administer their own advanced computer hardware . The accuracy of structures refined by cMDFF and ReMDFF has been confirmed by standard error analysis protocols , both in terms of quality of fit as well as in terms of the quality of the model . The results clearly show that sidechain refinements through MDFF produce accurate placement of Cα and Cβ atoms and modeling of the associated dihedrals . Beyond the standard error analysis protocols , which apply to single static structures , the quality of both the fit and the model can be evaluated by ensemble-based measures . An example demonstrated in this study pertains to the use of RMSF during MDFF refinement for simultaneously evaluating the quality of model , map , and fit . Furthermore , utilizing the fact that the inherent flexibility of a macromolecule is a key determinant of the achievable resolution of the corresponding map , RMSF values have been employed to identify the optimal amount of sharpening for a given map offering the highest contrast , as established typically through Guinier analysis . The RMSF computations in MDFF provide a viable means of determining per-residue B-factors . Altogether , interpretation of a map as being representative of an ensemble rather than a single model brings to light new ways of model validation . EM maps , including those at high-resolution , typically do not have a uniform local resolution ( Kucukelbir et al . , 2014 ) and contain low-resolution regions , such as flexible exterior or transmembrane segments ( Leschziner and Nogales , 2007 ) . The sub-5 Å EM maps have also been able to resolve proteins in multiple conformations ( Matthies et al . , 2016 ) . Both these type of maps will continue to benefit from accurate ensemble-based flexible fitting techniques for the foreseeable future . MDFF provides a natural method to model flexible regions; de novo models constructed for one conformational state can be flexibley fitted into the density of the other state ( s ) , thus avoiding the arduous task of model construction for all the conformational states for capturing a conformational transition process with cryo-EM .
MDFF requires , as input data , an initial structure and a cryo-EM density map . A potential map is generated from the density and subsequently used to bias a MD simulation of the initial structure . The structure is subject to the EM-derived potential while simultaneously undergoing structural dynamics as described by the MD force field . Let the Coulomb potential associated with the EM map be Φ ( 𝐫 ) . Then the MDFF potential map is given by ( 1 ) VEM ( r ) ={ζ ( Φ ( r ) −ΦthrΦmax−Φthr ) ifΦ ( r ) ≥Φthr , ζifΦ ( r ) <Φthr . where ζ is a scaling factor that controls the strength of the coupling of atoms to the MDFF potential , Φthr is a threshold for disregarding noise , and Φmax=max ( Φ ( r ) ) . The potential energy contribution from the MDFF forces is then ( 2 ) UEM=∑iwiVEM ( ri ) , where i labels the atoms in the structure and wi is an atom-dependent weight , usually the atomic mass . During the simulation , the total potential acting on the system is given by ( 3 ) Utotal=UMD+UEM+USS where UMD is the MD potential energy as provided by MD force fields ( e . g . CHARMM ) and USS is a secondary structure restraint potential that prevents warping of the secondary structure by the potentially strong forces due to UEM . A detailed description of the potentials arising in Equation 3 is given in Trabuco et al ( Trabuco et al . , 2008 , 2009 ) . After the MDFF and restraint potentials are created through the MDFF plugin of VMD ( Humphrey et al . , 1996 ) , the initial structure is rigid-body docked ( e . g . with Situs [Wriggers , 2010] ) into the density map . Prior to simulation , MDFF-specific parameters can be modified and include ζ and the subset of atoms to be coupled to the MDFF potential . The latter typically consists of all non-hydrogen atoms or backbone atoms and ζ is usually set to 0 . 3 . MDFF can be performed in various simulated conditions , including different temperatures and vacuum , membrane , and explicit or implicit ( Tanner et al . , 2011 ) solvent environments . The choice of parameters and conditions depends on the requirements of each specific case . For example , a highly polar molecule would be more accurately simulated in explicit solvent rather than in vacuum , but the computational cost would be much higher in this case . The MDFF simulation is run until the system has reached stationarity , as determined by RMSD; typical run times are nanoseconds . In cascade MDFF ( cMDFF ) , the initial structure is sequentially fitted to a series of potential maps of successively higher resolution , with the final potential map being the original one derived from the EM map . Starting with i=1 , the ith map in the series is obtained by applying a Gaussian blur of width σi≥0Å to the original potential map , such that σi decreases as the structure is fitted in the sequence i=1 , 2 , … , L , where L is the total number of maps in the series , so that σL≥0 \AA . One can intuitively understand cMDFF as fitting the simulated structure to an initially large and ergodic conformational space that is shrinking over the course of the simulation towards the highly corrugated space described by the original MDFF potential map . Figure 1 provides a visual representation of cMDFF . For a mathematical illustration , suppose that the original potential map can be written as a sum of Gaussians ( 4 ) VEM ( r ) =∑ncnG ( r;rn′ , sn ) , where cn is a weight , G ( 𝐫;𝐫n′ , sn ) is a Gaussian function centered at 𝐫n′ with half-width sn and evaluated at 𝐫 . The result of a Gaussian blur of half-width σi on VEM is ( see Appendix 1 - Section 1 for details ) ( 5 ) Vσi ( 𝐫 ) =∑ncnG ( 𝐫;𝐫n′ , sn2+σ2 ) . Hence , the half-width σi allows one to tune the characteristic width of the potential map through the half-widths of its component Gaussians sn2+σi2 . The initial fitting starts with a large σ1 , corresponding to a diffuse potential which allows much structural mobility , and proceeds along decreasing values of σi , corresponding to narrower potentials with steeper gradients , so that the structure is gradually settled into the original potential map , characterized by σL≥0 \AA . In practice , the series of cMDFF maps is generated from the original potential map using VMD’s volutil Gaussian blur tool . Optimal values for the first half-width σ1 and the change in σi from one map to the next are case-dependent . Values used in the present study were obtained through trial and error . In general , structures far from the ideal conformation benefit from a large σ1 ( >5 Å ) so as to explore a large conformation space . On the other hand , if the original map has a high resolution , small changes in σi ( <1 Å ) would allow a gradual convergence required to avoid being trapped in local potential minima . In our simulations , the change in σi is initially 1 Å . A concrete example is σ1=5 Å , σ2=4 Å , σ3=3 Å , σ4=2 Å , σ5=1 Å , σ6=0 Å . A second trial using changes of 0 . 5 Å was performed ( σ1=5 Å , σ2=4 . 5 Å , σ3=4 Å , … , σ10=0 . 5 Å , σ11=0 Å ) , and if the resulting structure of the second trial presented a better fit , then the first trial was discarded . Replica Exchange MD ( ReMD ) is an advanced sampling method that explores conformational phase space in search of conformational intermediates , which are separated by energy barriers too high to be overcome readily by fixed temperature simulations . Instead of working with a single , fixed MD simulation , ReMD carries out many simulations in parallel , but at different temperatures T1 < T2 < T3 < … where the lowest temperature T1 is the temperature of actual interest , typically , room temperature . The simulations of several copies of the system , the so-called replicas , run mainly independently , such that ReMD can be easily parallellized on a computer , but at regular time points the instantaneous conformations of replicas of neighboring temperatures are compared in terms of energy and transitions between replicas are permitted according to the so-called Metropolis criterion ( Sugita and Okamoto , 1999 ) . This way the highest temperature replicas overcome the energy barriers between conformational intermediates and through the Metropolis criterion moves the T1 replica benefits from it such that transitions between intermediates occur frequently . The application of the Metropolis criterion in the protocol guarantees that the conformations of the T1 replica are Boltzmann-distributed . ReMDFF extends the concept of ReMD to MDFF by simply differentiating replicas not by temperatures T1 < T2 < T3 < … , but by the half-width parameters σ1 > σ2 > σ3 > … . Numerical experiments showed that ReMDFF works extremely well as documented in the present study . As NAMD can parallelize ReMD well ( Jiang et al . , 2014 ) , it can do the same for ReMDFF , such that the enhanced sampling achieved translates into extremely fast MDFF convergence . At certain time instances replicas i and j , of coordinates 𝐱i and 𝐱j and fitting maps of blur widths σi and σj , are compared energetically and exchanged with Metropolis acceptance probability ( 6 ) p ( xi , σi , xj , σj ) =min ( 1 , exp ( −E ( xi , σj ) −E ( xj , σi ) +E ( xi , σi ) +E ( xj , σj ) kBT ) ) , where kB is the Boltzmann constant , E ( 𝐱 , σ ) is the instantaneous total energy of the configuration 𝐱 within a fitting potential map of blur width σ . Computational protocols for performing direct MDFF , cMDFF , and ReMDFF refinements of the two test systems , β-galactosidase and TRPV1 , are now outlined . The local resolutions of a density map can be computed with ResMap ( Kucukelbir et al . , 2014 ) and used within VMD to select the atoms of a structure that are contained in a range of resolutions found by the ResMap analysis . First , the local resolution map output by ResMap is loaded into VMD and then the interpvol keyword can be used to automatically select the atoms found inside the volume values specified , using interpolation . The average RMSF of each selection can then be calculated for a structure during the steps of a MDFF simulation after the initial fitting has occured and the structure has stabilized . In principle any criteria for atom selection can be used for RMSF analysis , though we use local resolution of the EM density here to illustrate the correlation between the two measurements . Additionally , we compute a global average RMSF of the entire structure . The ensemble-based nature of the RMSF analysis means that the quality metric is not dependent on a single structure , but instead a large family of structures can be employed as a better representative of the data . Ensemble-based analyses are a natural and powerful benefit of the MD-based nature of MDFF . RMSF analysis does not , however , require MDFF to be used as the method of refinement . In principle , any refinement method can be used to obtain the fitted model . A subsequent short MDFF simulation of the fitted model can then be performed to obtain the data necessary for the RMSF analysis . Incorporating advanced simulation techniques , such as multi-copy algorithms ( Jiang et al . , 2014 ) , into the MDFF protocol creates a more efficient and accurate computational strategy in cMDFF and ReMDFF . However , these advanced simulation techniques come with an added complexity in the setup and execution of the methods . The current implementation of these methods in NAMD ( Phillips et al . , 2005 ) and VMD ( Humphrey et al . , 1996 ) aim to automate the steps previously discussed . The MDFF Graphical User Interface ( GUI ) ( McGreevy et al . , 2016 ) can be used to set up cMDFF and ReMDFF simulations and provides default parameters , including the number and extent of smoothed maps used for the fitting , with which to run . The parameters for the smoothed maps and number of steps used per map are set heuristically based on previous experience and represent an adequate initial starting point . The GUI automatically generates each of the smoothed maps and converts them to potentials for use in the ReMDFF simulation . All parameters can be tuned by a user to adapt the protocols to their specific system and preference . The GUI produces a series of NAMD configuration files and scripts used for running the simulation , as well sorting and visualizing the results in VMD . Instruction on the use of MDFF , including the GUI , is given in the tutorial found at http://www . ks . uiuc . edu/Training/Tutorials/science/mdff/tutorial_mdff-html/ . Future development will allow for the automatic generation of the smoothed maps in NAMD at runtime . NAMD will also analyse the dynamics of the system to determine when the simulation has converged and move on to fitting to the next density map in the sequence in case of cMDFF calculations . Furthermore , advanced visualization and analysis techniques in VMD ( e . g . new graphical representations ) will be critical for properly understanding the RMSF analyses and to provide greater insight when examining the quality of a model . The use of advanced simulation techniques also comes with an added cost of computational requirements . Adapting the algorithms to best utilize available computational hardware is key when developing efficient methods . Fortunately , the cMDFF and ReMDFF methods and associated analysis algorithms are well suited to highly efficient software implementations on contemporary multi-core CPUs and graphics processing unit ( GPU ) accelerators . We observe that by storing the complete cascade resolution series in efficient multi-resolution data structures such as mip maps ( Williams , 1983 ) , the MDFF cascade algorithm can access a continuously variable resolution representation of the original cryo-EM density map , while making efficient use of CPU and GPU processors and memory systems ( Stone et al . , 2007 ) . The parallel nature of ReMDFF presents an opportunity for efficient , automated sampling of maps of varying resolution . However , to achieve the best efficiency , the simulations should be performed on multi-core CPUs with relatively high core counts ( i . e . at least 1 core per replica ) . Access to such multi-core computers could prohibit use of ReMDFF , however access to machines with the necessary hardware is easily achieved through cloud computing . The cloud computing model provides researchers with access to powerful computational equipment that would otherwise be too costly to procure , maintain , and administer on their own . A particular obstacle is that structural modeling often involves the use of different software suites , such as VMD ( Humphrey et al . , 1996 ) for simulation preparation and Situs ( Wriggers , 2010 ) for initial rigid-body docking or VMD , NAMD ( Phillips et al . , 2005 ) , and Rosetta ( Leaver-Fay et al . , 2011 ) for iterative refinement of models ( Lindert and McCammon , 2015 ) . Cloud platforms can easily bundle different software packages used in a modeling workflow to guarantee their availablity and interoperability on a standardized system . Through the cloud version of our MDFF program suite a user does not need to be aware of any of the above mentioned technical issues . To prove the feasibility of performing ReMDFF simulations on a cloud platform , we ran ReMDFF for the test system , carbon monoxide dehydrogenase ( PDB 1OAO:chain C ) , on the Amazon Web Services ( AWS ) Elastic Compute Cloud ( EC2 ) platform . For the purposes of testing ReMDFF on EC2 , we ran benchmarks on a variety of compute-optimized EC2 instance types . The details of the instance types can be found in Table 3 . We used the same 6 smoothed density maps as the previously discussed cMDFF and ReMDFF simulations in the Proof of Principle and , therefore , also 6 replicas . The files and information necessary to run ReMDFF on the test system using EC2 cloud computing resources are available at ( http://www . ks . uiuc . edu/Research/cloud/ ) . | To understand the roles that proteins and other large molecules play inside cells , it is important to determine their structures . One of the techniques that researchers can use to do this is called cryo-electron microscopy ( cryo-EM ) , which rapidly freezes molecules to fix them in position before imaging them in fine detail . The cryo-EM images are like maps that show the approximate position of atoms . These images must then be processed in order to build a three-dimensional model of the protein that shows how its atoms are arranged relative to each other . One computational approach called Molecular Dynamics Flexible Fitting ( MDFF ) works by flexibly fitting possible atomic structures into cryo-EM maps . Although this approach works well with relatively undetailed ( or ‘low resolution’ ) cryo-EM images , it struggles to handle the high-resolution cryo-EM maps now being generated . Singharoy , Teo , McGreevy et al . have now developed two MDFF methods – called cascade MDFF and resolution exchange MDFF – that help to resolve atomic models of biological molecules from cryo-EM images . Each method can refine poorly guessed models into ones that are consistent with the high-resolution experimental images . The refinement is achieved by interpreting a range of images that starts with a ‘fuzzy’ image . The contrast of the image is then progressively improved until an image is produced that has a resolution that is good enough to almost distinguish individual atoms . The method works because each cryo-EM image shows not just one , but a collection of atomic structures that the molecule can take on , with the fuzzier parts of the image representing the more flexible parts of the molecule . By taking into account this flexibility , the large-scale features of the protein structure can be determined first from the fuzzier images , and increasing the contrast of the images allows smaller-scale refinements to be made to the structure . The MDFF tools have been designed to be easy to use and are available to researchers at low cost through cloud computing platforms . They can now be used to unravel the structure of many different proteins and protein complexes including those involved in photosynthesis , respiration and protein synthesis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources",
"computational",
"and",
"systems",
"biology"
] | 2016 | Molecular dynamics-based refinement and validation for sub-5 Å cryo-electron microscopy maps |
Ca2+ influx through Orai1 channels is crucial for several T cell functions , but a role in regulating basal cellular motility has not been described . Here , we show that inhibition of Orai1 channel activity increases average cell velocities by reducing the frequency of pauses in human T cells migrating through confined spaces , even in the absence of extrinsic cell contacts or antigen recognition . Utilizing a novel ratiometric genetically encoded cytosolic Ca2+ indicator , Salsa6f , which permits real-time monitoring of cytosolic Ca2+ along with cell motility , we show that spontaneous pauses during T cell motility in vitro and in vivo coincide with episodes of cytosolic Ca2+ signaling . Furthermore , lymph node T cells exhibited two types of spontaneous Ca2+ transients: short-duration ‘sparkles’ and longer duration global signals . Our results demonstrate that spontaneous and self-peptide MHC-dependent activation of Orai1 ensures random walk behavior in T cells to optimize immune surveillance .
To initiate the adaptive immune response , T cells must make direct contact with antigen-presenting cells ( APCs ) in the lymph node , enabling T cell receptors ( TCRs ) to engage peptide-bound MHC molecules presented on the APC surface . Because cognate antigens are rare for any given TCR , many APCs must be scanned to identify those bearing cognate antigens . Thus , optimizing T cell motility to balance search sensitivity , specificity , and speed is crucial for efficient antigen search and proper immune function ( Cahalan and Parker , 2005; Krummel et al . , 2016 ) . Both cell-intrinsic and environmental factors have been proposed to regulate T cell motility within lymph nodes and peripheral tissues ( Miller et al . , 2002; Bousso and Robey , 2003; Mempel et al . , 2004; Mrass et al . , 2010 ) . T cell motility in steady-state lymph nodes under homeostatic conditions , referred to as ‘basal motility’ , has been likened to diffusive Brownian motion , resembling a ‘stop-and-go’ random walk that results in an overall exploratory spread characterized by a linear mean-squared displacement over time ( Miller et al . , 2002 ) . Subsequent studies defined a role of cellular cues in guiding T cell migration , such as contact with the lymph node stromal cell network or short-term encounters with resident dendritic cells ( Miller et al . , 2004; Bajénoff et al . , 2006; Khan et al . , 2011 ) . Whereas the basic signaling mechanisms for cell-intrinsic induction of random motility have been previously explored in fibroblasts and neuroblastoma cells ( Petrie et al . , 2009 ) , it remains unclear if such mechanisms apply in T cells . Upon T cell recognition of cognate antigen , TCR engagement results in an elevated cytosolic Ca2+ concentration that acts as a ‘STOP’ signal to halt motility and anchor the T cell to the site of antigen presentation ( Donnadieu et al . , 1994; Negulescu et al . , 1996; Dustin et al . , 1997; Bhakta et al . , 2005; Moreau et al . , 2015 ) . The predominant mechanism for increasing cytosolic Ca2+ in T cells is through store-operated Ca2+ entry ( SOCE ) , which is mediated by the molecular components STIM1 and Orai1 . TCR stimulation triggers depletion of intracellular Ca2+ stores in the endoplasmic reticulum ( ER ) , resulting in translocation of the ER-resident Ca2+ sensor STIM1 to specialized ER-plasma membrane ( PM ) junctions where Orai1 channels aggregate into puncta and activate to allow sustained Ca2+ influx ( Liou et al . , 2005; Roos et al . , 2005; Zhang et al . , 2005; Luik et al . , 2006; Vig et al . , 2006; Zhang et al . , 2006; Calloway et al . , 2009; Wu et al . , 2014 ) . Orai1 channel activity is crucial for immune function , as human mutations in Orai1 result in severe combined immunodeficiency ( SCID ) ( Feske et al . , 2006 ) . Additional roles of Orai1 have been defined in chemotaxis to certain chemokines and T cell homing to lymph nodes ( Greenberg et al . , 2013 ) ; actin cytoskeleton rearrangement ( Schaff et al . , 2010; Dixit et al . , 2011; Babich and Burkhardt , 2013; Hartzell et al . , 2016 ) ; migration during shear flow ( Schaff et al . , 2010; Dixit et al . , 2011 ) ; lipid metabolism ( Maus et al . , 2017 ) ; and dendritic spine maturation in neurons ( Korkotian et al . , 2017 ) . However , despite their contributions to other aspects of T cell function , no role has been identified for Orai1 channels in T-cell motility patterns underlying scanning behavior . In this study , we use human and mouse T cells to assess the role of Orai1 and Ca2+ ions in regulating basal cell motility . Expression of a dominant-negative Orai1-E106A construct was used to block Orai1 channel activity in human T cells , both in vivo within immunodeficient mouse lymph nodes ( Greenberg et al . , 2013 ) , and in vitro within microfabricated polydimethylsiloxane ( PDMS ) chambers ( Jacobelli , Friedman et al . 2010 ) . We use our genetically encoded Salsa6f tandem green/red fluorescent Ca2+ indicator ( Dong et al . , 2017 ) to monitor spontaneous Ca2+ signaling in human T cells migrating in confined microchannels in vitro . Finally , using a transgenic mouse strain expressing Salsa6f homozygously in Cd4+ T cells , designated as Cd4-Salsa6f ( Hom ) mice from here on , we show that Ca2+ signals occur in the absence of specific antigen as T cells crawl in the lymph node . Our results indicate that Ca2+ influx , activated intermittently through Orai1 channels , triggers spontaneous pauses during T-cell motility and fine-tunes the random-walk search for cognate antigens .
To study the role of Orai1 channel activity in T cell motility , we transfected human T cells with the dominant-negative mutant Orai1-E106A to selectively eliminate ion conduction through the Orai1 pore . The glutamate residue at position 106 in human Orai1 forms the selectivity filter of the Orai1 pore ( Prakriya et al . , 2006; Vig et al . , 2006; Yeromin et al . , 2006 ) , and because the Orai1 channel is a functional hexamer ( Hou et al . , 2012 ) , mutation of E106 to neutrally charged alanine completely inhibits Ca2+ permeation in a potent dominant-negative manner ( Greenberg et al . , 2013 ) . Using Fura-2 based Ca2+ imaging , we confirmed Orai1 channel block by E106A in activated human T cells transfected with either eGFP-tagged Orai1-E106A or empty vector for control . Thapsigargin-induced SOCE was greatly diminished in cells expressing eGFP-Orai1-E106A , referred to here as eGFP-E106hi T cells , compared to empty vector-transfected control cells ( Figure 1A ) . Ca2+ entry was also partially inhibited in a population of transfected T cells with minimal eGFP fluorescence referred to as eGFP-E106Alo cells . To confirm that eGFP-E106A inhibits T cell activation , we challenged transfected human T cells with autologous dendritic cells pulsed with the superantigen Staphylococcal enterotoxin B ( Lioudyno et al . , 2008 ) . T cell proliferation was markedly suppressed in eGFP-E106Ahi CD4+ and CD8+ T cells , but not in eGFP-E106Alo T cells ( Figure 1B ) . This shows that the residual Orai1 channel activity in eGFP-E106Alo T cells is sufficient for T cell activation and proliferation . Taken together , these experiments show that eGFP-tagged Orai1-E106A expression can serve as a robust tool to assess cellular roles of Orai1 channel activity and that transfected cells without detectable eGFP fluorescence can be used as an internal control . Orai1 function in human T cell motility was evaluated in vivo using a human xenograft model in which immunodeficient NOD . SCID . β2 mice were reconstituted with human peripheral blood lymphocytes , followed by imaging of excised lymph nodes using two-photon microscopy ( Greenberg et al . , 2013 ) . Reconstitution has been shown to produce a high density of human immune cells within the lymph nodes of immunodeficient mice ( Mosier et al . , 1988 ) , simulating the crowded migratory environment experienced by T cells under normal physiological conditions . Three weeks after reconstitution , human T cells were purified from the same donor , transfected , and adoptively transferred into the reconstituted NOD . SCID . β2 mice ( Figure 1—figure supplement 1 ) . Whereas control T cells transfected with eGFP showed robust expression and successfully homed to lymph nodes following adoptive transfer 24 hr post-transfection , eGFP-E106A transfected T cells did not home to lymph nodes in the same period , consistent with our previous study indicating that functional Orai1 channel activity is required for T cell homing to lymph nodes ( Greenberg et al . , 2013 ) . To circumvent the homing defect , we injected eGFP-E106A transfected T cells only 3 hr post-transfection , before the expression level of eGFP-E106A had become sufficiently high to block lymph node entry ( Figure 1C ) . To evaluate Orai1 function in T cell motility , we imaged human T cells within intact lymph nodes of reconstituted NOD . SCID . β2 mice by two-photon microscopy ( Figure 2A ) . We found that eGFP-E106Ahi T cells migrated with significantly higher average velocities than co-transferred , mock-transfected CMTMR-labeled T cells ( Figure 2B ) . Although both populations had similar maximum and minimum instantaneous cell velocities ( Figure 2C ) , eGFP-E106Ahi T cells traversed longer distances compared to CMTMR controls ( Figure 2D ) , and directionality ratios , a measure of track straightness , decayed more slowly ( Figure 2E ) indicating straighter paths when Orai1 channels were blocked . Orai1-blocked cells displayed shallower turn angles than controls ( Figure 2F ) . Furthermore , arrest coefficients , defined by the fraction of time that cell velocity was <2 µm/min , was six-fold lower for eGFP-E106Ahi T cells than for control T cells ( Figure 2G ) . These differences in motility suggest that the increase in average cell velocity caused by Orai1 block is not due to eGFP-E106Ahi T cells moving faster than control T cells , but rather due to a reduced frequency of pausing . Consistent with this interpretation , no eGFP-E106Ahi T cells with average velocities <7 µm/min were observed , unlike control T cells in which 23% of average velocities were <7 µm/min ( Figure 2H ) . To replicate our findings in a different immunodeficient mouse model , we repeated our human T cell adoptive transfer protocol using NOD . SCID mice depleted of NK cells . Lymph nodes in these mice are small and contain reticular structures but are completely devoid of lymphocytes ( Shultz et al . , 1995 ) . Similar to experiments on reconstituted NOD . SCID . β2 mice , eGFP-E106Ahi human T cells in NOD . SCID lymph nodes migrated with significantly elevated average velocities compared to control T cells ( Figure 2I ) , and exhibited lower arrest coefficients ( Figure 2J ) . Both eGFP-E106Ahi and control T cells migrated at lower speeds in the NK-depleted NOD . SCID model compared to the reconstituted NOD . SCID . β2 model . Because control human T cells in reconstituted NOD . SCID . β2 lymph nodes migrated at similar speeds to wild-type mouse T cells in vivo ( Miller et al . , 2002 ) , reconstitution results in a lymph node environment that more closely mimics normal physiological conditions . Furthermore , the greater effect of Orai1 block on T cell arrest coefficients in crowded reconstituted lymph nodes suggests that Orai1’s role in motility is more pronounced in crowded cell environments . To evaluate whether the pronounced effect of Orai1 channel block on the arrest coefficient in reconstituted lymph nodes was a result of environmental factors such as increased cellular contacts or increased confinement , we tracked human T cells in microfabricated PDMS chambers with cell-sized microchannels 7 µm high x 8 µm wide . These ICAM-1 coated microchannels simulate the confined environment of densely packed lymph nodes ( Jacobelli et al . , 2010 ) , while eliminating possible cell-extrinsic factors . Transfected human T cells were activated with plate-bound anti-CD3/28 antibodies and soluble IL-2 , then dropped into chambers and monitored by time-lapse confocal microscopy , using phase contrast to visualize eGFP-E106Alo T cells ( Figure 3A , B ) . Upon entry into microchannels , eGFP-E106Ahi T cells migrated with higher average cell velocities than eGFP-E106Alo T cells ( Figure 3C ) , similar to our in vivo findings from intact lymph node ( Figure 2 ) . To ensure that the observed difference in cell velocity was due to suppressed Orai1 channel function and not overexpression of Orai1 protein , we also tracked T cells transfected with eGFP-tagged wild-type Orai1 . Both eGFP-Orai1hi and eGFP-Orai1lo T cells migrated at the same average cell velocity ( Figure 3C ) , demonstrating that Orai1 channel overexpression , in itself , does not perturb T cell motility in microchannels . Since eGFP-E106Alo T cells have reduced Orai1 channel activity but still retain the same cell velocity as eGFP-Orai1 transfected T cells ( c . f . , Figures 1A and 3C ) , this suggests that partial Orai1 function is sufficient to generate normal pausing frequency in confined environments . The frequency distribution of cell velocities in vitro is comparable to our in vivo data: fewer GFP-E106Ahi T cells migrated with average cell velocities <7 µm/min as compared to eGFP-E106Alo T cells ( 11% vs 29%; Figure 3D ) . Furthermore , eGFP-E106Ahi T cells exhibited lower arrest coefficients ( Figure 3E ) and less variation in velocity than eGFP-E106Alo T cells ( Figure 3F ) . Although eGFP-E106Ahi T cells had lower arrest coefficients , the durations of their pauses were not significantly different than in eGFP-E106Alo T cells ( Figure 3G ) . Taken together , the reduced arrest coefficients in eGFP-E106Ahi T cells indicate that inhibition of Orai1 channel activity results in reduced frequency of pauses during T cell motility . These in vitro results confirm our in vivo findings and support the hypothesis that Orai1 activity intermittently triggers cell arrest , resulting in an overall decrease in motility within confined environments . Moreover , since our in vitro microchannel assay eliminates extrinsic cell-cell interactions , this indicates that Orai1 can be spontaneously activated to modulate T cell motility . To study the correlation between Ca2+ signals and T cell motility , human CD4+ T cells were transfected with Salsa6f , a novel genetically encoded Ca2+ indicator consisting of tdTomato fused to GCaMP6f , activated the T cells for 2 days with plate-bound anti-CD3/28 antibodies , then dropped into ICAM-1 coated microchambers . As previously shown ( Dong et al . , 2017 ) , Salsa6f is localized to the cytosol , with red fluorescence from tdTomato that reflects fluctuations in cell movement and very low baseline green fluorescence from GCaMP6f that rises sharply during Ca2+ signals ( Figure 4 ) . Salsa6f-transfected human T cells were tracked in both confined microchannels ( Figure 4A , Video 1 ) and the open space adjacent to entry into microchannels ( Figure 4C , Video 2 ) , to evaluate T cell motility under varying degrees of confinement . Intracellular Ca2+ levels were monitored simultaneously using the ratio of total GCaMP6f fluorescence intensity over total tdTomato fluorescence intensity ( designated as G/R ratio ) , enabling detection of a notably stable baseline ratio unaffected by motility artifacts in moving T cells while reporting spontaneous Ca2+ signals that could be compared to changes in motility ( Figure 4B , D , orange and black traces , respectively ) . Human T cells expressing Salsa6f migrating in confined microchannels exhibited sporadic Ca2+ signals as brief peaks unrelated to changes in cell velocity , or as more sustained periods of Ca2+ elevation associated with reduced cell velocity ( Figure 5A , B ) . To evaluate the correlation between T cell velocity and Ca2+ signals , we compared average T cell velocities during periods of sustained Ca2+ elevations to average velocities at baseline Ca2+ levels . T cell velocity decreased significantly when cytosolic Ca2+ was elevated above baseline ( 5 . 9 ± 0 . 1 µm/min vs . 10 . 0 ± 0 . 1 µm/min , p<0 . 0001; Figure 5C ) . Ca2+ signaling episodes that last for 30 s or longer accompany and appear to closely track the duration of pauses in cell movement . Comparison of instantaneous velocities with corresponding cytosolic Ca2+ signals ( G/R ratio ) by scatter plot revealed a strong inverse relationship: highly motile T cells always exhibited baseline Ca2+ levels , while elevated Ca2+ levels were only found in slower or arrested T cells ( Figure 5D ) . It is important to note that these Ca2+ signals and reductions in velocity occurred in the absence of any extrinsic cell contact or antigen recognition , indicating that Ca2+ elevations , like pausing and Orai1 activation , can be triggered in a cell-intrinsic manner . To compare the effects of Orai1 activity on the motility of T cells in a less confined environment , we also monitored T cell migration within the open space in PDMS chambers adjacent to entry into microchannels ( c . f . , Figure 4A , C ) . We reasoned that in this two-dimensional space with reduced confinement , T cells may not gain sufficient traction for rapid motility , and instead may favor integrin-dependent sliding due to increased exposure to the ICAM-1-coated surface ( Krummel et al . , 2014 ) . In addition , the same population of T cells could be tracked as they migrated into and along the confined microchannels , providing a valuable internal control . We found that eGFP-E106Ahi T cells migrated with similar velocities to eGFP-E106Alo T cells in the open space , but these eGFP-E106Ahi T cells still exhibited higher motility in the microchannels than eGFP-E106Alo T cells ( Figure 5E ) . Furthermore , Salsa6f-transfected T cells within the open space rarely produced Ca2+ transients ( c . f . , Figure 5D , F , top left quadrants , 13% of the time in microchannels vs 2% in open space ) , implying that Ca2+ elevations , and by extension , Orai1 channel activity , do not generate pauses when T cells are reliant on integrin binding for motility . Consistent with this , differentiated Th1 cells from Cd4-Salsa6f mice also showed similar instantaneous velocities and only rare Ca2+ transients when plated on open-field ICAM-coated coverslips ( Figure 5—figure supplement 1 ) . Taken together , these experiments establish a role for Orai1 channels and Ca2+ influx in modulating T cell motility within confined environments . Using Salsa6f , expressed in a Cd4Cre-dependent transgenic model we have reported that mouse T cells exhibit frequent transient Ca2+ signals ( ‘sparkles’ ) in homeostatic lymph nodes in the absence of specific antigen ( Dong et al . , 2017 ) . To further analyze the relationship between Ca2+ signaling and motility in detail within lymph nodes , we adoptively transferred homozygotic Cd4-Salsa6f ( Hom ) T cells into congenic mice and , using two-photon microscopy in explanted recipient lymph nodes , tracked the red tdTomato signal to establish cell position and the green CGaMP6f signal as a measure of cytosolic Ca2+ . First , to delineate any adverse effect of Salsa6f on homing and in situ motility of T lymphocytes , we co-injected equal numbers of Cd4-Salsa6f and Cd4Cre control cells into WT recipients ( Figure 6A ) . For simultaneous imaging and to normalize any dye toxicity , Cd4-Salsa6f and Cd4Cre T cells were labeled with CellTrace Yellow ( CTY ) and CellTrace Violet ( CTV ) , respectively . Comparable numbers of input cells were recovered from the subcutaneous lymph nodes after 18 hr ( Figure 6B ) . Two-photon imaging and tracking in lymph nodes showed typical stop and go motility and meandering cell tracks ( Figure 6C , D , Video 3 ) for both cell types . Instantaneous 3D velocities ( Figure 6E ) and mean track velocities ( Figure 6F ) were indistinguishable , as was the decay rate of directionality ratio ( Figure 6G ) . Furthermore , mean-squared displacement ( MSD ) time analysis showed random-walk behavior for both cell types with similar motility coefficients ( Figure 6H , I ) . Altogether , motility characteristics of Salsa6f T cells are indistinguishable from control T cells . To determine whether spontaneously occurring Ca2+ signals are correlated with motility , we transferred Cd4-Salsa6f cells alone into wild-type recipients and tracked red and green fluorescence intensities in the lymph nodes after 18 hr . Consistent with our previous observation , adoptively transferred T cells retained Salsa6f indicator in their cytosol , and Ca2+ signals were readily observed in motile Salsa6f+ T cells ( Figure 7A , Video 4 ) . We monitored the G/R ratios over time and observed a strong negative correlation between instantaneous cell velocity and Ca2+ levels ( Figure 7B ) . By examination of fluctuating cell velocity traces with corresponding G/R ratios , we found that the Ca2+ rise is clearly associated with a decrease in velocity ( Figure 7C and D , Video 5 ) . Notably , on average , peaks of Ca2+ transients precede the average cell velocity minimum , suggesting that spontaneous rise in intracellular Ca2+ levels leads to cell pausing ( Figure 7E ) . Imaging adoptively transferred T cells in recipient lymph nodes is an ideal approach to probe in vivo T-cell motility . However , this approach is limiting when it comes to identifying the abundance and duration of Ca2+ signaling events , because transferred cells label only a fraction of the lymph node ( <1% ) and longer imaging intervals are required to collect sufficient volume of 4D data ( >5 s ) . Therefore , to measure the endogenous frequency and duration Ca2+ transients , we imaged homeostatic Cd4-Salsa6f ( Hom ) lymph nodes at two frames per second . All endogenous T cells ( Cd4+ and Cd8+ ) are labeled with the Salsa6f probe in Cd4-Salsa6f lymph nodes because T cells go through the double-positive stage during development in the thymus . More than 800 Ca2+ transients were identified in a 300 × 300 µm area in a 10-min interval . We identified two types of Ca2+ transients: numerous small and brief spots ( sparkles ) ; and less frequent large , cell-wide transients ( Figure 8A , Video 6 ) . Consistent with our previous report ( Dong et al . , 2017 ) , most Ca2+ transients were localized to small regions of the cell and of short duration , spanning 2 µm2 in area ( Figure 8B ) and lasting about 2 s ( Figure 8C ) . Altogether , the strong association of Ca2+ transients with reductions in cell velocity leading to pausing , and the sheer number of Ca2+ transients in homeostatic lymph nodes suggest that cytosolic Ca2+ is a key regulator of basal cellular motility under steady-state conditions in the absence of specific antigen . Lymphocytes migrate in the immune dense micro-environment of secondary lymphoid tissues , constantly interacting with other immune cells , including resident antigen presenting cells ( Germain et al . , 2012 ) . Indeed , constant recognition of low levels of self-antigens through T cells receptor ( TCR ) -pMHC interactions is critical for maintaining sensitivity to foreign antigens ( Stefanová et al . , 2002 ) ; and deprivation ( >7 days ) of pMHC-II signals impairs T cell motility ( Fischer et al . , 2007 ) . To investigate whether Ca2+ signals in steady state lymph nodes are result of self-peptide recognition , we blocked MHC Class I and II signaling for 48 hr in Cd4-Salsa6f ( Hom ) lymph nodes . The number of cell-wide events was not significantly different ( p=0 . 06 ) , whereas the sparkle frequency was significantly decreased ( p=0 . 02 ) in MHC-blocked lymph nodes compared to isotype control ( ITC ) antibody treatment ( Figure 9A , D–G ) . There was also significant variation in the number of Ca2+ transients in ITC antibody and uninjected controls ( Coefficient of variation = 41% to 45% ) , which may be due to the presence of heterogeneous antigen presenting cells displaying varying amount of self-peptides during steady-state . Most notably , however , a significant number of Ca2+ transients remained even after MHC block , which we believe reflects a basal level of spontaneous Ca2+ activity independent of antigen recognition . In contrast , the intensity of individual Ca2+ transients in MHC blocked lymph nodes did not differ significantly from the ITC controls ( Figure 9B , C ) . Altogether , our data indicate that T cells display substantial spontaneous Ca2+ transients even in absence of self-peptide recognition , suggesting a role in regulating basal T lymphocyte motility .
In this study , we demonstrate that Orai1 channel activity regulates motility patterns that underlie immune surveillance . Human T cells expressing the dominant-negative Orai1-E106A construct migrated with higher average velocities than controls , both in reconstituted mouse lymph nodes in vivo and in confined microchannels in vitro . In particular , we found that the increase in average cell velocity was not due to an increase in maximum cell velocity , but to a reduced frequency of cell pausing accompanied by increased directional persistence , resulting in straighter paths . Human T cells demonstrate Ca2+ transient-associated and Orai1-dependent pauses in vitro within confined microchannels devoid of cell-extrinsic factors . Furthermore , we use a novel ratiometric genetically encoded Ca2+ indicator , Salsa6f , along with T cells from Cd4-Salsa6f ( Hom ) transgenic mice , to show that intermittent Ca2+ signals coincide with reduced cell velocity . Treatment of Cd4-Salsa6f ( Hom ) mice with MHC class-I and -II blocking antibodies substantially reduces but does not eliminate the frequent T cell Ca2+ transients seen in lymph nodes . Based on these findings , we propose that Orai1 channel activity regulates the timing of stop-and-go motility in T cells and tunes the search for cognate antigen in the crowded lymph node . Our Orai1-E106A construct derives its specificity and potency by targeting the pore residue responsible for the channel selectivity filter . Incorporation of Orai1-E106A likely blocks heterodimers of Orai1 and other channels , such as Orai2 or Orai3 , in addition to homomeric Orai1 channels . Very recent evidence demonstrates the existence of heteromeric channels composed of Orai1 and Orai2 in T cells ( Vaeth et al . , 2017 ) . These heteromers appear to simply reduce the flow of Ca2+ through the Orai1 channel without targeting additional signaling pathways . In the absence of contradictory evidence , we conclude that in T cells Orai1-E106A acts to produce an essentially complete functional knockdown of Orai1-mediated store-operated Ca2+ entry . Human T cells exhibit systematic changes in motility behavior after Orai1 block , including: increased average velocity , fewer pauses , increased directional persistence and decreased turn angles , and increased motility spread over time . Importantly , maximum and minimum instantaneous velocities are unchanged . For human T cells assessed in motility assays in vitro , similar Orai1-dependent changes are seen . Altered pausing behavior caused by Orai1 block is displayed by isolated single cells inside microchannels formed from photolithographically precise silicon master molds . Changes in motility are dependent upon confinement , as Orai1 block alters pausing under confined but not open-field conditions . The absence of changes to maximum and minimum velocities in vivo and open-field motility in vitro indicates that Orai1 block is not generally deleterious for cell health and movement , but instead acts upon subcellular mechanisms that are selectively employed during cell motility in confined spaces . Taken together , these systematic changes caused by Orai1 block reveal the presence of an Orai1-dependent cell motility program that is utilized frequently enough to be easily detected by changes in the motility characteristics of T cells in the lymph node . We used Cd4-Salsa6f ( Hom ) mice to track Cd4+ T cell Ca2+ signals in intact mouse lymph nodes and Salsa6f transient transfection to track Ca2+ signals in human T cells in vitro and in reconstituted mouse lymph nodes . In our companion paper , Dong et al . , 2017 , we show that Salsa6f expression in Cd4+ T cells is non-perturbing with respect to lymphocyte development , cellular phenotype , cell proliferation , and differentiation . Here , we demonstrate that homing to lymph nodes is unaffected , as are movement patterns within the lymph node , cell velocity , directional persistence , diffusive spread , and motility coefficient . We find that across cells , elevated Ca2+ levels are inversely correlated with instantaneous velocity , both in vitro and in vivo . In vivo , moving cells exhibit local Ca2+ signals that are strongly associated with pauses in motility . By inspection of movement patterns , turning is likely associated with Ca2+ signaling events as well , but this has not been established because most cells move outside of our shallow imaging field either before or after pausing . In many contexts , Ca2+ signaling has been shown not only to accompany , but also to cause cell arrest and loss of cell polarity , such as in T cells after activation by antigen ( Negulescu et al . , 1996; Dustin et al . , 1997; Wei et al . , 2007 ) . By averaging events , the peak of subcellular Ca2+ transients was found to precede the velocity minimum . This event order is consistent with Ca2+ causing pauses . While we do not show that the Ca2+ signals we observe emanate directly from Orai1 channels , taken together our data are consistent with Orai1 actively regulating cell motility by directly inducing a subcellular motility program that leads to cell arrest . Two-photon imaging indicated that the frequency of Cd4+ T cells Ca2+ transients varies widely between Salsa6f lymph nodes , even when events are normalized for different cell numbers . The origin of this variability is unclear but may result from differences in the distribution and functional properties of APCs within the imaging field . Treatment of MHC class-I and –II blocking antibodies substantially reduces but does not eliminate T cells Ca2+ transients . Clearly , a significant number of Ca2+ transients are caused by T cell-APC interactions that act through MHC proteins . Given that Orai1 motility events occur frequently as T cell migrate through the lymph node , and Ca2+ transients are associated with pauses in motility , we propose that spontaneously generated Orai1-dependent pauses and turns can be triggered by T cell-APC interaction through MHC proteins . However , we find evidence for MHC-independent triggering of Ca2+ signaling and Orai1 channel activation in the lymph node . Human T cells exhibit Orai1-dependent pauses in vitro when migrating as isolated cells in highly uniform microchannels . Salsa6f expression independently detects Ca2+ transients in isolated T cells moving within microchannels but not in T cells in adjacent open field portions of the same PDMS imaging chamber . In both cases , responses were produced in the absence of MHC proteins or APCs . Moreover , we note that , in paired experiments , treatment with MHC class-I and –II blocking antibodies leads to a reduced but notably consistent frequency of Ca2+ signaling events . Partial block would be expected to produce substantial variation , especially when combined with a variable input population . Taken together , these data point to the existence of not only MHC-independent Orai1 motility events , but also cell-intrinsic triggering of Orai1 . Of note , the apparently random nature of naïve T cell movement in the lymph node has led to the hypothesis that T cells use intrinsic and stochastic motility mechanisms to accomplish immune surveillance ( Wei et al . , 2003; Mrass et al . , 2010; Germain et al . , 2012 ) . In previous studies of Orai1 signaling , Orai1 activation has been placed downstream of extracellular ligand binding to cell surface receptors , integrating their input upon use-dependent depletion of Ca2+ from the ER ( Feske , 2007; Cahalan and Chandy , 2009 ) . While we expect signaling downstream of self-antigen detection to be the same as for cognate antigens ( Stefanová et al . , 2002 ) , at this point it is unclear which aspects of internal cell state might lead to cell-intrinsic opening of Orai1 channels and pauses in motility . Of particular interest is determining the step in the signaling cascade from phospholipase C to Orai1 that might be targeted by a novel cell-intrinsic activation pathway . Molecular candidates that underlie regulation of T cell motility by Ca2+ are less well defined . One clue to Orai1 action is the subcellular location of Ca2+ transients at the back T cells moving within intact lymph nodes , similar to the localization of Orai1 channels during movement in vitro ( Barr et al . , 2008 ) . Early studies demonstrated that immobilization and rounding of T cells bound to antigen-presenting B cells occurred via a calcineurin-independent pathway ( Negulescu et al . , 1996 ) . Ca2+-sensitive cytoskeletal proteins , such as myosin II or the actin bundling protein L-plastin , as good candidates for downstream effectors ( Babich and Burkhardt , 2013; Morley , 2013 ) . Like Orai1 , Myosin 1 g is selectively required for motility mechanisms under confined conditions ( Gérard et al . , 2014 ) . While Orai1 block reduces pausing but does not otherwise alter T cell velocity , Myo1g block increases pausing and causes cells to move faster . These differences in phenotype suggest that Orai1 and Myo1g act in different , and in part opposing , ways to control T cell motility . Immune surveillance requires balancing many factors associated with antigen search , including speed and sensitivity ( Friedl and Weigelin , 2008; Krummel et al . , 2016 ) . As moving T cells in the lymph node encounter APCs bearing antigen-MHC , they pause due to Ca2+-dependent mechanisms unleashed by Orai1 channel opening . These pauses likely ensure adequate time for TCR-antigen scanning by T cell-APC pairs . Our observations of T cell motility indicates that each T cell does not stop at every APC it encounters . Because of this movement pattern , Orai1 provides attractive point of regulation of immune surveillance . Increasing Orai1 activity might be expected to cause T cells to pause more frequently when encountering APCs , restricting the distance T cells move and offering increased opportunities for contact with nearby APCs . Alternatively , decreasing Orai1 activity leads to fewer pauses , greater directional persistence , fewer turns , and greater overall diffusive spread . In this way , Orai1 channel activity could tailor T cell excursions to match the density and reach of dendritic cells in the lymph node . Finally , our findings provide further evidence that during resting conditions , TCR interactions with self-MHC antigens drive continual but limited activation of downstream signaling pathways . We note some differences between our study and others involving Orai1 , STIM1 , and T cell motility . These differences might be accounted for , in part , by the expected consequences of our Orai1 dominant-negative approach: block of all three Orai isoforms , limited time for compensatory changes in cell function , and restriction of Orai block to T cells that are adoptively transferred after transfection . ( 1 ) Orai1/2 and STIM1/2 KOs have been reported to home to lymph node like wild type , unlike our results here and in a previous paper ( Greenberg et al . , 2013; Waite et al . , 2013; Vaeth et al . , 2017 ) . ( 2 ) Maximal T cell velocity in the lymph node requires the action of the integrin LFA-1 and the chemokine receptor CCR7 ( Davalos-Misslitz et al . , 2007; Katakai et al . , 2013 ) , which we have previously shown to be required for entry of T cells into lymph nodes and to act in an Orai1-dependent manner ( Greenberg et al . , 2013 ) . Based upon these findings , blocking Orai1 would be expected to reduce CCR7 and LFA-1 function during interstitial motility as well , resulting in a decrease in T cell velocity . Instead , we find the opposite: Orai1 block leads to an increase in average cell velocity . The absence of any Orai1-dependent change in maximum velocity strongly suggests that CCR7 and LFA-1 do not act through Orai1 during motility in the lymph node . Regardless , any motility effects of LFA-1 and CCR7 are more than compensated by the reduction in pauses caused by Orai1 E106A expression . ( 3 ) Previous studies using unconfined open field motility assays have excluded a role for Orai1 in T cell motility ( Svensson et al . , 2010; Kuras et al . , 2012 ) . Our experiments confirm that Orai1 block does not detectably affect unconfined motility; in contrast , our studies in reconstituted lymph nodes and in confined microchannels in vitro both exhibit Orai1-dependent effects . Others have shown that confined motility in vitro better recapitulates mechanisms of motility found in T cells in intact lymph nodes ( Jacobelli et al . , 2010; Krummel et al . , 2016 ) . In conclusion , we reveal the existence of an Orai1-dependent cell motility program that leads to pausing of T cells moving within lymph nodes . Imaging with the newly developed genetically encoded Ca2+ indicator Salsa6f identifies local transient Ca2+ signaling events with the expected characteristics of Orai1 Ca2+ signals . We provide evidence that Orai1-dependent pauses in T cells are triggered in at least two different ways: by self-peptide MHC complexes displayed on the surface of APCs and by a novel cell intrinsic mechanism within the T cells themselves . Together these mechanisms generate motility patterns that promote efficient scanning for cognate antigens in the lymph node .
NOD . Cg-PrkdcscidB2mtm1Unc/J ( NOD . SCID . β2 ) and NOD . CB17-Prkdcscid/J ( NOD . SCID ) mice obtained from Jackson Laboratory ( Stock #002570 and #001303 ) were housed and monitored in a selective pathogen-free environment with sterile food and water in the animal housing facility at the University of California , Irvine . NOD . SCID . β2 mice were reconstituted with human peripheral blood leukocytes ( PBLs ) as described previously ( Mosier et al . , 1988 ) . A total of 3 × 107 human PBLs were injected i . p . , and experiments were performed 3 weeks later . To inhibit NK cell activity , NOD . SCID mice were i . p . injected with 20 μL anti-NK cell antibody ( rabbit anti-Asialo GM1 , Wako Chemicals , Irvine , CA ) according to manufacturer's instructions 3–4 days before adoptive transfer of human T cells . Mice used were between 8 and 18 weeks of age . The mouse strain expressing Salsa6f selectively in T cells under Cd4-Cre recombinase is described in the comapanion manuscript Dong et al . , 2017 . In brief , LSL-Salsa6f ( tdTomato-V5-GCaMP6f ) mouse strain was generated in the C57BL/6N background , as described in the accompanying manuscript , and subsequently crossed to homozygotic Cd4Cre mice ( JAX #017336 ) to generate Cd4-Salsa6f ( Het ) mice expressing Salsa6f only in T cells . These mice were further bred to generate homozygotic Cd4-Salsa6f ( Hom ) mice for increased Salsa6f expression and fluorescence . Age- and sex-matched C57BL/6J mice from Jackson Laboratory ( stock #000664 ) were used as wild-type recipients of Cd4-Salsa6f ( Hom ) T cells . To block TCR-MHC interactions , 2 mg of anti-MHC II ( Clone Y3P ) and 2 mg of anti-MHC I ( Clone AF6-88 . 5 . 5 . 3 ) or 4 mg of IgG2a Isotype control ( Clone: C1 . 18 . 4 ) antibodies ( Bio X cell ) were injected into Cd4-Salsa6f ( Hom ) litter mates ( i . p ) 48 hr before imaging . Human PBMCs were isolated from blood of voluntary healthy donors by Histopaque-1077 ( 1 . 077 g/mL; Sigma , St . Louis , MO ) density-gradient centrifugation , and human T cells isolated using the appropriate EasySep T Cell Isolation Kit ( StemCell Technologies ) . Purified human T cells were rested overnight in complete RPMI , then transfected by nucleofection ( Lonza , Walkersville , MD ) , using the high-viability ‘U-014’ protocol . Enhanced green fluorescent protein ( eGFP ) -tagged wild-type Orai1 , eGFP-tagged Orai1-E106A mutant , Salsa6f ( tdTomato-V5-GCaMP6f construct ) , or empty vector control were transfected as indicated . Human T cells were used for experiments 3–48 hr after transfection . CMTMR control T cells were prepared by labeling with 10 μM CellTracker CMTMR dye ( Invitrogen , Carlsbad , CA ) for 10 min at 37°C . For in vivo imaging 10 million human T cells were injected into NOD . SCID . β2 or NOD . SCID mice as indicated . For in vitro imaging experiments , T cells were rested for 3–4 hr in complete RPMI , then washed and activated on plate-bound αCD3 and αCD28 ( Tonbo Biosciences , San Diego , CA ) in 2 . 5 ng/mL recombinant human IL-2 ( BioLegend , San Diego , CA ) , and imaged 24–48 hr after transfection . Single cell suspensions of mouse lymphocytes were prepared by mechanical dissociation of spleen and lymph nodes and passing through 70 µm filter . Cd4+ T cells were isolated using the EasySep T Cell Isolation Kit ( StemCell Technologies ) according to manufacturer's instructions . The purity of isolated cells was confirmed to be >95% by flow cytometry . To compare motility characteristics , Cd4-Salsa6f ( Hom ) and Cd4Cre control cells were labeled with 10 µM CellTrace Yellow or CellTrace Violet , respectively , for 20 min at 37°C . To measure Ca2+ during T cell motility , unlabeled Cd4-Salsa6f ( Hom ) T cells were adoptively transferred into wild-type recipients . A total of 3–10 million T cells were injected into recipient mice in adoptive transfer experiments ( i . v: tail-vein or retro-orbital ) . For confocal imaging on open-field ICAM-1-coated coverslips , Cd4+ T cells from Cd4-Salsa6f ( Het ) mice were differentiated into Th1 cells using 25 ng/mL rmIL-12 ( BioLegend ) , 10 µg/mL αmouse IL4 ( Biolegend ) for 4–6 days . Microchannel fluidic devices were fabricated by a soft lithography technique with PDMS ( polydimethylsiloxane; Sylgard Elastomer 184 kit; Dow Corning , Auburn , MI ) as described ( Jacobelli et al . , 2010; Gérard et al . , 2014 ) . PDMS base and curing agent were mixed 10:1 and poured onto the silicon master , then left overnight in vacuum . Once the PDMS was set , it was baked at 55°C for 1 hr and cooled at room temperature . The embedded microchambers were then cut from the mold , and a cell well was punched adjacent to entry into the channels . The PDMS cast and a chambered coverglass ( Nunc Lab-Tek , ThermoFisher , Grand Island , NY ) were activated for 2 min in a plasma cleaner ( Harrick Plasma , Ithaca , NY ) , bonded together , then incubated at 55°C for 10 min . Prepared chambers were stored for up to 1 month before use . Prior to imaging , microchambers placed in the plasma cleaner for 5 min under vacuum and 1 min of activation , then coated with 5 µg/mL recombinant human ICAM-1/CD54 Fc ( R and D Systems , Minneapolis , MN ) in PBS for at least 1 hr at 37°C . The microchambers were then washed three times with PBS , and T cells were loaded into cell wells ( 3−5 × 105 cells resuspended in 10 µL ) and incubated at 37°C for at least 1 hr before imaging . Two different Olympus confocal microscopy systems were used to image T cells in vitro . For experiments tracking T cell motility in microchambers , we used the self-contained Olympus Fluoview FV10i-LIV , with a 473 nm diode laser for excitation and a 60x phase contrast water immersion objective ( NA 1 . 2 ) . The FV10i-LIV contains a built-in incubator set to 37°C , together with a Tokai-Hit stagetop incubator to maintain local temperature and humidity . T cells were imaged in RMPI adjusted to 2 mM Ca2+ and 2% FCS , and mounted at least half an hour before imaging to allow for equilibration . Cells were imaged at 20 s intervals for 20–30 min , and the data analyzed using Imaris software . For Ca2+ imaging of Salsa6f transfected T cells , we used a Fluoview FV3000RS confocal laser scanning microscope , equipped with high-speed resonance scanner and the IX3-ZDC2 Z-drift compensator . Diode lasers ( 488 and 561 nm ) were used for excitation , and two high sensitivity cooled GaAsP PMTs were used for detection of GCaMP6f and tdTomato . Cells were imaged using the Olympus 40x silicone oil objective ( NA 1 . 25 ) , by taking four slice z-stacks at 1 . 5 µm/step , at 3 s intervals , for up to 20 min . Temperature , humidity , and CO2 were maintained using a Tokai-Hit WSKM-F1 stagetop incubator . Data were processed and analyzed using Imaris software . Multi-dimensional ( x , y , z , time , emission wavelength ) two-photon microscopy was employed to image fluorescently labeled lymphocytes in explanted mouse lymph nodes , as described ( Miller et al . , 2002; Matheu et al . , 2015 ) . The following wavelengths were used to excite single or combinations of fluorophores: 900 nm to excite eGFP and CMTMR; 800 nm to excite cell trace violet ( Thermofisher C34557 ) and cell trace yellow ( Thermofisher , C34567 ) ; 920 nm to excite tdTomato and GCaMP6f; Fluorescence emission was split by 484 nm and 538 nm dichroic mirrors into three detector channels , used to visualize CellTrace Violet or second harmonic signal generated from collagen in blue; GCaMP6f or eGFP-Orai1E106A transfected cells in green; tdTomato or CellTrace Yellow or CMTMR-labelled cells in red . For imaging , lymph nodes were oriented with the hilum away from the water dipping microscope objective ( Olympus 20x , NA 0 . 9 or Nikon 25x , NA 1 . 05 ) on an upright microscope ( Olympus BX51 ) . The node was maintained at 36–37°C by perfusion with medium ( RPMI ) bubbled with carbogen ( 95% O2/5% CO2 ) . For imaging of human T cells 3D image stacks of x = 200 μm , y = 162 μm , and z = 50 μm were sequentially acquired at 18–20 s intervals using MetaMorph software ( Molecular Devices , Sunnyvale , CA ) . For tracking adoptively transferred mouse T cells , 3D image stacks of x = 250 μm , y = 250 μm , and z = 20 or 52 μm ( Voxel size 0 . 48 μm x 0 . 48 μm x 4 μm ) were sequentially acquired at 5 or 12 s intervals , respectively , using image acquisition software Slidebook ( Intelligent Imaging Innovations ) as described previously ( Matheu et al . , 2015 ) . This volume collection was repeated for up to 40 min to create a 4D data set . For fast imaging of Cd4-Salsa6f ( Hom ) lymph nodes , we acquired 2DT images of 300 µm x 300 µm ( pixel size 0 . 65 × 0 . 65 μm ) every 0 . 5 s . For comparing Ca2+ transients in MHC blocking experiments , 3D image stacks of x = 350 μm , y = 350 μm , and z = 20 μm ( Voxel size 0 . 65 μm x 0 . 65 μm x 4 μm ) were sequentially acquired at 5 s intervals . Cell motility data were processed and analyzed using Imaris software ( Bitplane USA , Concord , MA ) . A combination of manual and automatic tracking was used to generate highly accurate cell tracks . The x , y , z coordinates of the tracks were used to calculate speed , M . S . D , directionality ratio , motility coefficients , and to plot tracks as described previously ( Gorelik and Gautreau , 2014; Matheu et al . , 2015 ) . Calcium transient ( sparkles and cell-wide ) analysis and estimation of duration was performed as described previously ( Dong et al . , 2017 ) . XYT data was processed to mask autofluorescent structures , and time was mapped on the Z axis for the purpose of Ca2+ transient identification . Ca2+ transients were identified in Imaris by a surface-based object identification approach , after manual thresholding of intensity , voxel size ( >10 ) and 2 s minimum duration . Objects were modeled as ellipsoids; X and Y diameter measurements of surfaces were used to calculated areas , and Z diameter ( time ) was used to estimate duration of Ca2+ transients . For MHC-block experiments to estimate the number and intensities of Ca2+ transients , we utilized maximum intensity projections from 6 Z stacks . Integrated intensities were normalized to standard deviations of the green channel for comparison of brightness of Ca2+ transients . Samples sizes were comparable to previous single cell analyses of motility ( Jacobelli et al . , 2010; Greenberg et al . , 2013; Gérard et al . , 2014 ) . Each experiment used separate isolations of human T cells from different donors . With the exception of instantaneous velocities in Figure 6C , each measurement corresponds to a different cell . Mean ± standard error of the mean was used as a measure of the central tendency of distributions . Video analysis was performed using Imaris software , Spots analysis was used for tracking of cell velocity and Volumes analysis was used for measuring total fluorescence intensity of GECI probes . To reduce selection bias in our analysis of motility and trajectory , all clearly visible and live cells were tracked from each video segment . The arrest coefficient is defined as fraction of time each cell had an instantaneous velocity <2 µm/min . The coefficient of variation was defined for each individual cell as the standard deviation divided by the mean of its instantaneous velocity . For Salsa6f imaging analysis , ratio ( R ) was calculated by total GCaMP6f intensity divided by total tdTomato intensity , while initial ratio ( R0 ) was calculated by averaging the ratios of the first five time points in each individual cell trace . Photobleaching of tdTomato fluorescence intensity ( 20–30% decline ) was corrected in ratio calculations , as a linear function of time . Figures were generated using Prism 6 ( GraphPad Software , San Diego , CA ) and Origin 5 ( OriginLabs , Northampton , MA ) . Due to the expectation that individual cells exhibit multiple motility modes , and to avoid assumptions concerning the shapes of motility distributions , non-parametric statistical testing was performed ( Mann-Whitney U test , unpaired samples , two-tailed , Spearman's rank correlation ) . Differences with a p value of ≤ 0 . 05 were considered significant: *p≤0 . 05; **p<0 . 01; ***p<0 . 005; ****p<0 . 001 . Similar distributions were compared using the Hodges-Lehmann median difference value and 95% confidence intervals under the assumption that the starting distributions had similar shapes . | To help protect the body from disease , small immune cells called T lymphocytes move rapidly , searching for signs of infection . These signs are antigens – processed pieces of proteins from invading bacteria and viruses – which are displayed on the surface of so-called antigen-presenting cells . To visit as many different antigen-presenting cells as possible , T cells move quickly from one to the next in an apparently random manner . How T cells are programmed to move in this way is largely unknown . The entry of calcium ions into cells triggers characteristic actions in many cells throughout the body . In T cells , calcium ions enter through Orai1 proteins that form calcium channels on the cell surface . Now , Dong , Othy et al . have asked whether calcium signals guide moving T cells as they search for antigens . Experiments with individual human T cells in small tubes showed that blocking the Orai1 calcium channels caused the T cells to move faster , because the cells paused less often . The same was seen when human T cells were transplanted into mice . These findings suggested that calcium signals may indeed guide the T cells’ movement , but actually being able to see the calcium signals in the cell would give a much clearer picture of what goes on . To achieve this , Dong , Othy et al . report , in a related study , how they genetically engineered mice to produce a calcium-sensitive reporter protein in their T cells . Using these new transgenic mice , Dong , Othy et al . could see calcium signals in the T cells before each of the T cell’s pauses . Further experiments showed that the calcium signals that control the cell’s movements are triggered both by contact with the antigen-presenting cells and internally within the T cells themselves . In another related study , Guichard et al . also conclude that contact with antigen-presenting cells causes calcium signals that control the responses of T cells . Seemingly random patterns of movement help T cells search for signs of infection , and these new findings reveal a basic part of how T cells are programmed to move in this way . A deeper understanding of T cell movement might allow this process to be controlled . In particular , this knowledge could lead to new treatments for autoimmune diseases , in which T cells incorrectly recognize the body’s own antigens as signs of an infection . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2017 | Intermittent Ca2+ signals mediated by Orai1 regulate basal T cell motility |
Morphogens regulate tissue patterning through their distribution in concentration gradients . Emerging research establishes a role for specialized signalling filopodia , or cytonemes , in morphogen dispersion and signalling . Previously we demonstrated that Hedgehog ( Hh ) morphogen is transported via vesicles along cytonemes emanating from signal-producing cells to form a gradient in Drosophila epithelia . However , the mechanisms for signal reception and transfer are still undefined . Here , we demonstrate that cytonemes protruding from Hh-receiving cells contribute to Hh gradient formation . The canonical Hh receptor Patched is localized in these cellular protrusions and Hh reception takes place in membrane contact sites between Hh-sending and Hh-receiving cytonemes . These two sets of cytonemes have similar dynamics and both fall in two different dynamic behaviours . Furthermore , both the Hh co-receptor Interference hedgehog ( Ihog ) and the glypicans are critical for this cell-cell cytoneme mediated interaction . These findings suggest that the described contact sites might facilitate morphogen presentation and reception .
Long-distance cell-cell communication is essential for development and function of multicellular systems . During embryogenesis morphogens are produced at a localized source and act at a distance , controlling the differential activation of target genes in a concentration-dependent manner ( Crick , 1970; Morgan , 1901; Rogers and Schier , 2011; Stumpf , 1966; Turing , 1952; Wolpert , 1969 ) . Thus , the graded distribution of morphogens , together with the ability of the receptor cells to respond specifically to different ligand concentrations , need to be tightly regulated . Increasing evidence supports active signal transport along specialized filopodia ( also called cytonemes ) ( Ramírez-Weber and Kornberg , 1999; Gradilla and Guerrero , 2013a , 2013b; Kornberg , 2014 ) , challenging models based on the free diffusion of morphogens . Here we keep investigating the mechanisms of gradient formation by active transport of morphogens across the epithelial surfaces . The Hedgehog ( Hh ) morphogen has a central role in many metazoan developmental processes . Hh is implicated in stem cell maintenance , axon guidance , cell migration and oncogenesis in a wide range of organisms ( Briscoe and Thérond , 2013 ) . Hh production , transport , release and reception must be kept under strict spatial and temporal control to fulfil its signalling function . Hh is also post-translationally modified by the addition of cholesterol ( Porter et al . , 1996 ) and palmitic acid ( Pepinsky et al . , 1998 ) , which promote its association with cell membranes making Hh transport potentially inconsistent with free morphogen diffusion . Hh gradient establishment has been extensively studied in the Drosophila wing imaginal disc epithelium , which is formed by anterior ( A ) and posterior ( P ) cell populations with different adhesion affinities . The P compartment cells produce Hh , which moves across the A/P compartment border to reach the Hh-responding cells in the A compartment . As Hh spreads away from the border , its concentration decreases , providing a graded signal that activates the different target genes that regulate imaginal disc development ( reviewed in Briscoe and Thérond , 2013 ) . In both wing disc and abdominal histoblasts , cytonemes from Hh-producing cells extend across its morphogenetic gradient ( Bischoff et al . , 2013 ) . Critically , there is a strong correlation between the extent of cytonemes from the P compartment and the graded response to Hh signalling in the A compartment . In vivo imaging of abdominal histoblasts showed that cytonemes extend and retract dynamically , and that Hh gradient establishment correlates with cytoneme formation in both space and time . These data support a model for Hh transport in which cytonemes act as conduits for morphogen movement mainly at the basal plane of the epithelium . Furthermore , we have shown that Hh is associated with vesicles transported along cytonemes ( Gradilla et al . , 2014 ) . The mechanisms for Hh signal transfer and reception , however , remain open questions . Here we show that cytonemes emanating from the Hh-receiving cells in the A compartment contribute to Hh reception and gradient formation . These cytonemes have similar dynamics than those emanating from the Hh-producing cells , falling between two different dynamic behaviours . We show that reception Hh signalling components localize to the signal-receiving cytonemes , including the glypicans Division abnormally delayed ( Dally ) and Dally-like ( Dlp ) , the adhesion molecule Interference hedgehog ( Ihog ) and the canonical Hh receptor Patched ( Ptc ) . Significantly , the spreading capacity of cytonemes is dependent on the glypicans present in the membranes of neighbouring cells . Thus , cytonemes cannot properly extend across Dally or Dlp mutant cells . In addition , cytonemes can cross smo ( smoothened ) and ptc mutant clones , which cannot internalize Hh , providing a bridging mechanism and allowing Hh delivery to adjacent wild type cells . Finally , we describe discrete cell-cell contact structures between Hh-sending and Hh-receiving cytonemes , where the morphogen may be transferred from one cytoneme to the other for its reception .
Hh-producing cells in the P compartment of the wing imaginal disc extend cytonemes that transport Hh to the A compartment cells and that are essential for the restricted distribution of Hh during Drosophila epithelial development ( Callejo et al . , 2011; Bilioni et al . , 2013; Bischoff et al . , 2013 ) . In addition , the Hh-receiving cells of the anterior compartment also extend cytonemes towards the Hh-secreting cells of the P compartment . Here we have characterized the cytonemes from the signal-receiving cells and investigated their role in Hh morphogen reception . In previous studies on Hh signalling filopodia in the abdominal histoblasts we showed that the P compartment generated highly dynamic protrusions that reached anteriorly the Hh-receiving cells ( Bischoff et al . , 2013 ) . The Hh-receiving cells also produce highly dynamic protrusions oriented towards the Hh-producing cells , easily visualized when expressing the actin-binding domain of moesin ( GMA ) fused to GFP ( Figure 1A , Video 1A ) . These GMA-labelled filopodia are less dynamic when they co-express Ihog ( Figure 1B , Video 1B ) , as was previously described for the Hh-producing histoblasts ( Bischoff et al . , 2013 ) . Here we show that both Hh-presenting and Hh-receiving histoblast cells emit protrusions with similar dynamics ( Video 1 and Video 2 ) . In a more detailed analysis of filopodia dynamics , we have been able to distinguish two different dynamic behaviours: one of filopodia that elongate and immediately retract , which we have classified as ‘triangle dynamics’ and another one with a ‘stationary’ interphase between the elongation and retraction phases , which we have classified as ‘trapezoid dynamics’ ( Figure 1—figure supplement 1; see Materials and methods ) . Both Hh-producing and Hh-receiving cell filopodia have similar values of average maximum extent , lifetime , elongation ( Ve ) and retraction ( Vr ) velocities ( Figure 1C–E , Figure 1—figure supplement 2 ) . 10 . 7554/eLife . 24045 . 003Figure 1 . Cytonemes emanating from the A compartment cells . ( A , B ) Dynamic behaviour of cytonemes emanating from the A compartment abdominal histoblasts , monitored by the actin cytoskeleton marker GMA-GFP alone ( the actin-binding domain of moesin fused to GFP ) ( A ) , or co-expressed with Ihog-RFP ( B ) , after 24 hr of expression using the ptc . Gal4 driver . Hh-receiving cells produce highly dynamic filopodia visualized by expressing GMA-GFP ( A ) , which are more stabilized when co-expressing Ihog-RFP ( B ) . ( C–E ) Violin plots represent filopodia maximum extent ( C ) , lifetime ( D ) , and elongation velocity ( Ve ) and retraction velocity ( Vr ) measurements ( E ) . Statistical analysis was done to compare the expression , for 24 hr , of GMA-GFP alone or co-expressed with Ihog-RFP in Hh-receiving cells ( ptc . Gal4 ) or Hh-producing cells ( hh . Gal4 ) . Coloured diamonds indicate the mean of the data and black lines the median . *p<0 . 05 , **p<0 . 01 , ***p<10−3 , ****p<10−4 . Scale bars represent 10 μm . ( F ) Interfering with A cytonemes extension ( by transient co-expression of UAS . ihog-RFP together with UAS . scar-RNAi , UAS . dia-RNAi , UAS . cpa RNAi or UAS . cher-RNAi for 30 hr using the ptc . Gal4 driver ) affects the spreading of the Hh gradient in the wing disc , monitored by Ptc and Ci expressions ( by transiently expressing the UAS-RNAi lines and not UAS . ihog-RFP for 30 hr ) . Graph representing the average of 5 discs in 3 independent experiments ( error bars represent SDs ) . ( G ) Two Ihog-RFP expressing clones induced in the wing pouch , one in the A compartment and another in the P compartment , in lateral , basal and transverse ( Z-stack ) sections . Note that very distant cells contact at the basal region through cytonemes oriented towards the A/P compartment border ( arrow ) . ( H ) A ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP wing disc after 24 hr at the restrictive temperature and stained with α-Hh antibody . Note that A compartment cytonemes expressing Ihog-RFP capture Hh produced by the P compartment cells ( arrows ) . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00310 . 7554/eLife . 24045 . 004Figure 1—figure supplement 1 . Phase models based on filopodia dynamics . ( A ) 2-Phases ‘triangle’ dynamics . Experimental data from a filopodium extent over time is represented on light blue; its ‘trapezoid’ model curve is represented in black and its ‘triangle’ model curve in grey . Notice that the filopodium dynamics are more similar to the ‘triangle’ model with an elongation and a retraction phase than to the ‘trapezoid’ model . ( B ) 3-Phases ‘trapezoid’ dynamics . Data from a filopodium extent over time is represented in light blue; its ‘trapezoid’ model curve is represented in black and its ‘triangle’ model curve in grey . Notice that the filopodium dynamics are more similar to the ‘trapezoid’ model with a stationary phase between the elongation and retraction phases than to the ‘triangle’ model . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00410 . 7554/eLife . 24045 . 005Figure 1—figure supplement 2 . Hh-receiving cells filopodia have similar dynamics to Hh-producing cells . ( A ) . Table showing the maximum extent ( E max ) , lifetime , and elongation ( Ve ) and retraction ( Vr ) velocities of filopodia from Hh-producing and Hh-receiving cells with and without Ihog overexpression . ( B–E ) Graphs representing GMA filopodia extent over time expressing GMA alone or co-expressed with Ihog in Hh-receiving cells ( B , D ) or in Hh-producing cells ( C , E ) respectively . ( D’ , E’ ) Five time frames from Video 1B ( D’ ) and Video 2B ( E’ ) . GMA filopodia with high levels of Ihog ( black arrows ) are less dynamic ( black curves ) than those with lesser levels ( grey arrows and grey curves ) or absence ( cyan arrows and cyan curves ) of Ihog . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00510 . 7554/eLife . 24045 . 006Figure 1—figure supplement 3 . Ihog overexpression in Hh-receiving cells leads to an extended Hh gradient . ( A , B ) . Lateral sections of a wild type , WT ( A ) and a ptc . Gal4>UAS . ihog-RFP ( B ) wing discs after 24 hr at restrictive temperature , co-labelled with α-Ptc and α-Ci antibodies . ( A’ , B’ ) Graphs showing plots of fluorescence intensity of Ptc and Ci shown in A ( A’ ) and B ( B’ ) . Note the extension of Ptc and Ci domains in response to Ihog overexpression in Hh-receiving cells . The data shown were consistent in at least ten independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00610 . 7554/eLife . 24045 . 007Video 1 . Dynamics of filopodia of A compartment abdominal histoblast cells expressing only GMA and co-expressing Ihog . ( A-B ) Abdominal histoblasts of pupae with the genotype w; ptc . Gal4 / tubGal80ts; UAS . GMA-GFP /+ ( A ) and w; ptc . Gal4 / tubGal80ts; UAS . GMA-GFP / UAS . ihog-RFP ( B ) . The actin-binding domain of moesin fused to GFP ( GMA-GFP ) was expressed during 24 hr in Hh-receiver cells to visualize actin-based filopodia dynamics , easily detected using the inverted grey-scale lookup table tool of Fiji ( A ) . Notice the highly dynamic filopodia emerging from A cells towards P cells . GMA-GFP ( B ) was co-expressed with Ihog-RFP ( B’ ) in Hh-receiver cells . Observe in the merge panel ( B’’ ) that Ihog-containing filopodia are stabilized while few filopodia with low or no Ihog levels detected are more dynamic . Histoblasts move up ( A ) and down ( B ) towards the dorsal midline , not shown because of the high magnification . Anterior is on the left . Pupae were around 30 hr APF ( after puparium formation ) . Movies of 30 min imaging with time intervals between frames of 2 min . Scale bars represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00710 . 7554/eLife . 24045 . 008Video 2 . Dynamics of filopodia from P compartment abdominal histoblast cells expressing only GMA and co-expressing Ihog . ( A-B ) Abdominal histoblasts of pupae with genotype w; tubGal80ts /+; hh . Gal4 / UAS . GMA-GFP ( A ) and w; tubGal80ts / UAS . ihog-RFP; hh . Gal4 / UAS . GMA-GFP ( B ) . GMA-GFP was expressed during 24 hr in Hh-producing cells to visualize actin-based filopodia dynamics , easily detected using the inverted grey-scale lookup table tool of Fiji ( A ) . Notice the highly dynamic filopodia emerging from P cells towards A cells . GMA-GFP ( B ) was co-expressed with Ihog-RFP ( B’ ) in Hh-producing cells . Observe in the merge panel ( B’’ ) that Ihog-containing filopodia are stabilized while few filopodia with low or no Ihog levels detected are more dynamic . Histoblasts move up ( A ) and down ( B ) towards the dorsal midline , and this is not shown because of the high magnification . Anterior is to the left . Pupae were around 30 hr APF ( after puparium formation ) . Movies of 30 min imaging with time intervals between frames of 2 min . Scale bars represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 008 The maximum extent of filopodia is similar with or without overexpression of Ihog ( Figure 1C , Figure 1—figure supplement 2 ) . This overexpression , however , changes cytoneme dynamics . Filopodia lifetime values are higher for filopodia with high levels of Ihog than those with low levels ( Figure 1D , Figure 1—figure supplement 2 ) . Hh-producing and Hh-receiving cell filopodia elongate and retract at similar rates ( Figure 1E , Figure 1—figure supplement 2A ) . In summary , Ihog overexpression leads to more stable cytonemes also in the A compartment cells . To further analyse the role of A compartment cytonemes , we manipulated actin cytoskeletal components implicated in filopodia formation within the wing imaginal disc , monitoring the expression of Ptc and Cubitus interruptus ( Ci ) as a read-out of Hh signalling . We knocked down SCAR , Diaphanous ( Dia ) , Filamin ( Cheerio , Cher ) or actin Capping protein alpha ( Cpa ) function in the A compartment using specific RNAi lines . In addition , we expressed these RNAis together with ihog-RFP to visualize cytonemes . The aberrant extension of cytonemes observed by knocking down these proteins in the Hh-receiving cells resulted in reduced Ptc and Ci expression domains ( Figure 1F ) , revealing a role of the A compartment cytonemes in regulating the Hh gradient . Thus , cytonemes control both Hh export from P compartment and Hh distribution and reception in the A compartment . Cytoneme-mediated long distant communication between Hh-producing and Hh-responding cells is visualized when Ihog-RFP marked cytonemes from the A and P compartments meet at the A/P boundary in the basal part of the epithelium ( Figure 1G , arrow; Video 3 ) . Anterior compartment cells overexpressing Ihog consistently extend basal cytonemes , which invade the P compartment and heavily accumulate Hh ( Figure 1H , arrow; Video 4 ) , leadingto an extended Hh gradient ( Figure 1—figure supplement 3 ) . Taken together , these results indicate that Hh-receiving cells generate dynamic cytonemes directed towards Hh-producing cells , and that these A cytonemes can influence the graded distribution of Hh . 10 . 7554/eLife . 24045 . 009Video 3 . Membrane contacts between A and P cell cytonemes . Z-stack , from apical to basal , of a wing disc showing the GRASP signal pattern between Hh-producing and Hh-receiving cells ( ptc . Gal4 , tub . Gal80ts>UAS . CD4-GFP1-10/ hh . LexA>LexAop . CD4-GFP11 wing disc after 24 hr at the restrictive temperature ) . Note the different GFP pattern in apical and basal sections . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 00910 . 7554/eLife . 24045 . 010Video 4 . Ihog overexpressing cytonemes emerging from the A compartment cells accumulate endogenous Hh produced by the P compartment cells . Z-stack , from apical to basal , of a ptc . Gal4 , tub . Gal80ts>UAS . ihogRFP wing disc after 24 hr at restrictive temperature , stained with α-Hh antibody . Note the increase of Hh levels at the basal A compartment cytonemes . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 010 To further study the involvement of cytonemes in Hh reception we blocked endocytosis specifically in Hh-receiving cells by expressing shiK44A , which encodes a dominant negative mutant of Dynamin ( aka Shibire , Shi ) . Expression of this mutant protein blocks the rapid internalization of the Hh-Ptc complex in signal-receiving cells . This approach enables the visualization of ligand and receptor at the site of signal reception ( Callejo et al . , 2011; Torroja et al . , 2004 ) . In this mutant background , Hh co-localizes with Ptc at the basal membrane of Hh-receiving cells , revealing the presence of filopodia-like structures oriented in the A-P axis ( Figure 2A , arrowheads ) . Extracellular staining of Ptc and Hh confirms this basal localization of Hh-Ptc complex for signal reception ( Figure 2A’ , arrow ) . Furthermore , cytonemes from Hh-producing cells ( labelled with Ihog-RFP ) invade the A compartment to contact the basal membrane of the Hh reception region ( labelled with Ptc ) in A cells ( Figure 2B arrowheads and 2B’ arrow ) . We can also detect the receptor Ptc localized in A cytonemes when its internalization is impeded in the Hh-receiving cells ( Figure 2C and C’ , arrow ) or co-localizing with P compartment Ihog-RFP labelled cytonemes ( Figure 2—figure supplement 1 ) . These results indicate that A compartment cells receive the Hh signal through their cytonemes , which also contain Ptc . 10 . 7554/eLife . 24045 . 011Figure 2 . Cytonemes from A and P compartment cells interact during Hh reception . ( A ) 3D view of a ptc . Gal4 , tub . Gal80ts>UAS . shiK44A wing disc after 12 hr at the restrictive temperature and double stained for Hh and Ptc . Note that Ptc colocalizes with Hh at the basal membranes of the Hh-receiving cells when endocytosis is frozen using a dominant negative form of Dynamin ( arrowheads ) . ( A’ ) Transverse section of an extracellular staining for both Hh and Ptc in a similar wing disc . Note the colocalization of Hh and Ptc in the most basal part of the disc ( arrows ) . ( a’ ) Magnification of A’ . ( B ) 3D view of a ptc . Gal4 , tub . Gal80ts>UAS . shiK44A / hh . LexA>LexAop . ihog-RFP wing disc after 12 hr at the restrictive temperature and labelled with α-Ptc antibody . Note Ihog-RFP localization in the P compartment cytonemes and Ptc in the A compartment cytonemes . ( B’ ) Transverse section of a similar wing disc stained for extracellular Ptc . Note the colocalization of Ihog and Ptc in the most basal part of the epithelium ( arrows ) . ( b’ ) Magnification of B’ . ( C , C’ ) Basal ( C ) and transverse ( C’ ) sections of a ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP>UAS . shiK44A wing disc after 12 hr at the restrictive temperature and stained with α-Ptc antibody . ( c’ ) Magnification of C’ . Note that when endocytosis is frozen , Ptc is located in A compartment cytonemes labelled with Ihog-RFP ( arrow ) . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01110 . 7554/eLife . 24045 . 012Figure 2—figure supplement 1 . Cytonemes from A and P compartment cells interact during Hh reception . ( A ) Basal section of a ptc . Gal4 , tub . Gal80ts>UAS . shiK44A / hh . LexA>LexAop . ihog-RFP wing disc after 12 hr at the restrictive temperature before dissection and co-labelled with α-Hh and α-Ptc antibodies . Note Ihog-RFP accumulated in P compartment cytonemes and Ptc and Hh in A compartment cytonemes . ( A’ ) Graph showing a plot of fluorescence intensity of Ihog , Hh and Ptc along the basal surface of the wing disc in A . ( B ) A hh . LexA>LexAop . ihog-RFP wing disc showing Ihog-RFP in P cytonemes when internalization is not blocked in the A compartment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 012 We then expressed Hh-CD2 protein , which cannot be released from the plasma membrane and is therefore unable to form a concentration gradient ( Strigini and Cohen , 1997 ) . As expected Hh-CD2 localizes to basal cytonemes ( Figure 3—figure supplement 1A ) , since this form of Hh is permanently anchored to the plasma membranes . Therefore , the Hh-CD2 protein expression allowed us to visualize interactions between cytonemes emanating from the A compartment ( expressing Ihog-RFP ) and cytonemes from the P compartment ( expressing Hh-CD2 ) along their lengths ( Figure 3A , arrowheads; Figure 3—figure supplement 1B ) . Hh-receiving cytonemes accumulate the co-receptor Ihog at contact sites with Hh-CD2-presenting cytonemes ( Figure 3—figure supplement 1C and C’ , compare C with D ) . Noticeably , the Hh receptor Ptc is also localized in these contact sites between cytonemes from A and P compartment cells ( Figure 3B , arrows; Figure 3—figure supplement 1C’ and E ) . These results suggest that the Hh receptor complex binds Hh-CD2 protein at the basal membrane of Hh signal-receiving cytonemes , while blockage in Hh release and internalization results in Ptc also being retained in the cytoneme membrane ( Figure 3—figure supplement 1E , compare E with F ) . Taken together , these data indicate that Hh reception is mediated by contacts between Hh-sending and Hh-receiving cytonemes . 10 . 7554/eLife . 24045 . 013Figure 3 . Hh reception at contacts between Hh-sending and Hh-receiving cytonemes . ( A ) A dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . hh-CD2 wing disc grown for 24 hr at the restrictive temperature and labelled with α-CD2 antibody . Observe that Hh-receiving and Hh-CD2-presenting cytonemes interact along their lengths . Higher levels of Ihog are in those A cytonemes that interact with P cytonemes at both sides of the A/P compartment border ( white and yellow arrowheads ) , probably due to the higher stability of these cytonemes . ( B ) Basal section of a similar wing disc labelled with α-Ptc antibody . Observe the accumulation of Ptc and Ihog at these A cytonemes ( arrows ) , probably because Ptc does not internalize Hh-CD2 so that the whole internalization complex is accumulated at the Hh-receiving cytonemes in the A compartment . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01310 . 7554/eLife . 24045 . 014Figure 3—figure supplement 1 . Hh reception at contacts between Hh-sending and Hh-receiving cytonemes . ( A ) The membrane-tethered Hh-CD2 protein is localized in cytonemes in a tub . Gal80ts , hh . Gal4>UAS . hh-CD2 wing disc . ( B ) A compartment cytonemes interact with Hh-CD2-presenting cytonemes labelled with the actin marker lifeactin-GFP and with α-CD2 antibody in a dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . hh-CD2 , UAS-lifeactin-GFP wing disc . ( C ) Ihog-RFP is accumulated in A compartment cytonemes that interact with Hh-CD2-presenting cytonemes in a dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . hh-CD2 wing disc . Note Ihog accumulation in cytonemes crossing the A/P compartment border from A to P and from P to A . ( C’ ) Graph showing a plot of fluorescence intensity of Ihog-RFP , Ptc and Hh ( labelled with α-Hh and α-Ptc antibodies ) along the basal surface of a wing disc similar to the one in C . ( D ) A dpp . LexA>LexAop . ihog-RFP wing disc to show that Ihog-RFP is not significantly accumulated in the A compartment cytonemes when Hh-CD2 is not present . ( E ) Longitudinal section at the A/P compartment border of wing disc with the same genotype as in C , co-labelled with α-Hh and α-Ptc antibodies . Note the accumulation of Ptc , Ihog and Hh-CD2 at the most basal part of the disc epithelium ( arrows ) . ( F ) Longitudinal section at the A/P compartment border of a dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . hh-GFP wing disc to compare the behaviour of Hh-GFP with Hh-CD2 shown in E . All larvae ( A–F ) were grown 24 hr at the restrictive temperature before dissection . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 014 To further explore the mechanism of contact between A and P cytonemes we adapted the GRASP technique ( GFP Reconstitution Across Synaptic Partner ) , which was originally developed to image membrane contacts at neuronal synapses ( Feinberg et al . , 2008; Gordon and Scott , 2009 ) . We used UAS . CD4-GFP1-10 with the complementary CD4-GFP11 fragment regulated by LexAop ( LexAop . CD4-GFP11 ) . Membrane contacts between Hh-sending and Hh-receiving cells were imaged by expressing the complementary GFP fragments separately in each compartment using the required LexA and Gal4 drivers . As expected , GFP fluorescence was localized to the A/P border when complementary fragments were expressed at the same time in each compartment . The pattern of GFP fluorescence varies along the apico-basal axis , as both A and P cells extend cytonemes along the basal surface of the epithelium ( Figure 4A , B , B’; Video 5 ) . Cytoneme membrane contacts occur on both sides of the A/P compartment border , although predominantly on the A side ( Figure 4C ) . Furthermore , periodic annular structures are present along the length of overlapping cytoneme ( arrows in Figure 4D and E ) , while the GRASP signal also indicates that cytonemes overexpressing Ihog-RFP are indeed in contact with neighbouring cell membranes , presumably receiving cytonemes ( Figure 4E , arrows ) . Importantly , both Hh and Ptc co-localize in these basal membrane contacts ( detected by GRASP signal ) upon blockage of their uptake in a shiK44A mutant background ( Figure 4F , arrowheads ) . Therefore , we propose that these contacts are sites where Hh-Ptc interaction takes place all along the length of A and P cytonemes for signal reception ( see diagram Figure 4G ) . 10 . 7554/eLife . 24045 . 015Figure 4 . GRASP showing cytoneme-cytoneme interaction at the A/P compartment border . ( A ) 3D view of a ptc . Gal4 , tub . Gal80ts>UAS . CD4-GFP1-10 / hh . LexA>LexAop . CD4-GFP11 wing disc after 24 hr at the restrictive temperature . ( B , B’ ) A wing disc of the same genotype as in A co-labelled with α-Ptc and α-Hh antibodies . Note the GFP complementation at the A/P compartment border in a lateral ( B ) and a basal section ( B’ ) . ( C , D ) Basal sections showing GRASP fluorescence in similar wing discs . Cytonemes cross the A/P compartment border from A to P and from P to A ( C ) . Annular rings are visualized along these interacting cytonemes ( D , arrows ) . ( E ) A ptc . Gal4 , tub . Gal80ts>UAS . CD4-GFP1-10 / hh . LexA>LexAop . CD4GFP11>LexAop . ihogRFP wing disc after 24 hr at the restrictive temperature to visualize Ihog labelled cytonemes emanating from the P compartment . Note the GRASP signal in the circular structures attached to cytonemes ( arrows ) . ( F ) A ptc . Gal4 , tub . Gal80ts>UAS . CD4-GFP1-10>UAS . shiK44A / hh . LexA>LexAop . CD4-GFP11 wing disc after 12 hr at the restrictive temperature and co-labelled with α-Hh and α-Ptc antibodies . Note the colocalization of GRASP , Ptc and Hh in the same structures at the most basal part of the disc ( arrowheads ) . ( G ) Diagram depicting cytoneme interactions at the A/P compartment border . The green colour corresponds to the GRASP signal at the A/P compartment border and also at the specific sites for Hh reception along overlapping cytonemes . The data shown were consistent in at least four independent experiments with an average of 8 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01510 . 7554/eLife . 24045 . 016Video 5 . Membrane contacts between A and P compartment cells . Z-stack , from apical to basal , of a wing disc showing the GRASP signal pattern between Hh-producing and Hh-receiving cells ( of a ptc . Gal4 , tub . Gal80ts>UAS . CD4-GFP1-10/ hh . LexA>LexAop . CD4-GFP11 wing disc after 24 hr at the restrictive temperature ) . Note the different pattern in apical and basal sections . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 016 It has been previously described that clones of ptc−/− and smo−/− located at the A/P compartment border are unable to internalize Hh , but do not block Hh signal transmission to more anterior wild type cells ( Chen and Struhl , 1996 ) . Our cytoneme transport hypothesis predicts that Hh could be transported across mutant ptc−/− and smo−/− clones in cytonemes that act as conveyer belts for long distance transport . Accordingly , we observed that Hh-producing cells still spread cytonemes when Hh reception was prevented in the A compartment by knocking down either Ptc or Smo , or both co-receptors Ihog and Brother of Ihog ( Boi ) at the same time ( Figure 5—figure supplement 1 ) . In addition , cytonemes from both compartments were still formed upon the elimination of Hh ( Figure 5—figure supplement 2A–F ) using the hh thermosensitive allele ( hhts2 ) ( Ma et al . , 1993 ) . No apparent changes in Hh-receiving cytonemes were also observed when Hh reception was prevented in ptc−/− MARCM clones expressing Ihog-YFP ( Figure 5—figure supplement 2G ) . Therefore , the presence of the network of cytonemes is independent of Hh signaling . We then asked if cytonemes could facilitate the Hh transport across a ptc mutant territory . Hh signalling is fully activated in ptc−/− mutant clones due to the lack of normal Ptc repression upon Smo ( Chen and Struhl , 1996 ) . We show here that Hh protein can be detected in the wild type cells anterior to the ptc−/− mutant clone ( Figure 5A ) . Accordingly , with previous experiments ( Figure 5—figure supplement 1B ) , cytonemes emanating from Hh-producing cells orient normally towards the A compartment cells across a ptc mutant clone induced adjacent to the A/P border in the wing disc ( Figure 5B ) . Consistently , abdominal histoblasts of the P compartment produce cytonemes that cross a ptc−/− clone in their way towards the anterior cells ( Figure 5C , Video 6 ) . These cytonemes have similar dynamics ( maximum extent , lifetime and elongation and retraction velocities ) when crossing wild type or ptc−/− mutant territories ( Figure 5C and C’ , Figure 5—figure supplement 3 ) . Therefore , cytoneme dynamics is independent of the presence of the receptor Ptc . 10 . 7554/eLife . 24045 . 017Figure 5 . Cytonemes cross ptc−/− mutant clones at the A/P compartment border . ( A ) A ptc16 null clone in the A compartment abutting the A/P compartment border of a wing disc co-labelled with α-Hh and α-Ptc antibodies . Note that both Ptc and Hh proteins are detected anterior to the clone in the A compartment ( arrowheads ) . ( B ) P compartment cytonemes extend through a ptc−/− clone ( arrowheads ) . Lateral and basal sections of a ptc−/− clone ( absence of βGal ) in the A compartment with Ihog-RFP expression in the P compartment to visualize cytonemes ( FRT42D ptc16 / hh . Gal4 , tub . Gal80ts>UAS . ihog-RFP wing disc after 24 hr at the restrictive temperature ) co-labelled with α-βGal and α-Ptc antibodies . ( C ) First and last time frames from Video 6 displaying a lateral view of GMA-GFP signal to easily visualize cell perimeters together with a Z-projection of GMA-GFP where filopodia are shown ( top panels ) or with a lateral view of nuclear RFP to distinguish between wild-type ( magenta nuclei ) and ptc−/− mutant ( absence of magenta nuclei ) territories ( bottom panels ) . Scale bars represent 10 μm . ( C’ ) Graph showing extent distribution over time of GMA-GFP filopodia emanating from Hh-producing cells . Notice that there is no difference between filopodia crossing a wild-type ( magenta ) or a ptc−/− mutant ( grey ) clone territories . ( D ) A ptc16 clone induced in the A compartment abutting the compartment border in a dpp . LexA>LexAop . CD4-GFP11>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . CD4-GFP1-10 wing disc after 24 hr at the restrictive temperature and labelled with α-βGal antibody to identify the mutant clone ( absence of βGal ) . Note the GRASP signal is not visualized laterally ( asterisk ) . Note also that basal cytonemes cross the ptc−/− clone in region 1 but not in region 2 , and that the GRASP signal is restricted to region 1 and absent from region 2 ( arrowheads ) . E ) Another ptc−/− clone induced in a disc with the same genotype as in D showing the interaction between cytonemes from wild type cells anterior to the clone and cytonemes from the P compartment . Note the GRASP signal along basal cytonemes from wild type A compartment cells that traverse the ptc−/− clone ( arrows ) . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01710 . 7554/eLife . 24045 . 018Figure 5—figure supplement 1 . Cytonemes from Hh-producing cells are still formed in mutants that modify Hh reception . ( A , A’ ) A hh . LexA>LexAop . ihog-RFP wing disc labelled with α-Ci antibody to show Hh signalling ( A’ ) . ( B , B’ ) A ptc . Gal4 , tub . Gal80ts>UAS . ptc-RNAi; hh . LexA>LexAop . ihog-RFP disc labelled with α-Ci antibody to monitor the full activation of the Hh pathway in the absence of Ptc by the low levels of Ci ( B’ ) . Note that cytonemes remain similar independently of the presence of Ptc in the A compartment ( A , B ) . ( C , C’ ) A ptc . Gal4 , tub . Gal80ts >UAS . smo-RNAi / hh . LexA>LexAop . ihog-RFP wing disc stained with α-Smo antibody to monitor the degree of Smo knock down . Note that cytonemes are similar with ( A ) or without ( C ) the presence of Smo in the A compartment . ( D , D’ ) A ptc . Gal4 , tub . Gal80ts >UAS . ihog-RNAi>UAS . boi-RNAi / hh . LexA>LexAop . ihog-RFP wing disc co-labelled with α-Ptc and α-Ci antibodies to monitor the low levels of Hh signalling in the absence of Hh co-receptors Ihog and Boi . Note that cytonemes are similar with ( A ) or without ( D ) the presence of both Ihog and Boi in the A compartment . All larvae ( A–D ) were grown 30 hr at the restrictive temperature before dissection . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01810 . 7554/eLife . 24045 . 019Figure 5—figure supplement 2 . Cytonemes are still present in absence of Hh ligand . ( A ) A dpp . LexA>UAS . ihog-RFP wing disc stained withα-Ci antibody to monitor the activation of the Hh pathway . ( B ) A dpp . LexA>LexAop . ihog-RFP / hhts2 wing disc shows by the low levels of Ci in the A compartment the lack of activation of the Hh pathway . Note that the A compartment cytonemes are not affected in the absence of Hh . ( C ) An en . Gal4>UAS . ihog-RFP wing disc stained with α-Ci antibody . ( D ) An en . Gal4>UAS . ihog-RFP / hhts2 wing disc where the absence the Hh signal is monitored by the low levels of Ci activation . ( E ) An en . Gal4>UAS . CD4-tomato wing disc stained with α-Ci antibody . ( F ) An en . Gal4>UAS . CD4-tomato / hhts2 wing disc and stained with α-Ci antibody to check the absence the Hh signal . Note that the P compartment cytonemes labelled either by Ihog ( D ) or CD4 ( F ) are not affected by the absence of Hh . ( G ) MARCM ptc16 UAS . ihog-YFP clones induced in A ( marked with yellow line ) and P compartments of a wing disc and stained with α-Ptc antibody to see the knock down of Ptc and the localization of the anterior clone at the A/P compartment border . Observe that the absence of Ptc does not affect the formation of cytonemes autonomously . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 01910 . 7554/eLife . 24045 . 020Figure 5—figure supplement 3 . Manual tracking of filopodia arising from Hh-producing abdominal histoblasts cells crossing either a ptc mutant or a wild-type territories . ( A ) Table showing the maximum extent ( Emax ) , lifetime , and elongation ( Ve ) and retraction ( Vr ) velocities of filopodia from Hh-producing cells crossing wild type and ptc−/− territories to the A compartment . ( B , C ) Violin plots representing the filopodia Emax ( B ) and lifetime ( C ) . Notice that there are not significant differences between filopodia crossing ptc mutant or wild-type territories ( n . s . ) . ( D , E ) Violin plots representing the Ve ( D ) and Vr ( E ) as trapezoid ( T ) or triangle ( t ) models . ( F , G ) Percentage graphs representing the type of elongation ( F ) and retraction ( G ) velocities of filopodia crossing ptc mutant ( ptc−/− ) or wild-type ( WT ) territories . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 02010 . 7554/eLife . 24045 . 021Video 6 . Cytonemes from the P compartment abdominal histoblasts are able to cross ptc mutant clones and their dynamics do not differ from those crossing wild-type territories . ( A-C ) Abdominal histoblasts of a pupa with the genotype y w hs-Flipase122; FRT42D ptc16 / FRT42D ubiRFPnls; hh . Gal4 / UAS . GMA-GFP . GMA-GFP is expressed during all development at 18°C and ptc16 clones were induced doing a heat-shock of 1 hr at 37°C two days before imaging . ( A ) Merge of nuclear RFP ( magenta ) and a Z-projection of GMA-GFP ( inverted grey-scale lookup table ) . Hh-producing cell cytonemes labelled with the actin marker GMA ( inverted grey-scale lookup table ) have the same dynamics and cross normally wild-type ( nuclear RFP , magenta ) and ptc16 mutant clone ( absence of magenta ) territories . ( B ) Merge of the Z-projection of GMA ( inverted grey-scale lookup table ) and the lateral side ( green ) to show the morphology of the epithelium . ( C ) Merge of nuclear RFP ( magenta ) and GMA in a lateral view ( green ) . Here we visualize the ptc16 mutant clone ( absence of magenta ) anterior to the A/P boundary . Histoblasts move down towards the dorsal midline , and this is not shown because of the high magnification . Anterior is on the left . Pupa is around 30 hr APF ( after puparium formation ) . Movie of 30 min imaging with a time interval between frames of 30 s . Scale bars represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 021 In addition , GRASP signal indicates that cytonemes from the Hh-receiving cells reach the Hh-sending cytonemes close to the A/P compartment border ( Figure 5D , region 1 arrowhead ) . So that , GRASP signal is detected along these A cytonemes that cross the clone and interact with the P compartment cytonemes ( Figure 5E , arrows ) . However , in subapical regions of the epithelium , where cytonemes are too short to reach the P compartment , GRASP fluorescence is not detected ( Figure 5D , asterisk ) . It is also known that when a ptc−/− clone is too large , wild type anterior cells are not able to respond to Hh ( Chen and Struhl , 1996 ) . Noticeably , in this situation we observe that cytonemes extending from the A cells do not reach the P cells and no GRASP fluorescence is detected ( Figure 5D , region 2 arrowhead ) . In smo−/− clones originating in the A compartment , and monitored by low levels of Ci expression , the A/P compartment border is re-established in the position where Hh signalling is activated in cells anterior to the clone ( Blair and Ralston , 1997; Rodriguez and Basler , 1997 ) . As a result , A compartment smo−/− cells integrate into the P compartment and maintain a straight boundary at the nearest anterior cells where Hh is then received . The Hh visualized within smo−/− clones is restricted to the most basal region of epithelial cells , demonstrating again that Hh travels mostly basally ( Figure 6A , arrowhead ) ( Bilioni et al . , 2013; Bischoff et al . , 2013; Callejo et al . , 2011 ) . In addition , cytonemes emanating from cells anterior to the clone are loaded with Hh ( Figure 6B , arrowhead ) , and P cytonemes cross a smo−/− clone reaching the A/P compartment border ( Figure 6C , arrow ) . Accordingly , GRASP fluorescence is detected where the compartment border is re-established anterior to a smo−/− clone ( Figure 6D , arrow ) . 10 . 7554/eLife . 24045 . 022Figure 6 . Cytonemes cross smo−/− mutant clones located at the A/P compartment border . ( A ) A smo3 mutant clone in a wing disc co-labelled with α-Hh , α-Ci and α-βGal to mark the clone ( absence of βGal ) . The smo−/− clone has an A compartment origin as it expresses low levels of Ci and shows no Hh localization in apical section . Interestingly , in a basal section of this disc Hh is visualized in Ci expressing cells ( arrowhead ) , indicating that Hh moves basally and not apically through a smo3 mutant clone originated in the A compartment . ( B ) A smo3 mutant clone induced in a wing disc that also expresses Ihog-RFP ( smo3 FRT40 / ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP ) after 24 hr at the restrictive temperature to visualize cytonemes emanating from cells anterior to the clone . Note that Hh is present in cytonemes from cells located anterior to the clone ( arrowheads ) . ( C ) A smo3 mutant clone ( absence of GFP ) induced in the A compartment abutting the A/P compartment border in a disc expressing ihog-RFP in the P compartment ( smo3 FRT40A / hh . Gal4 , tub . Gal80ts>UAS . ihog-RFP wing disc after 24 hr at restrictive temperature before dissection ) . Note that cytonemes emanating from P compartment cells cross along the smo−/− mutant clone ( arrows ) . ( D ) A smo3 clone of A compartment origin , identified by low levels of Ci expression , induced in a background to visualize the GRASP signal ( smo3 FRT40A / ptc . Gal4 , tub . Gal80t >UAS . CD4-GFP1-10 / hh . LexA>LexAop . CD4-GFP11 wing disc after 24 hr at the restrictive temperature ) . Note that cells of the clone are integrated in the P compartment and the GRASP signal is located anterior to the smo−/− clone ( arrow ) . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 022 All together these results indicate that Hh can be transported and received by cytonemes from P and A compartment , respectively , to cross ptc−/− and smo−/− mutant territories and re-establish the morphogen gradient . Heparan sulphate proteoglycans ( HSPGs ) are key players in cell-matrix interaction , cell-cell signalling , and long-range Hh function ( Lin , 2004; Tabata and Takei , 2004 ) . We have previously demonstrated that P compartment cytonemes do not normally extend along mutant HSPG clones located at the A/P compartment border ( Bischoff et al . , 2013 ) . Here we analyse the potential role of HSPG glypicans , in cytoneme function . The Drosophila glypicans Dally and Dlp are required for both Hh presentation and reception ( Ayers et al . , 2010; Bilioni et al . , 2013; Han et al . , 2004a; Yan et al . , 2010 ) . We observe that extension of Ihog-RFP expressing cytonemes is significantly reduced when glypicans levels are lowered by knocking down Dally or Dlp , in the opposite compartment ( Figure 7A–C , arrows; Figure 7—figure supplement 1A–D , arrows ) . In agreement with this observation , cytonemes are rarely detected crossing through dally and dlp double mutant clones ( Figure 7D , arrowhead ) . In addition , cytonemes emanating from the A compartment cells accumulate Dally and Dlp present in the neighbouring P compartment cells ( Figure 7E , arrow; Bilioni et al . , 2013 ) . This effect is observed by expressing Ihog-RFP and eliminating Dlp in the same compartment; Dlp of one compartment accumulate in labelled cytonemes emanating from other one ( Figure 7F , arrow ) . This dependence on the glypicans present in the neighbouring cells is also observed for the P compartment cytonemes ( Figure 7—figure supplement 1E and F , arrows ) . In summary , all these results demonstrate the requirement of glypicans for cytoneme spreading and their specific role in the cytoneme-mediated interplay between Hh-producing and Hh-receiving cells . 10 . 7554/eLife . 24045 . 023Figure 7 . Interaction with glypicans is required for cytoneme stabilization by Ihog in Hh-receiving cells . ( A , B , C ) Ihog-RFP labelled cytonemes arising from the A compartment cells ( A , dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts ) are dependent on the glypicans Dlp ( B , dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . dlp-RNAi ) and Dally ( C , dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . dally-RNAi ) levels in the P compartment ( arrows ) . All larvae were grown 30 hr at the restrictive temperature before dissection . ( D ) A dally32 dlp20 FRT2A double mutant clone ( absence of GFP ) induced in the P compartment and touching the A/P compartment border in a wing disc that expresses ihog-RFP in the receiving cells ( ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP ) to visualize cytonemes . Note the loss of cytoneme visualization crossing the clone ( arrowhead ) . ( E ) Endogenous Dlp is accumulated in A compartment cytonemes expressing Ihog-RFP ( ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP ) after 30 hr at the restrictive temperature ( arrows ) . ( F ) Wing disc showing that the endogenous Dlp accumulated in A compartment cytonemes belongs to the P compartment cells ( arrows ) , since Dlp has been knocked down in the A compartment ( ptc . Gal4 , tub . Gal80ts>UAS . ihog-RFP>UAS-dlp-RNAi ) after 30 hr at the restrictive temperature before dissection . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 02310 . 7554/eLife . 24045 . 024Figure 7—figure supplement 1 . Trans interaction with glypicans is required for cytoneme stabilization by Ihog in Hh-producing cells . ( A , B , C ) Ihog-RFP labelled cytonemes arising from the P compartment cells ( A , hh . LexA>LexAop . ihog-RFP / tub . Gal80ts ) are dependent on the glypicans Dlp ( B , ptc . Gal4>UAS . dlp-RNAi , hh . LexA>LexAop . ihog-RFP / tub . Gal80ts , ) and Dally ( C , ptc . Gal4>UAS . dally-RNAi hh . LexA>LexAop . ihog-RFP / tub . Gal80ts ) levels in the A compartment ( arrows ) . All larvae were grown 30 hr at the restrictive temperature before dissection . ( D ) Graph representing the maximum extent of cytonemes emanating from the P compartment when A compartment cells are mutant either for Dlp or Dally compared with the control wing disc ( same genotypes as those show in A , B and C , respectively ) ( average of 5 discs in 3 independent experiments; error bars represent SDs ) . Note that the effect on cytoneme abrogation in the P compartment is weaker than in case of cytoneme abrogation in the A compartment ( Figure 7 ) . This result is because in this case Ihog is induced during the whole development ( hh . LexA is not repressed by Gal80 ) and the expression of the RNAi to knock down Dlp and Dally is under ptc . Gal4 that is target of the Hh pathway , which is compromised when Ihog is overexpressed in the P compartment ( Bilioni et al . , 2013 ) . ( E ) A hh . LexA>LexAop . ihog-RFP wing disc shows the accumulation of the endogenous Dlp in cytonemes extending from P to A compartment ( arrows ) . ( F ) A hh . Gal4 tub . Gal80ts >UAS . ihog-RFP>UAS . dlp-RNAi wing disc shows that Dlp accumulated in the P compartment cytonemes belongs to the A compartment cells ( arrows ) , because the endogenous Dlp of the P compartment has been knocked down . Larvae in E and F were for 30 hr at restrictive temperature before dissection . The data shown were consistent in at least three independent experiments with an average of 5–10 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 024 We then analysed the interaction between confronting cytonemes at the A/P compartment border when Ihog is overexpressed in one compartment and either Dally or Dlp overexpressed in the other one . Remarkably , we observe an accumulation of Ihog and glypicans along the membranes of the Hh-sending and Hh-receiving interacting cytonemes ( Figure 8A–D , arrowheads ) . Moreover , these contact sites present well-defined annular structures containing both Ihog and the glypicans Dally and Dlp ( Figure 8A’ and B’ , arrows ) similar to the ones visualized by GRASP ( Figure 4D and E , arrows ) . These results show again that Hh-sending and Hh-receiving cytonemes touch-contact along their lengths and , importantly , reveal a role for an Ihog/glypican interaction in trans during this cell-cell interface . 10 . 7554/eLife . 24045 . 025Figure 8 . Cytonemes from A and P compartments contact by Ihog-glypicans trans interaction . ( A ) Basal P compartment cytonemes expressing Ihog-RFP interact with A compartment cytonemes expressing Dally-GFP ( ptc . Gal4 , tub . Gal80ts>UAS . dally-GFP / hh . LexA>LexAop . ihog-RFP ) at both sides of the A/P compartment border ( white and yellow arrowheads ) . ( A’ ) Enlargement of a similar disc showing annular rings , labelled by Dally-GFP and Ihog-RFP , associated with interacting cytonemes ( arrows ) . ( B ) Basal A compartment cytonemes expressing Ihog-RFP interact with P compartment cytonemes expressing Dally-GFP ( dpp . LexA>LexAop . ihog-RFP / tub . Gal80ts , hh . Gal4>UAS . dally-GFP ) at both sides of the A/P compartment border ( white and yellow arrowheads ) . ( B’ ) Enlargement of a similar disc showing ring structures associated with interacting cytonemes , labelled by Dally-GFP and Ihog-RFP ( arrows ) . ( C ) 3D view of a similar wing disc showing this interaction between basal A and P cytonemes at both sides of the A/P compartment border ( white and yellow arrowheads ) . ( D ) 3D view of a ptc . Gal4 , tub . Gal80ts >UAS . dlp-GFP / hh . LexA>LexAop . ihog-RFP wing disc . P compartment cytonemes expressing Ihog-RFP interact with A compartment cytonemes expressing Dlp-GFP at both sides of the A/P border ( white and yellow arrowheads ) . The larvae from A-D were grown for 24 hr at the restrictive temperature before dissection . The data shown ( A–D ) were consistent in at least three independent experiments with an average of 5 discs in each experiment . Bars , 10 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 025
In a previous study we described that the P compartment cytonemes were critical for the Hh gradient establishment acting as conduits for Hh transport ( Bischoff et al . , 2013 ) . Here we have extended this work and showed that the cytonemes emanating from signal-receiving cells in the A compartment also have a critical role in Hh gradient formation . Therefore , both A and P compartment cytonemes weigh in the formation of the Hh gradient . The two sets of cytonemes make contact through the basal side of the epithelium and show similar values of average maximum extent , lifetime , and elongation and retraction velocities . A and P cytonemes contact predominantly on the anterior side of the compartment boundary . Furthermore , receiving cytonemes include components of the Hh reception complex ( Ptc , Ihog , Dlp and Dally ) , supporting their signalling role; consistently , we find that manipulating their actin dynamics modifies Hh graded signalling . As implied in our earlier study , signal transmission might be dependent on cytoneme stability and the frequency with which A and P cytonemes make contact . More Hh would be delivered to A cells close to the A/P compartment border than to those further from the signal source . In agreement with this hypothesis , GRASP fluorescence analysis suggests that A cytonemes from cells close to the A/P boundary establish contact primarily with the cytonemes from adjacent P compartment cells , where Hh would also be delivered by further P cells cytonemes . The sum of these spatial interactions would then result in high levels of Hh signalling close to the A/P border . The longest cytonemes from cells further from the A/P compartment border would interconnect to deliver the reduced levels of Hh signalling in more distant A cells . We analysed the behaviour of confronting A and P cytonemes with mutated or overexpressed Hh signalling components . Interestingly , the expression in the P compartment of membrane-tethered Hh ( Hh-CD2 ) , which results in a non graded signalling response ( Strigini and Cohen , 1997 ) , alters the receiving cytonemes cargo by hindering Ptc internalization . Despite that Hh receptor Ptc is usually rapidly internalized and not visualized at the plasma membrane , both Ptc and Ihog accumulate in A compartment cytonemes at contact sites with Hh-CD2 expressing P compartment cytonemes . Thus , this form of Hh attached to membranes plausibly affects the reception process , potentially explaining the absence of a signalling gradient and further confirming that Hh release from the presenting membranes is necessary . The mechanism by which Hh is liberated from plasma membrane anchoring lipids in Drosophila remains unknown . However , in strong support of a Hh shedding process , proteolytic removal of both lipidated peptides releases active Sonic hedgehog ( Shh ) from the surface of cultured cells ( Ohlig et al . , 2011 ) . On the other hand , Ihog trans interaction with glypicans in cytoneme-mediated cell-cell contacts has a critical role for cytoneme dynamics . The spreading capacity of Ihog overexpressing cytonemes is dependent on the glypicans presence in the membranes of neighbouring cells . Cytonemes protruding from one compartment aberrantly extend over the other compartment when is mutant for dally and dlp . This is in agreement with our previous results showing that the Tout-velu ( Ttv ) and Brother of Ttv ( Btv ) proteins ( needed for the HSPGs synthesis ) are required to stabilize Ihog cytonemes ( Bischoff et al . , 2013 ) . In addition , Dally or Dlp overexpressing cytonemes from one compartment seem to be stabilized by Ihog overexpressing cytonemes from the other one , with which they associate along their length by trans interaction between Ihog and glypicans . Thus , cytonemes from one compartment can contribute to the stability of the cytonemes from the other one with which they contact . However , to elucidate the specific mechanism of this altered cytoneme behaviour ( impairment in elongation , stability or retraction processes ) further in vivo analysis is required . Taken together , these results establish that dynamic interactions between A and P compartment cytonemes play an important role in the formation of the Hh gradient . In conjunction components for both reception and presentation of Hh critically influence the performance of this cytoneme interaction . The membrane-bound GRASP fluorescence shows that A and P compartment cytonemes overlap in a region spanning the A/P compartment boundary . Within this region , direct ‘kissing’ contacts appear along the membranes of interacting A and P cytonemes to connect the two sets of cytonemes . These contact points are associated with annular structures , which may represent more stable links and are reminiscent of synaptic buttons . Similar ‘synaptic contacts’ have been previously described between air sac primordium ( ASP ) and wing disc cells for Dpp morphogen transference ( Roy et al . , 2014 ) . ASP cells receive Dpp by emitting long cytonemes that contact with the target cell of the wing disc by their tips . This process was visualized by the same membrane-bound GRASP . Although Dpp uses a cytoneme-mediated contact-dependent mechanism for signalling at distance , the reception takes place where Dpp-receiving cytoneme tip contacts with the Dpp source cell body , in contrast with the cytoneme-cytoneme contacts here described for Hh . Nevertheless , the ‘synaptic contacts’ described for Dpp reception in ASP might be conceptually comparable with the ‘kissing rings’ distributed along the length of overlapping Hh-sending and Hh-receiving cytonemes . We have previously demonstrated that Hh moves along cytonemes in multivesicular bodies ( MVB ) to be secreted as extracellular vesicles known as exosomes ( Gradilla et al . , 2014 ) , which raises intriguing questions about the mechanism of interaction between exosomal Hh and its receptor Ptc in the A cytonemes . Interestingly , Ptc and Hh can be detected co-localizing with cytoneme contact sites , suggesting that they could represent specific sites for Hh reception and internalization . These data strengthen our proposal of cytoneme-mediated Hh morphogen distribution and suggest that Hh containing exosomes could be received at intimate contacts between P and A cytonemes . Finally , we have shown that cytonemes are able to cross ptc−/− and smo−/− clones , enabling the Hh transport from P to A cells for Hh signalling reestablishment in wild type cells located anterior to the clones . In agreement , it is known that the Hh signal can cross small mutant clones of ptc and smo to reach the adjacent wild type territory . Similar behaviour , has been described for ptc−/− or smo−/− clones induced in the A abdominal histoblast nests and attributed to different cell polarity and cell affinity ( Struhl et al . , 1997 ) . Therefore , Hh transport and reception by cytoneme-mediated cell-cell contacts can explain the Hh signalling in wild type cells located anterior to ptc−/− and smo−/− clones . In summary , the data strengthen our proposal of cytoneme-mediated Hh morphogen distribution and suggest that Hh containing exosomes could be received at intimate contacts between P and A cytonemes . The described intimate interaction between these two sets of cytonemes at membrane contact sites , where ligand and receptor complex colocalize , may facilitate morphogen reception .
A description of mutations , insertions and transgenes is available at Fly Base ( http://flybase . org ) . The following mutants and transgenic strains were used: tub . Gal80ts , hs-Flp122 ( Bloomington Drosophila Stock Centre ( BDSC ) , Indiana , USA; http://flystocks . bio . indiana . edu ) , hhts2 ( Ma et al . , 1993 ) , shits1 ( Grigliatti et al . , 1973 ) , dally32 ( Franch-Marro et al . , 2005 ) , dlp20 ( Franch-Marro et al . , 2005 ) , smo3 ( Nüsslein-Volhard and Wieschaus , 1980 ) and ptc16 ( Nakano et al . , 1989 ) . The following Gal4 and LexA drivers were used for ectopic expression experiments using the Gal4/UAS ( Brand and Perrimon , 1993 ) and LexA/LexAop ( Yagi et al . , 2010 ) systems: hh . Gal4 ( Tanimoto et al . , 2000 ) , ptc . Gal4 ( Hinz et al . , 1994 ) , dpp . LHG ( Yagi et al . , 2010 ) ( referred as dpp . LexA ) , hh . LexA ( generated at CBMSO department of development and differentiation ) , en . Gal4 ( expressed exclusively in the P compartment was a gift from Christian Dahman ) . The pUAS-transgene and LexAop-transgene strains were: UAS . hh-GFP ( Torroja et al . , 2004 ) , UAS . dlp-GFP ( Han et al . , 2004b ) , UAS . dally-GFP ( Eugster et al . , 2007 ) , UAS . shiK44A ( Moline et al . , 1999 ) , UAS . ihog-YFP and UAS . ihog-RFP ( Callejo et al . , 2011 ) , UAS . GMA-GFP ( Bloor and Kiehart , 2001 ) , UAS . CD4-tdTomato ( Han et al . , 2011 ) , UAS . CD4-GFP1-10 and LexAop-CD411 ( a gift from Kristin Scott ) , UAS . hh-CD2 ( Strigini and Cohen , 1997 ) , UAS . lifeactin-GFP ( BDSC 35544 ) , UAS . scar-RNAi ( BDSC 36121 ) , UAS . dia-RNAi ( BDSC 33424 ) , UAS . Cpa-RNAi ( Vienna Drosophila RNAi Center ( VDRC ) , Vienna , Austria , v16731 ) , UAS . cher-RNAi ( BDSC 35755 ) , UAS . ptc-RNAi ( BDSC 28795 ) , UAS . smo-RNAi ( BDSC 27037 ) , UAS . ihog-RNAi ( VDRC 102602 ) , UAS . boi-RNAi ( VDRC 108265 ) , UAS . dally-RNAi ( VDRC 14136 ) , UAS . dlp-RNAi ( VDRC 106578 ) . The LexAop . ihog-RFP construct used ihog-RFP from the pTWR vector ( Callejo et al . , 2011 ) , which was introduced into the pLOTattB plasmid ( Lai and Lee , 2006 ) carrying the lexA operator ( LexAop ) . Transgenic strains were recovered using standard protocols . Transient expression of transgenic constructs used the tub-Gal80ts; Gal4 and LexA systems with fly crosses maintained at 18°C , with inactivation of the Gal80ts repressor for 16–40 hr at restrictive temperature ( 29°C ) before dissection . The transgene actin<CD2<Gal4 ( Pignoni and Zipursky , 1997 ) was used to generate random ectopic clones of the UAS lines . Larvae of the corresponding genotypes were incubated at 37°C for 10 min to induce hs-Flp-mediated recombinant clones . Mutant clones were generated by hs-Flp-mediated mitotic recombination . Larvae were incubated at 37°C for 45 min at 48–72 hr after egg laying ( AEL ) . The genotypes were: Figure 5A: y w hs-Flp122; FRT42D ptc16 / FRT42D arm . lacZ . Figure 5B: y w hs-Flp122; FRT42D ptc16 / FRT42D arm . lacZ; UAS . ihog-RFP/ tub . Gal80ts , hh . Gal4 . Figure 5C: y w hs-Flp122; FRT42D ptc16 / FRT42D ubi . RFPnls; UAS . GMA-GFP / hh . Gal4 . Figure 5D , E: y w hs-Flp122; FRT42D ptc16 , UAS . CD4-GFP1-10 , LexAop . CD4-GFP11/ arm . lacZ FRT42D; dpp . LexA , tub . Gal80ts / LexAop . ihog-RFP , hh . Gal4 . Figure 6A: y w hs-Flp122; smo3 FRT40A / arm . lacZ FRT40A . Figure 6B: y w hs-Flp122; smo3 FRT40A , ptc . Gal4 / arm . lacZ FRT40A; UAS . ihog-RFP / tub . Gal80ts . Figure 6C: y w hs-Flp122; smo3 FRT40A / arm . lacZ FRT40A; UAS . ihog-RFP / tub . Gal80ts , hh . Gal4 . Figure 6D: y w hs-Flp122; smo3 FRT40A / arm . lacZ FRT40A , UAS . CD4-GFP1-10 , LexAop . CD4-GFP11; dpp . LexA , tub . Gal80ts / hh . Gal4 . Figure 7D: y w hs-Flp122; ptc . Gal4 , tub . Gal80ts / UAS . ihogRFP; dally32 dlp20 FRT2A / ubi GFP FRT2A . Video 6: y w hs-Flp122; FRT42D ptc16 / FRT42D ubi . RFPnls; UAS . GMA-GFP / hh . Gal4 . Immunostaining was performed according to standard protocols ( Capdevila and Guerrero , 1994 ) . Imaginal discs from third instar larvae were fixed in 4% paraformaldehyde ( PF ) in PBS for 20 min at room temperature ( RT ) and permeabilized in PBS 0 , 1% Triton ( PBT ) before incubating with PBT 1% BSA for blocking ( 1 hr at RT ) and primary antibody incubations ( overnight at 4°C ) . Incubation with fluorescent secondary antibodies ( 1/400 ThermoFischer ) was performed for 1 hr at RT and then washing and mounting in mounting media ( Vectashield ) . Primary antibodies were used at the following dilutions: rabbit polyclonal anti-Hh ( α-Hh ) ( Bilioni et al . , 2013 ) , 1:500; mouse monoclonal α-Ptc ( Capdevila and Guerrero , 1994 ) , 1:150; rat monoclonal α-Ci ( a gift from B . Holmgren ) , 1:20 , mouse monoclonal α-Dlp ( Lum et al . , 2003 ) , 1:30; mouse monoclonal α-Smo ( Hybridome bank ) , 1:30; rabbit polyclonal α-βGal ( from Jackson laboratories ) , 1/1000 . The protocol for the extracellular labelling using α-Hh and α-Ptc antibodies is described in ( Torroja et al . , 2004 ) . Imaginal discs from third instar larvae were dissected on ice , transferred immediately to ice-cold M3 medium containing α-Hh ( 1:30 dilution ) and α-Ptc ( 1:10 dilution ) antibodies and incubated at 4°C for 1 hr . The incubation with the primary antibody under these 'in vivo' conditions , without detergents prior to fixation , prevented antibody penetration of cells . Imaginal discs were then washed in ice-cold PBS , fixed in PBS 4% PF at 4°C , washed in PBT and incubated with secondary fluorescent antibody as above . Laser scanning confocal microscope ( LSM710 Zeiss ) was used for confocal fluorescence imaging of imaginal discs . ImageJ software ( National Institutes of Health ) was used for image processing and for image analysis . Gradient lengths were determined as described in Bischoff et al . ( 2013 ) . To quantify the maximum extent of cytonemes , we measured the ten longest protrusions in the wing pouch , from the A/P border to the tip , using the line tool in ImageJ . Then , the average length of the ten longest cytonemes was used for the quantitative analysis . Imaging of pupal abdominal histoblasts was done using a chamber as described in Seijo-Barandiarán et al . ( 2015 ) . Hh signalling filopodia from histoblasts of dorsal abdominal segment A2 were filmed using 40x magnification taking Z-stacks of around 30 μm of thickness with a step size of 1 μm every 2 min ( Video 1 and Video 2 ) or 30 s ( Video 6 and Video 7 ) using a LSM710 confocal microscope . All movies were analysed with Fiji and displayed at a rate of 7 frames per second . All imaged pupae developed into pharate adults and hatched normally . 10 . 7554/eLife . 24045 . 026Video 7 . Filopodia and ptc mutant region manual tracking using MTrackJ . ( A-C ) Abdominal histoblasts of a pupae with a y w hs-Flipase122; FRT42D ptc16 / FRT42D ubiRFPnls; hh . Gal4 / UAS . GMA-GFP genotype . GMA-GFP expressed during all development and two days ptc16 clones ( induced by heat-shock of 1 hr at 37°C ) . ( A ) Z-projection of GMA-GFP using the inverted grey-scale lookup table of ImageJ . ( B ) Z-projection of GMA-GFP and the tracking of filopodia using the MTrackJ plugin of Fiji , where the fist number refers to the number of filopodium tracked and the second name refers to the base ( 1 ) or the tip ( 2 ) . Each filopodium has a colour and the coloured lines are the base and tip trajectories . Notice that also two dying larval epithelial cells are tracked ( orange tracks #165 . 1 and #165 . 2 ) . ( C ) Nuclear RFP ( magenta ) wild-type nuclei and FRT42D ptc16 mutant clone ( absence of magenta ) tracked using the first and last region where there is no RFP-positive nuclei just anterior to the A/P boundary ( blue tracks #164 . 1 and #164 . 2 ) . Histoblasts move down towards the dorsal midline , and this is not shown because of the high magnification . Anterior is on the left . Pupa is around 30 hr APF ( after puparium formation ) . Movie of 30 min imaging with a time interval between frames of 30 s . Scale bars represent 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 24045 . 026 Signalling filopodia were tracked using MTrackJ plugin of ImageJ ( https://imagescience . org/meijering/software/mtrackj/ ) . We took base ( track#1 ) and tip ( track#2 ) point ( x , y ) coordinates in each frame for each signalling filopodium ( cluster ) and coloured them by cluster , showing filopodia in different colours . Signalling filopodia were analysed in a 30 min time-window in all movies . We discarded filopodia not clearly visible within the 30 min movies due to tissue movement or surrounding dying larval epithelial cells ( as shown in Video 6 ) . The tracking was done in a region of 49 μm x 76 μm for experiments expressing GMA in Hh-producing ( N = 4 pupae ) or Hh-receiving ( N = 4 pupae ) cells , and for experiments co-expressing GMA and Ihog in Hh-producing ( N = 4 pupae ) or Hh-receiving ( N = 4 pupae ) cells . Tracking was restricted to a region of 79 μm x 116 μm for the ptc−/− mutant clone experiment ( N = 1 pupa ) , where filopodia from the ptc−/− mutant territory and the wild-type territory was defined by the area references shown in Video 7 , and used for statistical analysis . Using the ( x , y ) coordinates of base and tip points of each filopodium given by the MTrackJ we calculated the elongation velocity , retraction velocity and lifetime as we describe below in ‘Filopodia dynamics models description’ . We also calculated filopodium extent through this formula: Filopodium extent ( E ) : E= ( xtip−xbase ) 2+ ( ytip−ybase ) 2 We have distinguished two populations of filopodial dynamic behaviours ( Figure 1—figure supplement 1 ) : We used R ( http://www . R-project . org ) to perform the Shapiro-Wilk normality test and pooled data of the studied variables was compared between different genotypes using the Mann-Whitney-Wilcoxon test of homogeneity of variances . For statistical analysis of Emax comparing Hh producing and receiving filopodia dynamics we used pooled data comparing 104 filopodia of 4 ptc . Gal4>UAS . GMA-GFP pupae , 100 filopodia of 4 hh . Gal4>UAS . GMA-GFP pupae , 47 filopodia of 4 ptc . Gal4>UAS . GMA-GFP , UAS . ihog-RFP pupae and 75 filopodia of 4 hh . Gal4>UAS . GMA-GFP , UAS . ihog-RFP pupae . Statistical analysis of 72 , 79 , 28 and 30 filopodia for Ve and 94 , 82 , 36 and 31 for Vr was done from ptc . Gal4>UAS . GMA-GFP; hh . Gal4>UAS . GMA-GFP; ptc . Gal4>UAS . GMA-GFP , UAS . ihog-RFP; and hh . Gal4>UAS . GMA-GFP , UAS . ihog-RFP pupae , respectively; because not all filopodia had an elongation or retraction phase during the recording period . For statistical analysis of Emax of Hh producing filopodia dynamics either crossing ptc−/− mutant or wild-type territories we compared 23 filopodia crossing ptc−/− mutant clone territory with 29 filopodia crossing wild-type territory of 1 pupa . For statistical analysis of Ve and Vr we used that data of 17 filopodia crossing wild-type territory and 14 filopodia crossing ptc−/− mutant territory in an elongation and a retraction phase . | When an embryo develops , it is critical that tissues and organs form properly and at the right time . For this , cells need to be able to communicate over long distances by using signalling molecules called morphogens . Morphogens disperse via extensions that protrude from the surface of a ‘source’ cell . Previous research has shown that these extensions called cytonemes can transport the morphogens to ‘receiver’ cells , and depending on the distance from the source , build a concentration gradient that will either be higher or lower . These gradients then help unspecialized cells to develop into different specialized ones . One of the key morphogens during the development is the Hedgehog protein . Researchers have previously shown that vesicles along cytonemes of cells that produce Hedgehog transport the morphogen to the receiver cells . However , until now it was unclear how the Hedgehog signals are transferred and received . Here , González-Méndez et al . – including researchers involved in the previous studies – investigated the cytonemes located on Hedgehog-receiving cells in the fruit fly . The results showed that these cytonemes are oriented towards the Hedgehog-producing cells and help to create a concentration gradient by varying their length . Moreover , the cytonemes from signal-producing and signal-receiving cells connect at specific sites that are distributed along their lengths . This suggests that the contact sites might help to transfer and receive the morphogens . Thus , the way cells communicate in other tissues of the body could be similar to how nerve cells communicate with each other in the brain . Our next challenges will be to fully understand how cytonemes transfer the Hedgehog signal . This could shed more light on how Hedgehog signaling can be controlled and modulated . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology"
] | 2017 | Cytoneme-mediated cell-cell contacts for Hedgehog reception |
Antigens ( Ags ) with multivalent and repetitive structure elicit IgG production in a T-cell-independent manner . However , the mechanisms by which such T-cell-independent type-2 ( TI-2 ) Ags induce IgG responses remain obscure . Here , we report that B-cell receptor ( BCR ) engagement with a TI-2 Ag but not with a T-cell-dependent ( TD ) Ag was able to induce the transcription of Aicda encoding activation-induced cytidine deaminase ( AID ) and efficient class switching to IgG3 upon costimulation with IL-1 or IFN-α in mouse B cells . TI-2 Ags strongly induced the phosphorylation of protein kinase C ( PKC ) δ and PKCδ mediated the Aicda transcription through the induction of BATF , the key transcriptional regulator of Aicda . In PKCδ-deficient mice , production of IgG was intact against TD Ag but abrogated against typical TI-2 Ags as well as commensal bacteria , and experimental disruption of the gut epithelial barrier resulted in fatal bacteremia . Thus , our results have revealed novel molecular requirements for class switching in the TI-2 response and highlighted its importance in homeostatic commensal-specific IgG production .
Ag-specific antibody production is essential for humoral immunity . After T-cell-dependent ( TD ) antigen ( Ag ) exposure , B cells are activated by interacting with cognate T cells and then proliferate , undergo class switching , and differentiate into plasma cells ( PCs ) and memory cells . In contrast , T-cell-independent ( TI ) Ags activate B cells without cognate help of T cells . TI type 1 ( TI-1 ) Ags engage Toll-like receptors ( TLRs ) in addition to the B-cell receptor ( BCR ) whereas TI-2 Ags extensively crosslink the BCR because of their highly repetitive structure ( Mond et al . , 1995 ) . In addition , some help from non-T cells , such as dendritic cells and innate lymphoid cells , may support B cells in a TI response ( Magri et al . , 2014; Balázs et al . , 2002 ) . Following activation , TI Ags induce proliferation , class switching , and antibody production by B cells . TI-2 Ags are large multivalent molecules , such as bacterial capsular polysaccharides and viral capsids , and thus antibody production against polysaccharides by the TI-2 response confers protection against disease ( Mond et al . , 1995; Lesinski and Westerink , 2001 ) . To date , however , the mechanism of B-cell activation in the TI-2 response is poorly understood as compared to that in TD and TI-1 responses . Engagement of the BCR by Ag activates various signaling cascades and promotes B-cell survival , proliferation and differentiation . The antibody production in a TI-2 response , but not in a TD response , depends upon proximal BCR signaling molecules such as Btk and BLNK ( Khan et al . , 1995; Xu et al . , 2000; Fruman et al . , 2000 ) , suggesting that BCR signaling plays a more critical role for B-cell activation in a TI-2 response than in a TD response . Given the highly repetitive structure of TI-2 Ags and the high demands for BCR signaling in TI-2 responses , TI-2 Ags seem to induce BCR signaling and the subsequent response more strongly than TD Ags . However , such functional differences between TD Ags and TI-2 Ags have not been investigated . Class ( isotype ) switching of the immunoglobulin ( Ig ) on B cells from IgM to either IgG , IgE , or IgA is caused by selective recombination of Igh constant region ( CH ) genes , namely class-switch recombination ( CSR ) . Selection of the CH gene to be recombined is determined by T-cell-derived cytokines in the TD response but the mechanism in the TI-2 response is not clear . Immunization with TI-2 Ags , including NP-Ficoll and bacterial polysaccharides , predominantly elicits IgG3 ( Perlmutter et al . , 1978; Slack et al . , 1980; Rubinstein and Stein , 1988 ) and IgG3 is required for protection against pneumococcal infection ( McLay et al . , 2002 ) . CSR absolutely requires activation-induced cytidine deaminase ( AID ) ( Muramatsu et al . , 2000 ) . AID deaminates deoxycytidine and introduces DNA double strand breaks ( DSBs ) in the switch ( S ) regions lying 5′ of the Cμ gene and the other targeted CH gene by triggering the DNA repair machinery ( Stavnezer et al . , 2008 ) . CSR proceeds through looping-out deletion of the DNA segment intervening between Sμ and the other S region from the chromosome and religation of the DSB-free ends in the two S regions . Consequently , it leads to replacement of the Cμ gene with a different CH gene downstream of the variable region exon in the Igh locus . In a TD immune response , stimulation with CD40L and cytokines , such as IL-4 and TGFβ , induce the expression of AID and subsequent CSR in the B cell ( Vaidyanathan et al . , 2014; Dedeoglu et al . , 2004 ) , while signaling through TLRs , BCR , and TACI induce the expression of AID and CSR in a TI-1 response ( Xu et al . , 2012 ) . On the other hand , the signaling and molecular demands for the induction of AID and CSR in the TI-2 response are far less understood , presumably due to the lack of in vitro studies that mimic a TI-2 response . Although TACI is required for IgG production in a TI-2 response ( von Bülow et al . , 2001 ) , stimulation of TACI alone induces the expression of AID and IgG production only modestly ( Castigli et al . , 2005 ) . Therefore , signaling through other receptors besides TACI seems to be required . As antibody production itself is disrupted in mice lacking proximal BCR signaling molecules ( Khan et al . , 1995; Xu et al . , 2000 ) , the requirement of BCR signaling and the downstream molecules for CSR remain unclear . Here , we report our finding that stimulation of the BCR with NP-Ficoll , a typical TI-2 Ag , triggered IgG CSR in the presence of secondary stimulation by IL-1 , IFNα , or TLR ligands in vitro and that protein kinase C ( PKC ) δ is critical for the Ag-mediated upregulation of AID and IgG production in the TI-2 response .
Considering the unique structure of TI-2 Ags , it is plausible that the engagement of the BCR with TI-2 Ags and TD Ags differently induces downstream signaling that leads to B-cell activation , although this idea has not been tested properly so far . We tested this in vitro by stimulating NP-specific B cells with the TD Ag NP-CGG or the TI-2 Ag NP-Ficoll to compare their ability to induce signaling , proliferation , and antibody production . NP-specific B cells were prepared from Igk−/− mice ( expressing only λ isotype light chain ) carrying a VH B1-8 knock-in gene encoding a VH region which binds to NP when coupled with a λ light chain . Although NP-CGG induced little proliferation and no IgM production , NP-Ficoll induced strong proliferation and IgM production , whereas neither induced IgG production ( Figure 1A and B ) . Thus , although NP-Ficoll alone can strongly activate B cells , additional stimulation seemed to be required for induction of class switching , similar to a previous report that anti-δ mAb/dextran and TLR ligands synergistically induced AID and CSR ( Pone et al . , 2012 ) . Indeed , NP-Ficoll induced IgG production , generation of IgG+ cells and Aicda transcription in the presence of TLR ligands , LPS , R-848 , or CpG , while NP-CGG did so marginally ( Figure 1—figure supplement 1A-C ) . Besides TI-1 Ags , such as TLR ligands , we sought to identify costimulating molecules that induce class switching in the TI-2 response , some of which were suggested previously ( Magri et al . , 2014; Balázs et al . , 2002 ) . Since IgG3 is the most dominant class-switched Ig isotype produced in a TI-2 response , we screened various cytokines for their ability to promote the production of IgG3 in the presence of NP-Ficoll , and identified IL-1α , IL-1β , and IFNα as efficient costimuli for IgG3 production ( Figure 1—figure supplement 1D and Figure 1C ) . These cytokines , together with NP-Ficoll , induced generation of IgG3+ cells ( Figure 1D ) , as well as IgG1+ and IgG2b+ cells to a lesser extent ( Figure 1—figure supplement 1E ) . In the presence of these cytokines , NP-Ficoll was far more potent than NP-CGG for IgG3 production , IgG3+ cell generation , Sµ-Sγ3 CSR ( detectable as the Iγ3-Cµ transcript from the switch circle DNA ) ( Xu et al . , 2012; Kinoshita et al . , 2001 ) , and the induction of Aicda transcription ( Figure 1C–E ) . NP-Ficoll or each of these cytokines alone could not induce those responses ( Figure 1C–E ) . Collectively , these results indicate that BCR signaling elicited by TI-2 Ag engagement is pivotal for the TI-2 B-cell response , namely induction of proliferation and antibody production , as well as potentiation of CSR to IgG . By Western blot analyses for BCR signaling using NP-specific B cells , we noticed that NP-Ficoll-induced phosphorylation of PKCδ tyrosine 311 ( Y311 ) , indicative of activation of the kinase ( Balasubramanian et al . , 2006 ) , more strongly than NP-CGG ( Figure 2A ) . PKCδ belongs to a novel PKC subfamily ( Salzer et al . , 2016 ) , whose function in the immune response remains elusive . Thus , we analyzed the role of PKCδ in the TI-2 response using PKCδ-deficient ( Prkcd−/− ) mice . B cells from Prkcd−/− mice normally proliferate and produce IgM when stimulated with NP-Ficoll alone ( Figure 2—figure supplement 1A , B ) . Upon costimulation with NP-Ficoll and IL-1α , IL-1β , and IFNα , Prkcd−/− B cells produced comparable levels of IgM but markedly reduced levels of IgG3 compared to Prkcd+/+ cells ( Figure 2B ) . As class switching occurs along with cell division ( Deenick et al . , 1999 ) , we analyzed the frequency of IgG3+ cells at each cell division using B cells labeled with CellTrace Violet ( CTV ) . While cell division was almost equivalent between these cells after costimulation with any of the cytokines and NP-Ficoll ( Figure 2—figure supplement 1C ) , the frequencies of IgG3+ cells barely increased at any points of cell division in Prkcd−/− cells in contrast to Prkcd+/+ cells ( Figure 2C ) . Therefore , PKCδ is necessary for generation of IgG3+ cells regardless of cell division . Accordingly , Igh locus CSR to IgG3 assessed by the circle Iγ3-Cµ transcript and the expression of Aicda transcripts were attenuated in Prkcd−/− B cells ( Figure 2D ) . Collectively , these results suggest that PKCδ mediates the expression of AID and class switching to IgG3 induced by BCR stimulation with TI-2 Ag and IL-1/IFNα costimulation . Prkcd−/− B cells also exhibited defects in the production of IgG and Aicda transcripts , but normal IgM production , upon stimulation with NP-Ficoll and TLR ligands ( Figure 2—figure supplement 1D , E ) , but no defects in the production of IgM , IgG , and Aicda transcripts , upon stimulation with TLR ligands alone ( Figure 2—figure supplement 1F , G ) . Together with the above results , PKCδ appears to be selectively required for TI-2-Ag-mediated BCR signaling to potentiate CSR to produce IgG . We next evaluated the contribution of PKCδ to IgG production in a TI-2 immune response using mice carrying loxP-flanked Prkcd alleles and Cd19-cre allele that lack Prkcd specifically in B cells ( referred to herein as Cd19cre/+ Prkcdff/f ) . We first examined the cellularity of mature B-cell subpopulations in such mice and their control ( Cd19cre/+ Prkcd+/+ ) . The numbers of follicular and marginal zone B cells in spleens were comparable between these mice , whereas those of splenic and peritoneal B1 cell populations were slightly increased in Cd19cre/+ Prkcdff/f mice ( Figure 3—figure supplement 1A ) . Thus , PKCδ is dispensable for the development of B cells , as previously described ( Mecklenbräuker et al . , 2002; Miyamoto et al . , 2002 ) . To analyze the role of PKCδ in a TI-2 response , we immunized Cd19cre/+ Prkcdff/f and the control mice with NP-Ficoll . Although the production of serum anti-NP IgM was slightly enhanced in Cd19cre/+ Prkcdff/f mice , that of anti-NP IgG3 was severely suppressed ( Figure 3A ) , and so was that of anti-NP IgG1 , IgG2b , and IgG2c ( Figure 3B ) . We then analyzed PCs in the spleen 1 week after immunization . Among the NP-specific PCs , the proportion and the number of IgG3- and IgG2b-producing PCs were severely decreased in Cd19cre/+ Prkcdff/f mice compared to the control mice , whereas the number of IgM+ PCs was comparable between the two groups ( Figure 3C ) . Therefore , PKCδ is required for the generation of IgG+ PCs and the subsequent production of IgG in the TI-2 response . The capsular polysaccharides of Streptococcus pneumoniae , such as pneumococcal polysaccharide serotype 3 ( PPS3 ) , are also classified as TI-2 Ags , and immunization with PPS3 is known to induce an Ag-specific IgG3 response ( McLay et al . , 2002 ) . Thus , we next immunized Cd19cre/+ Prkcdff/f and control mice with PPS3 . Although early production of anti-PPS3 IgM was modestly enhanced in Cd19cre/+ Prkcdff/f mice , production of anti-PPS3 IgG3 was ablated in these mice ( Figure 3D ) . Thus , PKCδ appears to be generally required for IgG production in response to a variety of TI-2 Ags . We next assessed whether PKCδ is also required for IgG production in a TD response . After immunization with NP-CGG , the production of anti-NP IgM was transiently enhanced in Cd19cre/+ Prkcdff/f mice at 1 week , whereas anti-NP IgG1 and IgG3 titers were normal in these mice ( Figure 3—figure supplement 1B ) . These data indicated that PKCδ signaling is required for IgG production in a TI-2 response , but not in a TD response . We next assessed whether PKCδ mediates IgG production through class switching in an in vivo TI-2 response . To discriminate Ag-specific B-cell responses , we transferred CTV-labeled naive B cells from B1-8hi CD45 . 1 mice into C57BL/6 ( B6 ) mice , which were immunized with NP-Ficoll on the next day and analyzed by flow cytometry 3 days later . About 30% of the splenic donor cells were IgG3+ in the mice transferred with control B cells , whereas only about 3% were IgG3+ in recipients of Cd19cre/+ Prkcdff/f B cells ( Figure 4A ) . IgG3+ cells emerged at the second cell division and their frequency was increased as the cell division proceeded in the cells of control mice , whereas the frequency of IgG3+ cells was extremely low in the cells of Cd19cre/+ Prkcdff/f mice at any point of cell division ( Figure 4A ) . Thus , PKCδ mediates the generation of IgG3+ cells in a manner unrelated to cell division . We next analyzed molecular events associated with CSR in sorted donor cells ( Figure 4—figure supplement 1A ) . The amounts of the Iγ3-Cµ circle transcripts and the Iµ-Cγ3 early postswitch transcripts were far less in Cd19cre/+ Prkcdff/f B cells compared to control cells , indicating fewer CSR events in the former in vivo ( Figure 4B ) . The expression of Iµ-Cµ and Iγ3-Cγ3 germline transcripts was comparable between the Cd19cre/+ Prkcdff/f B cells and the control cells , whereas the expression of Aicda was substantially reduced in the former ( Figure 4B ) . These results indicate that PKCδ mediates CSR to IgG3 through upregulation of AID mRNA , but not through the activation of Ig gene S regions in B cells responding to TI-2 Ags in vivo . To examine whether the reduction of AID is a primary reason for the impaired generation of IgG3+ cells from PKCδ-deficient B cells , we transduced AID into in vivo primed Cd19cre/+ Prkcdff/f B1-8hi B cells and transferred them into B6 mice that had been immunized with NP-Ficoll 1 day previously ( Figure 4—figure supplement 1B ) . The reconstitution of AID expression restored IgG3 class switching to a frequency comparable to the same cells reconstituted with PKCδ ( Figure 4—figure supplement 1C and Figure 4C ) . Collectively , these results indicate that PKCδ mediates class switching to IgG3 by upregulating the expression of AID in the TI-2 response . It has been shown that expression of Aicda gene is regulated by various transcriptional factors ( Vaidyanathan et al . , 2014; Tran et al . , 2010; Crouch et al . , 2007 ) . To assess the role of PKCδ in Aicda gene expression , we quantified expression levels of the genes encoding such transcription factors in PKCδ-sufficient and PKCδ-deficient B cells collected from mice immunized with NP-Ficoll . Among the genes tested , we found that the amount of Batf mRNA was markedly lower in Cd19cre/+ Prkcdff/f B cells than in control cells ( Figure 5A ) . The expression of BATF mRNA and protein was induced by in vitro stimulation with NP-Ficoll alone in PKCδ-sufficient B cells , but only marginally in PKCδ-deficient B cells ( Figure 5B and C ) , whereas BATF expression was not induced by NP-CGG ( Figure 5B and Figure 5—figure supplement 1A ) . IL-1α , IL-1β , or IFNα did not augment the phosphorylation of PKCδ nor induce Batf expression nor enhance the NP-Ficoll-induced Batf expression ( Figure 5B and Figure 5—figure supplement 1B-D ) . Collectively , these data indicate that PKCδ induces the expression of BATF downstream of the BCR in the TI-2 response . It was reported that BATF binds to a regulatory region of the Aicda gene to directly promote its expression and that IgG3 production against TNP-Ficoll was impaired in Batf−/− mice ( Ise et al . , 2011 ) . Therefore , we next asked whether the defect of BATF expression is responsible for the suppression of AID expression and IgG3+ cell generation in PKCδ-deficient mice . First , we knocked down BATF in B1-8hi B cells and transferred them into mice immunized with NP-Ficoll . Both the expression of Aicda and the frequency of IgG3+ cells were significantly decreased in BATF knockdown cells compared with the mock-transduced control cells ( Figure 5D and E ) . Conversely , forced expression of BATF in B1-8hi Cd19cre/+ Prkcdff/f B cells partially but significantly restored Aicda expression and the generation of IgG3+ cells in the recipient mice immunized with NP-Ficoll ( Figure 5F and G ) . Taken together , these data indicate that PKCδ mediates expression of AID and class switching to IgG3 through upregulation of BATF expression in B cells undergoing a TI-2 response . Recent works have revealed that commensal microbes induce an IgG response and confer protection against systemic bacterial infection ( Zeng et al . , 2016 ) . Among antibacterial IgG , IgG3 is most abundant and produced in a TI manner ( Ansaldo et al . , 2019; Koch et al . , 2016 ) . Therefore , we assessed the contribution of PKCδ in IgG-mediated antibacterial responses . To standardize the microbiota , we cohoused control and Cd19cre/+ Prkcdff/f mice over 4 weeks and serum antibodies against fecal bacteria were titrated . Production of serum antibacterial IgM , IgG1 , and IgG2b was not changed significantly , but that of IgG3 was severely impaired in Cd19cre/+ Prkcdff/f mice ( Figure 6A ) . Antibacterial IgG2c was undetectable in both mouse groups ( data not shown ) . Given that PKCδ was required for production of all IgG subclasses in the anti-NP TI-2 response ( Figure 3 ) , the PKCδ-independent production of antibacterial IgG1 and IgG2b may be attributable to TD responses , as reported for IgG1 ( Ansaldo et al . , 2019 ) , while antibacterial IgG3 is mainly produced by the TI-2 response . We further asked whether regulation of commensal bacteria is defective in the PKCδ-deficient mice . Dextran sodium sulfate ( DSS ) treatment is known to disrupt the gut epithelium and to allow intestinal bacteria to translocate throughout the body . Subsequently , it leads to fatal bacteremia in the absence of microbiota-specific IgG ( Zeng et al . , 2016 ) . After the treatment with DSS , Cd19cre/+ Prkcdff/f mice exhibited increased numbers of aerobic and anaerobic bacteria in the blood compared to control mice ( Figure 6B ) . Accordingly , the mortality of Cd19cre/+ Prkcdff/f mice was significantly higher than that of control mice ( Figure 6C ) . Collectively , these results suggest that IgG3 production by a TI-2 response via PKCδ prevents lethal bacteremia .
Although previous reports showed that proximal BCR signal molecules such as BLNK or Btk are necessary for B-cell activation and subsequent IgM and IgG production in the TI-2 response in mice ( Xu et al . , 2000; Ellmeier et al . , 2000 ) , it was not clear whether there are any specific BCR signaling pathways inducing CSR . Here , we demonstrated that BCR stimulation with a TI-2 Ag ( NP-Ficoll ) , but not a TD Ag ( NP-CGG ) , both sharing the same BCR epitope , promoted transcription of Aicda and induced IgG3 CSR in the presence of a secondary stimulation . BCR engagement with a TI-2 Ag induced the phosphorylation of PKCδ and PKCδ was required for upregulation of BATF expression , thereby mediated the induction of Aicda transcription and subsequent CSR to the IgG subclasses ( Figure 7 ) . By contrast , PKCδ was dispensable for TI-2 Ag-mediated B-cell proliferation , IgM production , as well as for B-cell development and the TD response . Thus , we have found for the first time , as far as we know , a BCR signaling molecule that is selectively required for CSR in the TI-2 response . TI-2 Ags such as NP-Ficoll can induce antibody response in the absence of major histocompatibility complex class II-restricted T cell help or MyD88-mediated TLR receptor signaling ( Gavin et al . , 2006; Hou et al . , 2011 ) . Thus , it has been unclear whether any costimulation is necessary for B-cell activation in a TI-2 response . We demonstrated that NP-Ficoll , but not NP-CGG , was able to activate NP-specific naive B cells to induce proliferation , antibody production , and potentiation for CSR in vitro . Thus , despite sharing the same antigenic epitope ( NP ) at a similar average number per molecule , these Ags appear to elicit disparate BCR signal transduction . After Ag binding , the BCR is clustered and recruited to lipid microdomains , in which signaling proteins are localized that trigger B-cell activation signaling ( Niiro and Clark , 2002 ) . Ag structure probably affects such BCR dynamics or membrane organization of signal components . Consistently , it was previously reported that a high-valency TI-2 Ag induces the formation of large BCR clusters in the lipid microdomains ( Puffer et al . , 2007 ) . We have revealed that , in the TI-2 response , BCR-downstream signaling through PKCδ is critical for CSR and generation of IgG . PKCδ is known to control Ag-induced tolerance in immature B cells ( Mecklenbräuker et al . , 2002; Limnander et al . , 2011 ) . However , a possible alteration of the mature B-cell repertoire in PKCδ-deficient mice is not attributable to the defective class switching to IgG3 , since the defect was obvious in the NP-Ficoll response by PKCδ-deficient , B1-8VH-knock-in B cells with a monoclonal anti-NP repertoire . Here , we demonstrated that PKCδ mediated the expression of AID to induce CSR . BCR engagement with a TI-2 Ag induced the expression of BATF via PKCδ and BATF was required for the transcription of the Aicda gene in a TI-2 response , as reported previously ( Ise et al . , 2011 ) . Forced expression of BATF in PKCδ-deficient B cells restored the expression of AID and IgG3 class switching significantly , but less effectively , than that of PKCδ itself . In this regard , various transcription factors have been reported to cooperatively induce the transcription of AID in the TD response ( Vaidyanathan et al . , 2014; Crouch et al . , 2007; Tran et al . , 2010 ) . Although we have not assessed the contribution of other transcription factors in the TI-2 response , since the expression of other candidate genes was intact in PKCδ-deficient B cells ( Figure 5A ) , PKCδ signaling might regulate other transcription factors in addition to BATF to fully upregulate the expression of AID to induce CSR in the TI-2 response . PKCδ contains several phosphorylation sites , whose phosphorylation pattern affects its enzymatic activity and possibly the selection of downstream targets ( Salzer et al . , 2016; Steinberg , 2004 ) . Although the phosphorylation of Thr 505 and Ser 647 of PKCδ is known to be important for its enzymatic activity , phosphorylation of these sites was constitutive in unstimulated naive B cells and did not increase after NP-Ficoll stimulation ( data not shown ) . This observation and a previous report that PKCδ exerts proapoptotic activity in peripheral B cells through nuclear translocation , unless BAFF signaling reverses it ( Mecklenbräuker et al . , 2004 ) , indicate a constitutive PKCδ activity in B cells irrespective of Ag stimulation . Augmented IgM production after immunization with TI-2 as well as TD Ags in PKCδ-deficient mice ( Figure 3A and D and Figure 3—figure supplement 1B ) may reflect the lack of the general proapoptotic function of PKCδ . On the other hand , phosphorylation of PKCδ at Tyr 311 , shown here to be selectively augmented by TI-2 Ag stimulation , is known to alter its intracellular localization and the substrate specificity ( Rybin et al . , 2004; Steinberg , 2004 ) and to be involved in various cellular responses ( Balasubramanian et al . , 2006; Nakashima et al . , 2008 ) . Thus , to understand the mechanism of TI-2 Ag-specific BCR signaling that induces the expression of BATF and AID , it would be necessary to identify a specific substrate of the phosphorylated PKCδ at Tyr 311 . In contrast to TI-2 response , PKCδ was not required for IgG production in the TD response , like other proximal BCR molecules ( Khan et al . , 1995; Xu et al . , 2000 ) . In vitro stimulation with a TD Ag alone induced only a marginal level of phosphorylation of PKCδ and other signaling molecules ( Figure 2A and data not shown ) and did not induce proliferation , IgM production , and the expression of BATF . Accordingly , we assume that the TD Ag-mediated BCR signaling cannot exceed an activation threshold to induce gene expression leading to functional B-cell response including class switching . Instead , other stimuli such as CD40L would meet the demand for B-cell activation in TD response . Supporting this view , it was reported that BCR signaling does not enhance the class switching induced by CD40 signaling ( Pone et al . , 2012 ) . The synergistic effect of TLR and BCR on the induction of AID has been shown previously ( Pone et al . , 2012 ) . Here , we found that IL-1α/β , IFNα , and TLR ligands costimulate B cells with TI-2 Ag to induce Aicda expression and class switching to IgG3 and perhaps other IgG subclasses , while other cytokines tested barely induced IgG3 production in vitro . In the context of TD response , the recruitment of NF-κB and STAT6 to the 5′ upstream enhancer region of Aicda locus , in addition to BATF recruitment to the 3′ downstream region , is required for the transcription of Aicda ( Tran et al . , 2010; Vaidyanathan et al . , 2014 ) . Since the in vitro stimulation by TI-2 Ag induced the expression of BATF but not of Aicda without the additional stimulation , such costimulation may signal the recruitment of transcription factors to the 5′ upstream enhancer region . In this regard , it was reported that IL-1 and IFNα induce the transcription of Aicda through activation of NF-κB in hepatocytes during hepatitis B virus infection ( Watashi et al . , 2013 ) . IL-1 and IFNα have also been reported to activate STAT proteins , if not STAT6 ( Biffi et al . , 2019; Du et al . , 2007 ) . However , further study would be necessary to understand the overall mechanism for the induction of Aicda expression in B cells in the TI-2 response . Consistent with our in vitro data , previous reports showed the contribution of IL-1α/β , IFNα , and TLR ligands in several types of TI-2 responses: IgG production upon immunization with NP-Ficoll was dampened in IL-1α/β double-deficient mice ( Nakae et al . , 2001 ) , and IgG2c production with NP-Ficoll plus poly ( I:C ) was dependent on the IFNα receptor on B cells ( Swanson et al . , 2010 ) . While IgG production against NP-Ficoll did not require MyD88-mediated TLR signaling ( Gavin et al . , 2006; Hou et al . , 2011 ) , IgG production against pneumococcal polysaccharides requires TLR signaling ( Sen et al . , 2005 ) . Compared to NP-Ficoll , physiological TI-2 Ags are more complex ( Snapper , 2006 ) and may contain pathogen-associated molecular patterns ( PAMPs ) that mediate TLR signaling . Therefore , any of IL-1α/β , IFNα , and TLR ligands could function as costimulation for an in vivo TI-2 response , depending on the type of Ag . Our data revealed that PKCδ is generally required for IgG production in response to TI-2 Ags such as NP-Ficoll and PPS3 . Given the predominant role of IgG3 in antipneumococcal responses ( McLay et al . , 2002 ) , PKCδ seems to be needed for protection from pneumococcal infection . Furthermore , we found that PKCδ is required to produce IgG3 against commensal bacteria in the steady state and the prevention of bacteremia after epithelial barrier disruption . As commensal bacteria have been shown to contain various Ags and stimulate multiple immune pathways ( Belkaid and Harrison , 2017 ) , our result suggests that some commensal bacteria express TI-2 Ag-like repetitive structures that elicit IgG3 production . It has been shown that such IgG against symbiotic bacteria also plays a protective role in systemic infection by pathogens ( Zeng et al . , 2016 ) . Taken together , IgG production in the TI-2 response seems to be critical for regulation of various types of bacteria . Several reports uncovered the potential contribution of the dysregulated microbiome in SLE pathogenesis together with genetic risk factors ( Silverman , 2019 ) . Translocation of a gut pathobiont to the liver and other systemic tissues causes a lupus-like disease in genetically autoimmune-prone mice owing to induction of a systemic type I interferon response and autoantibody production ( Manfredo Vieira et al . , 2018 ) . Prevention of this bacterial translocation and autoimmune response by vaccination against the pathobiont indicated a role of antibodies for the prevention . Thus , besides the defect of B-cell tolerance ( Miyamoto et al . , 2002; Mecklenbräuker et al . , 2002 ) , the low serum antibacterial IgG3 in PKCδ-deficient mice may lead to the pathobiont invasion and autoantibody production through intact TD response . Loss-of-function PKCδ mutations in humans also cause SLE-like autoimmunity ( Salzer et al . , 2016 ) . Some of these patients show reduced IgG-positive B cells in the peripheral blood and have recurrent infections for unknown reasons ( Kiykim et al . , 2015; Salzer et al . , 2013; Kuehn et al . , 2013 ) . Taken together , PKCδ-mediated IgG production by the TI-2 response appears to be critical also in humans for host defense against certain bacteria and the regulation of autoimmunity .
C57BL/6NCrSlc ( B6 ) mice were purchased from Japan SLC . All the following mice were backcrossed to the B6 or B6-CD45 . 1 strain: B1-8flox/+ mice ( Lam et al . , 1997 ) , B1-8hi mice ( Shih et al . , 2002 ) , Igk–/– mice ( Chen et al . , 1993 ) , Cd19cre/+ mice ( Rickert et al . , 1995 ) , and Prkcd–/– mice ( Miyamoto et al . , 2002 ) . Prkcd fl/fl mice on the B6 background were developed by Drs . Niino , Shioda , and Sakimura ( Niino et al . , 2021 ) and purchased from the RIKEN BioResource Center ( RBRC06462 ) . Mice were immunized i . p . with 100 µg of NP46-Ficoll ( F-1420; Biosearch Technologies ) , 1 µg of PPS3 ( 169 X; American Type Culture Collection ) or 100 µg of NP40-CGG in alum ( Haniuda et al . , 2016 ) unless otherwise noted . For flow cytometry of spleen cells , mice were immunized i . v . with 100 µg of NP46-Ficoll . Sex-matched 7–14-week-old mice were used for all experiments . All mice were maintained in the Tokyo University of Science ( TUS ) mouse facility under specific pathogen-free conditions . Mouse procedures were performed under protocols approved by the TUS Animal Care and Use Committee ( Approval No . S19017 and S20011 ) . Spleen cells were stained with a cocktail of biotinylated Abs for CD4 , CD8 , CD43 , CD49b , Ter119 , and Streptavidin Particles Plus DM , from which naive B cells were purified by magnetic negative sorting using the IMag system ( BD Biosciences ) and MACS system ( Miltenyi Biotec ) , as described previously ( Nojima et al . , 2011 ) . B cells were cultured in RPMI-1640 medium ( Wako ) supplemented with 10% heat-inactivated fetal bovine serum , 10 mM HEPES pH 7 . 5 , 1 mM sodium pyruvate , 50 mM 2-mercaptoethanol , 100 U/ml penicillin , and 100 mg/ml streptomycin ( GIBCO ) . Typically , B cells were cultured at 2 × 105 /ml in the presence of the following stimuli at the indicated doses , unless otherwise noted: NP46-Ficoll ( 10 ng/ml ) , NP40-CGG ( 10 ng/ml ) , LPS ( 1 µg/ml , L2880; Sigma ) , R-848 ( 1 µg/ml , tlrl-r848; InvivoGen ) , CpG ODN 1826 ( 1 µg/ml , tlrl-1826; InvivoGen ) , IL-1α ( 1 ng/ml , 211–11 A; Pepro Tech ) , IL-1β ( 1 ng/ml , 211-11B; Pepro Tech ) , or IFNα ( 100 ng/ml , 752802; Biolegend ) . Concentrations of cytokines used in Figure 1—figure supplement 1D are shown in Supplementary file 1 . To produce retrovirus , pSIREN- or pMXs-based plasmids were cotransfected together with pVSVG into Plat-E cells ( kindly provided by T . Kitamura , University of Tokyo ) using PEI Max ( Mw 40 , 000 , 24765-1; Polysciences ) . The virus-containing supernatant was harvested 2 days after transfection . For retroviral transduction , B cells were preactivated in vivo: B1-8hi mice were injected i . p . with 50 µg of NP-Ficoll , and then B cells were purified from the spleens of these mice on the next day . These B cells were mixed with the virus-containing supernatant and spin infected at 2000 rpm , 37°C for 90 min with 10 mg/ml DOTAP Liposomal Transfection Reagent ( 11202375001; Sigma ) . One day later , the cells were harvested and 5 × 105 cells were transferred into B6 mice that had been immunized i . v . with NP-Ficoll on the previous day . This strategy is summarized in Figure 4—figure supplement 1B . The proliferation assay was performed as described previously ( Fukao et al . , 2014 ) . Naive B cells were cultured at 5 × 104 cells/well in 96-well plates for 72 hr , with the last 8 hr in the presence of [3H] thymidine ( 1 mCi/well , NET027001MC; PerkinElmer ) . Incorporated [3H] thymidine was counted by a BetaPlate scintillation counter ( Wallac , Gaithersburg , MD ) . Single-cell suspensions from spleen or peritoneal cavity were prepared , red blood cells were lysed with ammonium chloride buffer and then cells were incubated with anti-CD16/32 Ab ( 2 . 4G2 ) to block FcγRs . Cultured B cells were collected in MACS buffer ( phosphate-buffered saline [PBS] supplemented with 0 . 5 % BSA , 2 mM ethylenediaminetetraacetic acid [EDTA] ) at the indicated days of culture . Cells were stained with Abs and reagents on ice ( for splenocytes ) or at room temperature ( for cultured B cells ) . For the staining of IgM , IgG , and NP-binding Ig , Fixation/Permeabilization Solution Kit ( 554714; BD Biosciences ) was used according to the manufacturer’s protocol to detect total ( surface and intracellular ) proteins , after surface staining of other molecules . NP-binding Ig was stained with NP14-BSA-Alexa Fluor 647 ( Haniuda et al . , 2016 ) . Dead cells were stained with Fixable Viability dye eFluor 506 ( 65-0866-18; Thermo Fisher Scientific ) before cell fixation and excluded from analysis . All samples were analyzed using FACSCanto II , FACSAria II or III ( BD Biosciences ) with FlowJo software ( Tree Star , Inc ) . Naïve B cells were resuspended in PBS at 5 × 106 cells/ml and labeled with 5 µM of CTV Cell proliferation Kit ( C34557; Thermo Fisher Scientific ) at 37 °C for 20 min according to the manufacturer’s protocol . Collected cells were analyzed by flow cytometry as described above . Cell divisions were determined using the proliferation platform of FlowJo . 1 × 106 CD45 . 1 B1-8hi naive B cells were transferred into B6 mice , which were then immunized i . v . with NP-Ficoll on the next day . Donor B cells were magnetically enriched from pooled spleens of recipient mice using APC-conjugated anti-CD45 . 1 and anti-APC MicroBeads ( 130-090-855; Miltenyi Biotec ) , with the MACS system ( Miltenyi Biotec ) . After enrichment , cells were further stained with respective Abs and sorted using FACSAria II or III ( BD Biosciences ) as shown in Figure 4—figure supplement 1A . Rv-transduced donor B cells were enriched as described above and sorted as shown in Figure 4—figure supplement 1C . TRI Reagent ( T9424; Sigma ) or RNeasy Micro ( 74004; QIAGEN ) was used to isolate total RNA from B cells . cDNA was generated from total RNA using ReverTra Ace ( TRT-101; TOYOBO ) with an oligo ( dT ) 20 primer ( 18418020; Thermo Fisher Scientific ) according to the manufacturer’s protocols . For the analysis of a small number of cells , cDNA was generated from cell lysates with SuperPrep II Cell lysis & RT Kit for qPCR ( SCQ-401; TOYOBO ) according to the manufacturer’s protocols . Quantitative real-time PCR ( qPCR ) was performed using Thunderbird SYBR qPCR Mix ( QPS-201; TOYOBO ) with the 7500 fast Real-time PCR system or Quant-Studio 3 ( Applied Biosystems ) . For quantification of gene expression levels , each sample was normalized to the expression of a control housekeeping gene , Gapdh or Rps18 . The relative fold change in expression of each gene compared to a control sample , set as 1 . 0 , was calculated with the 2-ddCT method . Primers used in this study are listed in Supplementary file 1 . The germline and postswitched transcripts were analyzed with the following primer sets: germline Iµ-Cµ transcripts: Iµ Fw1 and Cµ Rv; Iγ3-Cγ3 transcripts: inner Iγ3 Fw and Cγ3 Rv; postswitched Iµ-Cγ3 transcripts: Iµ Fw2 and Cγ3 Rv . For the measurement of the circle Iγ3-Cµ transcript , cDNA was generated with external Cµ Rv primer from total RNA as described above , and preamplified with external Iγ3 Fw primer and Cµ Rv primer using GoTaq Green Master Mix ( M712B; Promega ) . Preapplication products were purified with QIAquick PCR Purification Kit ( 28104; QIAGEN ) and circle Iγ3-Cµ transcript was quantified with inner Iγ3 Fw primer and Cµ Rv primer by qPCR as described above . Specific amplification of the circle Iγ3-Cµ transcript was confirmed by analyzing the sequence of the qPCR products in advance . The expression of the Iγ3-Cµ transcript was normalized to the expression of Rps18 in cDNA generated with oligo ( dT ) 20 primer from the same RNA sample . Concentrations of total IgM , IgG , or IgG subclasses and of Ag-specific Igs ( where indicated ) were assessed by titration of culture supernatants or sera by enzyme-linked immunosorbent assay ( ELISA ) . Total and NP-specific antibodies were measured as described previously ( Fukao et al . , 2014; Nojima et al . , 2011 ) , with NP13 . 6-BSA used for coating plates for the latter . PPS3-specific IgM and IgG3 were detected using 96-well plates coated with PPS3 . Bacteria-specific antibody was detected as described previously ( Zeng et al . , 2016 ) . Heat-killed fecal bacteria isolated from Prkcd+/+ Cd19cre/+mice and Prkcdf/f Cd19cre/+mice cohoused at least for 4 weeks were mixed and used for plate coating . Cells were lysed with 1% NP-40 lysis buffer or RIPA buffer ( 40 mM Tris–HCl pH 7 . 5 , 150 mM NaCl , 1% NP-40 , 1 % sodium deoxycholate , 0 . 1 % SDS , and 1 mM EDTA ) supplemented with protease and phosphatase inhibitors . Lysates were sonicated and mixed with sample buffer and dithiothreitol and boiled . Lysates were resolved on sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride ( PVDF ) membranes ( Millipore ) , followed by immunoblotting as previously described ( Haniuda et al . , 2020 ) . Prkcd+/+ Cd19cre/+mice and Prkcdf/f Cd19cre/+ female mice were cohoused at 1:1 ratio from 4 weeks of age . After at least 4 weeks , cohoused mice were administered 3 % DSS ( 9011-18-1; MP Biomedicals ) in drinking water for 7 days and then switched to regular water . On the day of the last DSS administration , blood samples were collected from mice and cultured on Brain Heart Infusion Agar ( 221570; BD Bioscience ) under aerobic or anaerobic conditions for 24 hr to measure CFU . Biological replication is derived from multiple biological samples ( mouse or cell ) . Technical replication consisted of multiple samples derived from one biological sample . All statistical analyses were performed using GraphPad Prism eight software . Comparisons between two groups were performed by a two-tailed unpaired Student’s t-test , Welch’s t-test ( in case F-test is significant: p < 0 . 05 ) , or multiple t-test ( for grouped data ) . Comparisons between multiple groups were performed by one or two-way analysis of variance with Tukey’s multiple comparison . Survival of DSS-treated mice was analyzed by Kaplan–Meier survival plot using log-rank test . In all cases , *p < 0 . 05; **p < 0 . 01; ***p < 0 . 001; ****p < 0 . 0001; ns , not significant ( p > 0 . 05 ) . | When the human body faces a potentially harmful microorganism , the immune system responds by finding and destroying the pathogen . This involves the coordination of several different parts of the immune system . B cells are a type of white blood cell that is responsible for producing antibodies: large proteins that bind to specific targets such as pathogens . B cells often need help from other immune cells known as T cells to complete antibody production . However , T cells are not required for B cells to produce antibodies against some bacteria . For example , when certain pathogenic bacteria coated with a carbohydrate called a capsule – such as pneumococcus , which causes pneumonia , or salmonella – invade our body , B cells recognize a repetitive structure of the capsule using a B-cell antigen receptor . This recognition allows B cells to produce antibodies independently of T cells . It is unclear how B cells produce antibodies in this situation or what proteins are required for this activity . To understand this process , Fukao et al . used genetically modified mice and their B cells to study how they produce antibodies independently of T cells . They found that a protein called PKCδ is critical for B cells to produce antibodies , especially of an executive type called IgG , in the T-cell-independent response . PKCδ became active when B cells were stimulated with the repetitive antigen present on the surface of bacteria like salmonella or pneumococcus . Mice that lack PKCδ were unable to produce IgG independently of T cells , leading to fatal infections when bacteria reached the tissues and blood . Understanding the mechanism behind the T cell-independent B cell response could lead to more effective antibody production , potentially paving the way for new vaccines to prevent fatal diseases caused by pathogenic bacteria . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"immunology",
"and",
"inflammation"
] | 2021 | Protein kinase Cδ is essential for the IgG response against T-cell-independent type 2 antigens and commensal bacteria |
RNA splicing is an essential part of eukaryotic gene expression . Although the mechanism of splicing has been extensively studied in vitro , in vivo kinetics for the two-step splicing reaction remain poorly understood . Here , we combine transient transcriptome sequencing ( TT-seq ) and mathematical modeling to quantify RNA metabolic rates at donor and acceptor splice sites across the human genome . Splicing occurs in the range of minutes and is limited by the speed of RNA polymerase elongation . Splicing kinetics strongly depends on the position and nature of nucleotides flanking splice sites , and on structural interactions between unspliced RNA and small nuclear RNAs in spliceosomal intermediates . Finally , we introduce the ‘yield’ of splicing as the efficiency of converting unspliced to spliced RNA and show that it is highest for mRNAs and independent of splicing kinetics . These results lead to quantitative models describing how splicing rates and yield are encoded in the human genome .
Transcription of eukaryotic genes produces precursor RNA molecules that are processed by splicing . Splicing is a two-step reaction that results in the removal of introns from precursor RNA and the formation of mature RNA with joined exons . Splicing is catalyzed by the spliceosome , a dynamic ribonucleoprotein machine assembled from small nuclear ribonucleoprotein complexes ( snRNPs ) and several non-RNP factors ( Herzel et al . , 2017; Wahl et al . , 2009 ) . In metazoa , the majority of the introns are excised by the major spliceosome , whereas a minority is removed by the minor spliceosome ( Will and Lührmann , 2011 ) . Spliceosomes are recruited through conserved RNA elements , the 5´-splice site at the exon-intron border ( donor site ) , the 3´-splice site at the intron-exon border ( acceptor site ) , and the branch point , which is followed by a polypyrimidine track ( Coolidge et al . , 1997; Turunen et al . , 2013 ) and located ~18–40 nucleotides ( nt ) upstream of the acceptor site ( Ruskin and Green , 1985; Ruskin et al . , 1985; Taggart et al . , 2012 ) . Additional RNA sequences such as splicing enhancers and silencers are found in both introns and exons and can influence the choice between splice sites ( Zhu et al . , 2001 ) . In recent years , the intricate mechanisms of splicing were investigated with a combination of structural and functional studies ( Mayerle and Guthrie , 2017; Shi , 2017; Wilkinson et al . , 2018; Will and Lührmann , 2011 ) . The spliceosome assembles in a stepwise manner , adopting different intermediates with varying composition , conformation , and interactions between RNAs and proteins ( Will and Lührmann , 2011 ) . Assembly of the major spliceosome begins with recognition of the donor site by U1 snRNP ( Kondo et al . , 2015; Lerner et al . , 1980; Seraphin and Rosbash , 1989; Zhuang and Weiner , 1986 ) . The U2 auxiliary factor then binds to the poly-pyrimidine track and the acceptor site ( Valcárcel et al . , 1996; Zamore and Green , 1989; Zamore et al . , 1992 ) generating the E complex . U2 snRNP then binds the branchpoint , resulting in the A complex ( Konarska and Sharp , 1986; Perriman and Ares , 2010 ) . In the A complex , U1 snRNA binds the donor site and U2 snRNA binds the branchpoint , rendering the branchpoint adenosine base available for interaction with the acceptor site ( Berglund et al . , 2001; Query et al . , 1994 ) . Binding of the U4/U5/U6 tri-snRNP leads to the B complex ( Bertram et al . , 2017a ) , which is first activated ( Bact ) and then converted to the catalytically active B* complex . In the B* complex , the donor site is positioned close to the branchpoint in a RNA network formed between the precursor RNA and U2 , U5 and U6 snRNAs ( Zhang et al . , 2018 ) . The activated spliceosome catalyzes intron removal in two steps , which are both transesterification reactions . In the first step , the 2´-hydroxyl group of the branchpoint adenosine serves as a nucleophile to attack the donor site and to generate a cleaved 5´-exon and the lariat intermediate . This leads to the C complex ( Zhan et al . , 2018 ) that is then rearranged to form the C* complex , which is catalytically active to carry out the second step . In the C* complex , the ends of exons to be joined are juxtaposed ( Bertram et al . , 2017b; Zhang et al . , 2017 ) , and this enables the 3´-hydroxyl group of the last nucleotide of the 5´-exon to attack the acceptor site , leading to exon ligation and excision of the intron lariat . The resulting P complex contains the ligated exons , which are subsequently released , completing the splicing process ( Bai et al . , 2017; Wilkinson et al . , 2017 ) . Taken together , extensive in vitro studies have strongly advanced our understanding of the splicing process , but the kinetics and mechanisms of splicing in vivo remain far less understood . Although biochemical assays show that splicing can occur in the absence of transcription , in vivo splicing happens mainly co-transcriptionally , when newly transcribed RNA is still attached to RNA polymerase II ( Pol II ) and chromatin ( for review see Alexander and Beggs , 2010; Bentley , 2014; Saldi et al . , 2016 ) . Furthermore , compromising Pol II transcription elongation increases alternative splicing ( de la Mata et al . , 2003; Dujardin et al . , 2014; Ip et al . , 2011; Pagani et al . , 2003 ) , providing evidence that an optimal elongation rate is essential for a co-transcriptional splicing ( Davis-Turak et al . , 2018; Fong et al . , 2014 ) . Native elongating transcript sequencing in human cells ( NET-seq ) indicates that splicing occurs soon after introns are synthesized ( Mayer et al . , 2015; Nojima et al . , 2015 ) . Further , the combination of single molecule intron tracking ( SMIT ) and long read sequencing in yeast shows that splicing is 50% complete when Pol II is 45 nt downstream the acceptor spice site ( Oesterreich et al . , 2016 ) . Despite these advances , the in vivo kinetics of splicing remain poorly understood . In particular , different estimates for splicing rates have been reported . Splicing rates have been measured for selected endogenous human genes with the use of live cell imaging ( Coulon et al . , 2014; Martin et al . , 2013; Rino et al . , 2014; Schmidt et al . , 2011 ) or with a combination of cellular RNA extraction and quantitative PCR ( Pandya-Jones and Black , 2009; Singh and Padgett , 2009 ) . Such studies led to very different splicing rate estimates , ranging from 15 to 30 s ( Huranová et al . , 2010; Martin et al . , 2013; Rino et al . , 2014 ) to 4 . 3-10 min ( Coulon et al . , 2014; Schmidt et al . , 2011; Singh and Padgett , 2009 ) per splicing event . These discrepancies may stem from the difference in methods used , which can introduce perturbations , from the selection of genes studied , and from the great variance in intron lengths between human genes . Other studies have estimated in vivo splicing rates globally with the use of RNA sequencing technologies . The short sequence reads collected from steady-state in vivo samples reflect RNA synthesis , splicing and degradation , which are entangled ( Wachutka and Gagneur , 2017 ) . In particular , the ‘splicing efficiency’ is typically defined by the ratio of spliced over unspliced RNAs at steady state ( Braberg et al . , 2013; Wilhelm et al . , 2008 ) . However , the same ratio has been successfully employed to study RNA stability , with the assumption that unspliced RNA levels reflect RNA synthesis and spliced RNA levels reflect the ratio of RNA synthesis over degradation ( Gaidatzis et al . , 2015; Zeisel et al . , 2011 ) . This ambiguity in definitions and interpretations questions the use of splicing efficiency and calls for alternative concepts and metrics . To overcome the limitations of steady-state transcriptome sequencing , one approach is to sequence new transcripts from chromatin-associated RNA fractions and compare them to cytoplasmic fractions , which led to splicing rate estimates from 43 s ( Davis-Turak et al . , 2015 ) to 15–120 min ( Bhatt et al . , 2012; Pandya-Jones et al . , 2013 ) per splicing event . An alternative method to investigate splicing kinetics in vivo is the use of metabolic RNA labeling with 4-thiouracil ( 4sU ) ( Dölken et al . , 2008; Rabani et al . , 2011; Rabani et al . , 2014; Windhager et al . , 2012 ) coupled to sequencing of the labeled RNA ( 4sU-seq ) . We previously combined 4sU-seq with kinetic modeling to obtain RNA synthesis , splicing , and degradation rates in the fission yeast S . pombe ( Eser et al . , 2016 ) . Others have used 4sU-seq and similar approaches to obtain median splicing rate estimates in human cells of 6 . 7 min ( Mukherjee et al . , 2017 ) or 14 min ( Rabani et al . , 2014 ) per splicing event . However , 4sU-seq also introduces biases when applied to the human system because the obtained data are artificially biased toward pre-existing 5’-regions of the RNA due to the length of human genes ( Schwalb et al . , 2016 ) . These RNA 5’-regions predate the labeling time and are generally already observed to be spliced by 4sU-seq , leading to potential errors in the rate estimates . As a result of these difficulties , in vivo splicing kinetics remain unclear , and individual rate estimates for the two steps of splicing are lacking . However , such information is highly desirable because it may be interpreted alongside the mechanistic information obtained in vitro to provide a better understanding of the splicing process . To study kinetics of splicing in vivo , we performed TT-seq ( Schwalb et al . , 2016 ) after different 4sU-labeling time points in human K562 cells . In contrast to 4sU-seq , the TT-seq protocol includes RNA fragmentation before 4sU-labeled RNA is purified and sequenced . This is a crucial step that eliminates 5’ regions of nascent RNAs that were already transcribed , and spliced , prior to incorporation of the label , and thus removes the 5’-bias . We developed a computational approach that estimates the metabolic rates of single phosphodiester bonds . This approach enabled uncoupled quantification of donor- and acceptor-specific kinetics and to relate these to the two transesterification reactions and to the contribution of single nucleotides in spliceosome intermediates defined by structural studies . Moreover , to calculate the amount of precursor RNA successfully spliced into mature RNA , we introduced the ‘splicing yield’ as the conversion efficiency of unspliced to spliced RNA . As a result , our analysis provides genome-wide metabolic rates for donor and acceptor splice sites and identifies RNA-RNA interactions in the spliceosome that could contribute to in vivo splicing kinetics . Furthermore , we provide genome-wide estimates of the splicing yield that is not biased by splicing kinetics . From this work emerges a comprehensive global view of splicing kinetics and yield in human cells .
To monitor global RNA metabolism in human cells , we performed TT-seq analysis in K562 cells after different times of RNA labeling with 4-thiouracil ( 4sU ) ( Figure 1A ) . We previously showed that such a labeling time series can estimate splicing rates in the yeast S . pombe ( Eser et al . , 2016 ) . We exposed K562 cells to 500 μM of 4sU for a labeling time of 2 , 5 , 10 , 15 , 20 , 30 , or 60 min , isolated RNA , and conducted both TT-seq and total RNA-seq ( Figure 1A ) . On average , we obtained 250 million and 55 million 150-nucleotide ( nt ) paired-end reads for each of the TT-seq and RNA-seq samples , respectively . We next mapped TT-seq and RNA-seq data to the human genome ( Materials and methods ) . The experiments were highly reproducible ( Figure 1—figure supplement 1A ) . Visual inspection of the mapped reads revealed strong TT-seq signals in transcribed regions covering both introns and exons ( Figure 1B ) , whereas RNA-seq data covered mainly exons ( Figure 1B , Figure 1—figure supplement 1B ) . The relative number of intronic reads in TT-seq data decreased with 4sU-labeling time , whereas the signal for exons increased . These observations were consistent with capture of newly synthesized precursor RNA because spliced introns are more rapidly degraded than exons that are maintained in mature , stable RNA . Thus , our time series data contained information about the kinetics of precursor RNA splicing that we exploited further . We first analyzed our TT-seq data for the occurrence of reads that are informative of precursor RNA splicing , that is reads spanning exon-exon boundaries ( ‘split reads’ ) and reads spanning exon-intron ( ‘donor’ ) and intron-exon ( ‘acceptor’ ) boundaries ( ‘non-split reads’ ) . Using a threshold of at least 10 split reads for an exon-exon boundary , we found 341 , 855 putative introns ( Materials and methods ) . The relative number of these non-split reads compared to split reads was highest after 2 min of 4sU-labeling and progressively decreased with longer labeling times , eventually converging to a similar coverage as in the RNA-seq samples ( Figure 1—figure supplement 1C ) . Coverage with split reads decreased with the distance to the transcription start site ( TSS ) in 4sU-seq data but not in TT-seq data ( Figure 1C , data from Schwalb et al . , 2016 ) , suggesting that the previously reported estimation of splicing rates with 4sU-seq ( Mukherjee et al . , 2017; Rabani et al . , 2014 ) , overestimated splicing rates for 5’ introns compared to 3’ introns . With the use of TT-seq we could avoid a 5’-bias in splicing rate estimations . About one half of the putative introns ( 177 , 322 ) mapped to splice sites that had been annotated in the database of transcribed regions GENCODE ( Materials and methods ) , whereas the other half did not ( 164 , 533 ) . Of these putative introns that had not been previously annotated as splice sites in GENCODE , more than 99% represented introns ending with the GT|AG canonical dinucleotides . Moreover , 21% represented new combinations of already annotated donors and acceptors ( Figure 1D ) . Furthermore , 18% map to a non-annotated donor site and to a previously annotated acceptor site , 22% to a non-annotated acceptor site and to a previously annotated donor site . Interestingly , another 38% mapped to both non-annotated donor sites and non-annotated acceptor sites . Of these , about one half was located within an annotated GENCODE gene ( intronic ) , whereas the other half was located in regions of the genome not annotated in GENCODE ( intergenic ) . Overall , this analysis indicates that the number of splice sites has previously been underestimated , in agreement with recent studies that integrated very large datasets of the public RNA-seq repository ( Nellore et al . , 2016 ) or studies that used full-length mRNA sequencing ( Anvar et al . , 2018 ) . Defining the kinetics of RNA synthesis , splicing , and degradation from short-read-based protocols is inherently ambiguous due to the many RNA species overlapping any genomic position , including precursor RNAs and multiple splice isoforms . In the future , quantitative and high-throughput full-length transcriptome sequencing may become available to improve the situation; however , co-transcriptional alternative splicing would still cause ambiguities . We have therefore shown it is accurate to analyze the metabolism of phosphodiester bonds rather than RNA species themselves ( Wachutka and Gagneur , 2017 ) . Following this idea , we modeled the steady-state rates of synthesis and degradation ( or equivalently cleavage ) of each of three different phosphodiester bonds individually: the exon-intron bond at the donor site , the intron-exon bond at the acceptor site , and the bond between the two joined exons after successful exon ligation to yield product RNA ( Figure 2A left , Appendix ) . We refer to these definitions throughout when we use the terms ‘splicing kinetics’ or ‘splice site kinetics’ . We then considered the metabolism of these three phosphodiester bond types at steady-state . Synthesis balances out degradation at steady-state for any molecular species , independently of the kinetics . The steady-state synthesis rate ( amount produced per unit of time ) and the steady-state degradation rate ( ratio of steady state amount by the steady-state synthesis rate ) are defined quantities without any assumption on the kinetics ( Appendix ) . Synthesis of the donor and acceptor bonds reflects precursor RNA synthesis . Cleavage of the donor bond is caused by either splicing or by precursor RNA degradation . The half-life of those donor bonds that get spliced depends on intronic transcription at least up to the branchpoint and on the first transesterification step . Cleavage of the acceptor bond is caused by either splicing or by precursor RNA degradation . The half-life of those acceptor bonds that get spliced is determined by the first and the second transesterification step but not by intronic transcription . Synthesis of the junction bond is the outcome of completed splicing . Cleavage of the junction bond indicates RNA degradation ( Materials and methods , Appendix ) . Our experimental design includes the injection of labeled and unlabeled spike-ins at constant concentrations in all samples , prior to the purification step . These spike-ins allowed for estimating the variations in sequencing depth as well as the overall newly synthesized RNA fraction of every sample ( Materials and methods ) . The unlabeled spike-ins also allowed estimating the amount of cross-contamination , that is of unlabeled RNAs that are purified and which can represent a large fraction of all RNA-seq reads at short labeling durations ( Materials and methods ) . These technical parameters estimated from the spike-ins read counts were then used as covariates to model expected read counts of all three types of bonds in each sample . Using these considerations , we fitted the abundance of each of these three types of bonds with a first-order kinetic model for a total of 162 , 134 donor , 177 , 543 acceptor and 156 , 825 junction bonds that showed at least 100 supporting reads across the full dataset ( Figure 2A right , Materials and methods , Appendix ) . Overall , TT-seq read counts agreed with the expected counts of our kinetic model ( Figure 2A central column , Figure 2—figure supplement 1A ) . The synthesis rates for donors and acceptors , and the product half-life inferred from distinct splice junctions ( Materials and methods ) agreed well , demonstrating the robustness of our approach ( Spearman rank correlation >0 . 33 for synthesis time downstream of the first exon , p<2 × 10−16 and Spearman rank correlation >0 . 76 for half-life , p<2 × 10−16 , variation of 180% fold for synthesis rate , and 32% fold for half-life , Materials and methods , Figure 2—figure supplement 1B ) . Variations were larger for synthesis rates because these estimates are in a first approximation proportional to the coverage in the short-labeled TT-seq samples and are therefore more sensitive to sequencing biases . In contrast , half-lives , which are in a first approximation proportional to the ratio of coverages in short-labeled TT-seq samples and in RNA-seq , better control for sequencing biases . We furthermore conducted extensive simulations to assess the performance and limitations of the fitting procedure to estimate the rates when the ground truth is known . We simulated counts based on the estimated distributions of synthesis rates , splicing half-times and half-life times based on the experimental data . Based on simulated data , our method leads to unbiased estimates of ground truth synthesis rates ( Appendix 1—figure 12 ) , splicing half-time and half-life time ( Appendix 1—figure 13 ) with high precision compared to the dynamic range . Also , we used simulations to explore how estimation accuracy is affected when using data with much lower read coverage or for extremely slow or fast rates . These simulations showed that lowering the total read coverage cut-off below 100 reads would lead to relative errors typically surpassing 100% ( median , Appendix 1—figure 23 ) . These simulations also showed that estimations of half-lives shorter or much longer than our labeling durations ( shorter than 1 min or longer than 3 days ) would lead to median error surpassing 100% ( Appendix 1—figure 24 ) . First-order kinetic models are simple models that grossly model the underlying biochemical processes . We also investigated two alternative models that potentially capture more complex kinetics . The first one is a delay differential equation model for donor bond kinetics that modeled the time to transcribe the intron up to the branchpoint with a delay , followed by first-order kinetics for the first transesterification step ( Appendix ) . Simulations indicated that identifying the parameters of this delay differential equation model is difficult ( Appendix 1—figures 2 , 20–22 ) because the data do not support distinguishing the contribution of transcription from the one of the first transesterification step . However , fitting a first order kinetic model on data simulated according to the delay differential equation model showed that the estimated donor bond half-life approximately equated the sum of the intronic transcription delay and the half-time of the first transesterification step ( Appendix 1—figure 5 , yet usually underestimating with a median relative level of 0 . 89 ) . The second alternative model is a coupled differential equation model for the junction bonds that modeled junction formation as the outcome of a first-order kinetics splicing process rather than as a constant . Simulations showed that the data did not allow to easily distinguish this coupled kinetics from first-order kinetics ( Appendix 1—figure 3 ) . Moreover , the junction bond half-life estimated by the first order kinetics model approximately equated the sum of the splicing half-time and of the half-life of the processed RNA ( Appendix 1—figure 7 , yet usually overestimating with a median relative level of 1 . 2 ) . Unless specifically mentioned , we used the first order kinetics model in the remaining analyses because of its robustness and its approximate equivalence with alternative models ( Appendix ) . Based on GENCODE annotation , the analyzed bonds mapped to 8 , 770 mRNAs , 162 RNAs antisense to protein-coding genes ( asRNA ) , 204 long intergenic non-coding RNAs ( lincRNA ) , and 290 other non-coding RNAs ( Figure 2—figure supplement 1C ) . When averaged within major isoforms ( Materials and methods ) , synthesis rates and half-lives of donors and acceptors ranged over two orders of magnitude ( 42-fold and 48-fold change , 90% equi-tailed range ) , whereas the junction bond half-life spanned only slightly over one order of magnitude ( 8 . 1-fold 90% equi-tailed range , Figure 2B–D ) . These major isoform aggregated rates generally agreed with previously reported splicing rates and RNA half-lives ( Figure 2—figure supplement 1D; Mukherjee et al . , 2017 ) . Moreover , mRNAs were spliced significantly faster ( median of 7 . 2 min ) than lincRNAs ( median of 11 min , p<2 × 10−16 ) and other non-coding RNAs ( Figure 2C ) . Also , mRNAs had junction bonds with the longest half-lives ( median of 316 min ) ( Figure 2D ) , consistent with previous studies ( Mukherjee et al . , 2017; Schlackow et al . , 2017; Schwalb et al . , 2016 ) . Similar conclusions can be reached using site-specific rates ( Figure 2—figure supplement 1E–J ) . The obtained apparent splicing times in the range of minutes agree with many previous estimates that were obtained using different methods , but argue against fast splicing , within less than a minute , that was suggested by some studies ( Carmo-Fonseca and Kirchhausen , 2014 ) . Intron length has been suggested to affect splicing kinetics ( Hicks et al . , 2010; Khodor et al . , 2011; Pai et al . , 2017; Proudfoot , 2003; Windhager et al . , 2012 ) , and we therefore investigated this further . First , our de novo annotation of introns is in agreement with a minimal intron length of about 80 nt ( Figure 3A ) , as expected from the spatial needs within the spliceosome ( Ruskin et al . , 1985; Wieringa et al . , 1984 ) . Second , we find that among introns shorter than 2 , 000 nt , acceptor and donor bond half-life showed similar distributions and decreased with increasing intron length ( Figure 3B ) . The reasons for why short introns are spliced more slowly than long ones remain to be investigated . It is possible that for longer introns the splice site definition by the following exon facilitates splicing and that there are less restraints for splicing for longer introns . This observation also strongly argues for pre-ordering of the spliceosome on the transcribing polymerase . Our analysis also reveals that donor and acceptor bond half-lives differ for long introns . Among introns longer than 2 , 000 nt , acceptor bond half-life plateaued at a median value of about 4 min , whereas donor bond half-life increased with intron length up to a median value of about 8 min for introns larger than 7 , 700 nt ( last septile ) . A possible explanation for this significant difference is that donor sites of long introns are transcribed much earlier than acceptor sites and splicing can only start when the intron is transcribed . Indeed , the donor bond half-life is determined by the elongation time needed to transcribe at least the branchpoint and by the first transesterification step , whereas the acceptor bond half-life is determined by both the time for the first transesterification step and for the second transesterification step to be completed . Assuming a maximum polymerase elongation velocity of 4 kb/min ( Fuchs et al . , 2014; Gressel et al . , 2017; Jonkers and Lis , 2015; Saponaro et al . , 2014; Veloso et al . , 2014 ) , we observed very few introns violating this predicted limit ( Figure 3C ) . This limit for donor bond half-life affects only a small proportion of all introns ( last septile ) so that , overall , there is no positive correlation between donor bond half-life and intron length ( Figure 3C ) . For shorter introns , the donor bond half-life and the acceptor bond half-life were similar ( Figure 3B ) , indicating that the second transesterification step is fast compared to the overall splicing kinetics . Another prediction of this model is that for slowly transcribed introns the donors should take longer to cleave than the acceptors . Consistent with this hypothesis , the median half-life was 2 . 5-fold ( p<2×10−16 ) longer for donor bonds than for acceptor bonds of first introns ( Figure 3D ) , which are known to be more slowly transcribed ( Danko et al . , 2013; Fuchs et al . , 2014; Jonkers and Lis , 2015; Saponaro et al . , 2014; Veloso et al . , 2014 ) . A small significant difference was also found for the last intron ( 1 . 4-fold , p<2×10−16 ) , which could reflect slower polymerases near the transcript end or different kinetics of splicing of the last intron ( Davis-Turak et al . , 2015; Rigo and Martinson , 2008 ) . In conclusion , these data show that donor half-life and thus the beginning of splicing is limited by transcription elongation for long introns . Taken together , our results are generally consistent with the co-transcriptional nature of splicing and reveal that the length of the intron influences splicing kinetics in at least two different ways . Whereas overall trends in splicing kinetics can be explained by global features such as intron length and polymerase elongation velocity , the kinetics of splicing also critically depend on the RNA sequence context around the donor , acceptor , and branchpoint . To gain insights into the sequence determinants for splicing , we built a linear model ( Materials and methods ) that allowed us to estimate changes in donor bond half-life as a function of single nucleotide changes relative to the consensus sequence . The single nucleotide model could explain 19% of the observed variance in log-transformed donor bond half-life and achieved a median relative error for individual sites of 150% , which is small compared to the dynamic range across sites spanning two orders of magnitude ( Figure 4A ) . This analysis showed that nucleotide deviations from the consensus splice-site increase donor bond half-life . These findings are consistent with evolutionary pressure for donor sequences optimized for fast splicing . In order to elucidate the contribution a single nucleotide change might have on interactions within the spliceosome , we compared our predicted single nucleotide effects with base interactions observed in three different spliceosome structures ( Figure 4B , bottom ) . Recognition of the donor site by U1 snRNP plays a crucial role during early spliceosome assembly . RNA-RNA interaction between precursor RNA and U1 snRNA are mainly stabilized through Watson-Crick interactions ( Figure 4C; Kondo et al . , 2015 ) . In our model , substitution of the highly frequent G at +1 or −1 from the donor site with a C resulted in an increase in bond half-life ( Figure 4B ) , likely because C cannot form a stable interaction with a C in the U1 snRNA , in agreement with previous in vitro studies ( Kondo et al . , 2015 ) . In contrast , at position +3 from the donor site , a change from A to G has only a minor effect on the bond half-life ( Figure 4B ) , likely because G can still form a non-canonical base pair with U in the U1 snRNA ( Kondo et al . , 2015 ) . These results suggest that interactions of the precursor RNA donor region with U1 snRNA contribute to the observed donor bond half-lives . After donor site recognition by U1 snRNP , the Prp28 RNA helicase mediates the exchange of U1 with U6 snRNP and the U4/U5/U6 tri-snRNP can stably bind the precursor RNA . In the resulting B complex , the U5 stem loop interacts with the three terminal nucleotides of the 5´-exon , whereas the U6 ACAGA helix is formed near the donor site ( Figure 4D; Bertram et al . , 2017a ) . Our results suggest that U5 interactions may contribute to the kinetics of donor cleavage . For example , an A in the position −3 relative to the donor site leads to faster donor bond half-life supposedly because this enables base-pairing with U5 snRNA in the extended precursor RNA-U5 snRNA duplex . Completion of step-one results in the C complex that is then converted to the activated C* complex , which contains the two exon ends in close proximity for the step two reaction . RNA duplexes are formed between the intron region close to the donor site and U6 snRNA and between the branch site region and U2 snRNA ( Figure 5A ) ( Zhang et al . , 2017 ) . We also found that interactions in the C* complex were predictive of donor bond half-life ( Figure 4D ) . In particular , changes in the branchpoint adenine and at all positions −4 to +3 of the branchpoint show kinetic effects , except for the positions −1 and +2 that are predicted to not contribute to donor bond half-life ( grey highlighting in Figure 4D ) . In agreement with the structural data , these nucleotides are also the only two nucleotides in the vicinity of the branchpoint that do not interact with U2 in the C* complex ( Zhang et al . , 2017 ) . When compared to each other , the precursor RNA nucleotides interacting with snRNAs during 5’ splice site recognition showed the strongest effects on donor bond half-life , followed by nucleotides interacting in the B complex , and to a lesser extent the nucleotides interacting in the C* complex ( Figure 4—figure supplement 1A ) . Positions with no predicted contact in these structures showed least effects ( Figure 4—figure supplement 1A ) . These observations support our kinetic modeling , but also argue that the structurally characterized spliceosomal complexes represent functional states . Taken together , variation in the in vivo kinetics for the donor cleavage can in part be rationalized with early interactions of precursor RNA with U1 snRNA , and with later U5 snRNA interactions observed in structures of the B and Bact complexes . Moreover , the stability of the C* complex appears to also affect donor bond half-life , possibly because it prevents the reverse reaction of donor site cleavage ( Tseng and Cheng , 2008 ) . Since several precursor RNA positions are involved in different types of interactions in different splicing intermediates , the observed overall kinetics of donor cleavage reflect a combination of distinct microscopic rates , which cannot be distinguished by our in vivo approach . Furthermore , not all observed effects of nucleotide changes could be explained with available structures . For example , the first nucleotide of the downstream exon ( acceptor +1 position ) was important for donor cleavage kinetics . Although it remains unclear why , this could be related with co-transcriptional recruitment and recycling of splicing factors , maybe favored by Pol II 3´splice-site pausing , similar to that suggested in Aitken et al . ( 2011 ) . We also built a regression model predicting log-transformed acceptor bond half-life from sequence ( Figure 5B , 20% of variance explained , median relative error of 150% ) . Single nucleotide changes around the acceptor site generally had larger effects on acceptor bond half-life , reflecting effects on step two kinetics , whereas changes around the donor site had greater effects on donor bond half-life , reflecting effects on step one kinetics ( Figure 5C ) . The post-catalytic complex ( P complex ) , which is specific to step two splicing reaction , is not yet structurally characterized in human . Nevertheless , nucleotide changes that influence base pair interactions reported for the P-complex in S . cerevisiae ( Bai et al . , 2017 ) showed stronger effects on the acceptor bond half-life than for the donor bond half-life ( Figure 4—figure supplement 1A ) . Nucleotide positions in the precursor RNA that are not involved in base pair interaction with snRNAs in B-type and C* complex structures were irrelevant for predicting acceptor bond half-life ( grey highlighting in Figures 4B and 5C , feature selection , Materials and methods ) . Most nucleotides showed similar effects in the donor and acceptor bond half-life models but some noticeable differences were observed between them ( Figure 5D ) . Our results indicate that a non-canonical G branchpoint does not affect acceptor bond half-life but increases donor bond half-life . We also observed that the predominant G at the donor −1 nucleotide leads to fast donor cleavage kinetics but to slow acceptor cleavage kinetics , maybe because this interferes with positioning of the neighboring +1 donor nucleotide that serves as a nucleophile during step two . Despite this disadvantage in acceptor cleavage kinetics , the donor −1 position is predominantly G , presumably because this improves donor site recognition by base pairing with a C in U1 snRNA as described above . Taken together , available structural information on the spliceosome help to rationalize some of the effects of base changes around splice sites . Even though the contributions of several mechanistic processes to the observed kinetics cannot be disentangled , our results reveal which nucleotide positions around splice sites are critical for fast splicing kinetics . Splicing is modulated by auxiliary factors , including serine/arginine-rich proteins and hnRNPs ( heterogeneous nuclear ribonucleoproteins ) , that bind to regulatory motifs around the splice sites ( Fu and Ares , 2014; Matlin et al . , 2005; Wang and Burge , 2008 ) . We therefore aimed to identify putative regulatory motifs and to quantify their contribution to splicing kinetics . We derived two extended models for donor and acceptor bond half-life including the single nucleotide effects in the core regions investigated so far and all 65 , 536 possible RNA octamers in four extended intronic regions , 100 nt downstream of the donor site , 100 nt upstream of the acceptor site , and 100 nt upstream of the branchpoint and the region between branchpoint and acceptor site ( Figure 6A , Materials and methods ) . We also included intron length and GC content , but did not include exonic regions , because this would require isoform annotations . Because of the very large number of octamers , we used a feature selection method ( Lasso regression ) , which yielded 551 octamers jointly predicting donor bond half-life and 2 , 319 octamers jointly predicting acceptor bond half-life ( Materials and methods ) . Compared to the single nucleotide models , the extended models substantially increased the proportion of variance explained from 19% to 26% for the log-transformed donor bond half-life and from 20% to 29% for the log-transformed acceptor bond half-life ( Figure 6B and C ) . These proportions of variance increased when we restricted the analysis to junctions of major isoforms , showing that the results are not over-estimated due to double counting of donors and acceptors belonging to multiple exon junctions ( donor bond half-life model by 1 . 3% and the acceptor bond half-life model by 1 . 4% ) . Cumulatively , the proportions of variance explained of these non-overlapping regions largely exceed the proportion of variance explained by the joint model , indicating widespread co-occurrence of splicing-regulatory sequences across introns . The improved prediction of the extended models over the single nucleotide models is mostly attributable to the octamers of the extended intronic regions . The largest number of predictive octamers identified by the donor bond half-life model was found in the 5’ donor site region ( Figure 6—figure supplement 1A ) . This set of octamers was the most predictive feature for donor bond half-life individually ( 19% of the variance ) and the feature with the largest impact on variance when dropped from the joint model ( 3% of the variance ) . In the acceptor bond half-life model , the regions flanking the branchpoint and acceptor site contained most predictive octamers ( Figure 6—figure supplement 1A ) and associated with largest proportion of variance explained ( Figure 6C ) . Moreover , the effects of octamers on bond half-life were of similar order of magnitude than the effects of single nucleotides in donor , acceptor and branchpoint sites for both categories ( Figure 6—figure supplement 1B , median effect for octamer 1 . 4% , median effect for nucleotide 5 . 4% ) . The drop of proportion of variance explained when a feature was removed from the joint models were small ( between 0 . 0 and 3 . 4 , Figure 6B and C ) indicating of substantial correlation between the features . These correlations could be technical in the case of overlapping regions , or the result of co-evolution . Altogether , these results show that the octamers in the extended intronic regions contribute to splicing kinetics . To identify putative regulatory factors that could bind to the predicted RNA octamers , we scored the octamers binding affinities to the 159 human RNA-binding proteins of the ATtRACT database ( Giudice et al . , 2016 ) . We found 258 octamers predicted by the donor bond half-life model ( 47% versus 42% of non-selected octamers , p=0 . 017 , Fisher test ) associating with 69 RNA-binding proteins and 1 , 039 octamers identified by the acceptor bond half-life model ( 45% versus 42% of non-selected octamers , p=0 . 007 , Fisher test ) associating with 99 RNA-binding proteins motifs ( Figure 6D , Figure 6—figure supplement 1C , relative position weight matrix score >0 . 9 and selecting for the 5% highest absolute scores , Materials and methods ) . Our results suggest that several serine/arginine-rich and hnRNP proteins ( Supplementary file 5 ) regulate donor and acceptor bond half-life in both positive and negative fashions , depending on the location of their binding site . Octamers associated with the binding site of the polypyrimidine tract-binding protein Ptpb1 are predictive of short donor bond half-life when present between branchpoint and acceptor site but they prolong the donor bond half-life when located near the donor site ( Figure 6D ) . The remaining octamers may reflect cis-regulatory elements bound by splicing factors that remain to be characterized . To address the evolutionary conservation of the identified octamers , we aligned them to conserved sequences across 99 mammalian and other vertebrate genomes . Except for octamers predicted to affect donor bond half-life in the region 100 nt downstream of the donor site , the remaining ones show significantly higher phylogenetic conservation compared to a negative control of random octamers ( Figure 6E ) , providing evidence of their biological significance . Cleavage of the phosphodiester bonds at the donor and acceptor sites can lead to ligation of the two exon ends , thus completing splicing . However , cleavage of these bonds may also be non-productive in the sense that exon ligation can fail and RNA may be degraded after cleavage . To account for this , we defined the ‘splicing yield’ as the proportion of precursor RNA successfully converted into spliced RNA ( Figure 7A ) . A splicing yield of 1 means that all precursor RNAs that are synthesized are also successfully spliced , whereas a splicing yield less than one means that only a fraction of the precursor RNA is converted to spliced product . We estimated the splicing yield using the junction bonds modeled with the coupled model and the acceptor bonds modeled with first-order kinetics , because alternative kinetic models or using the donor bonds led to systematic biases ( Appendix ) . Hence , we did not computationally constrain our splicing yield estimates to be bounded by 1 . Due to estimation errors of the synthesis rates , yields sometimes turn out to be greater than 1 . We found that the splicing yield across sites was much higher for mRNAs ( median = 1 . 2 , Figure 7B ) than for antisense RNAs ( median = 0 . 2 ) , lincRNAs ( median = 0 . 3 ) , and other non-coding RNAs ( median = 0 . 5 ) , suggesting that degradation pathways are competing with splicing more intensively for non-coding RNAs than for coding RNAs . Moreover , the higher yield of mRNAs compared to lincRNAs also held when stratifying by cumulative read coverage across all samples and by half-life ( Appendix 1—figures 26 and 27 ) , two possible confounders associating with synthesis rate estimation biases in simulations ( Appendix ) . Furthermore , splicing yield was the same for the 139 , 344 ( 99 . 9% ) introns harboring the canonical terminal dinucleotides GU and AG recognized by the major spliceosome ( U2-type ) than for the 182 ( 0 . 1% ) introns harboring the terminal dinucleotides AU and AC recognized by the minor spliceosome ( U12-type ) ( Figure 7C ) . Although introns targeted by the minor spliceosome showed two-fold slower donor and acceptor bond half-lives compared to those targeted by the major spliceosome ( Figure 7D , Figure 7—figure supplement 1B ) , the minor spliceosome nonetheless reached similar splicing yields . To analyze how the nature of the splice sites contributes to splicing yield , we built a model that allow us to predict splicing yield based on sequence ( Figure 7E ) . Similar to the effects on bond half-lives ( Figure 4B , Figure 5C ) , deviations from the consensus sequence led to lower splicing yield . Furthermore , nucleotides near a splice site showed stronger effects than more distant ones , suggesting that the early recognition of donor and acceptor splice sites is a determinant for splicing yield . Taken together , these results indicate that rate and yield are distinct aspects of splicing that may have evolved independently and that the sequence around splice sites determines both the rate and the yield of splicing .
RNA splicing is an essential step of eukaryotic gene expression , but the in vivo kinetics of this two-step process and its dependence on transcription remain poorly understood . Here , we have coupled a metabolic RNA labeling time series to TT-seq analysis of new and total RNA to investigate RNA metabolism in human cells . We have then used kinetic modeling based on a definition of RNA metabolic rates at the level of individual phosphodiester bonds to provide rate estimations for cleavage of phosphodiester bonds at donor and acceptor splice sites in human introns . The obtained splice site cleavage rates , expressed as donor and accept bond half-lives , are free of ambiguities introduced by other methods and are related to the independent contributions of the two splicing steps in vivo . The donor and acceptor bond half-lives were found to be generally in the range of minutes , although we cannot exclude that we are missing a small population of quickly spliced introns . The donor and acceptor bond half-lives were found to depend on intron length , on the nucleotide sequence surrounding splicing sites , including the branchpoint , and on flanking octamer sequences that may bind regulatory factors . This is consistent with a complex relationship between the splicing machinery and its nuclear environment , in which splicing rates can be influenced not only by RNA sequence but also by gene structure and chromatin landscape ( Davis-Turak et al . , 2018; Davis-Turak et al . , 2015 ) . In addition , we define the yield of successful splicing and show that it differs dramatically between different RNA classes . Our results also provided insights into the nature and evolution of co-transcriptional splicing . Previous studies of splicing kinetics in mouse ( Rabani et al . , 2014 ) and fission yeast S . pombe ( Eser et al . , 2016 ) all found that splicing is faster for shorter introns . A recently published study in Drosophila ( Pai et al . , 2017 ) demonstrated that splicing is faster for intron-defined , short introns ( 60–70 nt ) , whereas for exon-defined introns , splicing was faster for introns longer than 2 , 944 nt , suggesting a complex relationship between splicing kinetics and intron length . We find here that exon-defined splicing in human cells is fastest for introns with a length of around 2 , 000 nt , whereas short introns ( <140 nt ) on average take about twice as long to be spliced . High spatial and temporal resolution kinetics data coupled with focused analyses of these very short introns remain to be performed for other species to understand how universal these observations are . Another observation we made was that longer human introns ( >10 , 000 nt ) show increased donor bond half-life , apparently because donor cleavage requires RNA polymerase II to first transcribe the intron . This effect of transcription-limited splicing is observed at ~10% of human precursor RNA introns , whereas most human introns are short enough so that their splicing kinetics are not limited by transcription . How the polymerase elongation rate depends on the nature of the intron and how this influences splicing remains to be investigated . The observed relationship between transcription and splicing of coding RNAs extends to non-coding RNAs . We found that spliced coding RNAs and spliced non-coding RNAs showed similar RNA synthesis rates ( Figure 2C ) , whereas previous studies reported considerably slower synthesis rates for mainly unspliced , non-coding RNAs ( Mukherjee et al . , 2017 ) . Our definition of splice site-specific RNA cleavage rates also allowed for a comparison of kinetic information in vivo with detailed structural knowledge of the spliceosome in different functional states obtained in vitro . Structure-based interpretation of our nucleotide-resolution kinetic results highlights the importance of interactions between snRNAs in the spliceosome and the precursor RNA substrate and provides guidance for interpreting interactions in structures of the spliceosome yet to be obtained . Our results suggest that the predicted interactions between the precursor RNA and snRNAs in the P complex , which have presently only been obtained in yeast ( Bai et al . , 2017; Wilkinson et al . , 2017 ) may be similar in human . We show that different types of RNA-RNA interactions observed at various stages of the process are related to the splice site cleavage time . In particular , RNA-RNA interactions must be of high enough affinity to allow for sufficiently specific recognition of splice sites , yet the affinities must be in a range that also allows for rapid conversion between subsequent states , which can show strongly altered RNA-RNA interactions . However , many processes and factors contribute to the observed apparent splicing rates , and these must be disentangled in the future . Taken together , we have analyzed the metabolism of individual donor and acceptor splice sites in vivo and provided quantitative models for how RNA splicing kinetics may be encoded in the human genome . As we looked at a single growth condition , our models are derived from comparisons across genes and essentially reflect the affinity of precursor RNAs to the core splicing machinery . However , the experimental and computational methodology presented here could be applied to different cell types or under dynamic responses to reveal and quantify the role of splicing regulatory factors and of their related binding sites . Another interesting future direction is the modeling of alternative splicing , which is understood to be the outcome of competitions of alternative donor or acceptor sites with various strengths . Our distinct models of donor and acceptor site kinetics may help to build up such quantitative competition models . Eventually , quantifying the contribution of individual bases to splicing rates , backed by structural and functional studies , may explain the numerous contributions of splicing to the genetics of rare ( López-Bigas et al . , 2005 ) and common ( Li et al . , 2016 ) diseases .
K562 cells were obtained from DSMZ ( DSMZ no . : ACC-10 ) and grown in RPMI 1640 medium ( Thermo Fisher Scientific , 31870–074 ) supplemented with 10% heat-inactivated fetal bovine serum ( Thermo Fisher Scientific , 10500–064 ) and 2 mM GlutaMAX ( Thermo Fisher Scientific , 35050087 ) at 37°C and 5% CO2 . Cells were routinely verified to be free of mycoplasma contamination using Plasmo Test Mycoplasma Detection Kit ( InvivoGen , rep-pt1 ) . K562 cells were authenticated at the DSMZ Identification Service according to standards for STR profiling ( ASN-0002 ) . TT-seq was performed as described ( Schwalb et al . , 2016 ) , with minor modifications . Specifically , 2 . 5 × 107 cells from two biological replicates were used for each time point . Cells were exposed to 500 µM of 4-thiouracil ( 4sU , Carbosynth , NT06186 ) for 2 , 5 , 10 , 15 , 20 , 30 , 60 min at 37°C and 5% CO2 . Cells were harvested by centrifugation at 600 g for 2 min at 37°C . Cell pellets were lysed in 5 mL of QIAzol ( Qiagen ) and 150 ng of RNA spike-ins mix were added to each sample . RNA spike-ins were produced in house , based on ERCC-RNA sequences ( sequences of spike-ins are described in Supplementary file 6 ) . RNA spike-ins were produced as described ( Schwalb et al . , 2016 ) . RNAs were extracted using QIAzol according to the manufacturer’s instructions . RNAs were sonicated to obtain fragments of <6 kbp using AFAmicro tubes in a S220 Focused-ultrasonicator ( Covaris Inc , parameters: 10 s , peak power 100 , cycles 200 , duty cycle 1% ) . The quality of RNAs and the size of fragmented RNAs were checked using Fragment Analyzer . 1 μg of each of the sonicated RNAs was stored at −80°C as total RNA ( RNA-seq ) and later eluted with miRNAeasy Micro Kit ( Qiagen , 217084 ) together with 4sU-labeled purified RNAs . 4sU-labeled RNAs were purified from 300 µg of each of the fragmented RNAs . Biotinylation and purification of 4sU-labeled RNAs was performed as described ( Dölken et al . , 2008; Schwalb et al . , 2016 ) . Biotinylated 4sU-labeled RNAs were separated from unlabeled RNAs with streptavidin beads ( Miltenyi Biotec , Bergisch Gladbach , Germany ) and eluted in 100 mM DTT as described in Dölken et al . ( 2008 ) and Schwalb et al . ( 2016 ) . 0 . 3M sodium acetate was added to 4sU-labeled purified RNAs and to total RNAs prior RNA extraction . RNAs were extracted and eluted using miRNAeasy Micro Kit ( Qiagen , 217084 ) . The on-column DNAse I treatment ( Qiagen , 79254 ) was performed for 15 min at 25°C . Prior to library preparation , total RNAs and 4sU-labeled purified RNAs were quantified using Qubit . Enrichment of 4sU-labeled versus unlabeled RNAs was analyzed by RT-qPCR using oligonucleotides amplifying selected regions of 4sU-labeled and unlabeled spike-ins ( sequences of oligonucleotides are described in Supplementary file 6 ) . Only 4sU-labeled purified samples showing ΔΔCt changes from 4 to 6 were subjected to library preparation ( total RNAs were used as a control for normalization ) . 100 ng of input RNA was used for strand-specific library preparation according to the Ovation Universal RNA-seq System ( NuGEN ) . Libraries were prepared using random hexamer priming only . The size-selected libraries were analyzed on a Fragment Analyzer before sequencing on the Illumina HiSeq 4000 . Paired-end 150 bp reads with additional 6 bp of barcodes were obtained for each sample . Reads were aligned using STAR version 2 . 5 . 0a ( Dobin et al . , 2013 ) in single pass mode . The genome Index was built against the full GENCODE version 24 annotation and the hg38 ( GRCh38 ) genome assembly ( Human Genome Reference Consortium ) using 150 bp overhang size . Additional specified parameters were alignSJDBoverhangMin 2 , chimSegmentMin 15 , chimScoreMin 15 , chimScoreSeparation 10 , and chimJunctionOverhangMin 15 . The aligned reads were filtered for duplicates using Picard tools version 2 . 5 . 0 ( https://broadinstitute . github . io/picard/ ) using the option MarkDuplicates REMOVE_DUPLICATES = true . In average , each TT-seq sample yielded about 250 M reads and each RNA-seq sample about 55 M reads . For each sample , ~90% of the reads could be uniquely mapped to the reference genome . The duplication ratio was estimated to 55% by FastQC ( https://www . bioinformatics . babraham . ac . uk/projects/fastqc/ ) . Using the rCube package ( https://github . com/gagneurlab/rCube ) , all split reads ( containing N stretches in Cigar string ) were extracted to create a database of potential introns ( ~341 k ) . The obtained introns were classified relative to annotated introns and genetic elements from the GENCODE annotation ( version 24 obtained from https://www . gencodegenes . org/releases/24 . html ) . For each intron three characteristic counts were calculated: The numbers of reads starting in the upstream exon and extending into the intron ( ‘donor’ ) , the number of reads starting in the intron and extending into the downstream exon ( ‘acceptor’ ) , and all split reads matching the introns coordinates ( ‘junction’ ) . The reads were filtered using a bam quality score of 255 . Reads having secondary alignment flag were discarded . To estimate the sample normalization factors Fj that account for variations in sequencing depth as well as the overall newly synthesized RNA fraction and the fraction of cross-contamination χj of non-labeled reads in the TT-seq data , we modeled the expectation of counts Eij of spike-in i in sample j using a statistical model similar to the one of Schwalb et al . ( 2016 ) . ( 1 ) Eij=Fjpij ( χj-δiχj+δi ) χj is set to 1 for all RNA-seq samples , δi is 0 for labeled spike-ins and 1 for unlabeled spike-ins . The parameter pij is the condition and spike-in specific extraction probability . The difference with ( Schwalb et al . , 2016 ) is to allow the parameter pij to be condition-specific ( TT-seq or RNA-seq ) , which turned out to model better cross-contamination of unlabeled RNA in the short duration TT-seq libraries . We set pij=pik for all sets of j and k belonging to either RNA-seq samples , TT-seq samples or if i belongs to a labeled spike-in . We assumed read count data to follow a negative binomial distribution with a common dispersion parameter for all data . The model parameters and the dispersion parameter were fitted as generalized linear model using maximum likelihood . For each detected intron i we modeled the concentrations ci , l of each of three characteristic bonds ( donor , acceptor , junction ) independently following a first order kinetic rate equation . Without loss of generality , we consider in the following just one of the three equations - the other two behave the same . ( 2 ) ddtci ( t ) =αi-βici ( t ) We assume that all newly synthesized RNAs are labeled . The concentration of labeled bonds , assuming an initial concentration of 0 , follows: ( 3 ) cit , labeled=αiβi ( 1-e-tβi ) Also , the old , non-labeled RNA decays exponentially as ci ( t , unlabeled ) =αiβie−tβi . Using the normalization factor Fj of sample j , the labeling time tj and χj the cross-contamination of unlabeled RNAs in the purified fraction , the concentration can be mapped to its expected count Ei , j: ( 4 ) Ei , j=Fjαiβi ( 1 +e−tjβi ( χj − 1 ) ) forTT-seq;Ei , j=FjαiβiforRNA-seq We modeled read counts using the negative binomial distribution , a count distribution often used for RNA-seq data because it captures sampling noise and further sources of variations . The kinetic parameters αi , βi are estimated by maximizing the log likelihood l=∑i , j log ( NBki , jEi , j ( αi , βi ) , θ , where ki , j are the observed counts , using the BFGS numerical optimization algorithm and using the dispersion parameter obtained from the spike-ins analysis . The optimization was initialized 10 times with independent random parameters; the final solution comprises the median of all αi , βi over the different runs to compensate for numerical instabilities . We removed all donors , acceptors , junctions with too few counts ( ∑jki , j<100 ) from the modeling . Using the table in Figure 2—figure supplement 1A we map the rates α , β to the characteristic kinetic parameters of donor , acceptor , and junction . The whole modeling approach was implemented in R and is available as apackage called rCube ( https://github . com/gagneurlab/rCube ) . Because the donor and acceptor bond half-life models work on a logarithmic scale , we present model errors as multiplicative errors given by the equation median ( exp ( |log ( yy^ ) | ) ) , with y as the observation and y^ the prediction . More details about kinetic models are provided in the Appendix . We applied the software Salmon ( Patro et al . , 2017 ) ( index kmer size = 31 ) to all RNA-seq samples and mapped them against the full transcriptome of the GENCODE ( Ver . 24 ) annotation . For each gene , we selected the isoform with the maximum mean TPM value across all RNA-seq samples as the major isoform . The major isoform was only used in the analyses in Figures 2B–D , 3D and 7B . Elsewhere , analyses were only relying on individual junction annotations . To estimate relative uncertainty of the kinetic parameters in a conservative way we assumed that all donors and acceptors of the major isoform of a given GENCODE gene shared the same synthesis rate equal to the transcription rate of the gene . We further assumed that all products ( ‘junctions’ ) shared the same half-life equal to the mature RNA half-life . Because noise of these rate estimates is typically multiplicative , we computed the standard errors of the logarithm of these rates and reported relative uncertainties as the exponential of these standard errors . Alignment , counting and estimation of normalization and cross-contamination factors of the RNA-seq data sets of Schwalb et al . ( 2016 ) was done as described above for our data . Counts for 4sU-seq and TT-seq was normalized using Ki , j^=Ki , jFj−χjKi , RNA-seqFRNA-seq , where i denotes the split / unsplit reads as shown if Figure 2A for each intron of the major transcripts and j is the sample ( 4sU- / TT-seq and replicate ) . Both replicates were pooled together . Due to the limited availability of experimental branchpoint measurements , the prediction algorithm LaBranchoR ( Paggi and Bejerano , 2018 ) was utilized to predict branchpoint positions within introns . We applied the model within the kipoi framework ( http://kipoi . org/ ) to score the last 100 nucleotides of each intron and took the nucleotide with the maximum score for being the utilized branchpoint . The results were validated using experimental data of Mercer et al . ( 2015 ) where available . We identified nucleotide positions not predictive of donor or acceptor bond half-life and estimated the effect of the remaining single nucleotides on the donor bond half-life and on the acceptor bond half-life by regression . To this end , we modeled log-transformed half-lives of the donor bonds ( and with a separate model of the acceptor bonds ) as a weighted sum of each of the 20 nucleotides upstream or downstream of the donor , acceptor site and branchpoint , as well as of the GC frequency of the whole intron , the donor site , and the acceptor site . In this linear model , the reference sequence was chosen to be the consensus sequence so that the coefficients can be interpreted as the effects of substituting a consensus nucleotide to an alternative nucleotide . Lasso regression ( Tibshirani , 1996 ) is a regularized linear regression method that can estimate some of the coefficients to be exactly 0 and that is therefore often used to select explanatory variables . We performed Lasso regression as implemented in glmnet ( Friedman et al . , 2010 ) , choosing the largest shrinkage parameter at which the mean squared error ( MSE ) was within one standard error of the minimal MSE using 10-fold cross-validation . For donor bond half-life , as well as for acceptor bond half-life , the Lasso regression fit led to several nucleotide coefficients to be exactly 0 . We then removed all the nucleotide positions where all single nucleotide effects had a coefficient equal to 0 . Next , we estimated the single nucleotide effects of all remaining positions as well as the effect of GC frequency of the whole intron , the donor site , and the acceptor site on log-transformed donor and acceptor bond half-lives using ordinary least squares regression . Images of spliceosome structures ( PDB code 4PJO , 5O9Z , 5XJC ) were drawn using Pymol ( https://pymol . org/ ) . The number of occurrences of all 65 , 536 nucleotide octamers in the regions 15–100 nt downstream of the donor site , 100 nt upstream of the branchpoint , all nucleotides between the branchpoint and the five nt upstream of the acceptor site and 5–100 nt upstream of the acceptor site were counted allowing for two mismatches . The 15 nt immediately downstream of the donor site or 5 nt upstream of the acceptor site were excluded from the octamer search space because they were already incorporated in the single nucleotide model . Regions extending in the upstream or downstream exon were cropped to keep them within the intron . The base 2 logarithm of octamer pseudo-counts log2 ( count +1 ) were used as covariates together with the GC frequency of the intron and the GC frequency of each region . The log-transformed donor/acceptor bond half-lives were the response variable . Lasso regression was applied to each region independently with 5-fold cross-validation to choose the optimal shrinkage parameter and select potential significant octamers . In a second step all selected octamers of each region were used together with the single nucleotide model as well as the GC frequency of the different regions , intron length and whether an intron is the first within the major transcript in a joint model to refine the selection of octamers ( Lasso 10-fold cross-validation ) . We compared each octamer to all reported PWMs with at least 5 nucleotides of the ATtRACT database and calculated the ratio between the probability of the best matching position ( PWM-score ) and the highest possible probability for any octamer ( RPM-score , Cook et al . , 2011 ) . Each octamer was padded with an equal number of ‘N’s at both sides if the PWM was longer than the octamer . We ranked all matches based on their RPM-score and kept only the best 5% for each PWM and removed afterwards all matches with a RPM-score less than 0 . 9 . The remaining matches were considered as hits . To calculate the phylogenetic conservation score for each octamer , we retrieved the PhastCons 100-way track ( http://hgdownload . cse . ucsc . edu/goldenpath/hg38/phastCons100way/ ) , which reports conservation across 99 vertebrates aligned to the human genome , and extracted the mean of all nucleotides for all matching positions . Octamers of the region 100 nt downstream of the donor site found to be predictive for donor site or acceptor bond half-life were also searched in the region 100 nt downstream of the donor site . Octamers of the region 100 nt upstream of the branchpoint or acceptor site or between the branchpoint and the acceptor site found to be predictive for donor or acceptor bond half-life were jointly searched in the region 100 nt upstream of the acceptor site , since these three regions were strongly overlapping . We also included as list of 2000 random octamers to estimate the background distribution in the same regions . We define the splicing yield of donor ηdonor and acceptor ηacceptor as follows: ( 5 ) ηdonor=∑{acceptor}αjunction ( donor , acceptor ) αdonorηacceptor=∑{donor}ajunction ( donor , acceptor ) aacceptorwhere αdonor and αacceptor denote the synthesis rates of the donor site and of the acceptor site phosphodiester bond , respectively , and αjunctiondonor , acceptor denote the synthesis rate of the spliced exon-exon phosphodiester bond utilizing the specified donor and acceptor . Since the first-order kinetic model does systematically underestimate junction synthesis rates and overestimate donor synthesis rates , we switched to the alternative kinetic models to estimate these rates . However , since the first-order kinetic model is more robust and the acceptor kinetics do not include a delay we used the first-order kinetic model for the estimation of the acceptor synthesis rate . We defined the intron splicing yield η as the acceptor site splicing yield because its estimation is more robust compared to the donor site splicing yield . All the code used for counting donor site , acceptor sites , and junction reads as well as estimating the kinetic rates is available in the R package rCube ( https://github . com/gagneurlab/rCube; Wachutka et al . , 2017 ) . The single nucleotide model is shared in the model repository Kipoi ( http://kipoi . org/models/CleTimer/; Avsec et al . , 2019 ) . The sequencing data and processed files were deposited in NCBI Gene Expression Omnibus ( GEO ) database under accession code GSE129635 . | Genes are portions of DNA that carry the instructions to build proteins . In particular , they are formed of segments called exons , which contain the protein-building information , and of non-coding segments known as introns . Exons and introns alternate within a gene . To create a given protein , the cell first uses an enzyme , Polymerase II , to copy the entire related gene – including introns and exons – into a molecule of ribonucleic acid , or RNA . As the gene is copied , a machine called the spliceosome comes onto the RNA molecule to remove the introns and create the final RNA template used to produce proteins . The spliceosome works by recognizing specific sequences that signal the border between introns and exons . Once the machine is bound to these ‘splice sites’ on each side of an intron , it brings the two neighboring exons close together and cuts out the intron . The two ends of the exons are then attached together . Previous studies have measured how fast introns are removed , but it remained unclear how long it takes to cut individual splice sites genome-wide . To address this question , Wachutka , Caizzi et al . combined a mathematical approach with a biochemical method that purifies newly made RNA in human cells . The experiments showed that it only took a few minutes to cut most splice sites . Cutting splice sites that bordered very long introns was slower , presumably because the Polymerase II took longer to produce these introns . In addition , the genetic sequences of the splice sites affected the time it took to remove the introns: some made it harder for the spliceosome to recognize where to cut , but others made it easier . Mistakes in removing introns from RNA can lead to producing abnormal proteins , and many diseases such as cystic fibrosis and Duchenne muscular dystrophy can be caused by such errors . In particular , small changes in the sequences at the splice sites or in the surrounding areas can create problems when it comes to eliminating introns . Decrypting the dynamics of intron cutting and removal may give scientists new insight into the molecular causes of cystic fibrosis and many other genetic disorders . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"computational",
"and",
"systems",
"biology",
"tools",
"and",
"resources",
"genetics",
"and",
"genomics"
] | 2019 | Global donor and acceptor splicing site kinetics in human cells |
The restricted host tropism of hepatitis C virus ( HCV ) remains incompletely understood , especially post-entry , and has hindered developing an immunocompetent , small animal model . HCV replication in non-permissive species may be limited by incompatibilities between the viral replication machinery and orthologs of essential host factors , like cyclophilin A ( CypA ) . We thus compared the ability of CypA from mouse , tree shrew , and seven non-human primate species to support HCV replication , finding that murine CypA only partially rescued viral replication in Huh7 . 5-shRNA CypA cells . We determined the specific amino acid differences responsible and generated mutants able to fully rescue replication . We expressed these mutants in engineered murine hepatoma cells and although we observed increases in HCV replication following infection , they remained far lower than those in highly permissive human hepatoma cells , and minimal infectious particle release was observed . Together , these data suggest additional co-factors remain unidentified . Future work to determine such factors will be critical for developing an immunocompetent mouse model supporting HCV replication .
Every year , an estimated 3–4 million individuals become newly infected with hepatitis C virus ( HCV ) ( Westbrook and Dusheiko , 2014 ) , with 60–80% going on to join the population of approximately 71 . 1 million who are chronically infected ( Blach et al . , 2017 ) . A hepatotropic virus , HCV has only been shown to robustly infect in vivo human and chimpanzee hepatocytes . This limited host tropism has proven problematic in developing an animal model that is not only ethically and financially sound but also immunocompetent . Such a model would allow the study of the underlying immunopathogenesis of HCV infection as well as the development and testing of vaccine candidates . Most efforts to generate such an in vivo model have focused on mice , which are amenable to genetic manipulation and have many well-established research tools built around their use . Overcoming the natural imperviousness of murine hepatocytes to HCV has required adjustments at multiple stages of the virus life cycle . Barriers at the level of entry ( McCaffrey et al . , 2002; Park et al . , 2009 ) in murine cells could not be overcome until the identification of the four canonical HCV human entry factors – claudin-1 ( CLDN1; Evans et al . , 2007 ) , occludin ( OCLN; Liu et al . , 2009a; Ploss et al . , 2009 ) , CD81 ( Pileri et al . , 1998 ) and scavenger receptor class B member 1 ( SCARB1; Scarselli et al . , 2002 ) . Of these , human CD81 and human OCLN were the minimal factors needed for viral entry ( Ploss et al . , 2009 ) . Alternatively , infecting with HCV adapted to utilize murine CD81 could also successfully overcome this initial obstacle ( Bitzegeio et al . , 2010 ) . Once inside murine cells , HCV faces another block at the level of replication , which initial studies circumvented through the use of selectable subgenomic HCV replicons ( McCaffrey et al . , 2002; Park et al . , 2009; Uprichard et al . , 2006 ) . Murine cells did appear to support assembly and release of infectious particles , albeit at low levels , once NS2 and the structural HCV proteins were provided in trans and the murine or human ortholog of the apolipoprotein ApoE was overexpressed ( Long et al . , 2011 ) . Further efforts to improve HCV replication in a murine context have relied on disruption of innate immune responses ( Aly et al . , 2011; Anggakusuma et al . , 2015; Chang et al . , 2006; Frentzen et al . , 2014; Lin et al . , 2010; Nandakumar et al . , 2013 ) , with robust replication and completion of the HCV life cycle proving difficult following infection unless a selectable genome is used ( Vogt et al . , 2013 ) . These poor levels of replication could be due to the lower compatibility of murine orthologs of vital replication host factors with the viral replication machinery or the absence of proteins that normally facilitate these interactions in human cells . Although numerous intracellular host factors have been implicated in HCV replication , extensive experimental evidence has only been provided for three host factors: cyclophilin A ( CypA ) ( Kaul et al . , 2009; Yang et al . , 2008a ) , phosphatidylinositol four kinase IIIα ( PI4KA ) ( Berger et al . , 2009; Borawski et al . , 2009; Reiss et al . , 2011; Tai et al . , 2009; Trotard et al . , 2009 ) and microRNA-122 ( miR-122 ) ( Jopling et al . , 2005; Lanford et al . , 2010; Machlin et al . , 2011 ) . miR-122 is highly conserved between humans and other species , including mice , making its expression alone unlikely to explain the weak replication of HCV in murine cells . In an effort to systemically dissect the impact of such replication co-factors during infection , we focused in this study on CypA and how it may contribute to the restricted host range of HCV . CypA is a cytosolic 18 kDa peptidyl-prolyl cis-trans isomerase ( PPIase ) and a part of the biologically ubiquitous cyclophilin enzyme family ( Fischer et al . , 1989 ) , the members of which were first characterized in mammals by their common ability to bind the immunosuppressive drug cyclosporin A ( CsA ) and their shared cyclophilin-like domain ( CLD ) which catalyzes the cis-trans isomerization of proline residues ( reviewed in Marks , 1996 ) . CypA overexpression has been implicated in a wide variety of human diseases , ranging from cancer to atherosclerosis ( reviewed in Nigro et al . , 2013 ) , and it has a demonstrated role in the life cycles of multiple viruses besides HCV ( de Wilde et al . , 2018; Frausto et al . , 2013; Li et al . , 2016; Phillips et al . , 2015; Tian et al . , 2010; von Hahn and Ciesek , 2015; Watashi and Shimotohno , 2007; Zhou et al . , 2012 ) . Early work showed that CsA had an inhibitory effect on HCV in chronically infected chimpanzees , but it was not until subsequent in vitro CypA knockdown experiments and dose-response assays with CsA derivatives that CypA was specifically recognized as critical to HCV replication ( Chatterji et al . , 2009; Ciesek et al . , 2009; Coelmont et al . , 2009; Kaul et al . , 2009; Liu et al . , 2009b; Yang et al . , 2008b ) . These studies showed that CypA’s relevance to HCV replication was intimately linked to its PPIase activity , as the introduction of point mutations in the PPIase active site led to impaired viral replication ( Chatterji et al . , 2009; Kaul et al . , 2009; Liu et al . , 2009b ) . Individuals exhibiting an HCV non-permissive phenotype were shown to express a rare homozygosity at any of three SNP sites in the coding region of CypA – but not in the enzymatic active site – that subsequent in vitro work showed resulted in markedly decreased levels of intracellular CypA ( von Hahn et al . , 2012 ) . Despite its known importance , the exact mechanism by which CypA facilitates HCV replication remains poorly characterized . Interactions between CypA and several HCV proteins have been demonstrated , including the RNA-dependent RNA polymerase ( RdRP ) NS5B ( Chatterji et al . , 2009; Fernandes et al . , 2007; Robida et al . , 2007; Yang et al . , 2008b ) , NS5A ( Anderson et al . , 2011; Coelmont et al . , 2010; Foster et al . , 2011; Grisé et al . , 2012; Hanoulle et al . , 2009; Nag et al . , 2012; Verdegem et al . , 2011 ) and NS2 ( Ciesek et al . , 2009; Kaul et al . , 2009 ) , but how CypA’s binding , PPIase activity and the viral polyprotein are precisely intertwined remains to be understood . The specific impact that cross-species differences in CypA might have on HCV replication and the restricted host tropism of this virus remains an open question and one the present study sought to address . Here , we examined the ability of CypA from diverse species , some of which could serve as feasible small animal models for HCV , to facilitate HCV replication . We found that murine CypA , relative to human CypA , is less proficient at facilitating HCV replication due to differences at the amino acid level and that overexpression of human CypA can increase replication in an engineered murine hepatoma line .
Knowing the critical role of human CypA in facilitating HCV replication , we first examined the conservation of CypA at the amino acid level across diverse species , focusing on those with promise to serve as biomedical research models and/or closely related to humans . As observed in mice , previous in vivo studies have suggested that several non-human primate ( NHP ) species – including cynomolgus , Japanese , and rhesus macaque; African green monkey; and Chacma and doguera baboons – appear resistant to HCV infection ( Abe et al . , 1993; Bukh et al . , 2001 ) . In contrast , more recent work in vitro demonstrated that primary hepatocytes from rhesus macaques ( PRMH ) ( Scull et al . , 2015 ) as well as hepatocyte-like cells derived from pigtailed macaque induced pluripotent stem cells ( iPSCs ) could support the HCV life cycle ( Sourisseau et al . , 2013 ) . Importantly , pharmacological-mediated suppression of innate immune responses via Jak inhibition enhanced viral replication in PRMH ( Scull et al . , 2015 ) . Additionally , albeit with limited evidence , tree shrews have also been demonstrated as a potential platform for studying HCV infection ( Amako et al . , 2010; Tong et al . , 2011; Xie et al . , 1998; Xu et al . , 2007 ) . Thus , in the present study we compared the amino acid similarity of CypA from great apes ( human , chimpanzee , bonobo , gorilla , orangutan ) , Old World monkeys ( rhesus macaque , pigtailed macaque , olive baboon ) , a New World monkey ( squirrel monkey ) , tree shrew , and mouse ( Figure 1a ) . Human , chimpanzee , bonobo , gorilla , olive baboon , and rhesus macaque CypA are 100% identical at the amino acid level and for subsequent experiments the human CypA ( hCypA ) CDS was used as the representative sequence for these six species . Multiple studies have shown that pigtailed macaques are predominantly homozygous for an insertion of the CypA exon at the TRIM5 locus , resulting in a chimeric TRIM5-CypA transcript ( Brennan et al . , 2008; Liao et al . , 2007; Newman et al . , 2008 ) , which we used for our experiments . Having identified these differences between CypA orthologs , we then compared their respective abilities to support HCV replication in a Huh7 . 5 cell line stably expressing an shRNA against endogenous human CypA ( Huh7 . 5-shRNA CypA ) ( von Hahn et al . , 2012 ) ( Figure 1—figure supplement 1 ) . Cells were transduced with a bicistronic lentivirus to express the CypA ortholog and a GFP-ubiquitin-neomycin resistance ( GUN ) fusion protein . The bicistronically expressed GFP provides a straightforward means to monitor protein expression indirectly . We deliberately chose not to add an epitope tag onto the different CypA variants to avoid impacting function . The percentage of GFP+ cells as determined by flow cytometry indicated >60% transduction efficiency ( Figure 1—figure supplement 2a ) . We also assessed protein expression by western blot using two different antibodies with different CypA antigen specificities – commercial antibodies with listed reactivity for human and mouse CypA were available but not for any of the other species under examination ( Figure 1—figure supplement 2b , Supplementary file 1 ) . All the orthologs were readily detected at the expected size of ~17 kDa except for squirrel monkey CypA , the signal for which was <2 times that of the background levels in the nontransduced Huh7 . 5-shRNA CypA cells , and pigtailed macaque CypA . In the latter case , the TRIM5-CypA fusion protein was expected at ~51 kDa but no signal was observed with either antibody . The transduced cells were subsequently infected with the HCV reporter virus Jc1-Gluc ( Marukian et al . , 2008 ) at an MOI of 0 . 1 . Levels of Gaussia luciferase in the culture supernatants were thus used as a proxy for assessing HCV replication to compare the rescue efficiencies of the CypA orthologs . As expected , expression of human CypA in Huh7 . 5-shRNA CypA cells increased HCV replication by more than two logs relative to the non-rescued cells at five days post-infection ( dpi ) ( Figure 1b ) . Of the orthologs tested , only orangutan CypA , which differs from human CypA by a single amino acid , was capable of rescuing HCV replication at levels similar to human . Compared to human CypA ( normalized to 100% ) , mouse CypA could still facilitate HCV replication but at levels ~3–4% of those observed for human , that is ca . 30-fold decrease . Tupaia CypA , which was well expressed by western blot , did not significantly increase HCV replication above the levels observed in the parental Huh7 . 5-shRNA CypA cells . It remains possible that the lack of HCV replication in cells transduced with squirrel monkey and pigtailed macaque CypA is due to these proteins not being properly expressed . However , as the transduction efficiency of our constructs expressing pigtailed macaque and squirrel monkey CypA was robust and we were able to detect all other CypA variants utilized in this study by western blot , it is more likely that the antibody reactivity for these two specific orthologs is weaker . As known blocks in the viral life cycle at the level of entry have been well characterized in murine hepatocytes , we aimed to further understand how murine CypA might affect viral replication . Thus , still in a human context , we sought to determine how the six amino acid differences between murine and human CypA contributed to the decreased ability of murine CypA to facilitate HCV replication in Huh7 . 5-shRNA CypA cells . ‘Murinized’ human CypA and ‘humanized’ murine CypA mutants were generated whereby each of the six differing amino acids were changed one at a time to their murine or their human counterpart , respectively ( Figure 2a ) , and transduction ( Figure 2—figure supplement 1 ) as well as CypA expression confirmed ( Figure 2—figure supplement 2 , Supplementary file 1 ) . None of these differences fell in the CsA-binding site of CypA ( Figure 2b ) . Human CypA was more sensitive to changes in the amino acid sequence , with significant decreases in rescue efficiency for all single residue changes tested ( Figure 2c ) . Mouse CypA demonstrated a greater ability to facilitate HCV replication , at least in a human cell context , when either residues 12 , 14 , 52 or 76 were altered , with levels of replication comparable to those observed in the presence of human CypA ( Figure 2d ) . As three of the residues that differ between mouse and human CypA are clustered together ( residues 11 , 12 and 14 ) , we also constructed and tested mutants triply ‘humanized’ or ‘murinized’ at these positions . Indeed , we observed a striking reversal of phenotype for both constructs , with ‘humanized’ murine CypA able to rescue HCV replication at levels comparable to human and vice versa for ‘murinized’ human CypA ( Figure 2e ) . Since humanizing residue 52 in murine CypA also resulted in a strong and significant increase in HCV replication , we combined the S52C mutant with each of the individual mutants T11A , A12V , and D14G as well as with the triply humanized mouse mutant T11A/A12V/D14G to ascertain whether there was an additional increase in HCV replication ( Figure 3a , Figure 3—figure supplement 1 , Supplementary file 1 ) . Mutating residue 52 , even in the triple mutant , did not have a significant synergistic effect for any of the mutants tested ( Figure 3b ) . Compared to mutant T11A , mutant T11A/S52C did demonstrate increased rescue efficiency , but upon performing further statistical tests to those shown in the figure , this was not statistically significant . As we observed a lower rescue efficiency of murine compared to human CypA in the Huh7 . 5-shRNA CypA cells , we considered the possibility that murine CypA may have a dominant negative effect on viral replication versus simply being incompatible . To test this , Huh7 . 5-shRNA CypA cells were dually transduced with the bicistronic mouse CypA lentivirus expressing eGFP used above and a monocistronic lentivirus containing a C-terminally triple FLAG-tagged hCypA ( Figure 4a , Figure 4—figure supplement 1 ) . The expression of the latter , with the expected shift in size , was confirmed via western blot ( Supplementary file 1 ) . These dually transduced cells , along with singly transduced controls , were then infected with Jc1-Gluc at an MOI of 0 . 1 and assessed five dpi by NS5A staining ( Figure 4b ) . We gated on cells that were highly dually positive for both mouse and human CypA and examined in this gate the fraction of HCV NS5A antigen-bearing cells . The presence of elevated human CypA along with murine CypA in the same cells did not result in a significant decline in infection , indicating no dominant negative effect . Although we readily observed replication of the Jc1-Gluc genome in our rescue lines by our luminometry readout , we also tested whether this replication was occurring in only the subset of cells initially infected by the inoculum or spreading across the culture over time ( Figure 5a ) . We took the parental Huh7 . 5-shRNA CypA cells plus the three rescue lines that displayed replication ( mouse , human , or triply humanized mouse CypA ) and infected them with Jc1-Gluc at an MOI of 0 . 1 . At three and five dpi , viral spread was assessed by NS5A staining ( Figure 5b ) , which significantly increased over time only for the +human CypA and +triply humanized mouse CypA cultures . Although significantly less compared to these two lines , the number of NS5A positive cells in the mouse rescue line was still significantly higher than that of the non-transduced Huh7 . 5-shRNA CypA cells . As the NS5A staining is less sensitive compared to the luminometry assay , we wanted to further confirm that infectious particle production was occurring and thus contributing to viral spread . Supernatants from the infected CypA rescue lines were also collected at three and five dpi , applied to naïve Huh7 . 5 cells and replication assessed three dpi by luminometry ( Figure 5c ) . As expected , the supernatants collected three dpi from all rescue lines resulted in lower replication in Huh7 . 5 cells compared to the five dpi supernatants , indicating an increase in infectious particle production over time . Supernatants collected from parental Huh7 . 5 shRNA CypA cells did not exhibit an increase in infectious particles over time , as replication levels in Huh7 . 5 cells did not significantly increase following infection with the five dpi supernatant . However , there clearly was still some infectious particle production occurring as the level of replication was at least a log above background . We next moved into a murine context to see how overexpression of murine CypA , human CypA or our triply humanized/murinized mutants might impact HCV replication . To this end , we generated murine Hep56 . 1D hepatoma cells expressing via lentiviral transduction a variety of factors already established as important to the HCV life cycle: the four HCV human entry factors discussed above ( OCLN , CLDN1 , SCARBI , and CD81 ) ; miR-122 to aid in replication; and SEC14L2 , which is absent in hepatoma cells but well-expressed in primary human hepatocytes and has allowed for the in vitro replication and low level viral particle production of normally non-permissive genotypes and clinical isolates of HCV ( Saeed et al . , 2015 ) . This line , termed Clone 8 , we also transduced with murine ApoE ( mApoE ) , which as described above serves in viral packaging and release ( Frentzen et al . , 2014; Long et al . , 2011 ) , to form Clone 8 + ApoE cells ( Figure 6a ) . Expression of all these factors was verified by a combination of flow cytometry , western blot and RT-qPCR ( Figure 6b–d ) and the replicative kinetics of Jc1-Gluc assessed over six days , with Huh7 cells serving as a positive control ( Figure 6e ) . Replication was consistently highest in the Huh7 cells , with the difference between the Clone 8/Clone 8 + ApoE and Huh7 cells increasing over time till by six dpi there was an approximately three-log difference in luciferase activity . The levels of replication in the parental Hep56 . 1D cells traced that of mock , with no de novo replication observed after the first wash of cells at two dpi . The addition of mApoE to the Clone 8 cells did not have an impact on the replication kinetics . Human , mouse , and the triply humanized/murinized CypA constructs were then transduced into the Clone 8 + ApoE cells ( Figure 7a ) and the replication assessed once more by luminometry following Jc1-Gluc infection ( Figure 7b ) . At six dpi , supernatants from Hep56 . 1D cells showed negligible luciferase activity as expected . The Clone 8 and Clone 8 + ApoE cells reached similar levels to one another , both about two logs higher than the parental line . The addition of any of the CypA constructs tested resulted in a significant increase in luciferase activity , with human or triply murinized human CypA the greatest , albeit by a small but significant margin . Based off these differences in replication , we then assessed whether there was any difference in infectious particle production . Supernatants from infected Hep56 . 1D and the Clone 8 lines were collected six dpi and used to infect naïve Huh7 . 5 cells where replication was then assessed three dpi by luminometry ( Figure 7c ) . While the supernatants collected from Clone 8 , Clone 8 + ApoE + human CypA and Clone 8 + ApoE + triply humanized mouse CypA cells did result in significantly higher replication in Huh7 . 5 cells compared to the Hep56 . 1D supernatants , the increase remained on average less than a log .
Although advances in available therapies , even cures , for HCV are striking , preventive measures , such as a vaccine , have still not been developed . In the United States alone over the past decade , there has been a more than 133% increase in acute HCV incidence , which is strongly associated with the domestic opioid epidemic ( Zibbell et al . , 2018; Zibbell et al . , 2014; Zibbell et al . , 2015 ) . Even with treatment , which has its own financial and logistical limitations , reinfection can still occur and is likely among high-risk individuals such as injection drug users and HIV-positive men who have sex with men ( Falade-Nwulia and Sulkowski , 2017; Hill et al . , 2017; Salazar-Vizcaya et al . , 2018 ) . Thus , HCV research is still greatly needed and immunocompetent models necessary more than ever for vaccine development . Due to the limited host range of HCV , research has focused on finding an immunocompetent small animal model without the financial and ethical restrictions associated with chimpanzees . At the heart of such research is understanding the basic virology and host restrictions of HCV , chipping away at the known barriers to the viral life cycle in a murine context such as at the level of entry ( Ding et al . , 2017; Dorner et al . , 2011; Ploss et al . , 2009 ) and viral particle assembly and release ( Frentzen et al . , 2014; Long et al . , 2011; Vogt et al . , 2013 ) . Viral replication in murine cells has consistently been low , and increasing replication has relied on selectable subgenomic replicons ( Long et al . , 2011 ) or disrupting cell-intrinsic antiviral responses ( Aly et al . , 2011; Anggakusuma et al . , 2015; Frentzen et al . , 2014; Long et al . , 2011; Nandakumar et al . , 2013; Vogt et al . , 2013 ) . Similarly , to observe even low levels of HCV replication in vivo using transgenic mouse models , blunting of innate immune signaling is also needed ( Dorner et al . , 2013; Vogt et al . , 2013 ) . Although the entire HCV life cycle can be completed with robust replication in xenorecipient mice engrafted with human or stem cell-derived hepatocytes ( Bissig et al . , 2010; de Jong et al . , 2014; Kosaka et al . , 2013; Mercer et al . , 2001; Meuleman et al . , 2005; Tesfaye et al . , 2013; Washburn et al . , 2011 ) , the immunocompromised nature of these mice precludes studying immune responses . It may not be possible to completely close this gap between levels of HCV replication in human versus other small animal model hepatocytes , but we sought to determine how species-specific differences in CypA – one of the few host factors experimentally validated as essential for viral replication – might contribute to these observed differences . We argue that such work has an impact on the development of an immunocompetent small animal model for studying HCV infection , which remains a critical need . In the present study , we assessed the ability of CypA orthologs from a variety of species to support HCV replication in a human context . Besides chimpanzee , bonobo , gorilla , olive baboon and rhesus macaque , which are 100% identical to human CypA at the amino acid level , all the remaining orthologs tested varied from hCypA by no more than six amino acids . Recent work has shown that single nucleotide polymorphisms in the hCypA gene can abrogate HCV infection in primary human hepatocytes ( von Hahn et al . , 2012 ) . The ability of these single missense mutations in the hCypA coding sequence to dramatically reduce HCV replication lent credence to the possibility that the small number of amino acid differences that differentiate hCypA from the orthologs we tested might have important functional significance in the context of HCV replication . In our system , the only other unique CypA ortholog able to rescue HCV replication during infection with Jc1-Gluc to the same extent as human was orangutan , which differs by only one amino acid . Tree shrew , pigtailed macaque and squirrel monkey CypA did not significantly increase replication compared to the non-transduced Huh7 . 5-shRNA CypA cells . Although we were not able to robustly detect squirrel monkey CypA and observed no signal for the pigtailed macaque TRIM5-CypA fusion protein by western blot , these two orthologs could still be expressed as commercial antibodies with confirmed specificity for these two orthologs are not available . This is especially pertinent as the antibodies we used had listed specificity for human and/or mouse . Compared to CypA from these two species , the pigtailed macaque and squirrel monkey orthologs differed by multiple residues that could impact antibody binding . Tree shrew CypA was readily detected with either of the two antibodies tested , so its failure to rescue HCV replication is more conclusive . It is important to underscore that our observations for tree shrew ( and for squirrel monkey and pigtailed macaque if they are in fact expressed ) do not mean that this species could not support HCV replication if its CypA ortholog was expressed in its native cellular environment . The amino acid differences in CypA might be best suited for interaction with species-specific factors for which the human ortholog is not an apt replacement . Indeed , recent work has demonstrated that HCV can replicate efficiently in pigtailed macaque stem cell-derived hepatocyte-like cells after overcoming the CD81- and OCLN-related barriers to viral entry ( Sourisseau et al . , 2013 ) , although whether the TRIM5-CypA fusion tested here was expressed in these particular cells is unknown . Likewise , rhesus macaque primary hepatocytes were also shown to support HCV infection in vitro and in vivo , albeit only at low levels ( Scull et al . , 2015 ) . Of course , it is still possible that a given CypA ortholog , even in its native context , would fail to properly interact directly with the necessary HCV proteins to facilitate replication . For example , the T73I difference between human and squirrel monkey CypA ( if indeed expressed ) could have a meaningful impact on such interactions as it is adjacent to G72 , one of the CsA-binding residues . In contrast , murine CypA was the only other ortholog that demonstrated some ability to rescue HCV replication . This was especially striking as mouse was the most evolutionarily distant species we examined and the murine CypA ortholog had the most differences in amino acid sequence compared to human . As we tested , this was not due to a dominant negative effect of murine CypA on replication as HCV could still replicate efficiently in the presence of both human and murine CypA . Thus , we created a series of mutants where we interchanged the six differing residues between human and mouse CypA . We were able to confirm expression of these mutants by western blot with at least one of the antibodies we tested . We cannot adequately control for the varying antibody reactivity , so comparing the expression of different CypA constructs with each other is difficult as lower detected expression could indicate lower antibody reactivity . This is highlighted by the differences between the two antibodies we tested , with one indicating a more than 100-fold higher expression for a CypA variant compared to the non-transduced Huh7 . 5-shRNA CypA cells while the other antibody reached only 20-fold . Since we could detect all of the CypA mutants by western blot , expressed them all in the same backbone construct , and observed high transduction efficiency , conceivably all the transduced cells would be able to express at least some level of the CypA variant . Of the six amino acids differing between the two orthologs , single ‘humanization’ of mCypA residues 12 , 14 , 52 , and 76 had the greatest impact , resulting in increased rescue . Conversely , ‘murinization’ of hCypA residues led to significant decreases in rescue ability across all residues tested . Strikingly , simultaneously altering the three residues clustered together that differed between human and mouse ( positions 11 , 12 and 14 ) resulted in a gain-of-function phenotype for murine CypA and a loss-of-function phenotype for human CypA . The gain-of-function phenotype was not further exacerbated in the triply humanized murine CypA by introducing the S52C mutation as well , which as a single mutant had shown rescue efficiency similar to the individual mutants at positions 12 and 14 . None of the differences between human and mouse CypA , or indeed for any of the CypA orthologs , overlapped with residues of the active site/CsA-binding site of CypA , which as further described below has been characterized as important for interaction with HCV proteins . This raises the intriguing possibility that this cluster of three residues plays a role in maintaining these interactions with the HCV replication machinery and/or associating with host factors that help facilitate such interactions . The location of the three residues on the face of the protein opposite the CsA-binding/active site makes it unclear whether changing these residues leads to a loss of PPIase activity that would explain the accompanying change in phenotype . Identifying these potential conformational changes and/or host factors would provide novel insights into the still unclear mechanism of action by which CypA promotes HCV replication . Furthermore , we tested in our system the impact of mouse , human or triply humanized mouse CypA on infectious particle production and viral spread over time . Even though spread over time was not detectable by NS5A staining in the murine CypA rescue cells , the more sensitive luminometry assay still demonstrated infectious particles were being released into the supernatant . Thus , in our system , the HCV life cycle is still being completed and the virus able to propagate beyond the cells initially infected with the inoculum . Although informative , these findings were all in the context of a human cell and thus whether or not this ‘humanization’ of mCypA would have the same effect in murine cells was unclear . In the same vein , it was not known if exogenous expression of human CypA in a murine context would boost HCV replication . As briefly alluded to above , there have been numerous efforts to engineer murine cells for studying HCV . Previous efforts using subgenomic replicons indicated replication could occur in murine embryonic fibroblasts ( MEFs ) at least to some extent and could be enhanced with the addition of miR-122 ( Jopling et al . , 2005; Lin et al . , 2010 ) and deletion of IRF3 ( Lin et al . , 2010 ) . Hep56 . 1D cells also support replication of subgenomic replicons and infectious particle release was achieved by using a selectable replicon with HCV core , E1 , E2 , p7 and NS2 provided in trans ( Long et al . , 2011 ) . The low levels of infectious particle release initially observed could be further increased by expressing either murine or human ApoE ( endogenous expression in Hep56 . 1D cells is low ) . However , in this same study , replication of transfected full-length genomes ( selectable or non-selectable ) was poor . Replication of subgenomic replicons was also enhanced in murine liver tumor ( MLT ) cells with the addition of miR-122 and disruption of IFN receptors ( Anggakusuma et al . , 2015 ) or in MEFs with blunted innate immune responses ( Nandakumar et al . , 2013 ) . Similarly , replication of full-length HCV RNA following transfection in primary murine hepatocytes remained low unless MAVS , IRF1 or IFNAR — all important players in cell-intrinsic antiviral responses — were knocked out ( Aly et al . , 2011; Nandakumar et al . , 2013 ) . Transfecting full-length genomes into MAVS-/- MLT cells expressing miR-122 and the human HCV entry factors resulted in robust replication and infectious particle production; however , in performing infections , replication declined and infectious particle release fell below background levels ( Frentzen et al . , 2014 ) . The only instance of viral infection in murine cells where strong replication and particle production were demonstrated relied on the use of a blasticidin-selectable genome ( Jc1-bsd ) in immortalized Stat1 knockout MEFs ( iMEFs ) expressing the human HCV entry factors , miR-122 and mApoE ( Vogt et al . , 2013 ) . Unlike these previous efforts , we wanted to focus on the replication of a non-selectable , full-length genome during infection without exogenous disruption of cellular immune responses as this is most immediately relevant and critical for generating a successful HCV animal model . Towards this aim , we engineered murine Hep56 . 1D hepatoma cells to express a variety of factors to overcome the various blocks at the level of entry and replication . These so-called Clone 8 cells demonstrated sustained de novo replication compared to the parental Hep56 . 1D cells , regardless of whether mApoE was present . Addition of murine CypA resulted in a further increase in replication that was even greater for human CypA . The phenotypes of the triple mutants in this murine context were the reverse of those observed in Huh7 . 5-shRNA CypA cells , suggesting that the mechanism by which these mutants function could depend to some extent on the species specificity of the cellular environment , which will be an interesting area for future study . In spite of the elevated replication and presence of mApoE in our Clone 8 cells , infectious particle production was not even a log greater than the parental Hep56 . 1D cells . This is similar to previous observations that when infecting humanized , MAVS-/- MLTs with full-length genomes , infectious particle release is negligible and viral replication lower compared to transfecting in the genome ( Frentzen et al . , 2014 ) . This raises the question of whether viral replication must be enhanced further in order to achieve greater viral release and/or packaging or whether factors directly important to these final steps of the viral life cycle are still missing . The evidence that infection with selectable HCVcc , so that the population is made up of cells containing HCV genome ( Vogt et al . , 2013 ) , results in efficient particle release as does transfecting in RNA to ‘hit’ more cells ( Frentzen et al . , 2014 ) suggests the former . However , in both of these cases , it is unclear how much blunting the innate immune response additionally contributed to these observations . Furthermore , if there is ( a ) murine factor ( s ) that is having an inhibitory effect on particle release , as observed for murine tetherin’s block of HIV-1 particle release due to its resistance to Vpu-mediated degradation ( McNatt et al . , 2009 ) , more viral replication may be required to overcome this . The luciferase assay we used is extremely sensitive , capturing even low levels of replication that may simply be insufficient for robust infectious particle production . To enhance replication , other ( co- ) factors/adaptors may be necessary to facilitate the interactions of CypA and are simply missing in murine cells or are incompatible orthologs of such factors . As observed for PI4KA , it is also possible that the effects of CypA we observe in the hepatoma lines tested may differ in a primary hepatocyte context and/or with the use of patient HCV isolates ( Harak et al . , 2017 ) . Understanding the mechanistic basis for the differences we observed in the respective rescue abilities of the CypA orthologs and mutants we tested remains an important area for future research . Indeed , these mutants may prove useful for delineating both the direct and indirect interactions necessary for robust HCV replication . Extensive study of human CypA has determined interactions with multiple HCV proteins . A connection between CypA and the HCV RdRp is evidenced by mutations in NS5B linked to CsA resistance ( Fernandes et al . , 2007; Robida et al . , 2007; Yang et al . , 2008a ) and indications that NS5B can bind at the CypA active site ( Chatterji et al . , 2009; Yang et al . , 2008a ) , which might enhance the viral protein's RNA-binding capcacity as a result ( Nag et al . , 2012 ) . The interaction of CypA with the HCV phosphoprotein NS5A is far better characterized . Mutations specifically in domain 2 of NS5A were linked to CsA treatment resistance among patients ( Goto et al . , 2009 ) . CypA stably binds NS5A and this interaction is dependent on the protein’s PPIase activity ( Chatterji et al . , 2010b ) , which promotes the isomerization of multiple conserved proline residues in domains II and III of NS5A , contributing to the protein’s proper folding ( Coelmont et al . , 2009; Foster et al . , 2011; Grisé et al . , 2012; Hanoulle et al . , 2009; Verdegem et al . , 2011 ) . As with NS5B , CypA is also believed to promote the RNA-binding capacity of NS5A ( Nag et al . , 2012 ) . Upon inhibition of CypA activity , lipid and protein trafficking patterns associated with the formation of lipid droplets and the VLDL synthesis pathway are altered ( Anderson et al . , 2011 ) . This has led to the hypothesis that CypA might facilitate the trafficking and assembly function of NS5A , which is known to localize to lipid droplets during the assembly of progeny virions ( Nag et al . , 2012 ) . Furthermore , whether CypA expression or just its PPIase activity is inhibited , formation of double membrane vesicles ( DMVs ) – a prevalent component of the membranous web derived from the host endoplasmic reticulum ( ER ) which serves as scaffolding for viral genome replication ( Paul et al . , 2013; Romero-Brey et al . , 2012 ) – is abrogated ( Chatterji et al . , 2015 ) . A portion of host CypA is already localized to ER compartments , which could be advantageous for later HCV replication complex formation in these same regions ( Chatterji et al . , 2010a ) . Within these DMVs , HCV replication complexes can then more safely assemble outside of the purview of host nucleic acid sensors and other defenses . Indeed , treatment of infected cells with cyclophilin inhibitors leads to changes in the ER that temporarily prevent re-infection ( Chatterji et al . , 2016 ) . Finally , an interaction between CypA and the viral assembly factor and protease NS2 has also been suggested , perhaps impacting the cleavage of NS2 from NS3 by the NS2-NS3 protease ( Ciesek et al . , 2009 ) and/or the overall proteolytic processing of the HCV polyprotein ( Kaul et al . , 2009 ) . It will be important to determine if the mutants and orthologs we examined here maintain or disrupt these known interactions or instead participate in novel , undocumented interactions . In this study , we focused on the contribution of CypA to the host restriction of HCV , but undertaking these same experiments with the other well-described HCV replication factor PI4KA is also of great interest . The percent amino acid similarity between human PI4KA and its orthologs in the species we examined in this study is lower compared to CypA , ranging from 83 . 8% ( chimpanzee ) to 99 . 5% ( rhesus macaque ) for PI4KA ( Supplementary file 2 ) versus 97–100% for CypA . However , assessing the PI4KA orthologs for all of these species is challenging , as a strong candidate for a PI4KA ortholog is either absent for a given species ( i . e . tree shrew ) or is a possible ortholog by our own sequence homology search but not formally annotated ( i . e . gorilla and bonobo ) . Some annotations for a predicted PI4KA ortholog are also more than 100 amino acids shorter than that of human , which is 2102 amino acids . Thus , determining the ‘true’ PI4KA ortholog for a given species requires additional corroboration . The engineered murine hepatoma line we have generated will serve as a ready platform for testing the potential of additional human factors or humanized murine factors to increase the replication capacity of HCV in these cells . If augmentation of viral replication is seen in vitro , genetic engineering approaches could be utilized to generate mice expressing these human orthologs or ‘humanized’ alleles . Such genetic humanization approaches would finally create the possibility to study HCV infection and pathogenesis in an immunocompetent small animal model , informing future vaccine development and testing .
Huh7 . 5 , Huh7 and 293 T cells were generously provided by Charles Rice ( Rockefeller University , NY ) and Huh7 . 5 . 1 cells as a kind gift from Frank Chisari ( The Scripps Research Institute , CA ) . Huh7 . 5 cells expressing a shRNA against either human CypA ( Huh7 . 5-shRNA CypA ) or an irrelevant target ( Huh7 . 5-shRNA irrel ) were graciously provided by Thomas von Hahn and Sandra Ciesek ( Hannover Medical School , Germany; University of Duisburg-Essen , Germany ) ( von Hahn et al . , 2012 ) . Huh7 cells were obtained from the American Tissue Culture Collection ( ATCC ) and Hep56 . 1D cells from CLS Cell Lines Service GmbH ( Eppelheim , Germany ) . The Hep56 . 1D-derived "Clone 8" cells were generated as described below . All cells have been authenticated and are clear of mycoplasma contamination . All cell lines were maintained in Dulbecco’s modified Eagle medium ( DMEM ) ( Thermo Fisher ) supplemented with 10% ( v/v ) fetal bovine serum ( FBS ) ( Omega Scientific ) . To select for the shRNA , Huh7 . 5-shRNA CypA and Huh7 . 5-shRNA irrel cells were maintained under blasticidin selection at 5 ug/mL ( BioVision ) . Cells transduced with lentivirus expressing the different CypA orthologs ( see below ) were selected for with Geneticin ( G418 ) ( Teknova ) at 750 µg/mL . All infections were performed in the absence of antibiotics . Cells were maintained at 37°C in a 5% ( v/v ) CO2 , 20% ( v/v ) O2 environment . The monoclonal mouse anti-NS5A 9E10 antibody was generously provided by Charles Rice ( Rockefeller University , NY ) . The following commercial primary antibodies were used: rabbit anti-FLAG for flow cytometry ( 1:1500 , #14793S , Cell Signaling Technology ) ; rabbit anti-human SEC14L2 for western blot ( 1:1000 , catalog #LS-B11733 , LifeSpan BioSciences , Inc ) ; mouse anti-human CypA for western blot ( 1 µg/ul , catalog #58144 , AbCam ) ; human CD81 conjugated to PE monoclonal for flow cytometry ( 1:200 , catalog #BDB555676 , BD Biosciences ) , rabbit anti-β actin for western blot ( 1:2000 , catalog #4970S , Cell Signaling Technologies ) , mouse anti-β actin for western blot ( 1:1000 , catalog #3700S ) and rabbit anti-human/mouse/rat/monkey CypA ( catalog #2175S , Cell Signaling Technologies ) . The following commercial secondary antibodies were used: goat anti-mouse Alexa 647 ( 1:250 , catalog #A-21235 , Invitrogen ) for flow cytometry; goat anti-rabbit Alexa 700 for flow cytometry ( 1:250 , catalog #A-21038 , Invitrogen ) ; goat anti-mouse Dylight 800 for western blot ( 1:10 , 000 , catalog #SA535521 , Thermo Fisher Scientific ) ; and goat anti-rabbit Dylight 680 for western blot ( 1:10 , 000 , catalog #35568 , Thermo Fisher Scientific ) . A multiple sequence alignment of the following CypA amino acid sequences was performed in MacVector ( v . 12 . 7 . 4 ) using the ClustalW multiple sequence alignment ( v1 . 83 ) : human ( NCBI Reference Sequence NP_066953 . 1 ) , chimpanzee ( NCBI Reference Sequence XP_001148412 . 1 ) , bonobo ( NCBI Reference Sequence XP_008967123 . 1 ) , gorilla ( NCBI Reference Sequence XP_018886247 . 1 ) , orangutan ( NCBI Reference Sequence NP_001126060 . 1 ) , olive baboon ( NCBI Reference Sequence XP_003896076 . 1 ) , rhesus macaque ( NCBI Reference Sequence NP_001027981 . 1 ) , pigtailed macaque TRIM5-CypA ( Genbank AGA83499 . 1 ) , squirrel monkey ( NCBI Reference Sequence XP_003923963 . 1 ) , mouse ( NCBI Reference Sequence NP_032933 . 1 ) and tree shrew ( NCBI Reference Sequence XP_006166088 . 1 ) . The additional amino acid alignment for PI4KA in Supplementary file 1 was performed using ClustalW in the same manner as for CypA . Since not all species , such as bonobo and gorilla , had genes annotated as PI4KA , sequences were retrieved from Ensembl by examining orthologs specifically for the human PI4KA gene ( ENSG00000241973 ) . In the case of one-to-many orthologs , the one with the highest whole genome alignment ( WGA ) coverage score ( calculated by Ensembl ) was used to access the accompanying amino acid sequence . The PI4KA ortholog gene sequences from all species used for the CypA alignment had WGA coverage scores of 99+ ( maximum score = 100 ) except for tree shrew , which was not included in the alignment due to its poor WGA score of 72 . 23 and less than 70% sequence identity . Indeed , many residues listed for the putative tree shrew ortholog were ‘X . ’ The accession IDs for the amino acid sequences used in the alignment are shown in Supplementary file 1 . The bicistronic lentiviral vector pWPI-IRES-GUN expressing both the human CypA open reading frame ( ORF; NCBI Reference Sequence NM_021130 . 4 ) and an IRES-regulated green fluorescent protein ( GFP ) -ubiquitin-neomycin resistance ( GUN ) fusion protein was kindly provided by Thomas von Hahn ( Hannover Medical School; Germany ) . As human , rhesus macaque , bonobo , gorilla , olive baboon , and chimpanzee CypA were 100% identical at the amino acid level , despite variation in the nucleic acid sequence , human CypA was used as the representative for these five other species and thus served as a proxy for the functional phenotype of the other orthologs . The CDS utilized for the remaining species of interest are as follows: orangutan CypA ( NCBI Reference Sequence NM_001132588 . 1 ) , tree shrew CypA ( NCBI Reference Sequence XM_006166026 . 2 ) , mouse CypA ( NCBI Reference Sequence NM_008907 . 1 ) , squirrel monkey CypA ( NCBI Reference Sequence XM_003923914 . 2 ) , and pigtailed macaque TRIM5-CypA ( GenBank Sequence JX865267 . 1 ) . The orangutan CypA ORF , which differs by one amino acid from human ( F8L ) , was generated by PCR mutagenesis with primers PU-O-3432 and PU-O-3428 ( Table 1 ) that simultaneously made the amplified coding region compatible for In-Fusion HD Cloning ( Takara Bio ) . The ORFs of tree shrew , mouse , squirrel monkey , and pigtailed macaque orthologs were synthesized as gBlock gene fragments ( Integrated DNA Technologies ) containing overlapping regions with the pWPI vector for subsequent In-Fusion HD Cloning . For all constructs , the pWPI-hCypA-IRES-GUN vector was digested with BamHI and SpeI to remove the hCypA and the CypA ortholog ORFs subsequently cloned in using In-Fusion HD Cloning ( Takara Bio ) . All plasmids were confirmed by sequencing and restriction enzyme digest . The generation of ‘murinized’ hCypA and ‘humanized’ mCypA single mutants expressing the analogous residue at one of the six amino acid positions differentiating the two orthologs was performed using PCR site-directed mutagenesis with the QuikChange XL Site-Directed Mutagenesis kit ( Agilent Technologies; Santa Clara , CA ) as outlined in the user manual . For the ‘murinized’ hCypA single mutants , the following primer pairs were utilized in the PCR mutagenesis reactions: PU-O-3853 and PU-O-3854 ( A11T ) , PU-O-3855 and PU-O-3856 ( V12A ) , PU-O-3857 and PU-O-3858 ( G14D ) , PU-O-3859 and PU-O-3860 ( C52S ) , PU-O-3861 and PU-O-3862 ( K76R ) , PU-O-3863 and PU-O-3864 ( A159S ) ( Table 1 ) . For the ‘humanized’ mCypA single mutants , the following primer pairs were utilized in the PCR mutagenesis reactions: PU-O-3871 and PU-O-3872 ( T11A ) , PU-O-3873 and PU-O-3874 ( A12V ) , PU-O-3875 and PU-O-3876 ( D14G ) , PU-O-3877 and PU-O-3878 ( S52C ) , PU-O-3879 and PU-O-3880 ( R76K ) , PU-O-3881 and PU-O-3882 ( S159A ) ( Table 1 ) . To generate the ‘murinized’ hCypA and ‘humanized’ mCypA triple mutants expressing the analogous residues at positions 11 , 12 , and 14 , the hCypA and mCypA residue 14 single mutants were transferred into pUC19 using primers PU-O-3851 and −3852 for human; primers PU-O-3851 and −4211 for mouse . The additional mutations at residues 11 and 12 were then introduced simultaneously using the QuikChange Multi Site-Directed Mutagenesis kit ( Agilent Technologies; Santa Clara , CA ) with primers PU-O-4138 and PU-O-4139 for the triply ‘murinized’ hCypA and primers PU-O-4140 and PU-O-4141 for the triply ‘humanized’ mCypA . The mutant regions were then PCR amplified from pUC-19 using primers PU-O-3424 and −3429 for mouse and PU-O-3424 and −3432 to be cloned back into pWPI . For the additional murine mutants shown in Figure 3 , the S52C mutation was introduced into the single mutants at residues 11 , 12 and 14 as well as the 11/12/14 triple mutant using primers PU-O-3877 and PU-O-3878 with the QuikChange XL Site-Directed Mutagenesis kit . C-terminal 3X-FLAG-tagged hCypA was made by first generating a Gblock gene fragment ( Integrated DNA Technologies ) for the C-terminal 3X-FLAG coding sequences with a glycine-linker sequence and appropriate backbone sequence overlap for subsequent In-Fusion HD Cloning ( Takara Bio ) . The hCypA ORF was PCR amplified using primers PU-O-4494 and −4495 , with the stop codons in the open reading frames removed to allow for the production of the C-terminal 3X-FLAG fusion . As before , these amplified regions were inserted into the pWPI backbone digested with BamHI and SpeI using In-Fusion HD Cloning ( Takara Bio ) . SEC14L2 was amplified from cDNA generated from cell lysates of human fetal liver cells ( HFLCs ) using primers PU-O-1755 and −1756 and cloned into pShuttle-CMV . SEC14L2 was then amplified by PCR with added restriction enzyme sites ( XhoI at 5’ end , BamHI 3’ end ) using primers PU-O-1943 and −1944 and cloned by Gibson Assembly ( New England BioLabs ) into pLVX-IRES-Puro ( Clontech ) that had been digested with XhoI and BamHI . Lentiviral particles containing the various pWPI-CypA constructs were produced by Xtremegene HP DNA transfection reagent ( Roche Applied Science; Indianapolis , IN ) -mediated co-transfection of HEK293T cells seeded twelve hours prior to transfection ( 4 . 4E6 cells per 10 cm poly-L-lysine-coated tissue culture dish ) with 4 µg of the appropriate pWPI-CypA plasmid , 4 µg of HIV gag-pol , and 0 . 57 µg of the G protein of vesicular stomatitis virus ( VSV-G ) per transfection reaction . Supernatants were harvested at 24 , 48 , and 72 hr post-transfection , stored at 4°C and then passed through 0 . 45 µm membrane filters ( Millipore; Darmstadt , Germany ) . Polybrene ( final concentration of 4 μg/mL ) ( Sigma-Aldrich ) and HEPES ( final concentration of 2 mM ) ( Gibco ) were added to all lentiviral supernatants which were aliquoted and stored at −80°C . All lentiviral transductions were performed via spinoculation with cells seeded at a concentration of 2E5 cells per well in a six well format 24 hr prior to transduction . Cell confluency at the time of transduction was 30–40% , and 2 mL of undiluted lentivirus was added to each well . Plates were spun at 37°C , 2 hr , 2000 rpm . Media replaced with 10% FBS DMEM 6 hr post-spinoculation . Transduction efficiency was assessed via flow cytometry for all constructs on a BD LSRII flow cytometer ( BD Biosciences ) with the exception of pLVX-SEC14L2-IRES-puro ( western blot was used to confirm transduction due to the lack of a fluorescent marker ) . All flow cytometry data was processed in FlowJo Software version 10 . 4 . 2 ( FlowJo , LLC ) . HCV RNA and subsequent viral stocks were produced as previously described ( Lindenbach et al . , 2005 ) . In brief , viral RNA was produced via in vitro transcription of an XbaI-linearized Jc1 ( p7nsGluc2A ) plasmid ( Marukian et al . , 2008 ) using the T7 RiboMAX Express Large Scale RNA Production kit ( Promega ) as outlined in the user manual . Viral RNA was purified using the Qiagen RNeasy Mini Kit ( Qiagen ) following manufacturer’s instructions , and quality control was performed by gel electrophoresis to ensure no significant RNA degradation . Viral RNA stocks were stored as 5 µg aliquots at −80°C . RNA was electroporated into Huh7 . 5 . 1 cells The pellet was resuspended in the appropriate volume of cold DPBS to achieve a concentration of 1 . 5E7 cells/mL . 6E6 cells were then electroporated in a 2 mm path length electroporation cuvette ( BTX Harvard Apparatus; Holliston , MA ) with 5 µg of viral RNA using an ECM 830 Square Wave Electroporation System ( BTX ) at the following settings: five pulses , 99 µs per pulse , 1 . 1 s pulse intervals , 860V . Following a ten-minute incubation at room temperature , the electroporated cells were seeded into p150s and maintained in 5% FBS DMEM . Media was changed one day post-electroporation , and supernatants were collected daily for six days and stored at 4°C . The pooled supernatants were passed through a 0 . 22 µm vacuum filter and subsequently concentrated to ~100 mL in an EMD Millipore Stirred Cell ( Cole-Parmer ) . The TCID50/mL ( Reed and Muench , 1938 ) of concentrated virus was determined after one freeze-thaw by limiting dilution assay . HCV infections of Clone 8 cells and CypA rescue experiments were conducted in a 24 well format with 3E4 cells seeded per well 12 hr pre-infection . Infections were conducted in triplicate wells using cell-culture produced Jc1-Gluc virus produced as described above . Viral inoculum was applied for 6 hr at which time the wells were washed twice with PBS and the media changed to 10% FBS DMEM . At no time during infections were antibiotics used . For kinetics experiments , 50 µL of supernatant were taken daily from each well for up to seven days and stored in 96 well plates at −20°C . As indicated in figures , when washes were performed , supernatant was collected immediately before a single wash with PBS and immediately after the subsequent replacement with fresh medium . Viral replication was quantified by measuring the luciferase activity of the supernatant using the Luc-Pair Renilla Luciferase HS Assay Kit ( GeneCopoeia ) and a Tristar2 LB 942 Multimode Microplate Reader ( Berthold Technologies ) according to manufacturer’s instructions . For assessing infectious particle production in supernatants , infections were performed in a 96 well format , with 6 . 4E3 Huh7 . 5 cells seeded per well 12 hr pre-infection . At the time of infection with supernatant , media was removed from the wells and replaced with 50 ul of the respective supernatant ( freshly collected without any freeze-thaws ) . The inoculum was removed and fresh media added to the wells six hpi . Supernatants from these infected Huh7 . 5 cells were then collected three dpi to assess Gaussia luciferase activity by lumionmetry . For NS5A staining , trypsinized cells were pelleted , fixed with 4% paraformaldehyde ( PFA ) ( Sigma-Aldrich ) and permeabilized in 0 . 1% ( w/v ) Saponin and 1% ( v/v ) FBS in DPBS . Pellets were subsequently incubated for 1 hr at room temperature with murine-produced Clone 9E10 primary antibody specific for NS5A ( Lindenbach et al . , 2005 ) , kindly provided by Dr . Charles Rice ( The Rockefeller University ) , diluted 1:8000 in FACS buffer ( 1% FBS ( v/v ) in DPBS ) . Cells were washed with DPBS and then incubated at 4°C for 30 min in the dark with goat anti-mouse Alexa 647 secondary antibody ( diluted 1:250 , catalog #A-21235 , Invitrogen ) . Cells were subsequently pelleted , washed once with FACS buffer and then analyzed in FACS buffer on a BD LSRII flow cytometer ( BD Biosciences ) . All flow cytometry data were processed in FlowJo Software version 10 . 4 . 2 ( FlowJo , LLC ) . Cell pellets were lysed for 5 min on ice in RIPA buffer ( 50 mM Tris , pH 7–8; 150 mM NaCl , 0 . 1% SDS ( v/v ) , 0 . 5% sodium deoxycholate ( v/v ) , 1% Triton X-100 ( v/v ) ) containing protease inhibitor cocktail ( Sigma-Aldrich , P-8340; 1:250 dilution ) . Lysates were spun down for 10 min , 12 , 000 rpm , 4°C . The resulting supernatant was mixed with 6X Laemmli buffer ( 375 mM Tris pH = 6 . 8 , 10% SDS , 50% Glycerol , 10% beta-mercaptoEtOH , 0 . 03% Bromo blue ) and heated for 5 min at 98°C along with PageRuler protein ladder marker ( Thermo Scientific ) . The samples were separated on a 10% or 12% ( wt/vol ) SDS-polyacrylamide gel in running buffer ( diluted from a 10X stock containing 30 . 3 g Tris , 144 g glycine , 10 g SDS in 1 L of ddiH2O ) at 150V for 60 min . The resolved proteins were transferred onto a 0 . 2 µm nitrocellulose membrane ( Bio-Rad Laboratories; Hercules , CA ) in transfer buffer ( 10X stock containing 30 . 3 g Tris , 144 g Glycine , 4 g SDS in 1 L ddiH2O; diluted to 1X in ddiH2O with 20% MeOH ( v/v ) ) for 1 hr at 18V . Membranes were blocked for at least 30 min in DPBS containing 5% milk ( wt/vol ) , washed twice with 1X TBS containing 0 . 5% ( v/v ) Tween ( TBS-T ) , and then incubated for 1 hr at room temperature or overnight at 4°C with primary antibodies as listed in the figure legends ( see ‘Antibodies’ section above for specific product information ) in TBS-T . Membranes were washed three times in TBS-T , incubated for 30 min in the dark at room temperature with the appropriate secondary antibody ( goat anti-mouse secondary antibody Dylight 800 , Thermo Fisher Scientific , #SA535521 , diluted 1:10 , 000 in TBS-T; goat anti-rabbit secondary antibody Dylight 680 Thermo Fisher Scientific , #35568 , diluted 1:10 , 000 in TBS-T ) and then washed three more times in TBS-T . All membranes were visualized on an Odyssey CLx Imaging System ( LI-COR Biotechnology; Lincoln , NE ) . Membranes that were re-probed were stripped in Restore PLUS Western Blot Stripping Buffer ( Thermo Fisher Scientific , #46430 ) for 15 min at room temperature , washed twice with TBS-T , and then blocked and incubated with antibody as described above . Where performed , signal intensity for bands of interest was determined using LI-COR Image Studio Software ( version 4 . 0 ) . Note that even if the contrast and brightness levels of a membrane image are adjusted within the software , the raw intensity values remain unchanged and were used for all quantifications . Hep56 . 1D cells were transduced with lentivirus produced from the following plasmids: pTRIP-human CD81 ( Flint et al . , 2006 ) , pTRIP-Venus/YFP-human OCLN ( Ploss et al . , 2009 ) , pTRIP-Cerulean/CFP-human CLDN1 ( Evans et al . , 2007 ) , pTRIP-mKate human SCARBI ( Vogt et al . , 2013 ) , pLVX-IRES-human SEC14L2-puro ( see ‘Plasmid construction’ section above ) and pTRIP-miR122 ( Vogt et al . , 2013 ) . Single cell-sorting was then performed , gating on highly YFP+/mKate+ cells , which were subsequently expanded and assessed for expression of the other transduced factors by a mixture of flow cytometry ( all the entry factors ) , western blot ( SEC14L2 ) and RT-qPCR ( miR-122 ) . miRNA was extracted from the cells using the miRNeasy mini kit ( Qiagen ) following manufacturer’s directions for mature miRNA isolation . cDNA was subsequently produced using the miScript II RT kit ( Qiagen ) and then quantified by real-time PCR using the miScript SYBR green PCR kit all according to the manufactuer’s directions . miR-122 expression was normalized to the snRNA RNU6B using the Hs_miR-122a_1 miScript Primer Assay and Hs_RNU6-2_11 miScript Primer Assay ( Qiagen; compatible for mouse and human ) , respectively . The final clone worked with throughout this paper was termed ‘Clone 8’ . These cells were subsequently transduced with lentivirus containing pTRIP-mApoE-tagRFP ( Vogt et al . , 2013 ) and various pWPI-CypA-GUN constructs ( see above ) prior to experiments , with tagRFP and eGFP expression , respectively , assessed by flow cytometry . The six residues differing between human and murine CypA were substituted in the structure of human CypA ( PDB ID 1CWA ) ( Mikol et al . , 1993 ) using the tools AutoSub and RelabelChains from AmberUtils by William D Lees ( https://github . com/williamdlees/AmberUtils ) run using Python ( v 2 . 7 . 10 ) . The resultant structure was visualized and labeled in PyMOL ( v 2 . 2 . 0 ) ( Schrodinger , 2015 ) . All statistical analyses , as described in figure captions , were done using GraphPad Prism software ( v . 6 . 0 ) . | Hepatitis C is a life-long disease that begins when a virus infects the cells of the liver . Although the infection is curable , it is expensive to treat , and there is not yet a vaccine to prevent the disease . This is largely because the virus that causes hepatitis C , also known as HCV , naturally only infects humans and chimpanzees , which has made it difficult to generate an effective animal model for developing a vaccine . Mice are frequently used as a model for studying disease and can be genetically altered to allow HCV to enter their liver cells . However , once HCV enters mouse cells , it struggles to replicate . As a result , an infection does not develop , and the immune system’s response to the virus cannot be fully explored . Replication of HCV in humans relies on a protein called cyclophilin A , or CypA for short . Now , Gaska et al . have set out to improve current animal models for HCV by investigating whether HCV can also use CypA from other species , including mice , to replicate . Gaska et al . showed that the mouse form of CypA could help HCV replicate in human liver cells with lower than normal levels of CypA , but only very poorly . Editing the mouse gene for CypA to be more like the human version resulted in higher HCV replication . Putting variants of CypA into the liver cells of mice , which do not normally replicate HCV at high levels , led to an increase in HCV replication . However , replication of HCV in mice was still far lower than in human liver cells , suggesting that the mouse model system could be improved by learning more about which proteins interact with CypA . Injection drug use – one of the main ways hepatitis C spreads – is becoming increasingly common because of the growing opioid epidemic in many countries . A clinically relevant animal model that supports hepatitis C virus infection would be an important milestone towards a vaccine that could prevent the continued spread of this disease . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"microbiology",
"and",
"infectious",
"disease"
] | 2019 | Differences across cyclophilin A orthologs contribute to the host range restriction of hepatitis C virus |
Hepcidin is the master regulator of systemic iron homeostasis . Derived primarily from the liver , it inhibits the iron exporter ferroportin in the gut and spleen , the sites of iron absorption and recycling respectively . Recently , we demonstrated that ferroportin is also found in cardiomyocytes , and that its cardiac-specific deletion leads to fatal cardiac iron overload . Hepcidin is also expressed in cardiomyocytes , where its function remains unknown . To define the function of cardiomyocyte hepcidin , we generated mice with cardiomyocyte-specific deletion of hepcidin , or knock-in of hepcidin-resistant ferroportin . We find that while both models maintain normal systemic iron homeostasis , they nonetheless develop fatal contractile and metabolic dysfunction as a consequence of cardiomyocyte iron deficiency . These findings are the first demonstration of a cell-autonomous role for hepcidin in iron homeostasis . They raise the possibility that such function may also be important in other tissues that express both hepcidin and ferroportin , such as the kidney and the brain .
As a constituent of hemoproteins , iron-sulphur proteins and other functional groups , iron is essential for cellular functions . Conversely , excess iron participates in cytotoxic Fenton-type chemical reactions . Thus , both iron deficiency and iron overload are detrimental to the cell . Therefore , the healthy functioning of tissues requires tight control of intracellular iron levels . These in turn are dependent both on cellular homeostatic pathways controlling iron uptake , usage , and storage , and on systemic pathways controlling iron levels in the plasma . In mammals , cellular iron homeostasis is controlled by the Iron Regulatory Proteins IRPs . Intracellular iron levels control the degradation of IRP2 and the conformational switch that confers the RNA-binding function of IRP1 . IRPs in turn control the levels of iron uptake proteins such as transferrin receptor 1 ( TfR1 ) and divalent metal transporter ( DMT1 ) , and the iron storage protein ferritin ( Rouault and Klausner , 1997; Rouault , 2006 ) . Systemic iron homeostasis is controlled by the hepcidin/ferroportin axis at the sites of iron entry into the circulation . Ferroportin ( FPN ) , which is encoded by the Solute Carrier Family 40 Member 1 ( Slc40a1 ) , is the only known mammalian iron export protein and mediates iron release into the circulation from duodenal enterocytes , splenic reticuloendothelial macrophages and hepatocytes , the sites of iron absorption , recycling and storage respectively ( Ganz , 2005; Donovan et al . , 2005 ) . FPN-mediated iron release is antagonized by the hormone hepcidin , also known as hepcidin antimicrobial peptide ( HAMP ) . Produced primarily in the liver , hepcidin binds to and induces internalization of FPN , thereby limiting iron release into the circulation and its availability to peripheral tissues ( Nemeth et al . , 2004; Qiao et al . , 2012 ) . The importance of the HAMP/FPN axis is illustrated by diseases of systemic iron overload such as hemochromatosis and β-thalassaemia , where hepcidin production is impaired ( Camaschella , 2005; Musallam et al . , 2012 ) , and in anaemia of chronic disease where hepcidin production is inappropriately elevated ( De Falco et al . , 2013; Nemeth and Ganz , 2014 ) . Other than the liver , hepcidin is also found in tissues with no recognized role in systemic iron homeostasis , including the heart ( Merle et al . , 2007 ) , the brain ( McCarthy and Kosman , 2014 ) , the kidney ( Kulaksiz , 2005 ) and the placenta ( Evans et al . , 2011 ) . The function of this extra-hepatic hepcidin remains unknown , but one hypothesis is that it is involved in local iron control . Relevant to this hypothesis are our recent findings that FPN is also present in the heart , that it is essential for cardiomyocyte iron homeostasis and that its cardiomyocyte-specific deletion leads to fatal cardiac iron overload in mice ( Lakhal-Littleton et al . , 2015 ) . Therefore , we hypothesised that cardiac HAMP regulates cardiac FPN , and that such regulation is important for local iron control and for cardiac function . To test this hypothesis , we generated two novel mouse models; the first with a cardiomyocyte-specific deletion of the Hamp gene , and the second , with cardiomyocyte-specific knock-in of Slc40a1 C326Y , that encodes a FPN with intact iron export function but impaired HAMP binding ( Schimanski et al . , 2005; Drakesmith et al . , 2005 ) . We studied cardiac function and iron homeostasis longitudinally in both models and report that both develop fatal cardiac dysfunction and metabolic derangement as a consequence of cardiomyocyte iron deficiency . This occurs against a background of otherwise normal systemic iron homeostasis . Both cardiac dysfunction and metabolic derangement are prevented by intravenous iron supplementation . Our findings demonstrate that , at least in the cardiomyocyte , endogenously-derived HAMP plays an essential role in cellular , rather than systemic iron homeostasis . It does this through the autocrine regulation of cardiomyocyte FPN . Disruption of this cardiac HAMP/FPN leads to fatal cardiac dysfunction . Currently , there is considerable clinical interest in strategies that target the HAMP/FPN axis for the treatment of systemic iron overload and iron deficiency . Our findings suggest that these strategies may additionally alter cardiac iron homeostasis and function . Other than the heart , both HAMP and FPN are also found in the brain , kidney and placenta ( McCarthy and Kosman , 2014; Kulaksiz , 2005; Evans et al . , 2011; Rouault , 2013; Moos and Rosengren Nielsen , 2006; Bastin et al . , 2006; Wolff et al . , 2011 ) . A pertinent question is the extent to which our findings in the heart extend to those tissues .
Hamp mRNA levels were approximately 30 fold lower in the adult mouse heart than in the liver ( Figure 1A ) . Next , we examined the regulation of cardiac Hamp mRNA and HAMP protein in response to dietary iron manipulation , having first established that this dietary manipulation altered cardiac and liver iron levels ( Figure 1—figure supplement 1 ) . In both tissues , Hamp mRNA levels were decreased by the iron-deficient diet ( Fe 2–5 ppm ) and increased by the iron-loaded diet ( Fe 5000 ppm ) ( Figure 1A ) . At the protein level , while changes in hepatic HAMP protein mirrored changes in its mRNA levels , cardiac HAMP protein was increased by the iron-deficient diet and unaffected by the iron-loaded diet ( Figure 1B ) . 10 . 7554/eLife . 19804 . 003Figure 1 . Hepcidin expression and regulation in the heart . ( A ) Relative Hamp mRNA expression in heart and liver of adult C57BL/6 mice , under control conditions and after provision of low or high iron diets . *p=0 . 047 , 0 . 001 respectively relative to control hearts , †p = 0 . 006 , 0 . 019 respectively relative to control livers . ( B ) Corresponding immunohistochemical staining for HAMP in heart and liver . ( C ) Relative Hamp mRNA expression in primary adult mouse cardiomyocytes cultured under control conditions or in presence of FAC or DFO . *p=0 . 023 , 0 . 001 and 0 . 014 respectively relative to control . †p = 0 . 024 , 0 . 037 , 0 . 016 and 0 . 037 respectively relative to control at the same timepoint . ( D ) Corresponding HAMP protein levels in supernatants of primary cardiomyocytes . DFO treatment was carried alone ( DFO ) or presence of Furin inhibitor ( DFO+CMK ) . *p=0 . 002 , 0 . 020 , 0 . 028 , 0 . 014 , 0 . 015 respectively relative to control at the same timepoint . ( E ) Relative Hamp expression in heart and liver of 3 month old Hampfl/fl and Hampfl/fl;Myh6 . Cre+ mice . *p=0 . 018 relative to cardiac Hamp in Hampfl/fl controls . ( F ) Corresponding immunohistochemical staining for HAMP in heart and liver . All values are plotted as mean ± SEM . Scale bar = 20 µm . n = 3 per group unless otherwise stated . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00310 . 7554/eLife . 19804 . 004Figure 1—figure supplement 1 . Cardiac and liver iron following dietary iron manipulation . Total elemental iron levels in heart and liver of C57BL/6 mice , under control conditions and after provision of low ( Fe 5 ppm ) or high iron ( Fe 5000 ppm ) diets from weaning for six weeks . *p=0 . 037 and 0 . 033 respectively relative to control heart , †p = 0 . 010 and 0 . 005 relative to control liver . n = 3 . Data are represented as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00410 . 7554/eLife . 19804 . 005Figure 1—figure supplement 2 . Furin regulation by iron . ( A ) Relative Furin mRNA in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC . *p=0 . 004 , 0 . 001 and 0 . 001 respectively relative to control at the respective timepoint . ( B ) Relative Furin mRNA in hearts of mice provided control diet or iron-deficient diet ( 2–5 ppm ) or iron-loaded diet ( 5000 ppm ) from weaning for six weeks . *p=0 . 015 relative to control diet . n = 3 . Data are plotted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00510 . 7554/eLife . 19804 . 006Figure 1—figure supplement 3 . Relative Hamp mRNA expression in cardiomyocytes following treatment with Furin inhibitor CMK . Relative Hamp mRNA expression in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC , in presence or absence of Furin inhibitor CMK . n = 3 . Data are plotted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00610 . 7554/eLife . 19804 . 007Figure 1—figure supplement 4 . HAMP in supernatants of primary cardiomyocytes . HAMP protein was measured by ELISA in supernatants of primary adult cardiomyocytes , derived from Hamp fl/fl or Hamp fl/fl;Myh6 . Cre+ mice and cultured under control conditions or in presence of FAC or DFO . n = 3 . Data are plotted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00710 . 7554/eLife . 19804 . 008Figure 1—figure supplement 5 . Confirmation of HAMP antibody specificity . ( A ) Staining with HAMP antibody ( Abcam ab30760 ) in liver and heart of C57BL/6 mice is completely abrogated by co-incubation with hepcidin-25 blocking peptide ( Abcam ab31875 ) . Scale bar = 20 µm . ( B ) Loss of HAMP staining in Hampfl/fl;;Myh6 . Cre+ hearts ( Figure 1F ) is consistent with the antibody detecting HAMP1 and not HAMP2 , because Hamp2 mRNA expression is not altered in Hampfl/fl;Myh6 . Cre+ mice relative to Hampfl/fl controls , either with control diet or iron-deficient diet ( six weeks from weaning ) . *p=0 . 007 and 0 . 047 relative to Hampfl/fl control under respective diet . n = 3 per group . Values are plotted as mean ± SEM . N . S=not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 00810 . 7554/eLife . 19804 . 009Figure 1—figure supplement 6 . HAMP detection by ELISA unaffected by FAC and DFO . HAMP standard from mouse hepcidin ELISA ( E91979Mu , Uscn ) was diluted either in the kit’s own standard diluent , or in unconditioned growth medium alone , or containing 100 µmol/L DFO or 500 µmol/L FAC . ELISA was performed as per manufacturer’s instructions . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 009 To explore further the regulation of cardiac hepcidin by iron , we isolated primary adult cardiomyocytes from mice , then carried out a timecourse treatment with the iron chelator desferroxamine DFO , or with ferric citrate FAC . Under control conditions , Hamp mRNA was upregulated following addition of fresh cardiomyocyte culture medium ( cardiomyocytes are cultured for 2 hr in serum-free medium prior to this ) . Relative to control cardiomyocytes at the respective timepoint , Hamp mRNA was increased by FAC from 4 hr of treatment , and decreased by DFO at 4 , 8 and 16 hr of treatment ( Figure 1C ) . HAMP protein , measured by ELISA in supernatants was also increased following addition of fresh medium . Relative to control cardiomyocytes at the respective timepoint , HAMP protein in supernatants was increased by DFO as early as 2 hr , but remained unchanged by FAC at all timepoints ( Figure 1D ) . Thus , the direction of response of the Hamp mRNA and HAMP protein to iron levels in vitro mirrored the responses seen in vivo . Next , we aimed to understand the mechanisms underlying increased HAMP secretion in DFO-treated cardiomyocytes . To this end , we investigated the role of the prohormone convertase Furin , which in hepatocytes , is required for cleavage of the propeptide and the release of the mature HAMP peptide ( Valore and Ganz , 2008 ) . Furin expression has been reported in the heart ( Beaubien et al . , 1995; Ichiki et al . , 2013 ) , and we also found that its expression was upregulated in the hearts of mice provided an iron-deficient diet and in cardiomyocytes treated with DFO ( Figure 1—figure supplement 2 ) . When we measured HAMP in supernatants of cardiomyocytes treated with the Furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone ( CMK ) , we found no increase in HAMP release following DFO treatment ( Figure 1D ) . We confirmed that Hamp mRNA levels in cardiomyocytes were not altered by CMK treatment ( Figure 1—figure supplement 3 ) . Together , these results indicate that Furin upregulation mediates increased HAMP secretion from iron-deficient cardiomyocytes . Having established that hepcidin is found in cardiomyocytes , we then aimed to define its function . To this end , we generated cardiomyocyte-specific Hamp knockout mice Hampfl/fl;Myh6 . Cre+ by crossing in-house conditional Hamp floxed ( fl ) mice with mice transgenic for Cre recombinase under the control of cardiomyocyte-specific promoter Myosin Heavy Chain 6 ( Myh6 . Cre+ ) . Hamp mRNA ( Figure 1E ) and HAMP protein ( Figure 1F ) in the hearts of Hampfl/fl;Myh6 . Cre+ mice were significantly reduced compared to the hearts of Hampfl/fl controls . Furthermore , compared to Hampfl/fl cardiomyocytes , levels of HAMP protein in the supernatants of cardiomyocytes from Hampfl/fl;Myh6 . Cre+ mice were either greatly reduced or undetectable , both under baseline conditions and following treatment with DFO or FAC ( Figure 1—figure supplement 4 ) . Near complete ablation of the cardiac Hamp mRNA and HAMP protein in Hampfl/fl;Myh6 . Cre+ mice confirmed that cardiomyocytes were the primary site of hepcidin expression in the heart . Liver Hamp mRNA ( Figure 1E ) and HAMP protein ( Figure 1F ) were not different between Hampfl/fl;Myh6 . Cre+ and Hampfl/fl controls , consistent with the cardiac-specific nature of Hamp gene deletion . Also consistent with this cardiac-specific deletion , Hampfl/fl;Myh6 . Cre+ mice had normal levels of liver iron stores and circulating markers of iron homeostasis when compared to Hampfl/fl controls , demonstrating that loss of cardiac hepcidin did not affect systemic iron homeostasis . In addition , circulating HAMP levels were not reduced in the serum of Hampfl/fl;Myh6 . Cre+ mice , suggesting that cardiac hepcidin does not contribute significantly to circulating HAMP levels ( Table 1 ) . 10 . 7554/eLife . 19804 . 010Table 1 . Indices of systemic iron in six month old Hampfl/fl and Hampfl/fl;Myh6 . Cre+ mice . n = 6 per group . All values are shown as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 010Hampfl/fl Hampfl/fl;Myh6 . Cre+ liver total elemental iron ( ng/mg tissue ) 96 . 3 ± 12 . 2 88 . 7 ± 19 . 2 liver ferritin ( µg/mg total protein ) 0 . 65 ± 0 . 04 0 . 64 ± 0 . 05 serum iron ( µmol/L ) 28 . 60 ± 7 . 20 31 . 50 ± 8 . 40 serum ferritin ( mg/L ) 1 . 81 ± 0 . 04 1 . 88 ± 0 . 29 hemoglobin ( g/L ) 122 . 7 ± 11 . 5 116 . 0 ± 11 . 9 serum hepcidin ( µg/L ) 23 . 5 ± 7 . 6 23 . 9 ± 10 . 40 To determine the effects of loss of cardiac hepcidin , we first assessed the cumulative survival of Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls over a period of 52 weeks . Significantly greater mortality was observed amongst Hampfl/fl;Myh6 . Cre+ mice , with only 29% of animals surviving to 52 weeks , compared with 90% of Hampfl/fl controls . The median survival of Hampfl/fl;Myh6 . Cre+ mice was 28 weeks , whereas the majority of Hampfl/fl controls were still alive at 52 weeks ( Figure 2A ) . 10 . 7554/eLife . 19804 . 011Figure 2 . Fatal cardiac abnormalities in Hampfl/fl;Myh6 . Cre+ mice . ( A ) Cumulative survival of Hampfl/fl;Myh6 . Cre+ mice ( n = 50 ) and Hampfl/fl littermate controls ( n = 47 ) over 52 weeks . ( B ) Representative H and E longitudinal heart sections from a six month old Hampfl/fl;Myh6 . Cre+ mouse and Hampfl/fl littermate control . ( C ) Representative WGA cardiac staining from a six month old Hampfl/fl;Myh6 . Cre+ mouse and Hampfl/fl littermate control . ( D ) Quantitation of cardiomyocyte size based on WGA staining in six month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . *p=0 . 001 relative to Hampfl/fl littermate controls . ( E ) Relative expression of the hypertrophic gene markers Myh7 and Nppb in hearts of 6 month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . *p=0 . 001 , 0 . 047 for the respective gene relative to Hampfl/fl littermate controls . ( F ) Representative cardiac in-situ TUNEL staining from a six month old Hampfl/fl;Myh6 . Cre+ mouse and Hampfl/fl littermate control . ( G ) Quantitation of percentage of apoptotic cardiomyocytes based on in-situ TUNEL staining in six month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls *p=0 . 001 relative to Hampfl/fl littermate controls . ( H ) Representative midventricular Cine MR images of hearts from Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl controls at 3 , 6 and 9 months of age . ( I–K ) Cine MRI measurements of LV lumen , at end-systole ( LVES ) , end-diastole ( LVED ) , and of ejection fraction ( LVEF ) in Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls at three months ( n = 8 per group ) , six months ( n = 11 per group , *p=0 . 043 for LVES , 0 . 047 for LVED and 0 . 020 for LVEF ) and nine months ( n = 5 per group , *p=0 . 044 for LVES , 0 . 042 for LVED and 0 . 034 for LVEF ) . p values are relative to Hampfl/fl controls of the respective age . All Values are plotted as mean ± SEM . n = 3 per group unless otherwise stated . Scale bar = 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01110 . 7554/eLife . 19804 . 012Figure 2—source data 1 . Source data file for Figure 2I , J and K . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 012 Six-month old mice were sacrificed for assessment of cardiac morphology , which showed gross enlargement of the left ventricle ( LV ) in Hampfl/fl;Myh6 . Cre+ hearts compared to Hampfl/fl controls ( Figure 2B ) . Assessment of cardiomyocyte size by wheat germ agglutinin ( WGA ) staining confirmed that Hampfl/fl;Myh6 . Cre+ cardiomyocytes were hypertrophied ( Figure 2C–D ) . This was accompanied by upregulation of expression of hypertrophic gene markers myosin heavy chain ( Myh7 ) and natriuretic peptide precursor ( Nppb ) ( Figure 2E ) . TUNEL staining for in-situ detection of cell death also showed significantly greater apoptosis in the hearts of Hampfl/fl;Myh6 . Cre+ mice than in Hampfl/fl controls ( Figure 2F–G ) . To characterise further the phenotype caused by loss of cardiac hepcidin , we used cine MRI in anaesthetised Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls at 3 , 6 and 9 months of age . Mid-ventricular cine MR images showed no differences between the two genotypes at 3 months of age . At 6 and 9 months of age , cine MR images showed marked enlargement of the LV in Hampfl/fl;Myh6 . Cre+ mice compared to Hampfl/fl controls ( Figure 2H ) . Formal quantitation of cardiac parameters by cine MRI confirmed enlargement of the LV lumen in Hampfl/fl;Myh6 . Cre+ mice , both at end-systole ( LVES ) ( Figure 2I ) and at end-diastole ( LVED ) ( Figure 2J ) , accompanied by a decrease in LV ejection fraction ( LVEF ) from 62% to 42% ( Figure 2K ) . Other parameters of cardiac performance were not significantly altered between mice from the two genotypes ( Table 2 ) . Taken together , histological examination of the hearts and cine MRI studies indicated that Hampfl/fl;Myh6 . Cre+ mice developed fatal LV dysfunction with reduced LVEF . 10 . 7554/eLife . 19804 . 013Table 2 . Non-LV parameters of cardiac function are not altered between Hampfl/fl and Hampfl/fl;;Myh6 . Cre+ mice . Cine MRI measurements of cardiac function in Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl controls at three months ( n = 8 per group ) , six months ( n = 11 per group ) and nine months ( n = 5 per group ) of age . Values are shown as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 013 3 months6 months9 monthsHamp fl/fl Hamp fl/fl;Myh6 . Cre+ Hamp fl/fl Hamp fl/fl;Myh6 . Cre+ Hamp fl/fl Hamp fl/fl;Myh6 . Cre+ Average mass ( mg ) 70 . 51 ± 7 . 47 70 . 38 ± 5 . 33 72 . 59 ± 5 . 38 82 . 54 ± 11 . 32 78 . 30 ± 4 . 92 83 . 07 ± 5 . 44 RVED lumen ( µl ) 31 . 37 ± 2 . 55 26 . 81 ± 2 . 30 33 . 12 ± 3 . 19 30 . 63 ± 2 . 35 38 . 39 ± 3 . 88 39 . 95 ± 3 . 00 RVES lumen ( µl ) 7 . 43 ± 0 . 79 5 . 23 ± 0 . 67 8 . 70 ± 1 . 22 8 . 34 ± 1 . 06 13 . 36 ± 2 . 21 15 . 51 ± 2 . 47 RVEF ( % ) 76 . 42 ± 1 . 32 79 . 88 ± 2 . 80 73 . 66 ± 2 . 50 73 . 42 ± 2 . 22 65 . 20 ± 4 . 36 61 . 95 ± 3 . 53 Stroke volume ( µl ) 25 . 47 ± 1 . 99 22 . 73 ± 2 . 11 25 . 70 ± 2 . 70 23 . 76 ± 1 . 85 29 . 79 ± 2 . 30 29 . 84 ± 1 . 59 Cardiac output ( ml/min ) 10 . 34 ± 1 . 06 9 . 62 ± 0 . 90 10 . 72 ± 1 . 07 10 . 53 ± 1 . 05 11 . 46 ± 1 . 15 11 . 99 ± 1 . 38 Heart Rate ( bpm ) 404 . 07 ± 18 . 69 426 . 90 ± 17 . 94 419 . 64 ± 19 . 05 436 . 14 ± 14 . 30 384 . 84 ± 28 . 55 400 . 97 ± 41 . 20 Heart/body weight ratio x10002 . 80 ± 0 . 18 3 . 09 ± 0 . 26 2 . 51 ± 0 . 23 2 . 98 ± 0 . 37 2 . 36 ± 0 . 28 2 . 88 ± 0 . 27 Such changes in cardiac performance could not be attributed to Cre recombinase toxicity in the heart as we have previously shown that Myh6 . Cre+ mice have normal cardiac function compared to wild type littermate controls ( Lakhal-Littleton et al . , 2015 ) . We examined FPN protein in the hearts of Hampfl/fl;Myh6 . Cre+ mice , and found that FPN protein was markedly upregulated compared to Hampfl/fl controls ( Figure 3A ) , consistent with the idea that loss of cardiac HAMP was acting through upregulation of cardiomyocyte FPN . In order to test whether cardiac dysfunction arose from upregulation of cardiomyocyte FPN , we engineered mice where the Slc40a1 gene harbours a conditional cardiac-specific C326Y point mutation , which confers HAMP-resistance while conserving the iron export function of FPN ( Schimanski et al . , 2005; Drakesmith et al . , 2005 ) . We confirmed that the Slc40a1 C326Y fl allele produced the C326Y transcript specifically in the heart ( Figure 3—figure supplement 1 ) and that Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice did not exhibit changes in systemic iron indices ( Table 3 ) . As seen in Hampfl/fl;Myh6 . Cre+ mice , cardiomyocyte FPN was indeed upregulated in Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice ( Figure 3B ) . 10 . 7554/eLife . 19804 . 014Figure 3 . The role of cardiomyocyte FPN . ( A–B ) Immunohistochemical staining for FPN in the hearts of three month old Hampfl/fl;Myh6 . Cre+ , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and respective controls . ( C ) Cumulative survival of Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice ( n = 36 ) and Slc40a1 C326Yfl/fl littermate controls ( n = 31 ) over 52 weeks . ( D ) Representative H and E-stained longitudinal heart sections from a six month old Slc40a1 C326Yfl/fl;Myh6 . Cre+ mouse and a Slc40a1 C326Yfl/fl control . ( E ) Representative WGA cardiac staining from a six month old Slc40a1 C326Yfl/fl;Myh6 . Cre+ mouse and Slc40a1 C326Yfl/fl control . ( F ) Quantitation of cardiomyocyte size based on WGA staining . n = 3 per group . *p=0 . 001 relative to Slc40a1 C326Yfl/fl controls . ( G ) Relative expression of Myh7 and Nppb in hearts of 6 month old Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and Slc40a1 C326Yfl/fl controls . n = 3 per group . *p=0 . 032 , 0 . 044 for the respective gene relative to Slc40a1 C326Yfl/fl controls . ( H ) Representative cardiac TUNEL staining from a six month old Slc40a1 C326Yfl/fl;Myh6 . Cre+ mouse and Slc40a1 C326Yfl/fl control . ( I ) Quantitation of percentage of apoptotic cardiomyocytes based on TUNEL staining , n = 3 per group . *p=0 . 0003 relative to Slc40a1 C326Yfl/fl controls . ( J–L ) Cine MRI measurements of LVES , LVED and LVEF in Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and Slc40a1 C326Yfl/fl controls at three months ( n = 6 per group ) , six months ( n = 6 per group , *p=0 . 003 for LVES , 0 . 043 for LVED and 0 . 001 for LVEF ) and nine months ( n = 5 per group , *p=0 . 033 for LVES , 0 . 047 for LVED and 0 . 023 for LVEF ) . P values are relative to Slc40a1 C326Yfl/fl controls of the same age . ( M ) Percentage Fe55 efflux in cardiomyocytes from Hampfl/fl;Myh6 . Cre+ mice , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and respective controls , in presence or absence of HAMP peptide . *p=0 . 018 , 0 . 006 and 0 . 007 respectively ( N ) Elemental iron levels in cardiomyocyte fractions ( CF ) from the hearts of Hampfl/fl;Myh6 . Cre+ mice , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and their respective controls . n = 4 per group . *p=0 . 032 , 0 . 044 , 0 . 047 and 0 . 031 respectively . ( O–P ) Relative TfR1 ( *p=0 . 038 , 0 . 001 ) and Fpn ( *p=0 . 039 , 0 . 047 ) expression in hearts of 3 month old Hampfl/fl;Myh6 . Cre+ mice , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and their respective controls . All values are plotted as mean ± SEM . Scale bar = 50 µm . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01410 . 7554/eLife . 19804 . 015Figure 3—source data 1 . Source data file for Figure 3JK and L . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01510 . 7554/eLife . 19804 . 016Figure 3—figure supplement 1 . Characterisation of Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice . Total mRNA was extracted from the liver and heart of an adult a Slc40a1 C326Yfl/+;Myh6 . Cre+ mouse , reverse transcribed using primers for exon 7 of the Slc40a1 mRNA transcript . Products were sequenced to confirm successful heterozygous expression of the C326Y transcript in the heart but not in the liver . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01610 . 7554/eLife . 19804 . 017Figure 3—figure supplement 2 . Elemental iron levels in total hearts of Hampfl/fl;Myh6 . Cre+ mice , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice and their respective controls at three months and 6 months of age . n = 4 per group . Data are plotted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01710 . 7554/eLife . 19804 . 018Table 3 . Characterisation of Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice . Indices of iron status in Slc40a1 C326Yfl/fl and Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice at six months of age ( n = 4 per group ) . Values are shown as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 018 Slc40a1 C326Yfl/fl Slc40a1 C326Y fl/fl; Myh6 . Cre+ liver total elemental iron ( ng/mg tissue ) 92 . 77 ± 21 . 30 84 . 00 ± 26 . 00 liver ferritin ( µg/mg total protein ) 0 . 87 ± 0 . 06 0 . 92 ± 0 . 05 serum iron ( µmol/L ) 27 . 30 ± 5 . 20 29 . 60 ± 7 . 20 serum ferritin ( mg/L ) 2 . 10 ± 0 . 04 2 . 20 ± 0 . 15 hemoglobin ( g/L ) 125 . 70 ± 8 . 80 126 . 00 ± 12 . 30 serum hepcidin ( µg/L ) 25 . 90 ± 11 . 60 27 . 50 ± 8 . 40 Like Hampfl/fl;Myh6 . Cre+ mice , Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice also had increased mortality relative to their littermate controls ( Figure 3C ) . We then determined whether Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice developed the same phenotype of cardiac dysfunction as Hampfl/fl;Myh6 . Cre+ mice . Histologically , Slc40a1 C326Yfl/fl;Myh6 . Cre+ hearts from six month old mice also showed LV enlargement ( Figure 3D ) , hypertrophied cardiomyocytes ( Figure 3E–F ) , upregulation of hypertrophy markers Myh7 and Nppb ( Figure 3G ) and a greater degree of apoptosis compared to Slc40a1 C326Yfl/fl controls ( Figure 3H–I ) . When we measured cardiac performance by cine MRI , we found that Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice also developed LV dysfunction by 6 months of age , with a reduction in LVEF from 73% to 55% ( Figure 3J–L ) . The similarity between Hampfl/fl;Myh6 . Cre+ and Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice in terms of the nature and time course of cardiac dysfunction suggests a common mechanism of cardiac dysfunction involving upregulation of cardiomyocyte FPN . Therefore , we tested whether this upregulation of FPN resulted in increased iron efflux from cardiomyocytes . Iron Fe55 efflux was indeed significantly greater in cardiomyocytes isolated from Hampfl/fl;Myh6 . Cre+ and Slc40a1 C326Yfl/fl;Myh6 . Cre+ hearts than in cardiomyocytes isolated from their respective controls ( Figure 3M ) . Addition of exogenous mouse HAMP in the efflux medium inhibited the increase in Fe55 efflux from Hampfl/fl;Myh6 . Cre+ cardiomyocytes but not from Slc40a1 C326Yfl/fl;Myh6 . Cre+ cardiomyocytes , consistent with the HAMP-resistant mutation in Slc40a1 C326Yfl/fl;Myh6 . Cre+ cardiomyocytes . We hypothesised that upregulation of cardiac FPN and iron export in Hampfl/fl;Myh6 . Cre+ and Slc40a1 C326Yfl/fl;Myh6 . Cre+ hearts caused cardiomyocyte iron depletion . To test this hypothesis , we quantified iron levels both in total hearts and in the isolated cardiomyocyte fractions at 3 months and 6 months of age . While iron levels in total hearts were not significantly different between any of the genotypes ( Figure 3—figure supplement 2 ) , the iron content of the cardiomyocyte fraction was significantly lower in Hampfl/fl;Myh6 . Cre+ and in Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice than in their respective controls ( Figure 3N ) . Furthermore , expression of TfR1 mRNA was upregulated ( Figure 3O ) and Slc40a1 mRNA was downregulated ( Figure 3P ) in Hampfl/fl;Myh6 . Cre+ and Slc40a1 C326Yfl/fl;Myh6 . Cre+ hearts relative to their respective controls , consistent with a transcriptional response to intracellular iron deficiency ( Rouault and Klausner , 1997; Rouault , 2006; Ward and Kaplan , 1823 ) . Together these results demonstrate that loss of either cardiac hepcidin or hepcidin responsiveness in the heart results in upregulation of cardiomyocyte FPN , and that cardiomyocytes of Hampfl/fl;Myh6 . Cre+ and Slc40a1 C326Yfl/fl;Myh6 . Cre+ hearts are iron deficient as a result of upregulation of FPN-mediated iron export . As cardiomyocyte iron deficiency preceded the development of cardiac dysfunction , we hypothesised it is the cause of the cardiac phenotype in Hampfl/fl;Myh6 . Cre+ mice . To test this hypothesis , we treated Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl controls with fortnightly intravenous injections of ferric carboxymaltose solution containing 0 . 5 mg iron from three months of age , and confirmed the effects of this treatment on cardiac and systemic iron indices at 6 months of age ( Table 4 ) . At this timepoint , we performed cine MRI and found that the LV enlargement and the reduced LVEF , seen in untreated Hampfl/fl;Myh6 . Cre+ mice , were prevented in iron-treated Hampfl/fl;Myh6 . Cre+ mice ( Figure 4A–C ) . The transcriptional response to intracellular iron deficiency in untreated Hampfl/fl;Myh6 . Cre+ mice ( upregulation of TfR1 mRNA and downregulation of Slc40a1 mRNA relative to Hampfl/fl controls ) , was absent in iron-treated Hampfl/fl;Myh6 . Cre+ mice , consistent with correction of cardiomyocyte iron deficiency ( Figure 4D–E ) . Prevention of cardiac dysfunction in Hampfl/fl;Myh6 . Cre+ mice by intravenous iron treatment confirms the causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction in this setting . 10 . 7554/eLife . 19804 . 019Figure 4 . The role of cardiomyocyte iron deficiency and metabolic derangement in cardiac dysfunction . ( A–C ) Cine MRI measurements of LV lumen , at end-systole ( LVES , *p=0 . 048 ) , end-diastole ( LVED , *p=0 . 031 ) , and of ejection fraction ( LVEF , *p=0 . 004 ) in 6-month old untreated ( -iron ) and I . V iron-treated ( +iron ) Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . n = 5 per group . ( D–E ) Relative TfR1 ( *p=0 . 001 ) and Fpn ( *p=0 . 002 ) expression in hearts of 6-month old untreated ( -iron ) and I . V iron-treated ( +iron ) Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . n = 4 per group . ( F–H ) Enzymatic activities of Aconitase I ( *p=0 . 035 , 0 . 041 ) , Complex I ( *p=0 . 004 , 0 . 030 ) and Complex IV ( *p=0 . 003 , 0 . 026 ) in untreated ( -iron ) 3-month and 6-month old and in I . V iron-treated ( +iron ) 6-month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . n = 4 per group . ( I ) Representative EM micrographs of hearts from untreated ( -iron ) 3-month and 6-month old and I . V iron-treated ( +iron ) 6-month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . Scale bar = 2 µm . ( J–L ) Relative Hk2 ( *p=0 . 010 , 0 . 002 ) , Eno ( *p=0 . 014 , 0 . 021 ) and Ldha ( *p=0 . 003 , 0 . 001 ) expression levels in hearts of untreated ( -iron ) 3-month and 6-month old and in I . V iron-treated ( +iron ) 6-month old Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . n = 4 per group . NS=not significant . All values are plotted as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 01910 . 7554/eLife . 19804 . 020Figure 4—source data 1 . Source data file for Figure 4A , B and C . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 02010 . 7554/eLife . 19804 . 021Table 4 . Effect of intravenous iron treatment on iron indices . Total cardiac and liver elemental iron , serum iron and circulating HAMP in 6-month old untreated and I . V iron-treated Hampfl/fl;Myh6 . Cre+ mice and Hampfl/fl littermate controls . Treated mice were injected with 0 . 5 mg iron fortnighly from the age of 3 months . Tissues and serum were harvested 12 hr after the final injection . n = 5 per group . *p<0 . 05 relative to untreated Hampfl/fl mice . †p < 0 . 05 relative to untreated Hampfl/fl;Myh6 . Cre+ . Values are shown as mean ± SEM . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 021Hampfl/fl Hamp fl/fl;Myh6 . Cre+ untreatedtreated with I . V ironuntreatedtreated with I . V ironcardiac total elemental iron ( ng/mg tissue ) 82 . 2 ± 16 . 9 331 . 3 ± 21 . 5*74 . 9 ± 7 399 . 8 ± 68 . 5†liver total elemental iron ( ng/mg tissue ) 100 . 4 ± 11 2527 . 6 ± 27 . 63*96 . 3 ± 14 2258 . 2 ± 239 . 9†serum iron ( µmol/L ) 30 . 09 ± 6 . 37 74 . 48 ± 17 . 96*31 . 5 ± 6 . 9 80 . 12 ± 24 . 9†serum hepcidin ( µg/L ) 27 . 41 ± 6 . 7 237 . 3 ± 16 . 7*28 . 9 ± 9 . 4 209 . 8 ± 38 . 8† Having confirmed a causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction , we aimed to understand the mechanisms linking the two . Iron is a cofactor for several enzymes involved in metabolism ( Meyer , 2008; Sono et al . , 1996; Solomon et al . , 2003 ) , and metabolic derangement is a well-recognized precursor to cardiac dysfunction ( Belke et al . , 2000; Kakinuma et al . , 2000; Ashrafian et al . , 2007 ) . Iron deficiency has been reported to reduce the levels and/or activities of key metabolic iron-containing enzymes , in cell lines , in the hearts of mice with impaired cardiac iron uptake ( cardiac-specific TfR1 knockouts ) and in the hearts of mice fed an iron-deficient diet ( Dallman , 1986; Dhur et al . , 1989; Xu et al . , 2015; Oexle et al . , 1999 ) . Based on those studies , we postulated that iron deficiency in the cardiomyocytes of Hampfl/fl;Myh6 . Cre+ mice would also result in reduction in the activities of key metabolic iron-containing enzymes . To test this hypothesis , we measured the activities of the iron-sulphur containing enzyme Aconitase I as well as electron transport chain ( ETC ) complexes in cardiac lysates from Hampfl/fl;Myh6 . Cre+ hearts at 3 months and 6 months of age . We found that Aconitase I , Complex I and Complex IV activities were significantly reduced in Hampfl/fl;Myh6 . Cre+ hearts compared to Hampfl/fl controls at 3 months and 6 months of age , and that this reduction in activity was prevented in iron-treated 6-month old mice ( Figure 4F–H ) . As ETC activity is essential to mitochondrial function , we examined whether Hampfl/fl;Myh6 . Cre+ hearts had signs of mitochondrial failure , and whether such mitochondrial failure was prevented by intravenous iron supplementation . By electron microscopy ( EM ) , dilation of mitochondrial cristae was seen in Hampfl/fl;Myh6 . Cre+ hearts as early as 3 months of age and progressed further at 6 months of age . However , this was prevented in iron-treated Hampfl/fl;Myh6 . Cre+ mice , which had healthy-looking mitochondria at 6 months of age ( Figure 4I ) . Similar changes in mitochondrial morphology were seen in the hearts of 3 month old Slc40a1 C326Yfl/fl;Myh6 . Cre+ mice ( data not shown ) . Impairment of electron transport is known to drive glycolysis , as an alternative route of ATP production ( Schönekess et al . , 1997; Kato et al . , 2010 ) . Therefore , we examined the expression of a number of genes encoding glycolytic enzymes . We found that expression of genes encoding Hexokinase 2 ( Hk2 ) which catalyses the first step of glycolysis , Enolase ( Eno ) , which catalyses the penultimate step of glycolysis and of Lactate dehydrogenase A ( Ldha ) which catalyses the ultimate step of glycolysis were all significantly increased at 3 months and 6 months of age in hearts of untreated Hampfl/fl;Myh6 . Cre+ mice . This upregulation of glycolytic genes was not seen in iron-treated Hampfl/fl;Myh6 . Cre+ mice ( Figure 4J–L ) . These results demonstrate that reduction in the activities of key iron-containing metabolic enzymes , mitochondrial failure and upregulated glycolysis precede the development of cardiac dysfunction in Hampfl/fl;Myh6 . Cre+ hearts , and are prevented by intravenous iron treatment .
The major finding of this study is that cardiomyocyte hepcidin is required for autonomous cellular iron homeostasis . Loss of hepcidin responsiveness specifically in cardiomyocytes engendered the same effects as loss of cardiac hepcidin , demonstrating that cardiac hepcidin operates in an autocrine fashion by regulating cardiomyocyte FPN . A role in cellular iron homeostasis , and an autocrine mode of action for the HAMP/FPN axis have not been described previously in any other tissue . Indeed , hepcidin and FPN are better known to interact in an endocrine fashion , at the level of the gut , spleen and liver , to regulate systemic iron homeostasis . A second important finding is that this cardiac HAMP/FPN axis is essential for normal heart function , and that its disruption leads to ultimately fatal cardiac metabolic and contractile dysfunction , even against a background of intact systemic iron homeostasis . Metabolic and contractile dysfunction are preceded by cardiomyocyte iron deficiency and prevented by intravenous iron supplementation , indicating a causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction . This causal relationship has previously been demonstrated in studies using mice lacking cardiac TfR1 , in which cardiac iron deficiency also affects the heart against a background of otherwise intact systemic iron homeostasis ( Xu et al . , 2015 ) . Furthermore , the importance of metabolic derangement seen in our mouse models is also consistent with the findings in mice lacking the cardiacTfR1 , and in mice and rats fed iron deficient diets ( Dhur et al . , 1989; Xu et al . , 2015; Tanne et al . , 1994; Walter et al . , 2002 ) . Correction of metabolic and contractile dysfunction by intravenous iron treatment likely involves not only increased iron availability for uptake into cardiomyocytes , but also the effects of increased circulating HAMP on cardiomyocyte FPN . Surprisingly , we found that , both in vitro and in vivo , cardiac HAMP protein responded differently from its transcript to changes in iron levels . The current understanding of hepcidin regulation is based on studies of hepatic hepcidin . In that setting , release of the active mature HAMP peptide is dependent on cleavage of the propeptide by Furin ( Valore and Ganz , 2008 ) . We found that Furin inhibition increased iron export from Hampfl/fl but not from Hampfl/fl;Myh6 . Cre+ cardiomyocytes ( Appendix 1—figure 1 ) , demonstrating that cardiomyocytes secrete an active HAMP peptide in a Furin-dependent manner . Furthermore , we found that increased HAMP release from iron-deficient cardiomyocytes depended on Furin , and that cardiac Furin itself is upregulated by iron deficiency both in vitro and in vivo . The latter finding is consistent with the reported regulation of Furin by iron deficiency through Hypoxia-Inducible Factors HIFs ( Silvestri et al . , 2008 ) . These data suggest that differential regulation of Furin by iron may explain the divergent effects of iron on Hamp transcript and HAMP protein . Comprehensive studies using cardiac-specific knockouts of putative regulators will be required to explore formally the regulation or otherwise , of cardiac hepcidin by pathways known to regulate hepatic hepcidin . The upregulation of cardiac HAMP in mice fed an iron-deficient diet raises the possibility that it may be involved in protecting the heart in the setting of systemic iron deficiency . This hypothesis is supported by the finding that , when provided an iron-deficient diet , Hampfl/fl;Myh6 . Cre+ mice exhibited a greater cardiac hypertrophic response than their Hampfl/fl littermate controls ( Appendix 1—figure 2 ) . Intracellular iron levels are dependent both on cellular homeostatic pathways and on systemic iron availability in plasma . Therefore , the interplay between the cardiac and the systemic HAMP/FPN axes is important in determining cardiomyocyte iron levels . Some insight into this interplay is gained from comparing systemic and cardiac mouse models of disrupted iron homeostasis . It is interesting that ubiquitous Hamp knockout ( Lakhal-Littleton et al . , 2015 ) and ubiquitous Slc40a1 C326Y knock-in mice ( Appendix 1—figure 3 ) , both models of systemic iron overload , do not develop the cardiac dysfunction seen in cardiomyocyte-specific Hamp knockout and cardiomyocyte-specific Slc40a1 C326Y knock-in mice . This suggests that , while upregulation of cardiomyocyte FPN under conditions of normal iron availability ( cardiac-specific models described in this study ) is detrimental to cardiac function , it is protective under conditions of increased systemic iron availability ( systemic models ) . Previously , we also showed that deletion of cardiomyocyte FPN resulted in fatal cardiomyocyte iron overload , preventable by dietary iron restriction ( Lakhal-Littleton et al . , 2015 ) . Together , our studies demonstrate that iron levels within cardiomyocytes are a balance between cellular iron efflux which is regulated by the cardiac HAMP/FPN axis , and systemic iron availability which is regulated by the systemic HAMP/FPN axis ( Figure 5 ) . 10 . 7554/eLife . 19804 . 022Figure 5 . Interplay between systemic and cardiac iron HAMP/FPN axes . Cardiomyocyte iron content is determined by both systemic iron availability , which is regulated by liver HAMP , and by the cardiac HAMP/FPN axis , which regulates cardiomyocyte iron efflux . In the wild type heart , cardiac HAMP regulates the levels of cardiac FPN and iron release from cardiomyocytes . In this study , we have demonstrated that loss of cardiac HAMP ( cardiac Hamp KO ) or loss of cardiac HAMP responsiveness ( cardiac Slc40a1 C326Y KI ) result in cardiomyocyte iron deficiency due to increased cardiomyocyte FPN and iron release . Previously , we also demonstrated that loss of cardiomyocyte FPN caused cardiomyocyte iron overload . In these two sets of conditions , cardiomyocyte iron deficiency and cardiomyocyte iron overload cause cardiac dysfunction . We have also shown that upregulation of cardiac FPN occurs as a result of loss of either systemic HAMP or systemic HAMP responsiveness , and is protective against the otherwise detrimental effects of systemic iron overload . DOI: http://dx . doi . org/10 . 7554/eLife . 19804 . 022 Iron overload is detrimental to cardiac health , as demonstrated by iron overload cardiomyopathy in hemochromatosis and thalassemia major patients ( Gulati et al . , 2014 ) . Our model of cardiac iron homeostasis implies that the cardiac HAMP/FPN axis may have a modifying effect on the severity of iron-overload cardiomyopathy . Thus , it would be interesting to explore whether differences in the levels of cardiac FPN and HAMP , possibly due to different local stimulatory and suppressive signals ( e . g local inflammation , local ischemia ) , explain the reported lack of concordance between the degrees of cardiac iron overload and liver iron overload in a significant proportion of hemochromatosis and thalassemia major patients ( Anderson et al . , 2001; Noetzli et al . , 2008 ) . Iron deficiency is also detrimental to cardiac health . Indeed , systemic iron deficiency correlates with functional and molecular markers of disease severity in patients with chronic heart failure ( CHF ) ( Klip et al . , 2013; Comín-Colet et al . , 2013 ) , and also appears to contribute to the risk of death after an episode of acute heart failure ( Jankowska et al . , 2014 ) . Given the high prevalence of iron deficiency in patients with CHF , ranging between 30–50% ( Klip et al . , 2013; Nanas et al . , 2006; Ezekowitz et al . , 2003 ) , the European Society of Cardiology recently recommended the assessment of iron deficiency as a comorbidity in CHF . Furthermore , several clinical trials have now established the benefits of intravenous iron supplementation in CHF patients , with or without anaemia ( Ponikowski et al . , 2015; Anker et al . , 2009; Avni et al . , 2012 ) . The mechanisms underlying the anaemia-independent effects of iron deficiency and the benefits of intravenous iron in CHF patients are not fully understood . In light of the direct effect of cardiomyocyte iron deficiency on heart function , demonstrated in this and other studies , it would interesting to explore whether systemic iron deficiency in CHF patients is accompanied by cardiomyocyte iron deficiency , and whether correction of the later underlies the benefits of intravenous iron supplementation in non-anaemic patients . Another open question is whether disruption of the cardiac HAMP/FPN axis contributes to the pathophysiology of heart disease . It has been shown , in the rat model of myocardial infarction ( MI ) , that Hamp mRNA and HAMP protein are elevated in the ischemic myocardium during the acute phase ( Simonis et al . , 2010 ) . In humans , circulating HAMP was shown to be elevated in the serum within 4 hr of MI , although the tissue source of this hepcidin was not identified ( Suzuki et al . , 2009 ) . In addition , decreased cardiac HAMP expression has been reported in a transgenic mouse model of dilated cardiomyopathy , where the phenotype was ameliorated following transgenic overexpression of cardiac hepcidin ( Zhang et al . , 2012 ) . Thus , further studies are warranted in humans to explore formally the role of the cardiac HAMP/FPN axis in the aetiology of heart disease . Currently , there is considerable interest in targeting the HAMP/FPN axis for the treatment of iron overload and iron deficiency . Our studies suggest that such strategies may also impinge on cardiac iron homeostasis and function . Other than the heart , both FPN and hepcidin are also found in the brain , kidney and placenta ( McCarthy and Kosman , 2014; Kulaksiz , 2005; Evans et al . , 2011; Rouault , 2013; Moos and Rosengren Nielsen , 2006; Bastin et al . , 2006; Wolff et al . , 2011 ) . It would be important to establish whether our findings in the heart extend to these tissues .
All animal procedures were compliant with and approved under the UK Home Office Animals ( Scientific Procedures ) Act 1986 . Both males and females were used in experiments , with the respective littermate control being of a matching sex . The strategy for generating cardiac Hamp knockout mice is outlined in Appendix 1—figure 4 . Briefly , a targeting vector was designed to introduce a floxed Hamp allele into C57BL/6N mouse ES cells ( JM8F6 ) with exons 2 and 3 , which encode the majority of the peptide , flanked by LoxP sites and a line of floxed mice was generated by blastocyst injection of targeted ES cells , as previous described ( Lakhal-Littleton et al . , 2015 ) . Further breeding with a C57BL/6 Flp recombinase deleter mouse allowed removal of the Neomycin resistance cassette . Cardiac Hamp knockouts were then generated by by crossing Hampfl/fl animals with mice transgenic for Cre recombinase , under the control of cardiomyocyte-specific Alpha Myosin Heavy Chain ( Myh6 ) promoter ( B6 . FVB-Tg ( Myh6-cre ) 2182Mds/J ) . The subsequent breeding strategy was designed to produce cardiac Hamp knockouts and homozygous floxed controls ( Hampfl/fl;Myh6 . Cre+ and Hampfl/fl respectively ) in the same litter . The strategy for generating Cardiac Slc40a1 C326Y knock-in mice is outlined in Appendix 1—figure 5 , and further details are provided in the Appendix . Mice were anaesthetized with 2% isofluorane in O2 and positioned supine in a purpose-built cradle . ECG electrodes were inserted into the forepaws , a respiration loop was taped across the chest and heart and respiration signals were monitored using a custom-built physiological motion gating device . The cradle was lowered into a vertical-bore , 11 . 7 T MR system with a 40 mm birdcage coil ( Rapid Biomedical , Würzburg , Germany ) and visualised using a Bruker console running Paravision 2 . 1 . 1 . A stack of contiguous 1 mm thick true short-axis ECG and respiration-gated cine-FLASH images were acquired . The entire in vivo imaging protocol was performed in approximately 60 min . Image analysis was performed using ImageJ ( NIH Image , Bethesda , MD ) . Left ventricular volumes and ejection fractions were calculated from the stack of cine images as described ( Lakhal-Littleton et al . , 2015 ) . Unless otherwise stated , animals were provided with a standard rodent chow diet containing 200 ppm iron . In iron manipulation experiments , mice were given an iron-deficient diet ( 2–5 ppm iron; Teklad TD . 99397; Harlan Laboratories ) , or an iron-loaded diet ( 5000 ppm iron; Teklad TD . 140464 ) or a matched control diet ( 200 ppm iron; Teklad TD . 08713 ) from weaning for six weeks . Adult primary cardiomyocytes were isolated from eight week old C57BL/6 mice . Hearts were cannulated and mounted on a langendorff apparatus , then perfused using a liberase solution for 10 min . After filtration through a 400 µm gauze , cells were cultured in MEM medium containing Hanks salts , L-glutamine and antibiotics . Within 2 hr of cardiomyocyte culture , supernatants were replaced with fresh medium containing 10% Fetal calf serum , with 0 . 5 mmol/L ferric citrate ( FAC ) ( F3388 , Sigma Aldrich ) or 100 µmol/L desferroxamine ( D9533 , Sigma Aldrich ) for 8 hr . The Furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone ( CMK ) ( N1505 , Bachem ) was added at a concentration of 50 µmol/L for the duration of DFO and FAC treatment . Total RNA extraction and cDNA synthesis were carried out as previously described ( Lakhal-Littleton et al . , 2015 ) . Gene expression was measured using Applied Biosystems Taqman gene expression assay probes for Slc40a1 , Hamp , TfR1 , Myh7 , Nppb , Ldha , Hk2 , Eno and house-keeping gene β-Actin ( Life Technologies , Carlsbad , CA ) . The CT value for the gene of interest was first normalised by deducting CT value for β-Actin to obtain a delta CT value . Delta CT values of test samples were further normalised to the average of the delta CT values for control samples to obtain delta delta CT values . Relative gene expression levels were then calculated as 2-delta deltaCT . Tissues were prepared as described previously ( Lakhal-Littleton et al . , 2015 ) and stained with rabbit polyclonal anti-mouse HAMP antibody ( ab30760 , Abcam , RRID:AB_2115844 ) at 1/40 dilution , or rabbit polyclonal anti-mouse FPN antibody ( MTP11-A , Alpha Diagnostics , RRID:AB_1619475 ) at 1/200 dilution . Results of control experiments confirming the specificity of the HAMP antibody are shown in Figure 1—figure supplement 5 . HAMP was measured in mouse sera and in cardiomyocyte supernatants using a HAMP ELISA kit ( E91979Mu , USCN ) according to the manufacturer’s instructions . Results of control experiments confirming that in vitro treatments did not affect HAMP peptide detection by this ELISA kit are shown in Figure 1—figure supplement 6 . Formalin-fixed paraffin-embedded tissue sections were deparaffinised using Xylene , then rehydrated in ethanol . Slides were then stained for 1 hr with 1% potassium ferricyanide in 0 . 1 mol/L HCl buffer . Endogenous peroxidase activity was quenched , then slides were stained with DAB chromogen substrate and counterstained with haematoxylin . They were visualised using a standard brightfield microscope . Hearts were dissected and 0 . 5–1 mm3 slices were fixed by immersion for 2 hr in 2 . 5% glutaraldehyde in 0 . 1 mol/L cacodylate buffer and prepared for electron microscopy by standard methods . Briefly , cells were post-fixed in osmium tetroxide ( 1% w/v in 0 . 1 mol/L phosphate buffer ) , stained with uranyl acetate ( 2% w/v in distilled water ) , dehydrated through increasing concentrations of ethanol ( 70–100% ) and acetone and embedded in TAAB resin ( TAAB , Aldermaston , UK ) . Ultrathin sections ( 50–80 nm ) were prepared using a Reichert ultracut S microtome and mounted on 200 mesh nickel grids . Sections were lightly counterstained with lead citrate and uranyl acetate and examined with a Jeol transmission electron microscope ( JEM-1010 , JEOL , Peabody MA ) . Adult cardiomyocytes were isolated from mice of the desired genotype at 9 weeks of age as described above . Cardiomyocytes were then cultured in 24-well plates at equal densities for 16 hr before the efflux experiment was performed as described ( McKie et al . , 2000 ) . Briefly , after washing with three times PBS , cells were incubated for 30 min in 200 µl uptake solution ( 98 mmol/L NaCl , 2 . 0 mmol/L KCl , 0 . 6 mmol/L CaCl2 , 1 . 0 mmol/L MgCl2 , 1 . 0 mmol/L ascorbic acid , 10 mmol/L HEPES [pH 6 . 0] with Tris base , 50 μmol/L Fe55 [NEN , Boston , MA] ) , then washed three times with PBS and incubated for 30 min with efflux solution ( 98 mmol/L NaCl , 2 . 0 mmol/L KCl , 0 . 6 mmol/L CaCl2 , 1 . 0 mmol/L MgCl2 , 10 mmol/L HEPES [pH 7 . 4] with Tris base , 300 U/ml bovine ceruloplasmin ( cp ) ( Sigma ) and 40 μg/ml human apotransferrin ( tf ) [Sigma] ) , in the absence or presence of 0 . 5 µmol/L mouse HAMP peptide ( Peptides International ) . The efflux medium was then removed , the cells washed three times in ice-cold PBS and disrupted by incubation in 100 µl of 10% SDS solution for 10 min . The efflux solution and cell lysates were then transferred into scintillation vials for Fe55 counting . Where Furin inhibition was carried out , CMK was added to the culture medium at 50 µmol/L 2 hr before the efflux experiment was carried out . Ferritin concentration in serum and in liver lysates was determined using the ferritin ELISA kit ( ICL , Inc . Portland ) . Serum iron levels were determined using the ABX-Pentra system ( Horiba Medical , CA ) . Determination of total elemental iron in the heart was carried out by inductively coupled plasma mass spectrometry ( ICP-MS ) as described previously ( Lakhal-Littleton et al . , 2015 ) . Calibration was achieved using the process of standard additions , where spikes of 0 ng/g , 0 , 5 ng/g , 1 ng/g , 10 ng/g , 20 ng/g and 100 ng/g iron were added to replicates of a selected sample . An external iron standard ( High Purity Standards ICP-MS-68-A solution ) was diluted and measured to confirm the validity of the calibration . Rhodium was also spiked onto each blank , standard and sample as an internal standard at a concentration of 1 ng/g . Concentrations from ICP-MS were normalised to starting tissue weight . Following cardiac perfusion , hearts were dissected into small pieces in ice-cold Hanks buffer , and subject to collagenase P digestion at 37C for 1 hr ( 11213857001 , Roche Diagnostics ) . Following lysis of red blood cells , cell suspensions were passed through a 70 µm sieve , before being labelled using cardiomyocyte isolation kit ( 130-100-825 , Miltenyi Biotec ) . Separation of cardiomyocyte and non-cardiomyocyte fractions was carried out using MACS magnetic separation system according to the manufacturer’s instructions . Cardiac fractions were lysed immediately for ICP-MS analysis . Approximately 10 mg of frozen , crushed tissues were suspended in 200 µl of ice-cold KME buffer ( 100 mmol/L KCl , 50 mmol/L Mops , 0 . 5 mmol/L EGTA , pH 7 . 4 ) , then homogenized by rupturing with a TissueRuptor ( Qiagen , UK ) over ice . In a plastic cuvette , the cardiac lysate is mixed with assay buffer and slotted into a spectrophotometer . Details of assay buffers and of reaction procedure for each enzyme are detailed in the Appendix . Values are shown as mean ± standard error of the mean ( SEM ) . Comparison of iron indices , enzyme activities and parameters of cardiac function between groups was performed using Student’s T test . p values < 0 . 05 were deemed as indicating significant differences between groups . Where significant , exact p values for a figure panel are stated in the corresponding figure legend . No explicit power analysis was performed prior to the experiments to determine sample size , since we had no means to reliably estimate the size and variability of the effects of deleting hepcidin on parameters of cardiac function . For Cine MRI assessment of cardiac function , typically 5–11 animals of each genotype were used , with a matching number of littermate controls . For gene expression , iron quantitation and enzyme activity assays from mouse tissues , typically , 3–6 independent biological replicates and matching littermate controls were analysed . Since significant results were obtained from these set of experiments , no further animals were sacrificed . All 'n' values reported refer to independent biological replicates . | Many proteins inside cells require iron to work properly , and so this mineral is an essential part of the diets of most mammals . However , because too much iron in the body is also bad for health , mammals possess several proteins whose role is to maintain the balance of iron . Two proteins in particular , called hepcidin and ferroportin , are thought to be important in this process . Some ferroportin is found in the cells that line the gut ( where iron is absorbed into the body ) and is required to release this iron into the bloodstream . It is also found in the spleen , which is where iron is removed from old red blood cells so that it can be recycled . The liver produces hepcidin to control when ferroportin is active in the gut and spleen . Both hepcidin and ferroportin are also found in heart cells . In 2015 , a study reported that that heart ferroportin plays an important role in heart activity . However , it was not clear what role hepcidin plays in this organ . Now , Lakhal-Littleton et al . – including many of the researchers from the previous work – have genetically engineered mice such that they specifically lacked heart hepcidin , or had a version of ferroportin in their heart that does not respond to hepcidin . The experiments show that these changes caused fatal heart failure in the mice because ferroportin releases iron from heart cells in an uncontrolled manner . Lakhal-Littleton et al . were able to prevent heart failure by injecting the animals with iron directly into the bloodstream . These findings show that hepcidin produced outside the liver has a role in controlling the levels of iron in the body’s organs . Other organs such as the brain , kidney and placenta all have their own forms of hepcidin and ferroportin; further work could investigate the roles of these proteins . Finally , another challenge for the future will be to test whether new drugs that are being developed to block or mimic hepcidin from the liver have the potential to treat heart conditions in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"biochemistry",
"and",
"chemical",
"biology",
"cell",
"biology"
] | 2016 | An essential cell-autonomous role for hepcidin in cardiac iron homeostasis |
Striatal dysfunction plays an important role in dystonia , but the striatal cell types that contribute to abnormal movements are poorly defined . We demonstrate that conditional deletion of the DYT1 dystonia protein torsinA in embryonic progenitors of forebrain cholinergic and GABAergic neurons causes dystonic-like twisting movements that emerge during juvenile CNS maturation . The onset of these movements coincides with selective degeneration of dorsal striatal large cholinergic interneurons ( LCI ) , and surviving LCI exhibit morphological , electrophysiological , and connectivity abnormalities . Consistent with the importance of this LCI pathology , murine dystonic-like movements are reduced significantly with an antimuscarinic agent used clinically , and we identify cholinergic abnormalities in postmortem striatal tissue from DYT1 dystonia patients . These findings demonstrate that dorsal LCI have a unique requirement for torsinA function during striatal maturation , and link abnormalities of these cells to dystonic-like movements in an overtly symptomatic animal model .
Primary dystonia encompasses a group of sporadic and inherited disorders characterized by disabling involuntary twisting movements . This sole clinical feature of primary dystonia implies selective abnormalities of motor pathways . The absence of an overt neuropathological signature complicates identification of pathogenic brain structures and cell types , and underlies the widely accepted notion that primary dystonia results from abnormal functioning of a structurally intact central nervous system ( CNS ) . Several lines of evidence implicate the striatum as the major node of dysfunction in primary dystonia . Secondary dystonia–where symptoms result from CNS damage or exogenous pharmacological insult–is linked strongly to striatal ( especially putaminal ) damage ( Marsden et al . , 1985 ) . Therapeutic benefit from antimuscarinic drugs ( Burke et al . , 1986 ) also implicates striatal dysfunction , as striatal cholinergic interneurons play a poorly understood but important role in motor control . Striatal-associated behavioral ( Carbon et al . , 2011 ) and functional imaging abnormalities are present in primary dystonia ( reviewed in Pappas et al . ( 2014 ) ) , and altering basal ganglia output with deep brain stimulation therapy is an effective dystonia treatment ( Vidailhet et al . , 2013 ) . Despite this evidence , the key striatal cell type ( s ) that drive dystonic movements are unknown . Studies aimed at defining mechanistic features in primary dystonia primarily use rodent models of DYT1 dystonia , a neurodevelopmental disorder manifesting during childhood , and the most common inherited primary dystonia . DYT1 dystonia is caused by a dominantly inherited mutation of the TOR1A gene that impairs function of the encoded protein torsinA . TorsinA is an endoplasmic reticulum/nuclear envelope-localized AAA+ ATPase ( Ozelius et al . , 1997 ) implicated in protein quality control and nuclear membrane-localized functions ( reviewed in Dauer ( 2014 ) ) . Heterozygous Tor1aΔE/+ mice ( mimicking the human DYT1 genotype ) do not exhibit any overt abnormalities , while constitutive Tor1a knockout and homozygous ΔE knock-in mice both exhibit perinatal lethality ( Goodchild et al . , 2005; Tanabe et al . , 2012 ) . Transgenic mice overexpressing wild type or mutant torsinA do not exhibit overt motor abnormalities ( Sharma et al . , 2005 ) , but are used to explore striatal electrophysiological abnormalities linked to overexpression of mutant torsinA . These studies demonstrate that in DYT1 mutant transgenics , striatal large cholinergic interneurons ( LCI ) exhibit a paradoxical response to dopamine D2 receptor agonists that may be involved in abnormalities of corticostriatal plasticity ( Pisani et al . , 2006; Martella et al . , 2009; Sciamanna et al . , 2011; Grundmann et al . , 2012; Sciamanna et al . , 2012a , 2012b ) . The relationship of these abnormalities to dystonic movements is unclear , as they occur in rodent models both with and without abnormal movements . Conditional deletion of torsinA in single brain regions ( e . g . , cortex , striatum ) or cell types ( e . g . , cerebellar Purkinje cells , cholinergic neurons ) implicated in the disease causes subtle changes in motor function , but no overt abnormal movements ( Yokoi et al . , 2008 , 2011; Zhang et al . , 2011; Sciamanna et al . , 2012a ) . Overt twisting movements are only observed in DYT1 model mice where torsinA function is impaired in precursor cells giving rise to multiple neuronal cell types ( Liang et al . , 2014 ) . These results implicate the importance of developmental timing of torsinA loss of function and the potential involvement of multiple dysfunctional cell types in disease pathophysiology . These models exhibit focal neurodegeneration in a discrete set of sensorimotor structures , and together with human subject neuroimaging studies ( reviewed in Ramdhani and Simonyan ( 2013 ) ) , raise questions regarding the ‘normal structure , abnormal function’ hypothesis of primary dystonias . To further explore this structure-function question as well as the potentially important role for torsinA during the early development of corticostriatal circuitry , we developed a novel mouse model by deleting torsinA with Dlx5/6-Cre , which acts in progenitors of forebrain GABAergic and cholinergic neurons ( Monory et al . , 2006 ) . This model exhibits face , construct and predictive validity . These mice are initially normal , but exhibit overt motor deficits as juveniles , coincident with selective loss of striatal LCIs and related electrophysiological abnormalities . Similar to DYT1 patients , the abnormal twisting and clasping movements of these mice are reduced significantly with chronic antimuscarinic administration . Moreover , we identify cholinergic abnormalities in postmortem putamen from DYT1 subjects . These observations are the first to demonstrate the unique vulnerability of a specific striatal cell type to torsinA loss of function , and have important implications for the understanding of disease pathogenesis and the development of targeted therapeutics .
We conditionally deleted Tor1a from precursors of forebrain GABAergic and cholinergic neurons by crossing Dlx5/6-Cre and Tor1a ‘floxed’ mice ( Monory et al . , 2006; Liang et al . , 2014 ) . Using mT/mG and Rosa26 LacZ Cre-reporter lines ( Soriano , 1999; Muzumdar et al . , 2007 ) , we confirmed that Cre activity was restricted to forebrain structures ( striatum , cortex , globus pallidus , basal forebrain , reticular thalamic nucleus ) , and included both direct and indirect pathway-projecting striatal neurons ( Figure 1A ) . TorsinA immunohistochemistry confirmed the essentially complete deletion of torsinA protein from striatum , partial deletion from cortex ( reflecting loss from GABAergic interneurons ) , and sparing of the thalamus–with the exception of the inhibitory neurons of the reticular thalamic nucleus ( Figure 1B ) . Dlx5/6-Cre+;Tor1aflx/− mice ( herein Dlx5/6 conditional KO ‘Dlx-CKO’ ) are born in the expected Mendelian ratio and are indistinguishable initially from littermate controls , including normal postnatal weight gain ( Figure 1—figure supplement 1 ) . 10 . 7554/eLife . 08352 . 003Figure 1 . Conditional TorsinA deletion from forebrain GABAergic and cholinergic neurons causes dystonic-like movements in juvenile mice . ( A ) Dlx5/6-Cre expression is restricted to forebrain , as demonstrated by rosa26 LacZ and mT/mG reporter lines . ( B ) TorsinA immunohistochemistry demonstrates complete torsinA deletion in the striatum and partial deletion in the cortex . ( C ) Dlx-CKO mouse forebrain architecture appears normal ( Nissl ) and there is no evidence of gliosis ( GFAP , s100β , Iba-1 ) . ( D–E ) Gross striatal and cortical development appears normal . Cortical thickness: two-way ANOVA main effect of age F3 , 65 = 17 . 24; p < 0 . 0001 , genotype F1 , 65 = 0 . 35; p = 0 . 55 ) ; striatal volume: main effect of age F3 , 65 = 307 . 0; p < 0 . 0001; genotype F1 , 65 = 0 . 724; p = 0 . 39 . ( F ) The behavior of neonatal Dlx-CKO mice is normal . Negative geotaxis and forelimb suspension did not differ from littermate controls . Forelimb suspension: t-test t ( 92 ) = 0 . 753; p = 0 . 45 ) . ( G–H ) Dlx-CKO mice develop severe forelimb and hindlimb clasping at P15 ( Chi square test , Χ2 = 64 . 03; p < 0 . 0001 ) , and a subset exhibits severe trunk twisting . ( I ) Dlx-CKO mice develop an inability to hang from a wire grid at 1 month of age ( two-way ANOVA; main effect of genotype F1 , 269 = 16 . 63; p < 0 . 0001 , time F6 , 269 = 6 . 613; p < 0 . 0001; and interaction F6 , 269 = 2 . 285; p = 0 . 036 ) . Motor learning remains intact , as demonstrated by the accelerating rotarod test ( two-way ANOVA main effect of trial F9 , 324 = 38 . 27 p < 0 . 0001 , genotype: F1 , 36 = 3 . 591; p = 0 . 066 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 00310 . 7554/eLife . 08352 . 004Figure 1—figure supplement 1 . Postnatal weight gain is normal in Dlx-CKO mice . Dlx-CKO mice exhibit normal postnatal growth , minor differences in weight after weaning , and no differences after maturation ( two-way ANOVA main effect of age F5 , 230 = 1022 , p < 0 . 0001; genotype F1 , 230 = 7 . 903 , p = 0 . 005; interaction F5 , 230 = 9 . 76 , p < 0 . 0001; Sidak's multiple comparisons test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 00410 . 7554/eLife . 08352 . 005Figure 1—figure supplement 2 . Dlx-CKO mice exhibit normal motor function during gait . ( A–I ) results from digigait analysis . Genotype differences were present in stance width ( two-way ANOVA main effect of genotype F1 , 62 = 16 . 51 , p < 0 . 0001; limb F1 , 62 = 70 . 90 , p < 0 . 0001; Interaction F1 , 62 = 0 . 381 , p = 0 . 539 , Sidak's multiple comparisons test ) , but no other abnormalities were observed . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 00510 . 7554/eLife . 08352 . 006Figure 1—figure supplement 3 . Dlx-CKO mice are hyperactive . ( A ) Open field analysis of horizontal movements ( two-way ANOVA main effect of genotype F1 , 35 = 16 . 29 , p = 0 . 0003; time F11 , 385 = 12 . 72 , p < 0 . 0001; interaction F11 , 385 = 1 . 713 , p = 0 . 06 , Bonferroni's multiple comparisons test ) . ( B ) Open field analysis of vertical movements ( two-way ANOVA main effect of genotype F1 , 35 = 10 . 72 , p = 0 . 002; time F11 , 385 = 4 . 176 , p < 0 . 0001; interaction F11 , 385 = 1 . 07 , p = 0 . 37 , Bonferroni's multiple comparisons test ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 006 Nissl-stained brain sections of Dlx-CKO mice did not demonstrate gross or microscopic abnormalities of forebrain architecture , and immunostaining showed no evidence of reactive gliosis or neural injury ( Figure 1C ) . Cortex and striatal development was normal , as assessed by size measurements throughout development ( Figure 1D , E ) . Dlx-CKO pre-weaning motor function did not differ from littermate controls ( Figure 1F ) . These data indicate that initial development and postnatal maturation of forebrain motor circuitry occurs normally in the absence of torsinA in forebrain cholinergic and GABAergic neurons . Dystonia in humans is commonly exacerbated by action , and may occur exclusively in particular settings or during specific motor tasks ( e . g . , runner's dystonia , dystonic writer's cramp ) . We assessed motor function during gait and during tail suspension , when mice vigorously kick their limbs and attempt to attain an upright body posture . Gait analysis was largely unremarkable ( Figure 1—figure supplement 2 ) . Dlx-CKO mice exhibited normal behavior during tail suspension up to 14 days of age , but nearly 100% of the animals developed severe forelimb and hindlimb clasping behaviors beginning at 15–17 days of age ( Figure 1G ) . This abnormal behavior remained fixed for the duration of the animal's life , to at least 1 year of age ( 17 out of 18 Dlx-CKO mice clasped during tail suspension at 1 year ) . A subset of these mutants simultaneously developed severe abnormal twisting of the trunk ( ∼70% of mice; Figure 1H ) . Dlx-CKO mice also developed a defect in the ability to hang from a wire grid ( Figure 1I , upper panel ) that in some cases appeared related to abnormal hindpaw twisting . In contrast , motor learning and gross coordination appeared normal , as assessed by the ability to remain on an accelerating rotarod at 8 weeks of age ( Figure 1I , lower panel ) . Dlx-CKO mice are also significantly hyperactive in the open field ( Figure 1—figure supplement 3 ) . These observations demonstrate that torsinA loss of function in forebrain GABAergic and cholinergic neurons is sufficient to cause action-induced abnormal twisting movements . The onset of these abnormal movements during juvenile CNS maturation and their persistence into adulthood broadly resembles the natural history and symptomatology of DYT1 dystonia ( Dauer , 2014 ) . Chronic antimuscarinic administration is a common therapy for DYT1 dystonia ( Burke et al . , 1986 ) , prompting us to evaluate the ability of antimuscarinics to ameliorate the abnormal twisting movements in Dlx-CKO mice . Dlx-CKO mice were treated with once-daily injections of the antimuscarinic scopolamine ( 5 mg/kg , s . c . ) or saline for 10 days . The duration of forelimb clasping , hindlimb clasping , and trunk twisting ( Figure 2A ) was assessed on videos by blinded observers . Scopolamine-treated animals exhibited significantly less limb clasping and twisting than saline-treated animals throughout the treatment period ( Figure 2B; main effect of drug F1 , 141 = 36 . 17; p < 0 . 0001 , Sidak's multiple comparisons test ) . The symptomatic improvement appeared to depend on the continued presence of scopolamine , as behavioral benefit disappeared following a 3-week washout period . The antimuscarinic trihexyphenidyl ( THP ) is the most commonly used agent to treat DYT1 dystonia and is clinically validated ( Burke et al . , 1986; Jankovic , 2013 ) . Similar to scopolamine , THP ( 5 mg/kg , i . p . ) significantly reduced clasping , and this effect resolved following a 3 week washout ( Figure 2C; Main effect of drug F1 , 82 = 46 . 69; p < 0 . 0001 , Sidak's multiple comparisons test ) . These data indicate that antimuscarinics effectively reduce clasping and twisting behaviors , and support the predictive validity of Dlx-CKO mice for the study of DYT1 dystonia . 10 . 7554/eLife . 08352 . 007Figure 2 . Antimuscarinic drugs ameliorate clasping and twisting behaviors in Dlx-CKO mice . ( A ) Examples of forelimb clasping , hindlimb clasping , and trunk twisting that were evaluated during review of the videos by blinded raters . ( B ) Duration of clasping and twisting was significantly reduced by once-daily 5 mg/kg scopolamine administration ( tail suspension recorded 45 min following drug treatment; two-way ANOVA: main effect of drug F1 , 141 = 36 . 14; p < 0 . 0001 , Sidak's multiple comparisons test . n = 8 saline , n = 6 scopolamine . This study was also repeated in a second cohort ) . ( C ) Clasping and twisting duration was reduced by once-daily 5 mg/kg THP administration compared to saline-treated mice ( tail suspension recorded 45 min following drug treatment; two-way ANOVA main effect of drug F1 , 82 = 46 . 69 , p < 0 . 0001 , Sidak's multiple comparison test . n = 6 saline , n = 8 THP ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 007 To assess the neural substrate of motor dysfunction in Dlx-CKO mice , we examined major markers of striatal signaling . Western blot analyses of microdissected striatum demonstrated a significant reduction in choline acetyltransferase ( ChAT ) , but no significant alterations in glutamic acid decarboxylase ( GAD67 ) or tyrosine hydroxylase ( TH ) expression ( Figure 3A , B ) , suggesting a specific abnormality of cholinergic elements . Consistent with this possibility , the receptor tyrosine kinase TrkA , expressed specifically by striatal LCIs ( Sobreviela et al . , 1994 ) , was reduced by approximately 50% ( Figure 3A , C ) . In contrast , expression of the medium spiny projection neuron ( MSN ) marker DARPP-32 did not differ significantly between Dlx-CKO mice and littermate controls ( Figure 3A , D ) . 10 . 7554/eLife . 08352 . 008Figure 3 . Cholinergic-specific abnormalities in the striatum of Dlx-CKO mice . ( A ) Western blots of microdissected striatum from 10 week old control and Dlx-CKO mice for markers of cholinergic , GABAergic , and dopaminergic signaling . ( B–D ) Quantification of the western blots demonstrated a selective reduction of LCI markers choline acetyltransferase ( t-test: t ( 8 ) = 2 . 683; p = 0 . 013 ) and TrkA ( t ( 8 ) = 1 . 883; p = 0 . 048 ) . No differences were observed for markers of GABAergic or dopaminergic neurons ( GAD67; t ( 8 ) = 0 . 012; p = 0 . 99; TH; t ( 8 ) = 0 . 742; p = 0 . 47; DARPP-32; t ( 8 ) = 1 . 12; p = 0 . 29 ) . ( E ) Microdialysis and HPLC-MS analysis demonstrates a significant reduction of ACh in dorsal striatum of Dlx-CKO mice ( t-test: t ( 12 ) = 3 . 895; p = 0 . 002; data reported as dialysate concentration and represent the average of 3 fractions per animal following neostigmine perfusion; n = 6–8 probes/group from 4 mice/group ) . ( F ) Microdialysis followed by benzoyl chloride derivatization and analysis by LC-MS demonstrated no significant change in any dorsal striatal neurotransmitter examined ( basal values measured in absence of Acetylcholinesterase ( AChE ) inhibitors ) . Data represent the average of 5 basal collections per animal ( n = 7 probes/group from 4 mice/group and are normalized to control levels ( two-way ANOVA for genotype: F1 , 190 = 0 . 0206; p = 0 . 88 ) . ( G ) AChE histochemistry on fresh frozen brain sections demonstrates a significant reduction of striatal AChE in Dlx-CKO mice ( t-test; t ( 22 ) = 5 . 16; p < 0 . 0001 ) . Specificity of AChE reaction was confirmed using several methods ( Figure 3—figure supplement 1 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 00810 . 7554/eLife . 08352 . 009Figure 3—figure supplement 1 . AChE histochemistry is selective for AChE . Assay controls demonstrate that omitting substrate , substituting an alternate thiocholine substrate , or inhibiting AChE activity with neostigmine fully abolishes staining . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 009 To test if the alteration of cholinergic markers reflected abnormal cholinergic neurotransmission in vivo , we performed striatal microdialysis in awake , behaving mice ( Song et al . , 2012 ) . Levels of extracellular acetylcholine ( ACh ) were significantly reduced in Dlx-CKO mice ( ∼73% reduction from control levels , performed in the presence of the acetylcholinesterase ( AChE ) inhibitor neostigmine; 1070 ± 231 nM in control [n = 6] vs 290 ± 33 nM in KO [n = 8]; t12 = 3 . 89; p = 0 . 0021; Figure 3E ) . In contrast , the basal extracellular concentrations of 16 other neurotransmitters and metabolites did not differ significantly from controls in a separate microdialysis study ( n = 7 per group; Figure 3F ) . Consistent with the reduction of ACh , histochemical analysis demonstrated a significant reduction in striatal AChE activity , an effect that appeared most prominent in dorsolateral striatum ( Figure 3G , assay controls in Figure 3—figure supplement 1 ) . We previously reported a link between torsinA loss-of-function and developmental neurodegeneration ( Liang et al . , 2014 ) . To determine if cholinergic abnormalities in Dlx-CKO mice reflect loss of LCIs , we quantified the number of LCIs in control and Dlx-CKO mice at 10 weeks of age , after all abnormal behaviors are fully established . Unbiased stereological quantification of striatal ChAT-positive neurons demonstrated 40% reduction in LCI number in Dlx-CKO compared to control mice ( Figure 4A , B ) . To explore whether LCIs might be lost from a specific striatal sub-region reflecting a discrete circuit ( Alexander et al . , 1986 ) , we subdivided the dorsal striatum into four quadrants and examined cell density throughout its rostro-caudal extent ( Figure 4C ) . Cell loss was non-uniform , showing a clear predilection for dorsolateral ( motor ) striatum , and a rostro-caudal gradient of cell loss , with relative sparing of caudal regions . Strikingly , cell loss was roughly 6 times greater in dorsolateral compared to ventromedial striatum ( 57% vs 9% reduction; Figure 4C , D; Figure 4—figure supplement 1 ) . Cell loss did not appear to be selective for the patch or matrix striatal subregions . 10 . 7554/eLife . 08352 . 010Figure 4 . Large cholinergic interneurons are selectively Lost from the striatum of Dlx-CKO mice . ( A , B ) Stereological quantification of the number of ChAT-positive neurons in the striata of Dlx-CKO and littermate control mice ( t-test: t ( 23 ) = 5 . 87; p < 0 . 0001 ) . ( C ) Characterization of the topology of ChAT-positive cell loss in dorsal striatum . Significant decreases in ChAT-positive cells were observed only in the dorsal quadrants . Two-way ANOVA main effects of genotype: F1 , 56 = 38 . 17; p < 0 . 0001 and interaction: F3 , 56 = 6 . 405; p = 0 . 0008 . ( D ) Pseudocolor representation of the degree of ChAT-positive cell loss in the dorsal striatum of Dlx-CKO mice . ( E ) Stereological quantification of the number of VAChT-positive and large ( >20 μm diameter soma ) Nissl-stained cells . VAChT t ( 13 ) = 3 . 305; p = 0 . 005 , Nissl t ( 13 ) = 5 . 293; p = 0 . 0001 . ( F ) Stereological quantification of the number of striatal small/medium ( <20 μm diameter soma ) nissl-positive cells ( nissl+ , t ( 13 ) = 0 . 606; p = 0 . 549 ) , medium spiny neurons ( DARPP-32+: t ( 22 ) = 1 . 14; p = 0 . 266 ) , and SST- and PV-expressing inhibitory interneuron classes ( PV+: t ( 23 ) = 2 . 806 , p = 0 . 01 SST+: t ( 23 ) = 0 . 6865; p = 0 . 499 ) . ( G ) Stereological quantification of the number of ChAT-positive neurons in basal forebrain nuclei ( BFC—Basal Forebrain Complex , MS—Medial Septum , VDB—Vertical Limb of the Diagonal Band ) of Dlx-CKO and littermate control mice ( t ( 7 ) = 0 . 392; p = 0 . 706 ) . ( H ) Stereological quantification of the number of cortical SST- and PV-expressing inhibitory interneuron classes ( PV+: t ( 15 ) = 1 . 32; p = 0 . 206; SST+: t ( 15 ) = 1 . 18; p = 0 . 256 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 01010 . 7554/eLife . 08352 . 011Figure 4—figure supplement 1 . LCI cell loss is most prominent in dorsolateral striatum . Percent cell density reductions in striatal quadrants as compared to control striata . Cell loss occurred in a dorsal to ventral gradient . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 01110 . 7554/eLife . 08352 . 012Figure 4—figure supplement 2 . TorsinA is deleted from basal forebrain cholinergic neurons in Dlx-CKO mice . TorsinA and ChAT costains demonstrate torsinA expression in basal forebrain cholinergic projection neurons from control but not in Dlx-CKO mice . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 012 To address the possibility that the striatal abnormality reflects ChAT downregulation , we used stereology to quantify an independent marker of LCIs , vesicular acetylcholine transporter ( VAChT ) . Dlx-CKO striata again appeared to contain 40–50% fewer VAChT+ LCIs ( Figure 4E ) . To fully exclude the potential confound of phenotypic marker down regulation , we took advantage of the fact that cholinergic neurons are the largest striatal neurons , being approximately twice as large as GABAergic striatal neurons ( Kreitzer , 2009 ) . Stereological quantification of the number of Nissl-stained neuron profiles ≧20 µm demonstrated a ∼40% reduction in Dlx-CKO compared to littermate controls ( Figure 4E ) . To examine if cell loss was specific to LCIs , we used unbiased stereology to quantify other GABAergic and cholinergic cell types from which torsinA is deleted in Dlx-CKO mice . LCIs are the only non-GABAergic neurons in the striatum , so we first quantified the number of small and medium sized Nissl-stained striatal cells ( i . e . , ≤20 µm ) , and found no significant difference in their numbers ( Figure 4F ) . We next quantified well-characterized subpopulations of GABAergic interneurons in cortex and striatum . We found no significant abnormalities in the number of cortical or striatal fast-spiking ( marked by parvalbumin; ‘PV’ ) or low-threshold spiking ( marked by somatostatin; ‘SST’ ) interneurons ( Figure 4F , H ) . Similarly , the number of striatal MSNs , ( marked by DARPP-32 ) did not differ between 10 week-old Dlx-CKO and littermate striata ( Figure 4F ) . In contrast to striatal LCIs , there is a normal number of basal forebrain cholinergic neurons ( Figure 4G ) , despite the fact that these neurons also express Cre recombinase ( Sanchez-Ortiz et al . , 2012 ) and lack torsinA ( Figure 4—figure supplement 2 ) . We next explored the relationship between cholinergic cell loss and motor dysfunction . At P7 , when motor function is normal ( Figure 1G , I ) , there were normal numbers of ChAT+ neurons ( Figure 5A ) , and normal levels of ChAT and TrkA ( Figure 5—figure supplement 1 ) . Progressive loss of ChAT+ neurons occurred from 1 to 2 months of age , a time period partially overlapping with the onset of motor dysfunction . Numbers of ChAT+ neurons were not reduced further at the 6-month time point ( Figure 5A ) , a time when motor abnormalities similarly plateaued . 10 . 7554/eLife . 08352 . 013Figure 5 . Dlx-CKO LCIs degenerate during juvenile striatal maturation , coincident with the onset of abnormal twisting . ( A ) Stereological quantification of the number of ChAT-positive neurons in the striata of Dlx-CKO and littermate control mice at time points between postnatal day 7 and 168 . Two-way ANOVA main effects of age: F3 , 66 = 2 . 899; p = 0 . 04 , genotype: F1 , 66 = 33 . 74; p < 0 . 0001 , and interaction: F3 , 66 = 7 . 232; p = 0 . 0003; * represents time points where significant differences exist using Sidak's multiple comparison test . ( B , C ) Quantification of the number of ChAT-positive striatal neurons co-expressing cleaved caspase-3 between P10 and P24 in control and Dlx-CKO brain sections ( two-way ANOVA main effects of age F2 , 21 = 43 . 68; p < 0 . 0001 , genotype: F1 , 21 = 122 . 1; p < 0 . 0001 , and interaction F2 , 21 = 32 . 91; p < 0 . 0001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 01310 . 7554/eLife . 08352 . 014Figure 5—figure supplement 1 . ChAT and TrkA expression is normal at P7 . Western blot analysis demonstrates no differences in ChAT or TrkA levels at postnatal day 7 , a time point when no behavioral or cellular deficits are present . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 01410 . 7554/eLife . 08352 . 015Figure 5—figure supplement 2 . There are no differences in the number of non-cholinergic apoptotic striatal cells . Although there are significantly more apoptotic cholinergic interneurons , there are no differences in the number of non-cholinergic apoptotic cells , as measured by expression of cleaved caspase-3 ( two-way-ANOVA main effect of age F2 , 21 = 18 . 93 , p < 0 . 0001; genotype F1 , 21 = 2 . 371 , p = 0 . 13; interaction F2 , 21 = 0 . 04 , p = 0 . 96 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 015 The temporal and spatial pattern of cell loss suggests that LCIs degenerate during postnatal striatal maturation . To confirm LCI cell death rather than altered cellular phenotype , we co-stained striatal sections for ChAT and cleaved caspase-3 ( CC3; Figure 5C ) . We quantified the number of CC3+ and co-localized CC3/ChAT+ cells at 3 time points during striatal development . While CC3+ and CC3/ChAT+ cell numbers did not differ between mutant and littermate control mice at postnatal day 10 , Dlx-CKO mice exhibited significantly more CC3/ChAT+ co-localized cells at postnatal days 12–14 and 24 ( Figure 5B ) , precisely the time that abnormal movements emerge . There were no differences in the overall number of non-ChAT+ CC3 striatal cells ( Figure 5—figure supplement 2 ) . To determine whether there are abnormalities in the remaining LCIs , which could contribute the behavioral phenotype of Dlx-CKO mice , we examined the morphological and electrophysiological properties of these cells . Surviving LCIs exhibited significant cell soma hypertrophy ( Figure 6A ) . The delayed time course of this phenotype following cell loss and rightward shift of the cell size frequency histogram ( Figure 6B ) support the likelihood that surviving neurons are becoming larger . In contrast , PV+ , SST+ , or DARPP-32+ neurons in striatum and cortex showed no changes in soma size ( Figure 6C , D ) and MSN dendritic structure was normal , as assessed by Golgi-Cox staining and Sholl analysis ( Figure 6E–H ) . 10 . 7554/eLife . 08352 . 016Figure 6 . Surviving striatal LCIs exhibit cell soma hypertrophy . ( A ) Quantification of ChAT-positive cell soma area in dorsal striatum between postnatal day 7 and 168 ( two-way ANOVA significant main effect of genotype F1 , 60 = 12 . 51; p = 0 . 0008 and time F3 , 60 = 117 . 8; p < 0 . 0001 , Tukey's multiple comparison test ) . ( B ) Frequency histogram of cell soma area data at postnatal day 168 . ( C , D ) Cell soma area of striatal and cortical GABAergic interneurons and striatal MSNs at postnatal day 168 . ( E ) Example of Golgi-Cox-stained MSN and dendritic tree reconstruction . ( F–H ) Analysis of dendritic complexity ( n = 69 control , 25 Dlx-CKO neurons ) . No differences observed in average highest dendritic branch order ( one-way ANOVA F3 , 90 = 1 . 079; p = 0 . 36 ) , dendritic length ( one-way ANOVA F3 , 92 = 1 . 023; p = 0 . 386 ) , or intersections on sholl analysis ( two-way ANOVA F1 , 92 = 0 . 019; p = 0 . 89 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 016 Electrophysiological analyses of surviving striatal LCIs performed after the full extent of cell loss support the possibility that abnormalities of these cells may contribute to the behavioral phenotype of Dlx-CKO mice . Spontaneous firing rates and coefficients of variation were similar for LCIs between Dlx-CKO and controls in cell-attached patch clamp mode ( Figure 7A , B ) . However , cell capacitance measurements were significantly larger in Dlx-CKO LCIs than control ( Figure 7C ) when recorded at a membrane potential of −70 mV with a K-gluconate based internal recording solution , or using a cesium-methanesulfonate internal solution ( data not shown ) . This finding is consistent with morphological evidence of larger cell somata ( Figure 6A ) . The inputs to these larger cells also appear to be abnormal . Dlx-CKO LCIs displayed significantly more spontaneous inhibitory postsynaptic currents ( sIPSCs ) than control LCIs ( Figure 7D , E ) ( p = 0 . 006 ) . Although the mean spontaneous excitatory postsynaptic current ( sEPSC ) frequency also was greater for Dlx-CKO LCIs ( Figure 7D , F ) the difference from the control LCI mean was not statistically significant ( p = 0 . 17 ) . The sIPSC/sEPSC ratio was significantly higher in Dlx-CKO cells ( Figure 7G ) indicating that these cells receive abnormal synaptic input . 10 . 7554/eLife . 08352 . 017Figure 7 . Surviving LCIs exhibit changes in excitability and abnormal synaptic inputs . ( A ) Sample cell-attached recordings from tonically active control and Dlx-CKO LCIs . ( B ) Mean frequencies of spontaneous firing ( cell-attached ) and coefficients of variation from control and Dlx-CKO LCIs . ( C ) Capacitance , input resistance , time constant , and resting membrane potential values from recordings with K-gluconate internal solution . ( D ) Sample recordings of sIPSCs and sEPSCs . ( E ) Mean sIPSC frequency from Dlx-CKO LCIs was significantly greater than that of control LCIs ( p = 0 . 006 ) . ( F ) Mean sEPSC frequencies from both genotypes were similar . ( G ) Ratio of sIPSC/sEPSC indicates that Dlx-CKO LCIs received significantly more inhibitory inputs than control LCIs ( p = 0 . 05 ) . ( H ) Examples of typical responses of control and Dlx-CKO LCIs to injected current pulses . Control LCIs generated more action potentials . ( I ) Mean numbers of action potentials are significantly reduced in Dlx-CKO LCIs at higher injected currents ( two-way ANOVA with posthoc Bonferroni test , p < 0 . 001 ) . ( J ) Sample traces of evoked EPSCs in control and Dlx-CKO LCIs . ( K ) Peak amplitudes of evoked EPSCs were significantly larger in Dlx-CKO LCIs ( two-way ANOVA with posthoc Bonferroni test , p < 0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 017 Several abnormalities were also identified in the response of LCIs to various stimuli . Depolarizing current pulses ( 1 s duration , 25 pA increments ) produced significantly fewer action potentials in LCIs from Dlx-CKO than control at 150–200 pA current intensities ( Figure 7H , I; two-way ANOVA , interaction between current intensity and frequency of action potentials , p < 0 . 001 ) . These findings indicate that Dlx-CKO LCIs may be less excitable than LCIs from control mice . However , evoked EPSCs from Dlx-CKO LCIs were significantly larger than those of control LCIs ( Figure 7J , K ) at higher stimulation intensities ( two-way ANOVA , interaction between current intensity and response amplitude , p < 0 . 01 ) . A subpopulation of Dlx-CKO LCIs displayed very large responses ( >150 pA at 0 . 06 mA , 3/12 ) that were not observed in control LCIs . LCIs exhibiting large amplitude responses also had significantly larger membrane capacitances than cells that did not ( 134 . 8 ± 14 . 7 vs 97 . 0 ± 8 . 3 pF , p = 0 . 047 ) , suggesting that increased numbers of synapses on the larger somata may account for the increased response amplitude . In support of this possibility , current density measurements of evoked responses ( evoked response amplitude divided by cell capacitance ) were not significantly different at all stimulation intensities ( data not shown ) . These multiple disturbances of surviving LCIs function raise the possibility that the beneficial effect of anticholinergic agents may in part arise from suppressing their aberrant signaling . Our results suggest that the selective loss of LCIs may be a pathogenic event in DYT1 dystonia . We further explored this possibility by analyzing postmortem putamen from DYT1 subjects and controls . As DYT1 tissue is in very limited supply and in general not of sufficient quality to perform valid quantification of cell numbers , we analyzed whole cell lysates of this tissue for cholinergic ( TrkA , VAChT and AChE ) , GABAergic ( GAD67 ) , and dopaminergic ( TH ) markers ( Figure 8A; Table 1 ) . TrkA levels were significantly reduced in DYT1 putamen ( Figure 8B; approximately 84% reduction ) . Similarly , the normalized mean expression levels of VAChT and AChE were reduced by 66% and 50% respectively ( Figure 8B , C ) , but these differences did not reach statistical significance , likely because of the large variability between control subjects for these markers . There was also considerable variability in the expression levels of the biosynthetic enzymes for ACh , GABA , and dopamine . Mean ChAT expression levels were slightly higher in DYT1 patients as compared to control subjects ( Figure 8A , D ) . GAD67 and TH levels were comparable between control and DYT1 dystonia subjects ( Figure 8E , F ) . Consistent with our experimental studies , these results suggest dysregulation of cholinergic function and TrkA signaling in the putamen of DYT1 dystonia patients . 10 . 7554/eLife . 08352 . 018Figure 8 . DYT1 dystonia postmortem putamen displays selective reductions in cholinergic markers . ( A ) Western Blot analysis of postmortem putamen samples from 3 dystonia patients and 3 age-matched control subjects . ( B ) Significant reductions in TrkA expression ( t-test; t ( 4 ) = 4 . 413; p = 0 . 014 ) . ( B–F ) No significant alterations in AChE ( t ( 4 ) = 0 . 940; p = 0 . 400 ) , VAChT ( t ( 4 ) = 0 . 208; p = 0 . 208 ) , ChAT ( t ( df = 4 ) = 1 . 766; p = 0 . 152 ) , TH ( t ( 4 ) = 0 . 7459; p = 0 . 497 ) , or GAD67 ( t ( 4 ) = 0580; p = 0 . 593 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 01810 . 7554/eLife . 08352 . 019Table 1 . Human subject dataDOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 019Case I . D . numberLane #StatusAge ( years ) Cause of deathOther neuro-pathologySexPMI ( hrs ) Time in storageRaceBBID1001Control87Unknown–Female910 year 2 monthCaucasianUMB16192DYT187 . 8Stroke–Female2312 year 4 monthCaucasianBBID3843Control89Respiratory failure–FemaleNot recorded11 year 9 monthCaucasianUMB48774DYT190 . 3Stroke–Female26 year 6 monthCaucasianBBID7325Control91UnknownLacunar infarctions , cerebellar microinfarctions , modest nigral cell loss , lewy bodiesMale810 year 10 monthCaucasianUMB52006DYT188 . 8‘complications of disorder’–Female95 year 5 monthCaucasian
Our studies establish the first model of DYT1 dystonia that exhibits face , construct , and predictive validity . This model incorporates several features believed essential for disease pathogenesis , including torsinA loss of function and targeting of forebrain motor circuitry including GABAergic and cholinergic neurons . Similar to the natural history and therapeutic response of the human disease , Dlx-CKO mice develop overt , dystonic-like twisting movements during juvenile CNS maturation that are reduced significantly with antimuscarinic drugs . The onset of abnormal movements coincides with a selective degeneration of striatal LCIs . Dlx-CKO mice are one of a small number of DYT1 mutant mouse models that exhibit overtly abnormal twisting movements ( Liang et al . , 2014 ) , all of which exhibit loss of discrete populations of neurons critical for motor function . Clasping and twisting during tail suspension is observed in models of many neurological diseases , so are unlikely to be equivalent to dystonia . Nevertheless , this abnormal behavior is an overt manifestation of abnormal motor function linked to cell loss in Dlx-CKO and other symptomatic dystonia models ( Liang et al . , 2014 ) . This is the first demonstration of selective vulnerability of a striatal cell type to torsinA loss of function , findings supported by studies in postmortem putamen from DYT1 dystonia subjects . Our observations add to an emerging literature implicating CNS maturation-related vulnerability of discrete cell types in the pathogenesis of DYT1 dystonia , a disease mechanism with potentially important implications for future approaches to therapy . The onset and progression of Dlx-CKO motor abnormalities as juveniles , followed by a fixed defect that persists for life ( Figure 1 ) , broadly resembles the progression of DYT1 and other childhood-onset primary dystonias . Abnormal behavior onset and striatal LCI loss in Dlx-CKO mice coincides with the period of striatal circuit maturation when connections are established and projection- and interneurons become physiologically active and mature ( Hattori and McGeer , 1973; Tepper et al . , 1998; Lee and Sawatari , 2011 ) . Subjects that carry the DYT1 mutation but do not manifest dystonia by their mid-20's almost invariably remain asymptomatic for life ( Bressman et al . , 2000 ) . These clinical and experimental observations support a critical role for torsinA during CNS maturation , and suggest that torsinA deficiency during a defined developmental window is a key component of DYT1 pathogenesis . Previous torsinA loss-of-function models also demonstrate cell loss restricted to a neurodevelopmental window ( Liang et al . , 2014 ) . These observations suggest that CNS maturation , synaptogenesis , and associated processes may exert a unique stress on neuronal circuits . TorsinA is a AAA+ protein chaperone ( Ozelius et al . , 1997 ) with putative functions in protein quality control ( Chen et al . , 2010; Nery et al . , 2011 ) regulation of nucleocytoskeletal connections ( Gerace , 2004; Worman and Gundersen , 2006 ) , trafficking of membrane-bound proteins ( Torres et al . , 2004 ) , protein secretion ( Hewett et al . , 2007 , 2008 ) , and nuclear export of large ribonucleoprotein granules during synapse development ( Jokhi et al . , 2013 ) . Periods of synaptogenesis and circuit maturation are therefore periods in which torsinA function may be particularly important . Cell loss and behavior onset during a neurodevelopmental window is consistent with a critical period for torsinA function in CNS development . LCIs play an early role in striatal development , which may in part account for the differences observed between Dlx-CKO mice and previous models . LCIs are the first striatal neurons to become post-mitotic ( Phelps et al . , 1989 ) , and cholinergic signaling is believed to have important but poorly understood influences on the development of later-born striatal neurons ( Chesselet et al . , 2007 ) . In contrast to our findings using the Dlx5/6-Cre transgene , which is active prior to these events ( in progenitor cells ) , postnatal conditional deletion of Tor1a in cholinergic neurons does not cause overt abnormal movements or changes in striatal LCI number , morphology , and neurochemistry ( Sciamanna et al . , 2012a ) . Consistent with an important role for timing of deletion , germline heterozygous ΔE Tor1a knock-in mice exhibit increased LCI cell body area , though these mutants show no difference in the number of LCIs ( Song et al . , 2013 ) . Mice may require more marked torsinA loss of function to fully model cholinergic vulnerability , as the heterozygous ΔE Tor1a knock-in mice retain considerable torsinA function , and do not exhibit overt abnormal movements ( Tanabe et al . , 2012 ) . It is possible that non-cholinergic neuron populations in Dlx-CKO mice appear morphologically normal but exhibit functional abnormalities . However , conditional deletion of torsinA in MSNs , which constitute 95% of all striatal neurons , does not cause overt abnormal movements ( Yokoi et al . , 2011 ) . Current dogma holds that neurodegeneration does not occur in primary dystonia , but postmortem samples from DYT1 subjects are limited and have not been assessed quantitatively for abnormalities in the number of striatal neuronal cell types ( Paudel et al . , 2012 ) . Neuroimaging studies increasingly point to changes potentially consistent with cell loss ( reviewed in Ramdhani and Simonyan ( 2013 ) ) . Our studies in postmortem striatum from DYT1 subjects , the first to examine cell type-specific markers in this context , are consistent with our experimental findings pointing to selective cholinergic defect in this structure . Previous work in overtly symptomatic DYT1 models demonstrates loss of discrete cell types in other motor-related structures ( e . g . , loss of neurons of the deep cerebellar nuclei , with preservation of all other cerebellar cell types ( Liang et al . , 2014 ) . Our analysis of postmortem tissue was restricted to the striatum; future work is required to determine whether there is cell loss in extra-striatal areas corresponding to those identified in murine models . Postmortem analyses of DYT1 subjects suggests the presence of inclusion bodies in pedunculopontine cholinergic neurons , but did not report cell loss ( McNaught et al . , 2004 ) . These observations , and additional work ( Liang et al . , 2014 ) and reviewed by Dauer ( 2014 ) , suggest that abnormal protein quality control may represent a pathogenic event in DYT1 dystonia . Several lines of evidence are consistent with a potentially causal connection between LCI loss or dysfunction and abnormal twisting movements . Unilateral immunotoxin-mediated striatal cholinergic ablation causes ‘asymmetric concaving postures’ in mice ( Kaneko et al . , 2000 ) and LCI ablation also reduces the threshold for the development of motor tics ( Xu et al . , 2015 ) . LCIs are lost in a perinatal hypoxia-ischemia model of cerebral palsy-associated dystonia , mimicking the striatal pathology seen in that disease ( ‘status marmoratus’ ) ( Burke and Karanas , 1990 ) , which is also treated with anticholinergic drugs . Furthermore , imaging studies of idiopathic cervical dystonia subjects suggest deficits in the density of striatal cholinergic axon terminals ( Albin et al . , 2003 ) . LCIs innervate fast-spiking interneurons ( Koos and Tepper , 2002 ) , a cell type concentrated in the lateral striatal areas ( Berke , 2011 ) where LCI loss is greatest in Dlx-CKO mice . Fast-spiking interneurons contribute to action execution and suppression of unwanted movements ( Gage et al . , 2010 ) , and have been implicated in dystonic-like postures ( Gittis et al . , 2011 ) . Previous reports of DYT1 rodent models support a role between torsinA loss of function and aberrant physiological properties of LCIs ( Eskow Jaunarajs et al . , 2015 ) , but these models do not exhibit abnormal twisting movements or cell loss ( Sharma et al . , 2005; Pisani et al . , 2006; Martella et al . , 2009; Quartarone and Pisani , 2011; Sciamanna et al . , 2011 , 2012b ) . Abnormal clasping is present in a transgenic rat model of dystonia , but it is not known if striatal LCI number is reduced in these animals ( Grundmann et al . , 2012 ) . Loss of cholinergic neurons may be required for the remaining cells to become interconnected into an abnormal circuit . Recordings in striatal slices performed after maximal cell loss in Dlx-CKO mice demonstrate that surviving LCIs develop a complex set of intrinsic membrane and synaptic alterations . Dlx-CKO LCIs are less excitable when depolarized by intracellular current injections , yet display significantly larger evoked EPSCs , potentially reflecting increased excitatory synapses on the surviving ( larger ) neurons . Dlx-CKO LCIs also receive significantly enhanced inhibitory synaptic that could lead to aberrant firing when cells are challenged , as we have demonstrated in a related movement disorder , Huntington's chorea ( Holley et al . , 2015 ) . These multiple disturbances of surviving LCIs function raise the possibility that the beneficial effect of anticholinergic agents may in part arise from suppressing aberrant LCI signaling . Our findings add to the emerging literature demonstrating a unique requirement for torsinA function for cell viability during CNS maturation ( Liang et al . , 2014 ) , a concept that may alter approaches to the therapeutic targeting of DYT1 dystonia . These experiments also raise the possibility that unique classes of striatal LCIs may exist that vary in their molecular and functional features . Future studies aimed at identifying factors that modulate the vulnerability of these cells to torsinA loss-of-function may help to define such classes , advancing our understanding of striatal organization and function .
Tor1a floxed mice were generated as previously described ( Liang et al . , 2014 ) . Dlx5/6-Cre mice were obtained from Jackson laboratories ( Tg ( dlx6a-Cre ) 1Mekk/J; stock number 008199 ) and then maintained in our mouse colony at the University of Michigan . Conditional Tor1a null animals were generated with the following breeding scheme: Cre+ Tor1a+/− X Tor1aflx/flx , with four possible offspring genotypes: Tor1aflx/+ ( WT ) , Tor1aflx/− ( Flx control ) , Cre+ Tor1aflx/+ ( Cre control ) , and Cre+ Tor1aflx/− ( Dlx-CKO ) . Mice were genotyped for Tor1a and the Cre transgene using the primers and PCR programs previously described ( Liang et al . , 2014 ) . For electrophysiology experiments , a Chat ( BAC ) -eGFP allele ( strain B6 . Cg-Tg ( RP23-268L19-EGFP ) 2Mik/J; Stock Number 007902 ) was bred into the Dlx-CKO cross to allow for visualization of LCIs . Mice were housed 2–4 per cage using microisolation technique , maintained in a temperature- and light-controlled room , and provided with food and water ad libitum . Mice of all genotypes were housed together to prevent environmental bias . The University of Michigan Committee on the Use and Care of Animals ( UCUCA ) approved all experiments involving animals . Age and sex-matched littermate mice were used for all experiments . Mice were sacrificed with cervical dislocation , brains were rapidly removed , frozen over dry ice , and stored at −80°C . Fresh frozen brains were cut into 500 μm sections , were refrozen onto uncoated glass slides , and striatum samples were taken using a modified 16 gauge needle . Micropunch samples were placed into microcentrifuge tubes containing 100 μl lysis buffer ( Tris Buffered Saline containing 1% Sodium Dodecyl Sulfate , 0 . 1 mM phenylmethanesulfonyl fluoride , 1 mM Dithiothreitol , and Halt Protease Inhibitor Cocktail [Life Technologies product 87786] ) and were homogenized using a plastic plunger . Homogenates were centrifuged at 12 , 000 rpm for 5 min , pellets were discarded , and the supernatants were removed to a new tube . Bradford protein assay was performed and final lysates were prepared at 1 μg/μl , including sample-loading buffer ( 0 . 05% bromophenol blue , 0 . 1 M dithiothreitol , 10% glycerol , 2% SDS , and 5% β-mercaptoethanol ) , and were boiled for 5 min . 10 μg ( 10 μl ) protein samples and Dual Precision Plus protein standards were run on 4–15% Biorad Mini Protean TGX precast polyacrylamide gels , underwent wet transfer to 0 . 22 μm PVDF membranes in transfer buffer containing 10% methanol ( run for 2 hr at 400 mA at 4°C ) , and were processed for enhanced chemiluminescence as described below . Membranes were washed in tris buffered saline ( TBS ) containing 1% Tween-20 ( TBS-T ) , blocked for 30 min in 5% non-fat dry milk in TBS-T , and incubated in primary antibody overnight at 4°C ( see Table 2 for primary antibody details ) . Membranes were then washed in 5% milk/TBS-T , incubated for 1 hr in horseradish peroxidase-conjugated secondary antibody ( Table 2 ) , and rinsed in TBS-T . Bands were visualized using Supersignal West Pico , Dura , or Femto enhanced chemiluminescence substrates , underwent multiple exposures to Amersham hyperfilm ECL , and were developed and fixed with an x-ray film developer . Films were scanned using an Epson scanner , and band intensity was quantified in ImageJ . Serial dilutions of protein were examined for each antibody to determine optimal antibody dilutions before running experimental samples , and multiple exposures were examined to confirm that bands were in the linear range and not overexposed . 10 . 7554/eLife . 08352 . 020Table 2 . Antibodies used for immunohistochemistry and western blotsDOI: http://dx . doi . org/10 . 7554/eLife . 08352 . 020LevelAntigenHostConjugatedDilutionSourceIHCPrimaryTorsinARabbit–1:100Abcam ab34540PrimaryGFAPRabbit–1:2000Dako Z0334Primarys100βRabbit–1:2000Abcam ab41548PrimaryIba-1Rabbit–1:500Wako 019-19741PrimaryChATGoat–1:100Millipore AB144PPrimaryVAChTGoat–1:2000Millipore ABN100PrimaryDARPP-32Rabbit–1:300Cell Signaling #2302PrimaryPVMouse–1:500Swant #235PrimarySSTRabbit–1:500Abcam ab103790PrimaryCC3Rabbit–1:500Cell Signaling #9664Secondaryanti-mouseDonkeyAx4881:800Life Technologies A-31572Secondaryanti-mouseDonkeyAx5551:800Life Technologies A-21202Secondaryanti-mouseDonkeybiotin1:800Jackson Immunoresearch 115-065-003Secondaryanti-rabbitDonkeyAx4881:800Life Technologies A-21206Secondaryanti-rabbitDonkeyAx5551:800Life Technologies A-31572Secondaryanti-rabbitDonkeybiotin1:800Jackson Immunoresearch 711-065-152Secondaryanti-goatDonkeybiotin1:800Jackson Immunoresearch 705-065-003Western blotPrimaryChATRabbit–1:1000Abcam ab137349PrimaryGAD67Mouse–1:1000Millipore MAB5406PrimaryTHRabbit–1:2000Millipore AB152PrimaryActinMouse–1:6000Sigma A5316PrimaryTrkARabbit–1:4000Advanced Targeting Systems ABN03PrimaryDARPP-32Rabbit–1:2000Cell Signaling #2302PrimaryTorsinARabbit–1:10 , 000Abcam ab34540PrimaryCalnexinRabbit–1:20 , 000Enzo Life Sciences SPA-860PrimaryAChERabbit–1:200Santa Cruz sc-11409PrimaryVAChTGoat–1:1000Millipore ABN100PrimaryTrkA ( for human ) Rabbit–1:1000Cell Signaling #2505SecondaryAnti-goatRabbitHRP1:7500Pierce 31402Secondaryanti-mouseGoatHRP1:5000Jackson Immunoresearch 115-035-003Secondaryanti-rabbitGoatHRP1:10 , 000Jackson Immunoresearch 111-035-003 Custom made microdialysis probes with 1 mm polyacrylonitrile membrane length were implanted bilaterally into the dorsal striatum 24 hr before experiments with the following coordinates: Anteroposterior +1 . 1 mm , mediolateral +2 . 05 mm , dorsoventral −3 . 8 mm . On the day of experiments , artificial cerebrospinal fluid ( aCSF ) ( composition in mM: CaCl2 1 . 2; KCl 2 . 7 , NaCl 148 and MgCl2 0 . 85 ) was perfused through the microdialysis probe at 2 μl/min for 1 hr and then 1 μl/min for another hour for equilibration . For studies that only measured ACh concentrations ( Figure 3E ) , three 5-min fractions were collected per animal following neostigmine addition ( 50 μM ) to the aCSF perfusate . Dialysate samples were collected and analyzed following the addition of d4-ACh ( 20 nM ) as internal standard . For comprehensive neurochemical analysis of basal differences between genotypes ( Figure 3F ) , five 3-min fractions were collected per animal , and a benzoyl chloride derivatization scheme was employed ( Song et al . , 2012 ) . Briefly , 2 . 5 μl borate buffer ( 100 mM ) , 2 . 5 μl 2% benzoyl chloride in acetonitrile , and 2 . 5 μl internal standard solution was added to each dialysate sample prior to analysis . Following sample collection , a Thermo Scientific Accela HPLC ( Waltham , MA ) system automatically injected 5 µl of the sample onto a Waters ( Milford , MA ) HSS T3 reverse phase HPLC column ( 1 mm × 100 mm , 1 . 8 µm ) at 200 μl/min . For ACh analysis , a 2 min isocratic elution was employed ( 25/75 mobile phase A/B ) . Mobile phase A consisted of 10 mM ammonium formate and 0 . 15% formic acid . Mobile phase B was acetonitrile . Analytes were detected by a Thermo Scientific TSQ Quantum Ultra triple quadrupole mass spectrometer operating in multiple reaction monitoring mode . ACh was detected using the m/z transition 146‡87 while d4-ACh was detected using 151‡90 . For comprehensive neurochemical analysis , samples were analyzed as previously described but with a 6 min HPLC gradient ( Song et al . , 2012 ) . Mice were deeply anesthetized with a lethal dose of ketamine/xylazine and received transcardial perfusion of 0 . 01 M phosphate buffered saline ( PBS ) followed by 4% paraformaldehyde in 0 . 1 M phosphate buffer ( PB ) . Brains were postfixed in 4% paraformaldehyde for 2 hr and cryoprotected overnight in 20% sucrose in PB . Consecutive serial 40 μm brain sections through the forebrain were generated on a cryostat and stored in PBS . Free-floating brain sections were processed for fluorescence immunohistochemistry by washing in PBS containing 0 . 1% Triton-X-100 ( PBS-Tx ) , blocking in 5% normal donkey serum ( NDS ) , and incubating in primary antibody overnight at 4°C ( see Table 2 for primary antibody details ) . Sections were washed in PBS-Tx followed by 1 hr in secondary antibodies conjugated to Alexafluor 488 or Alexafluor 555 ( Table 2 ) . Brain sections were mounted onto gelatin-coated slides , coverslipped with prolong gold antifade mounting medium , and imaged under epifluorescence microscopy . Free-floating brain sections were processed for DAB staining by washing in PBS-Tx , blocking with 0 . 3% H2O2 followed by 5% NDS , and incubating overnight in primary antibody overnight at 4°C ( see Table 2 ) . Sections were washed with PBS-Tx , incubated in biotinylated secondary antibody for 1 hour ( Table 2 ) , followed by 2 hr in avidin-biotin-peroxidase complex ( Vectastain Elite ABC Kit Standard; PK6100 , Vector Laboratories , Burlingame , CA ) . Sections were exposed to 3 , 3′ diaminobenzidine using Sigmafast DAB tablets ( Sigma D4418 ) and were flooded with PBS to halt staining . Sections were mounted onto gelatin-coated slides , dried overnight , dehydrated in ethanol and xylenes , and were coverslipped with permount mounting medium . Staining for each experiment was performed in parallel by an investigator blinded to experimental group . Omitting the primary or secondary antibody from incubation prevented all staining . To process for Nissl staining , 40 μm brain sections were mounted onto gelatin-coated slides , were dried for 24 hr , incubated in successive decreasing concentrations of ethanol , followed by 3 min in 0 . 005% cresyl violet solution containing acetic acid , were dehydrated in ethanol followed by xylenes , and were coverslipped using permount mounting medium . Mice were anesthetized with ketamine/xylazine and sacrificed by cervical dislocation . Brains were removed , frozen over dry ice , and stored at −80°C . 25 μm fresh frozen brain sections were generated on a cryostat , were adhered to gelatin-coated slides , dried at room temperature , and stored at −80°C . Sections were stained as previously described ( Geneser , 1987 ) . Sections were dehydrated with ethanol followed by xylenes , and were coverslipped with permount mounting medium . Brain sections were imaged using brightfield microscopy , and optical density of striatal staining was determined using ImageJ software . Values from the anterior commissure white matter tract were used for a background subtraction value . AChE specificity was confirmed by omitting substrate , substituting butyrylcholinesterase substrates , or including the AChE inhibitor neostigmine in the incubation medium ( Figure 3—figure supplement 1 ) . Mice were anesthetized with ketamine/xylazine and sacrificed with cervical dislocation . Brains were removed and immediately processed using the FD Rapid GolgiStain Kit ( FD Neurotechnologies , Columbia , MD ) . Processed brains were frozen with dry ice-chilled isopentane , placed on dry ice , and 100 μm brain sections were generated on a cryostat . Brain sections were mounted onto gelatin-coated slides , stained according to the FD Rapid GolgiStain Kit , and coverslipped with permount mounting medium . Slides were observed under brightfield microscopy using a Zeiss Axiophot 2 microscope , first using a 5× objective lens . Striatal medium spiny neurons containing full golgi-cox impregnation without breaks along the dendrites , and no obstructions by neighboring cells were then used for analysis . Neurons were observed using a 63× objective lens , were traced , and reconstructed using Neurolucida software ( MBF Bioscience , Williston , VT ) , and dendritic complexity was determined with Sholl analysis . 94 neurons from 25 animals were used for this study . Frozen postmortem putamen samples from three DYT1 patients were provided from University College London . Frozen putamen samples from three control subjects stored at the Michigan Brain Bank were obtained from Dr Roger Albin . Subjects were chosen to control for age , sex , and postmortem interval ( see Table 1 ) . Small putamen samples were taken with a razor blade , avoiding white matter tracts , and homogenates were prepared as described above for mouse striatum tissue . 10 µg protein lysates were run on 4–20% Biorad gels Mini Protean TGX precast polyacrylamide gels , were transferred to PVDF membranes , and were stained as described above ( see Table 2 for antibody details ) . Bands were visualized using Supersignal West Pico , Dura , or Femto enhanced chemiluminescence substrates , underwent several exposures to Amersham hyperfilm ECL , and were developed and fixed with an x-ray film developer . Data are reported as mean ± SEM . Student's t-tests and Chi square tests were performed using Graphpad Prism software ( version 6 ) . One-way or two-way ANOVAs were performed using SPSS software ( version 22 ) , and post hoc Sidak's or Bonferroni's multiple comparisons tests were performed when significant main effects were observed ( p < 0 . 05 ) . All experiments were repeated at least once before effects were considered significant . | Dystonia is disorder of the nervous system that causes people to suffer from abnormal and involuntary twisting movements . These movements are triggered , in part , by irregularities in a part of the brain called the striatum . The most common view among researchers is that dystonia is caused by abnormal activity in an otherwise structurally normal nervous system . But , recent findings indicate that the degeneration of small populations of nerve cells in the brain may be important . The striatum is made up of several different types of nerve cells , but it is poorly understood which of these are affected in dystonia . One type of dystonia , which most often occurs in children , is caused by a defect in a protein called torsinA . Pappas et al . have now discovered that deleting the gene for torsinA from particular populations of nerve cells in the brains of mice ( including a population in the striatum ) causes abnormal twisting movements . Like people with dystonia , these mice developed the abnormal movements as juveniles , and the movements were suppressed with ‘anti-cholinergic’ medications . Pappas et al . then analyzed brain tissue from these mice and revealed that the twisting movements began at the same time that a single type of cell in the striatum—called ‘cholinergic interneurons’—degenerated . Postmortem studies of brain tissue from dystonia patients also revealed abnormalities of these neurons . Together these findings challenge the notion that dystonia occurs in a structurally normal nervous system and reveal that cholinergic interneurons in the striatum specifically require torsinA to survive . Following on from this work , the next challenges are to identify what causes the selective loss of cholinergic interneurons , and to investigate how this cell loss affects the activity within the striatum . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Forebrain deletion of the dystonia protein torsinA causes dystonic-like movements and loss of striatal cholinergic neurons |
Organ function depends on tissues adopting the correct architecture . However , insights into organ architecture are currently hampered by an absence of standardized quantitative 3D analysis . We aimed to develop a robust technology to visualize , digitalize , and segment the architecture of two tubular systems in 3D: double resin casting micro computed tomography ( DUCT ) . As proof of principle , we applied DUCT to a mouse model for Alagille syndrome ( Jag1Ndr/Ndr mice ) , characterized by intrahepatic bile duct paucity , that can spontaneously generate a biliary system in adulthood . DUCT identified increased central biliary branching and peripheral bile duct tortuosity as two compensatory processes occurring in distinct regions of Jag1Ndr/Ndr liver , leading to full reconstitution of wild-type biliary volume and phenotypic recovery . DUCT is thus a powerful new technology for 3D analysis , which can reveal novel phenotypes and provide a standardized method of defining liver architecture in mouse models .
The correct three-dimensional ( 3D ) architecture of lumenized structures in our bodies is essential for function and health . The cardiovascular system , lungs , kidneys , liver , and other organs depend on precisely patterned tubular networks . Several diseases are caused by , or result in , alterations in the 3D architecture of lumenized structures . Vascular architecture defects contribute to Alzheimer’s disease ( Klohs et al . , 2014 ) , opportunistic infections cause narrowing of bile ducts in liver ( De Angelis et al . , 2009 ) , and branching morphogenesis defects in the renal urinary system cause hypertension ( Short and Smyth , 2016 ) . In some pathologies , several lumenized structures are affected at once . Visualizing multiple tubular systems in tandem in 3D , in animal disease models , is necessary to allow investigation of how these systems interact in vivo in development , homeostasis , and disease . 2D histological sections remain the standard practice for all types of tissues . Recent advances in tissue clearing ( Chung et al . , 2013; Susaki et al . , 2014; Renier et al . , 2016 ) carbon ink injections ( Kaneko et al . , 2015 ) and high-end microscopy begin to address the need for , and benefits of , whole organ analysis . Importantly , organ systems often interact with one another and almost always are connected to blood supply . In order to study tissue spatial organization in development , disease , and regeneration , a 3D analysis of multiple networks is indispensable . A lack of markers , suitable antibodies , tissue autofluorescence and/or organ size often preclude the possibility for whole organ analysis . Radiopaque resin casting is an alternative approach that enables micro computed tomography ( µCT ) scanning , digitalization with full rotation and the possibility for both qualitative and quantitative analyses compatible with multiple imaging softwares . In this study , we focused on establishing a simple , robust , antibody-free and inexpensive method for whole organ visualization of lumenized structures . We built on previous work to image a single network using resin ( Masyuk et al . , 2003; Kline et al . , 2011; Walter et al . , 2012 ) and imaging with µCT . First , in order to visualize multiple structures , we tested different radiopaque substances to enhance the resin contrast , resulting in mixing of two MICROFIL resins with distinctive radiopacity . As proof of principle , we focused on studying the hepatic vascular and biliary systems of wild type and Jagged1 Nodder ( Jag1Ndr/Ndr ) ( Andersson et al . , 2018 ) mice , which are challenging to image by other methods due to liver size and autofluorescence ( Renier et al . , 2016 ) . We devised double resin-casting micro computed tomography ( DUCT ) to inject , image and digitalize two systems in tandem in 3D to gain a deeper insight into the organ recovery process . The subsequent analysis is performed using a custom-written MATLAB pipeline ( available and deposited in https://github . com/JakubSalplachta/DUCT; Hankeova , 2021 copy archived at swh:1:rev:6b0b0eb88bbaf9bfc4f8ee42cafa4c122866fbba ) as well as with ImageJ . Alagille syndrome ( ALGS ) is a congenital disorder affecting multiple organs , including the hepatic and cardiovascular systems ( Spinner et al . , 1993; Mašek and Andersson , 2017 ) . The disease is usually caused by mutations in the Notch ligand JAGGED1 ( JAG1 , OMIM: ALGS1 [Oda et al . , 1997; Li et al . , 1997; Gilbert et al . , 2019] ) or , less frequently , in the receptor NOTCH2 ( OMIM: ALGS2 [Spinner et al . , 1993; McDaniell et al . , 2006] ) , and is chiefly characterized by intrahepatic peripheral bile duct paucity ( Alagille et al . , 1975; Riely et al . , 1979 ) . Importantly , bile duct development is regulated by portal vein mesenchyme ( Hofmann et al . , 2010 ) , implying that the architectural relationship between liver cells and the vasculature affects opportunities for signaling cross-talk between these systems in liver . Some patients with ALGS spontaneously recover a biliary system ( Riely et al . , 1979; Fujisawa et al . , 1994 ) . Ductular reaction , aberrant biliary growth and/or trans-differentiation from hepatocytes can contribute to biliary recovery in ALGS and other cholestatic disorders ( Schaub et al . , 2018; Fabris et al . , 2007 ) . Furthermore , it has been reported that liver vascular architecture is affected in ALGS ( previously also known as arteriohepatic dysplasia or syndromic paucity of bile ducts [Hadchouel et al . , 1978] ) . It is thus clear that understanding and defining both biliary and vascular intrahepatic defects is essential for ALGS . DUCT is a versatile , reliable tool allowing standardized architecture analysis and definition of multiple lumenized trees on a whole organ level , facilitating systems insights . We demonstrated the applicability of DUCT by revealing the distinct morphological features that allow the de novo generated Jag1Ndr/Ndr adult biliary system to achieve wild-type biliary volume: ( 1 ) an increase in the number of central low generation branches and ( 2 ) profound tortuosity in the liver periphery . We confirmed these 3D findings in 2D sections from Jag1Ndr/Ndr mice and patients with ALGS , demonstrating that the new phenotypes identified with DUCT in the mouse model are representative of patient pathology . Using DUCT we also discovered novel phenotypes such as bile duct bridging between two portal veins , which would be misinterpreted as bile duct proliferation in 2D histological sections . Hence , 2D histological sections are not sufficient to understand the structural abnormalities of tubular networks .
In order to define and quantify the adaptive process resulting in a de novo generated biliary system in adult Jag1Ndr/Ndr mice , we investigated the spatial relationship of portal venous and biliary systems in normal and diseased liver in 3D . First , we compared double carbon ink injection ( Kaneko et al . , 2015 ) and whole mount immunofluorescence staining combined with tissue clearing using iDISCO+ ( Renier et al . , 2016 ) to assess the 3D architecture of the liver ( Figure 1—figure supplement 1A and B ) . Neither carbon ink injection nor iDISCO+ ( due to poor labeling of vascular network ) were suitable for dual 3D analysis of vascular and biliary networks . We therefore developed an alternative approach for 3D analysis: DUCT ( Figure 1A , Figure 1—figure supplement 1C ) followed by semi-automated segmentation generating 3D binary masks of these two systems . The binary masks were used directly for DUCT data volume analysis . For further quantification of the architectural parameters the binary masks were skeletonized and analyzed ( Figure 1B ) using a custom-written MATLAB pipeline , ImageJ , or qualitative visual assessment . DUCT outperformed ink injection and immunofluorescence in most aspects ( Figure 1—figure supplement 1D ) , from 3D analysis ( not possible with ink ) to analysis of lumenization ( not possible with immunofluorescence ) . One limitation , however , is that DUCT cannot visualize lumens smaller than 5 µm . Finally , to test whether DUCT can be applied to other organ systems , we visualized lung architecture by injecting the airways ( via the trachea ) and the vasculature ( via the pulmonary artery ) and 3D reconstructed the respiratory and vascular systems ( Figure 1—figure supplement 1E ) . In lungs , as in liver , the two lumenized systems can be clearly distinguished using DUCT . In all the DUCT liver experiments in this manuscript , we applied DUCT to the right medial lobe , and used the other lobes for sample-matched quality control ( Figure 1—figure supplement 2 ) . Our previous work revealed that Jag1Ndr/Ndr bi-potential hepatoblasts did not differentiate into cholangiocytes during embryonic development . Newborn Jag1Ndr/Ndr mice were jaundiced and displayed intrahepatic bile duct paucity . However , by adulthood Jag1Ndr/Ndr livers exhibited mature bile ducts , but with abnormal apical polarity ( Andersson et al . , 2018 ) . Using DUCT , we first investigated to what degree Jag1Ndr/Ndr mice can grow a biliary tree by postnatal day 15 ( P15 ) . We examined the biliary system in 7 Jag1Ndr/Ndr P15 pups and discovered a rudimentary or absent biliary tree in Jag1Ndr/Ndr pups , while there was a fully developed biliary tree in Jag1+/+ mice ( Figure 1C ) . The 3D reconstruction of both systems can be explored in separate channels or in tandem with rotation ( interactive PDFs , Figure 1—figure supplements 3 and 4 ) . We noted high variability in biliary outgrowth between the right medial lobe ( RML ) and the left lateral lobe ( LLL ) in Jag1Ndr/Ndr livers ( illustration of liver lobes Figure 1—figure supplement 5A ) . In the RML of five Jag1Ndr/Ndr pups , the biliary tree covered >5% of liver area , whereas in LLL only 1 Jag1Ndr/Ndr pup displayed >5% biliary tree coverage , while four Jag1Ndr/Ndr pups had no lumenized bile ducts . In Jag1+/+ pups the biliary network covered on average 75% of the liver area in both RML and LLL ( Figure 1—figure supplement 5B and C ) . The P15 Jag1Ndr/Ndr pups were cholestatic , and manifested increased levels of alkaline phosphatase ( ALP ) ( Figure 1D ) , aspartate aminotransferase ( AST ) , alanine aminotransferase ( ALT ) and decreased levels of albumin ( Figure 1—figure supplement 5D ) . Interestingly , two different groups were noted in Jag1Ndr/Ndr pups with regard to the total bilirubin levels , that is , 50% of the animals had highly increased and 50% had mildly increased total bilirubin ( p=0 . 0079 ) , while all Jag1+/+ mice displayed bilirubin levels below detection limits ( Figure 1—figure supplement 5D ) . We correlated the total bilirubin amount with the liver area covered by lumenized bile ducts and detected a strong negative correlation between these two factors ( Figure 1E , r = −0 . 8214 ) . We further sectioned the resin injected P15 liver and stained for the early biliary marker SOX9 ( SRY-Box transcription factor 9 ) ( note remaining resin in some portal veins ) . In Jag1+/+ central liver , lumenized bile ducts were clearly detected , whereas in all four Jag1Ndr/Ndr central livers no lumenized bile ducts were visible and the number of SOX9 positive cells varied between animals , poorly reflecting the total bilirubin levels ( Figure 1—figure supplement 5E ) . Next , we aimed to characterize and quantify the biliary architecture of the Jag1Ndr/Ndr de novo generated biliary system in adult mice ( Figure 1F ) . The 3D reconstruction of both systems can be explored in separate channels or in tandem ( interactive PDFs , Figure 1—figure supplement 6–11 ) . While Jag1+/+ mice demonstrated a stereotyped vascular and biliary architecture ( Figure 1—figure supplement 12A , B and C ) , Jag1Ndr/Ndr livers exhibited greater architectural variability ( Figure 1—figure supplement 12D , E and F ) . To quantify the degree of de novo biliary formation , we extracted the volume ( using the binary masks ) and diameters of the portal venous and biliary systems ( using the binary masks and skeletons ) . The total volume of the vascular and biliary trees was similar in Jag1Ndr/Ndr and Jag1+/+ mice ( Figure 1G , Figure 1—figure supplement 12H ) . There was a tendency toward a larger portal venous volume and smaller biliary volume in Jag1Ndr/Ndr mice , resulting in a trend toward a reduced BD:PV ( bile duct , portal vein ) volume ratio in Jag1Ndr/Ndr mice ( Figure 1—figure supplement 12H , p=0 . 0594 ) . Next , we investigated portal vein and bile duct diameters along the main branch . There was a tendency to an increase in the Jag1Ndr/Ndr portal vein diameter and a decrease in biliary diameter as a function of distance compared with Jag1+/+ mice ( Figure 1—figure supplement 12I ) . However , there was high variability in venous and bile duct diameter in the Jag1Ndr/Ndr hilar region . In Jag1+/+ liver , the BD:PV diameter ratio was consistently 1:3 in hilar and intermediate regions ( Figure 1H ) . This BD:PV diameter ratio was not preserved in the Jag1Ndr/Ndr livers ( Figure 1H ) . DUCT enabled 3D visualization of postnatal and adult biliary and vascular trees . The biliary network in Jag1Ndr/Ndr mice appeared postnatally , but with high heterogeneity between liver lobes and animals . Moreover , the liver area with lumenized bile ducts correlated with total bilirubin levels and disease severity at P15 . The adult segmented µCT data were analyzed for volume and inner diameter of two injected systems . Comparisons of the portal vein and bile duct diameters demonstrated a conserved portal vein – bile duct architectural relationship with a stereotype BD:PV 1:3 diameter ratio in Jag1+/+ liver . The adult Jag1Ndr/Ndr mice displayed a heterogeneous phenotype that nevertheless resulted in full restoration of biliary function ( Andersson et al . , 2018 ) via recovery of a wild-type biliary volume . The intrahepatic biliary tree forms by a tubulogenic process in which a heterogeneous , hierarchical fine mesh of connected cholangiocytes is refined to single larger conduits of bile ducts , resembling a branching tree ( Ober and Lemaigre , 2018; Tanimizu et al . , 2016 ) . In Jag1+/+ liver , the qualitative analysis of network connectivity revealed that the biliary system formed a continuous tree , branching outwards toward the periphery as expected ( Figures 1C , F and 2A left panel ) . In contrast , the Jag1Ndr/Ndr biliary system displayed some branches oriented from peripheral to hilar , with abrupt endings ( Figure 2A right panels , blue arrowheads , on average one abruptly ending BD per lobe ) . We confirmed the abruptly ending bile ducts in Jag1Ndr/Ndr livers in serial liver sections ( Figure 2B , bottom panels ) . The black arrowhead labels a well-formed bile duct , that ended bluntly in the following section ( +5 μm , blue arrowhead ) , and disappeared completely in the next section ( +10 μm ) . To determine whether the Jag1Ndr/Ndr biliary abnormalities were representative of pathology in patients with Alagille syndrome , we evaluated liver serial sections from whole liver explants ( patients with severe Alagille syndrome ( S-ALGS ) that underwent transplantation ) and biopsies obtained for clinical reasons in non-transplanted patients ( patients with mild Alagille syndrome ( M-ALGS ) ) . The liver function tests for individual patients are reported in Table 1 and representative liver sections are presented in Figure 2—figure supplement 1 . One patient ( M_ALGS-5 ) had been biopsied at multiple time points revealing paucity of bile duct at 2 . 5 months , a regenerating phase with hepatocytes expressing CK7 at 1 . 3 years and bile duct recovery at 4 . 7 years ( Figure 2—figure supplement 1D ) . We evaluated liver sections from patients with severe ( top panel ) and mild ( bottom panel ) Alagille syndrome for the presence of abruptly terminating bile ducts . A black arrowhead labels a well-formed bile duct that terminated in the subsequent section ( +5 μm , blue arrowhead ) , ( Figure 2C ) . In conclusion , DUCT facilitated qualitative assessments of the tubular networks , including connectivity and perfusion . Both the Jag1Ndr/Ndr and Alagille syndrome biliary systems displayed abruptly ending bile ducts , which may affect bile flow and shear stress . During embryonic development , cholangiocytes differentiate from hepatoblasts that are in contact with portal vein mesenchyme expressing Jag1 ( Ober and Lemaigre , 2018 ) . Whether postnatally de novo generated bile ducts arise adjacent to the portal vein , or whether they are less dependent on portal vein proximity has not yet been explored . We therefore analyzed the distance between the biliary and portal venous systems by calculating the surface distances using the MATLAB pipeline . Specifically , the surface distance was defined as the shortest length from biliary skeleton to the portal venous skeleton , minus radiuses of these systems at the defined points ( Figure 3A ) . Jag1+/+ bile ducts maintained a uniform distance to adjacent portal veins throughout the liver ( Figure 3B top panel , asterisk , 3C , 3D 100% of BDs within 0 . 5 mm of a PV ) . In contrast , Jag1Ndr/Ndr bile ducts did not maintain a uniform distance to the nearest portal vein ( Figure 3B bottom panel , double arrow , 3C , 3D 1 . 5% of BD are placed 0 . 5–1 . 26 mm away from a PV ) and sometimes traversed the parenchyma to join another portal vein branch ( Figure 3B bottom right panel , Figure 1—figure supplement 2 empty arrowheads ) . Both the increased BD-PV distance and parenchymal bile ducts were validated in histological sections , in which Jag1+/+ bile ducts were in close proximity to , or embedded in portal vein mesenchyme ( Figure 3E left panel , asterisk ) . In contrast , Jag1Ndr/Ndr bile ducts were confirmed to be present outside of the portal vein mesenchyme area ( Figure 3E middle panel , double arrow ) or even in the liver parenchyma close to the edge of the liver ( Figure 3E right panel ) . This phenotype , visualized in 2D sections , could resemble biliary proliferation , or ductular reaction , rather than a bridging structure , highlighting the importance of 3D imaging . Parenchymal bile ducts were also detected in liver samples from patients with severe and mild Alagille syndrome ( Figure 3F ) but not in control human liver . In sum , DUCT pipeline together with the MATLAB algorithm , measured the gap between surfaces of two resin injected systems to address the spatial relationship between them . Our data showing biliary cells in the parenchyma and bile ducts far from portal veins in Jag1Ndr/Ndr liver and liver from patients with Alagille syndrome thus suggest that postnatal bile duct formation does not rely on close proximity to portal vein mesenchyme and may occur independent of signals from portal vein mesenchyme . Portal vascular and biliary systems are ductal tree-like structures with numerous branches , which function to maximize the area of exchange between the tissue and its lumen . We evaluated portal venous and biliary branching using the DUCT pipeline and the MATLAB script to quantify the total number of vascular or biliary branch points . Branch points were identified using the 3D skeletons of each system , and categorized based on the number of incoming/outgoing branches ( classifying as bifurcations , trifurcations , or nodes with more than three branches ) ( Figure 1—figure supplement 12D and G ) . We did not identify any differences in the absolute numbers of branch points in Jag1+/+ and Jag1Ndr/Ndr systems , again suggesting that de novo biliary growth generally reconstituted a full-volume , well-branched biliary tree ( Figure 4—figure supplement 1 ) . During embryonic development , the biliary system is established alongside the portal venous system . This is reflected in the final architecture of the system , with bile ducts in close proximity to portal veins ( Figure 3 ) . A prediction based on this embryonic process and BD/PV dependency is therefore that bile ducts should invariably branch where portal veins branch . We extracted the coordinates for branch points in the biliary and portal venous systems from the corresponding 3D skeletonized data and calculated 3D Euclidean distances between biliary branch points and their nearest neighboring portal vein branch point ( Figure 4A ) . Indeed , Jag1+/+ bile ducts ( Figure 4B left panel , blue arrowhead ) branched adjacent to portal vein branch points ( magenta arrowhead , defined as within 0 . 5 mm ) . In contrast , Jag1Ndr/Ndr bile ducts ( Figure 4B middle panel , blue arrowhead ) branched further from portal vein branch points ( magenta arrowhead ) , or independent of portal vein branch points ( blue arrow in Figure 4B , and data in 4C; independence defined as distance >0 . 54 mm ) . On average , 1 . 3% of Jag1+/+ bile ducts branch points and 8 . 6% of Jag1Ndr/Ndr bile duct branch points were further than 0 . 5 mm from the nearest portal vein branch point ( Figure 4C ) . We analyzed consecutive histological liver sections to confirm branching morphology defects discovered using DUCT . Jag1+/+ bile ducts ( Figure 4D top panel , blue arrowhead ) indeed branched at the same point as portal veins branch ( magenta arrowhead ) . In contrast , in Jag1Ndr/Ndr liver bile ducts might bifurcate in the absence of portal vein branching ( Figure 4D bottom panel , blue arrow ) . We next asked whether similar branching phenotypes were present in healthy human liver or in patients with mild or severe Alagille syndrome . In normal human liver , biliary branching occurred close to portal vein bifurcation ( Figure 4E top panel , blue arrowheads , PV branching within 25 μm in this example , branching not shown ) . In patients with severe Alagille syndrome , biliary branching could be seen independent of portal vein branching ( Figure 4E middle panel , blue arrows ) , the nearest portal vein branch point for this bile duct was 13 sections hilar ( circa 65 μm earlier ) . We also detected independent bile duct branching in patients with mild Alagille syndrome ( Figure 4E bottom panel , blue arrows ) . In conclusion , DUCT revealed dual system 3D architectural phenotypes: ( 1 ) similar numbers of branch points in Jag1+/+ and Jag1Ndr/Ndr livers but ( 2 ) a greater distance between portal vein and bile duct branch points in Jag1Ndr/Ndr livers . Bile ducts in patients with Alagille syndrome displayed similar branching abnormalities , corroborating the architectural independence from portal vein patterning . Whether de novo bile duct formation occurs evenly throughout the liver or is more extensive in certain regions has not yet been quantitatively defined . One of the mechanisms by which the biliary system can regenerate is via abundant branching of the network ( Vartak et al . , 2016; Masyuk et al . , 2001 ) . We therefore employed the DUCT pipeline and ImageJ to perform a Strahler analysis ( 3D branching analysis based on generation number from the origin , see Material and methods section ‘Branching analysis’ for details on generation number calculation ) to address the branch length , number and distribution in specific liver areas . In order to define anatomy correctly for differently sized livers , the liver lobe data were separated into three equivalent regions using portal vein branch generation number and lengths as a proxy for hilum/intermediate/periphery . These regions were denoted region 1 ( R1 enriched for hilar region ) , 2 ( R2 , intermediate ) , and 3 ( R3 , peripheral-enriched ) ( Figure 5A ) . The average branch length of the portal vein was shorter in Jag1Ndr/Ndr livers , resulting in significantly smaller portal vein region sizes ( Figure 5B middle panel ) , reflecting the overall smaller size of the mice and livers . Using the regions defined by the portal venous system , biliary region size was similar in the 1 st and 2nd region of wild type and Jag1Ndr/Ndr mice , while region three was smaller in both animal groups due to resin not penetrating ducts < 5 μm ( Figure 5B right panel , Figure 5—figure supplement 1 ) . Two out of three Jag1Ndr/Ndr mice had a marked reduction in biliary region size in R3 , while one had an increase , reflecting the somewhat variable phenotype manifested by both the patients and the mouse model . Branching trees in biological systems have a stereotype structure in which branch lengths shorten with each branching generation from start ( R1 ) to end ( periphery , R3 ) ( Masyuk et al . , 2001; Hannezo et al . , 2017 ) . We analyzed the portal vein and biliary branch lengths within each region and found that portal vein segments shortened as expected with each generation in both Jag1+/+ and Jag1Ndr/Ndr livers ( Figure 5C middle panel ) . Jag1+/+ bile ducts followed the same stereotype branching principle , but Jag1Ndr/Ndr bile ducts branch lengths were significantly shorter in regions 1 and 2 , and uniform across the hierarchy of branches ( Figure 5C right panel ) . We further analyzed the distribution of number of branches in the three regions and discovered that , on average , 9% of Jag1+/+ portal vein branches fell into R1 , 66% into R2% and 25% into R3 , with a similar distribution in Jag1Ndr/Ndr livers ( Figure 5D middle panel ) . Based on biological tree structure , we would expect fewest biliary branches in R1 , an intermediate number in R2 , and most branches in R3 . However , resin penetration and lumen diameter precluded filling of all terminal portal vein branches in R3 . Mirroring the wild type portal venous branch distribution , Jag1+/+ biliary branch distribution was , on average , 16% in R1 , 66% in R2% and 18% in R3 . Jag1Ndr/Ndr biliary branch distribution was shifted , with 44% in R1 , 44% in R2% and 12% in R3 ( Figure 5D right panel ) . The distribution of biliary branches number was highly heterogeneous among the different Jag1Ndr/Ndr mice . DUCT pipeline is compatible with multiple image analysis software programs . Here , we used ImageJ to address the branching length over branching generations , and our data collectively indicated that low-generation number biliary segments were shorter and more numerous than expected , suggesting that ectopic regenerative branching and/or incorporation of de novo generated biliary cells , occurs in region 1 . However , peripheral branching , which may be undetectable if biliary diameters are under 5 μm , cannot be excluded . The biliary system can adapt to excessive amounts of bile by increasing its diameter or length ( Vartak et al . , 2016; Slott et al . , 1990 ) . One means of enlarging length is by duct convolution . We therefore investigated the length and tortuosity of the biliary and portal vascular trees using the DUCT pipeline in combination with the MATLAB algorithm ( whole system and main branch quantification ) and ImageJ ( R1 , R2 and R3 analysis ) . 3D reconstruction of portal vein vasculature and the biliary network revealed straight Jag1+/+ bile ducts ( Figure 6A top panel ) , whereas Jag1Ndr/Ndr bile ducts were tortuous ( Figure 6A bottom panel ) , especially in the liver periphery . We confirmed in histological liver sections that the Jag1Ndr/Ndr BDs were tortuous ( Figure 6B ) . We further assessed biliary tortuosity in patients with mild Alagille syndrome and found several tortuous bile ducts ( Figure 6C ) ; however , we did not detect any tortuous bile ducts in patients with severe Alagille syndrome ( data not shown ) . In order to quantify tortuosity , we calculated the actual ( curved ) length and theoretical ( chord ) length ( scheme Figure 6D ) . The curved and chord lengths of the entire system , and the main branch alone did not differ for portal venous or biliary systems in Jag1+/+ and Jag1Ndr/Ndr mice ( Figure 6—figure supplement 1A–G ) . The BD:PV ratio was not significantly different for curved ( Figure 6—figure supplement 1H ) or chord length ( Figure 6—figure supplement 1I ) . However , there was a 6% increase in overall portal venous tortuosity in Jag1Ndr/Ndr mice when the entire system was taken into account ( Figure 6E ) and biliary tortuosity was increased by 50% . Biliary tortuosity was greatest in the Jag1Ndr/Ndr liver periphery , with a 140% increase in R3 ( Figure 6E ) . The tortuosity of the main portal vein or main bile ducts branch analyzed alone was not significantly different ( Figure 6—figure supplement 1J and K ) , highlighting the importance of analyzing the entire tree to obtain comprehensive and accurate results . In summary , DUCT allowed analysis of curved and chord length measurements for the entire or defined regions of the injected trees . Jag1Ndr/Ndr bile ducts recovered wild-type lengths postnatally , with a pronounced increase in peripheral tortuosity .
Precisely defining the three-dimensional ( 3D ) architecture of healthy and diseased organs is a fundamental aspect of biology , and improved imaging methods would allow stricter characterization of animal models for human diseases . Until now , 3D liver analysis has been restricted by a lack of adequate tools and high auto-fluorescence of the tissue . Carbon ink injection is robust , but imaging is 2D , precluding 3D analysis of the architecture . In contrast , immunostaining and clearing allows 3D analysis , but success is variable and highly dependent on tissue fixation , antibody quality , penetrance and tissue autofluorescence . Here , we further advanced organ resin casting , which was previously used to analyze one system at a time ( Masyuk et al . , 2003; Kline et al . , 2011; Walter et al . , 2012 ) . We developed a simple , robust and inexpensive method ( DUCT ) for simultaneous visualization and digitalization of two lumenized systems in mouse to analyze organ architecture . DUCT is completely independent of antibody staining , endogenous fluorescent proteins and is not sensitive to tissue fixation . Unlike whole mount immunohistochemistry techniques , DUCT provides information about the lumen , internal diameter , perfusion and connectivity of the injected tree . The most important limitation of DUCT is that it cannot visualize structures with a diameter under 5 μm , due to resin viscosity . We showed that resin casting , segmentation and 3D representation can be used as input for further investigation by visual qualitative assessment , and for in depth analysis by imaging softwares such as ImageJ or custom written MATLAB scripts . The pipeline for imaging and segmentation followed by detailed customized quantification of cellular and architectural mechanisms of two tubular networks could serve as a standard for whole organ analysis in animal models , and can be further adapted for a specific applications . DUCT is based on radiopaque resin injection into multiple lumenized systems . First , identifying resins with sufficient contrast and low viscosity is crucial for scanning using computed tomography . Combining multiple resins with distinct contrasts can upscale the analysis to several networks simultaneously . In our study , it was imperative to use two fresh MICROFIL resins to obtain sufficiently distinct contrast to separate the two injected systems . Prolonged storage ( ~>3 months ) of the MICROFIL resin leads to resin precipitation and a significant decrease in the resin radiopacity ( for details see materials and methods ) . Future efforts to identify or develop differentially radiopaque substances would further accelerate the analysis pipeline . Second , in order to achieve a successful injection , an appropriate resin injection site for each system must be identified , and a suitable amount of pressure must be applied while avoiding bubble formation . Third , resin segmentation of the scanned sample is straightforward in well-injected , well-contrasted samples , but in samples with bubbles or poor contrast segmentation requires time-consuming manual correction and careful tracing of the resin throughout the whole organ to ensure a coherent network . Minor resin leakage can be digitally excluded during the image segmentation , but artefacts such as air bubbles in resin , non-homogenous resin contrast ( in our set up caused by mixing of blue and yellow MICROFIL ) and resin leakage due to lumen rupture ( probably caused by high pressure during injection ) slow down the analyses substantially ( Figure 7B ) . Thus , a well-chosen and well-injected resin are a prerequisite for efficient downstream analyses . Regarding instrumentation , the DUCT pipeline is not restricted to specific CT systems or acquisitions parameters , therefore samples can be imaged on any CT device with sufficient spatial resolution to study selected samples and their morphology . For further qualitative and quantitative analysis of the DUCT µCT data , robust computational power is necessary – and dedicated workstations with sufficient RAM memory ( >64 MB RAM ) are recommended , dependent on the volume of acquired data . Our study describes a complex spatial adaptation of the biliary tree to postnatal BD paucity in a mouse model for Alagille syndrome ( ALGS ) , with validation in samples from patients with ALGS . Based on our data and reported case studies , we propose a model in which Jag1 mutant intrahepatic bile ducts ( IHBDs ) did not form during embryonic development ( Andersson et al . , 2018; Alessandro et al . , 2007 ) but in some animals ( and some patients , Figure 2—figure supplement 1D ) , bile ducts grew after birth ( Dahms et al . , 1982 ) . The lumenized bile ducts formed from hilar to peripheral regions with different timings in the different liver lobes and in individual animals . The fully remodeled biliary system was tortuous , exhibited abrupt ends , and was hyper-branched in region 1 ( summarized in Figure 8 ) . Our findings in the mice , using DUCT , may help to explain the poor prognostic value of biopsies in patients with ALGS . Our results suggest that the lack of predictive value between peripheral bile duct paucity , observed in diagnostic biopsies , and phenotype severity in patients with ALGS ( Mouzaki et al . , 2016 ) may reflect differences in bile duct growth and presence/absence in hilar versus peripheral regions ( Figure 1C , Figure 1—figure supplement 5B , C , E ) . Mouzaki et al , showed that bile duct density was not predictive of outcome , instead , bilirubin levels , fibrosis , and cholestasis were correlated with disease presentation . In line with this , in the Jag1Ndr/Ndr pups the total bilirubin levels correlated well with 3D postnatal bile duct growth , which was apparent from 3D resin-injected whole lobe analysis , but did not correlate with bile duct paucity in 2D sections of central liver . A 40% regrowth of lumenized bile ducts in Jag1Ndr/Ndr pup was sufficient to reduce cholestatic burden to almost wild type levels , while not leading to a normal density of bile ducts in the periphery - emphasizing the importance of whole-liver architecture analyses . Identifying and quantifying architectural defects such as branch length differences , stochastic branching , tortuosity , differences in portal-biliary distance and blunt end bile ducts is very challenging in sections ( compare histology and 3D imaging in Figures 2–6 ) . We demonstrated here that DUCT is a powerful method for visualization and semi-automated quantitative analysis of two lumenized biological systems in vivo . DUCT could be applied to other tubular networks including blood vessels and bronchi in lung ( Figure 1—figure supplement 1E ) or blood vessels and urinary ducts in kidney ( Wagner et al . , 2011; Wei et al . , 2006 ) . DUCT has multiple advantages over ink injections and iDISCO+ , as 3D imaging with µCT avoids the drawbacks of tissue autofluorescence or poor antibody penetration . By injecting two resins into a single animal , it is possible to study the relation between these biological systems , which has not been previously reported . While experts in the field are careful to discriminate hilar and peripheral regions of the liver , carefully tracing organ structures for hundreds of micrometers in tissue sections is not standard practice and is a demanding endeavor . DUCT would be a suitable readout for testing drug compounds in mouse models for liver cholestatic diseases . With DUCT , it is now possible to map and quantify architecture of two networks in mouse models , setting the stage for an in depth understanding of how systems interact in health , disease , and regenerative processes .
All animal experiments were performed in accordance with Stockholm’s Norra Djurförsöksetiska nämnd ( Stockholm animal research ethics board , ethics approval numbers: N150/14 , N61/16 , N5253/19 , N2987/20 ) regulations . Animals were maintained with standard day/night cycles , provided with food and water ad libitum , and were housed in cages with enrichment . For postnatal day 15 ( P15 ) experiments , 10 wild type ( Jag1+/+ ) ( eight males and two females ) and 10 Jagged1 Nodder ( Jag1Ndr/Ndr ) littermate pups ( five males and five females ) were used for serum analysis . Within this group , nine Jag1+/+ and seven Jag1Ndr/Ndr mice were injected with resin . All 16 animals were analyzed in 3D , revealing extensive heterogeneity that would necessitate performing DUCT on a large number of animals to obtain significant quantitative data , while the bile duct paucity was obvious . 1 Jag1+/+ and 1 Jag1Ndr/Ndr pair was therefore scanned and rendered in 3D . From this group , four Jag1+/+ and four Jag1Ndr/Ndr left medial lobes were used for 2D liver sections and staining . Adult animals were between 4 . 5 and 6 . 5 months old . In total , 18 Jag1+/+ and 6 Jag1Ndr/Ndr animals were injected with resin for µCT . Quality control of injections ( Figure 1—figure supplement 2 ) was performed on all livers during method development until surgery and injection technique resulted in well-injected livers . Three Jag1+/+ and three Jag1Ndr/Ndr animals were used for the DUCT quantifications in adulthood . For liver histology , two Jag1+/+ and three Jag1Ndr/Ndr mice were used . For ink injections , nine Jag1+/+ mice were used ( four males and five females ) and for iDISCO+ four Jag1+/+ and four Jag1Ndr/Ndr mice were used ( six males and two females ) . For lung 3D resin casting five Jag1+/+ mice were used ( two males and three females ) . Samples were not blinded for investigation since the phenotype is overt and the genotype is therefore obvious to the experimenter . The animals were maintained on a mixed C57bl6J/C3HeN background . Jag1Ndr/+ ( Nodder ) mice were bred and genotyped as previously described ( Andersson et al . , 2018 ) . Collection of liver samples and clinical data from patients or donors was approved by the Swedish Ethical Review Authority ( 2017/269-31 , 2017/1394–31 ) . Samples from patients with severe Alagille syndrome ( four ) were obtained at time of liver transplant from extirpated liver . Samples were obtained with a consent to be used for research according to ethical permit 2017/269–31 . Samples were dissociated for primary cell culture e . g . organoids ( data not shown ) , and a matching sample was formalin fixed or fresh-frozen for comparative analyses . Liver tissue samples from patients with mild Alagille syndrome ( six ) were obtained for clinical follow-up purposes and were retrospectively analyzed . The liver material was obtained within the framework of clinical patient care can be analyzed retrospectively without the need for consent according to ethical permit 2017/1394–31 . Healthy controls ( two ) were left-over donor material , or from organ donation post-mortem . The liver function tests were obtained during routine biochemical analyses . MICROFIL ( Flow Tech Inc ) was prepared as follows . Yellow MICROFIL ( Y ) cat . #MV-122 was diluted with clear MICROFIL ( C ) cat . # MV-Diluent in 3:1 ( Y:C ) . Blue MICROFIL ( B ) cat . #MV-120 was diluted 1:1 ( B:C ) with clear MICROFIL . Diluted yellow MICROFIL was mixed with diluted blue MICROFIL 1:1 creating a green MICROFIL . Yellow MICROFIL was injected into common bile duct ( CBD ) or pulmonary artery ( PA ) . Green MICROFIL was injected into portal vein ( PV ) or trachea ( TA ) . 1 ml of diluted MICROFIL is mixed with 50 μl of hardener ( supplied by Flow Tech Inc ) prior injection . Postnatal day 15 ( P15 ) mice were sacrificed by decapitation and perfused through the heart with 3 ml of Hanks' Balanced Salt solution ( HBSS ) ( Life Technologies cat . # 14025092 ) . Adult mice were sacrificed by CO2 inhalation and perfused through the heart with HBSS for 3 min ( perfusion rate 5 ml / 1 min ) . For liver resin injections , the mice were perfused trough the left ventricle , for lung resin injections the mice were perfused though the right ventricle . Mice were sacrificed by CO2 inhalation and transcardially perfused with HBSS for 3 min ( perfusion rate 5 ml/1 min ) . Injection into CBD and PV . CBD and PV were accessed in the same way as described for MICROFIL injections . When injecting with ink , there is no need to tighten the CBD with silk suture as the ink is not leaking out . Black ink ( Higgins cat . #44032 ) was injected into the CBD until BDs on the surface were filled or resistance was met . White ink ( Higgins cat . #44032 ) was injected using PE50 tubing into the PV until blood vessels on the surface were filled or resistance was met . Liver was dissected out and separated into lobes . All lobes were cleared in BABB as described above . Liver ink images of right medial lobe were taken under stereomicroscope Stemi 305 ( Carl Zeiss Microscopy ) using PowerShot S3 IS camera ( Canon ) . Mice were anesthetized by isoflurane inhalation ( ~2% ) and transcardially perfused with HBSS for 3 min ( perfusion rate 5 ml/1 min ) and 10% neutral buffered formalin ( NBF ) for 5 min . Liver was dissected out and further immersion fixed with 10% NBF ON at 4°C . The next day liver was washed and kept in DPBS and separated into lobes . Right medial lobe was stained and cleared following the iDISCO+ protocol and imaged by light sheet microscope by Gubra ( Denmark ) . Fixed and washed samples were dehydrated in methanol/H2O gradient: 20% , 40% , 60% , 80% and 2 × 100% methanol , each step 1 hr at room temperature ( RT ) . The samples were bleached in cooled fresh 5% H2O2 in methanol ON at 4°C . The samples were subsequently rehydrated in methanol/PBS series: 80% , 60% , 40% , 20% , with 0 . 2% Triton X-100 , 1 hr each at RT . They were washed in PBS with 0 . 2% Triton X-100 ( PTx . 2 ) for 2 × 1 hr at RT . Samples were incubated in permeabilization solution at 37°C for 3 days . Blocking is carried out in blocking solution at 37°C for 2 days . The samples were incubated with primary antibody in PTwH/5%DMSO/3% donkey serum at 37°C for 7 days . They were washed in PTwH for 1 × 10 min , 1 × 20 min , 1 × 30 min , 1 × 1 hr , 1 × 2 hr and 1 × 2 days . Samples were incubated with secondary antibody in PTwH/3% donkey serum at 37°C for 7 days , followed by washes in PTwH: 1 × 10 min , 1 × 20 min , 1 × 30 min , 1 × 1 hr , 1 × 2 hr and 1 × 3 days . All steps were performed in tightly closed tubes to minimize evaporation and oxidation . 5 μm FFPE-liver ( mouse and human ) sections were deparafinized and rehydrated through consecutive baths of xylene ( cat . #28975 . 325 , VWR ) and isopropanol ( cat . #K50655934838 , Merck ) . Endogenous peroxidase was blocked by immersion of the slides in methanol ( cat . #322415 , Sigma-Aldrich ) containing 0 , 3% H2O2 ( cat . #H1009 , Sigma-Aldrich ) for 15 min and rehydration was finalized by rinsing the slides in tap water . Heat-induced epitope retrieval was done using citrate buffer ( PH 6 . 0 ) for 20 min in a pressure cooker . After blocking of the sections with 2% BSA ( cat . #A7906 , Sigma-Aldrich ) for 20 min , slides were incubated for 1 hr at 37°C with primary antibody ( antibodies used are listed in Key resources table ) . Anti-mouse ( cat . #G21040 , dilution: 1/1000 , Invitrogen ) or anti-goat ( Impress , cat . #MP7405 , Vector ) , respectively , HRP-coupled secondary antibody was applied for 30 min at 37°C and revealed with DAB for 30 s ( cat . #K3468 , Dako ) . After counterstaining with hematoxylin ( cat . #HX86014349 , diluted 1/5 , Merck ) , the sections were dehydrated in consecutive baths of ethanol ( cat . #20821 . 310 , VWR ) , isopropanol ( cat . #K50655934838 , Merck ) , and xylene ( cat . #28975 . 325 , VWR ) to finally be mounted with hardening medium ( Eukitt , cat . #03989 , Sigma-Aldrich ) . Whole mount liver images cleared with iDISCO+ were initially processed in ImageJ for maximum z-projection and segmentation . The images were filtered using the unsharp mask and integral image filter function . Images were next processed in Amira . In Amira images were filtered using the Gaussian filter and background detection correlation . Images were manually segmented . The manual segmentation was further traced using the autoskeleton function . The skeletons were further analyzed in Amira for length , volume and branching . Images of ink injected liver were proceeds for filament tracing . Bile duct and portal vein filament tracing was performed using Amira . The images were filtered using the unsharp mask and mean filter . The signal was manually segmented to remove artificial signal . The manual segmentation was further traced using the autoskeleton function . The skeletons were analyzed in Amira for length , volume and branching . For double ink injection ( Figure 1—figure supplement 1A ) the background was changed for esthetic purposes using the lasso tool in Adobe Photoshop . DUCT 2D slices were exported from MyVGL ( Volumegraphics ) and processed in ImageJ for maximum contrast and brightness . P15 MICROFIL injected and BABB-cleared left lateral and right medial lobe were analyzed in ImageJ . The total liver area was measured followed by the measurement of the liver area covered by MICROFIL injected bile ducts . The percentage of liver containing bile ducts was calculated . Blood from P15 pups was collected from the trunk after decapitation into 1 . 5 ml tubes . The serum was allowed to clot at room temperature . The blood was centrifuged for 15 min at 17 , 000 g at room temperature . The serum was stored at −80°C until analyzed . Serum was sent to the Swedish University of Agricultural Sciences for analysis of alanine aminotransferase ( ALT ) , alkaline phosphatase ( ALP ) , aspartate aminotransferase ( AST ) , albumin ( Alb ) , and total bilirubin . The system GE Phoenix v|tome|x L 240 ( GE Sensing and Inspection Technologies GmbH , Germany ) equipped with nanofocus X-ray tube ( 180 kV/15 W ) was used for the tomographic measurements that were carried out in the air-conditioned cabinet ( fixed temperature 21°C ) . The samples were adapted for this temperature before the measurement to prevent any thermal expansion effect . To prevent any sample motion during the scanning , the samples were placed in 15 ml Falcon tube , filled with 1% agarose gel . The tomographic reconstruction of acquired data was performed using GE phoenix datos|x 2 . 0 software . The voxel resolution was fixed for all the adult liver samples at 12 µm , except one ( sample #2401 , 8 µm ) . For all the P15 liver samples the voxel resolution was fixed at 6 . 5 µm and for the lung lobe sample at 8 µm . Detailed overview of used acquisition parameters is stated in Table 2 . Jag1+/+ and Jag1Ndr/Ndr data were tested for significant differences using multiple tests based on the type of experiment and data distribution . Student´s t-test ( Figures 1D , G and 3D , Figure 1—figure supplement 12H , Figure 6—figure supplement 1B–E , H–K ) . Kolmogorov-Smirnov test ( on raw data , graph depicts cumulative sum ) ( Figures 3C and 4C ) . Mann-Whitney test ( Figure 4—figure supplement 1 ) . Wilcoxon test ( Figure 1—figure supplement 5D ) . Two-way ANOVA ( Figures 1H and 5B–D , 6E , Figure 1—figure supplements 5C and 12I , Figure 6—figure supplement 1F and G ) followed by Sidak's multiple comparisons test . Spearman correlation ( Figure 1E ) . A p value below 0 . 05 was considered statistically significant . The statistical analysis was done in Prism 9 ( GraphPad ) . | Many essential parts of the body contain tubes: the liver for example , contains bile ducts and blood vessels . These tubes develop right next to each other , like entwined trees . To do their jobs , these ducts must communicate and collaborate , but they do not always grow properly . For example , babies with Alagille syndrome are born with few or no bile ducts , resulting in serious liver disease . Understanding the architecture of the tubes in their livers could explain why some children with this syndrome improve with time , but many others need a liver transplant . Visualising biological tubes in three dimensions is challenging . One major roadblock is the difficulty in seeing several tubular structures at once . Traditional microscopic imaging of anatomy is in two dimensions , using slices of tissue . This approach shows the cross-sections of tubes , but not how the ducts connect and interact . An alternative is to use micro computed tomography scans , which use X-rays to examine structures in three dimensions . The challenge with this approach is that soft tissues , which tubes in the body are made of , do not show up well on X-ray . One way to solve this is to fill the ducts with X-ray absorbing resins , making a cast of the entire tree structure . The question is , can two closely connected tree structures be distinguished if they are cast at the same time ? To address this question , Hankeova , Salplachta et al . developed a technique called double resin casting micro computed tomography , or DUCT for short . The approach involved making casts of tube systems using two types of resin that show up differently under X-rays . The new technique was tested on a mouse model of Alagille syndrome . One resin was injected into the bile ducts , and another into the blood vessels . This allowed Hankeova , Salplachta et al . to reconstruction both trees digitally , revealing their length , volume , branching , and interactions . In healthy mice , the bile ducts were straight with uniform branches , but in mice with Alagille syndrome ducts were wiggly , and had extra branches in the centre of the liver . This new imaging technique could improve the understanding of tube systems in animal models of diseases , both in the liver and in other organs with tubes , such as the lungs or the kidneys . Hankeova , Salplachta et al . also lay a foundation for a deeper understanding of bile duct recovery in Alagille syndrome . In the future , DUCT could help researchers to see how mouse bile ducts change in response to experimental therapies . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"stem",
"cells",
"and",
"regenerative",
"medicine",
"physics",
"of",
"living",
"systems"
] | 2021 | DUCT reveals architectural mechanisms contributing to bile duct recovery in a mouse model for Alagille syndrome |
RNA virus infections are detected by the RIG-I family of receptors , which induce type-I interferons through the mitochondrial protein MAVS . MAVS forms large prion-like polymers that activate the cytosolic kinases IKK and TBK1 , which in turn activate NF-κB and IRF3 , respectively , to induce interferons . Here we show that MAVS polymers recruit several TRAF proteins , including TRAF2 , TRAF5 , and TRAF6 , through distinct TRAF-binding motifs . Mutations of these motifs that disrupted MAVS binding to TRAFs abrogated its ability to activate IRF3 . IRF3 activation was also abolished in cells lacking TRAF2 , 5 , and 6 . These TRAF proteins promoted ubiquitination reactions that recruited NEMO to the MAVS signaling complex , leading to the activation of IKK and TBK1 . These results delineate the mechanism of MAVS signaling and reveal that TRAF2 , 5 , and 6 , which are normally associated with NF-κB activation , also play a crucial role in IRF3 activation in antiviral immune responses .
The innate immune system deploys germline-encoded pattern recognition receptors ( PRRs ) to detect pathogen invasion . Pathogen-associated molecular patterns ( PAMPs ) such as bacterial LPS and viral dsRNA are recognized by PRRs such as membrane-bound Toll-like receptors ( TLRs ) and cytosolic retinoic-acid-inducible gene-I ( RIG-I ) -like receptors ( RLRs ) ( Akira et al . , 2006 ) . Upon activation , the receptors trigger the production of type-I interferons ( e . g . , IFNα and IFNβ ) and other cytokines ( e . g . , TNFα and IL-6 ) to rapidly restrict the infection and further activate the adaptive immune system ( Pichlmair and Reis e Sousa , 2007; Stetson and Medzhitov , 2006 ) . RLRs include RIG-I , MDA5 , and LGP2 ( Yoneyama et al . , 2004; Yoneyama and Fujita , 2008 ) . All RLRs contain a DEAD/H-box RNA helicase domain , which binds to double-stranded RNA . RIG-I and MDA5 , but not LGP2 , harbor N-terminal tandem CARD domains that are crucial for triggering type-I IFN production . RIG-I also contains a C-terminal regulatory domain that specifically binds to RNA bearing 5′ triphosphate ( Hornung et al . , 2006; Pichlmair et al . , 2006; Cui et al . , 2008 ) . Upon binding to different viral RNA ligands , RIG-I and MDA5 activate another CARD-containing protein , mitochondrial antiviral signaling protein ( MAVS , also known as IPS-1 , VISA , and CARDIF ) , presumably through CARD–CARD interaction . MAVS in turn activates the transcription factors IRF3 and NF-κB , leading to interferon induction ( Kawai et al . , 2005; Meylan et al . , 2005; Seth et al . , 2005; Xu et al . , 2005 ) . Recent studies have provided insights into the mechanisms by which RIG-I and MAVS are activated upon viral infection ( Gack et al . , 2007; Zeng et al . , 2010; Kowalinski et al . , 2011; Luo et al . , 2011; Jiang et al . , 2012 ) . The sequential binding of RIG-I to viral RNAs and unanchored lysine-63 ( K63 ) polyubiquitin chains promotes RIG-I to form higher order oligomers , which then rapidly induce MAVS polymerization . The MAVS polymers recruit other MAVS molecules on the mitochondrial surface to form larger polymers through a prion-like mechanism ( Hou et al . , 2011 ) . These MAVS polymers potently activate the cytosolic kinases IKK and TBK1 through mechanisms that remain to be elucidated . The ubiquitin system has been shown to be important for RIG-I signaling at steps both upstream and downstream of MAVS . Upstream of MAVS , polyubiquitin binding to RIG-I is required for RIG-I activation as described above . Downstream of MAVS , K63-linked polyubiquitin and the E2 UBC5 are required for IRF3 activation ( Zeng et al . , 2009 ) . NEMO , the regulatory subunit of IKK and TBK1 , functions as a ubiquitin sensor in different pathways that lead to NF-κB and IRF3 activation ( Ea et al . , 2006; Wu et al . , 2006; Zhao et al . , 2007; Zeng et al . , 2009 ) . Additionally , several proteins , including TANK , SINTBAD , NAP1 , and STING ( also known as MITA , ERIS , and MPYS ) , have been found to associate with TBK1 ( Sasai et al . , 2006; Guo and Cheng , 2007; Ryzhakov and Randow , 2007; Ishikawa and Barber , 2008; Jin et al . , 2008; Zhong et al . , 2008; Sun et al . , 2009 ) . Several E3 ligases including TRAF3 , TRAF5 , cIAP1/2 , and MIB1/2 were proposed to regulate IRF3 activation downstream of MAVS ( Saha et al . , 2006; Mao et al . , 2010; Tang and Wang , 2010; Li et al . , 2011; Wang et al . , 2012 ) . In addition , LUBAC , an E3 ligase complex that synthesizes linear ubiquitin chains , has been shown to negatively regulate the RIG-I pathway ( Inn et al . , 2011; Belgnaoui et al . , 2012 ) . However , the biochemical mechanism of how ubiquitination regulates the activation of the protein kinases and transcription factors remains unclear . We have previously described a cell-free system that mimics cellular response to viral infection . To further dissect the mechanism of how MAVS propagates downstream signaling , we performed conventional purification and identified TRAF6 as an IRF3 activator in the presence of activated MAVS . Moreover , by knocking down TRAF6 in Traf2/5-deficient cells , we found that TRAF2 and TRAF5 act redundantly with TRAF6 to activate both IRF3 and NF-κB in response to virus . In addition , we provide evidence that the E3 ligase activity of TRAF6 is essential in the TRAF6-dependent activation of IRF3 and NF-κB , whereas the E3 ligase activity of TRAF2 is redundant with that of LUBAC in the Traf2-dependent pathway downstream of MAVS . Furthermore , mutations of the binding sites for TRAF2 , TRAF5 , and TRAF6 on MAVS abolished the ability of MAVS to activate downstream signaling after virus infection , without affecting its ability to form prion-like polymers . MAVS polymerization mutants , however , failed to recruit TRAFs . Finally , we found that through its ubiquitin-binding domains , NEMO formed a ubiquitination-dependent complex with TRAFs and MAVS , both in vitro and in cells . These results demonstrate a key role of TRAF2 , TRAF5 , and TRAF6 , which form a ubiquitin-dependent signaling complex with NEMO and the kinases , in propagating antiviral signaling downstream of MAVS .
We have previously established a cell-free IRF3 dimerization assay that mimics IRF3 activation in virus-infected cells ( Zeng et al . , 2009 ) . When crude mitochondrial fraction ( P5 ) from Sendai virus-infected HEK293T cells was incubated with cytosolic extracts ( S5 or S100 ) from uninfected cells , along with [35S]-IRF3 and ATP , IRF3 dimerization was detected by native gel electrophoresis followed by autoradiography ( Figure 1A , B ) . Similarly , IκBα phosphorylation was detected by immunoblotting with a phospho-IκBα specific antibody ( Figure 1B ) . P5 from virus-infected cells could also be replaced by purified MAVS without its transmembrane domain ( MAVSΔTM ) as previously described ( Hou et al . , 2011 ) . MAVSΔTM expressed and purified from E . coli forms functional polymers , thereby bypassing the requirement for its mitochondrial membrane localization ( Hou et al . , 2011 ) . In our previous study , HeLa S100 was separated by Q-Sepharose column into Q-A containing the flow-through and Q-B containing proteins eluted with 0 . 3 M NaCl ( Zeng et al . , 2009 ) . Both Q-A and Q-B were required to support IRF3 activation in our in vitro assay . The key factor in Q-A was identified as the ubiquitin E2 Ubc5 ( Zeng et al . , 2009 ) . 10 . 7554/eLife . 00785 . 003Figure 1 . TRAF2 , TRAF5 , and TRAF6 are important for IRF3 and IKK activation in vitro . ( A ) Diagram of differential centrifugation of cell homogenates . HEK293T cells infected with Sendai virus ( +SeV ) or mock treated ( −SeV ) were homogenized in hypotonic buffer , followed by sequential centrifugation to separate crude mitochondria ( P5 ) from cytosolic supernatant ( S5 and S100 ) . ( B ) IRF3 and NF-κB activation in vitro . Mitochondrial fraction ( P5 ) from Sendai virus-infected HEK293T cells or purified His6-tagged MAVS without transmembrane domain ( His6-MAVSΔTM ) was incubated with cytosolic extract ( S5 ) from uninfected cells in the presence of ATP and 35S-IRF3 . Dimerization of IRF3 was analyzed by native gel electrophoresis , followed by autoradiography . IκBα phosphorylation was analyzed by immunoblotting . ( C ) NEMO-interacting complex is required for IRF3 activation in vitro . GST-tagged NEMO without its N-terminal IKK-binding region ( GST-NEMOΔN ) was mixed with cytosolic extract from Nemo−/− MEF cells to collect GST-NEMOΔN pull down ( NEMOΔN PD ) . This material , or GST-NEMO , was incubated with cytosolic extract ( S100 ) from HeLa cells depleted of NEMO with a NEMO antibody . Activation of IRF3 was analyzed as described in ( B ) . ( D ) Reconstitution of IRF3 dimerization in vitro . NEMOΔN PD , His8-E1 , Ubc5c , His6-TRAF6 , ubiquitin , His6-MAVSΔTM , His8-IRF3 , and 35S-Flag-IRF3 were incubated together with ATP as indicated , followed by analysis of IRF3 dimerization . ( E and F ) TRAF6 is important for IRF3 and NF-κB activation by MAVS in vitro . Cytosolic extracts from wild-type or Traf6−/− MEF cells were incubated with His6-MAVSΔTM together with WT or mutant TRAF6 protein as indicated , followed by analysis of IRF3 dimerization ( E ) or IκBα phosphorylation ( F ) . T6RZC: TRAF6 containing the RING , zinc , and coiled-coil domains , with the TRAF-C domain replaced by a fragment of bacterial gyrase B . ( G ) TRAF2 and 5 are important for IRF3 activation by MAVS in vitro . Cytosolic extracts from WT or different TRAF deficient MEF cells were incubated with His6-MAVSΔTM , followed by IRF3 dimerization assay . ( H and I ) Either TRAF2 or TRAF5 rescues IRF3 and IKK activation by MAVS in the Traf2/5 DKO extract . Traf2/5 DKO extracts were supplemented with TRAF2 or TRAF5 proteins as indicated , together with His6-MAVSΔTM and ATP , followed by measurement of IRF3 dimerization and IκBα phosphorylation . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 00310 . 7554/eLife . 00785 . 004Figure 1—figure supplement 1 . TRAF6 and TRAF2/5 are IRF3 activators . ( A ) Wild-type and IKKα/IKKβ double knockout MEF cells were infected with VSV for the indicated time . IRF3 dimerization was analyzed by immunoblotting . ( B ) Flag-tagged NEMO WT and ΔN were tested for their ability to rescue IRF3 dimerization in Ikbkg−/− ( Nemo−/− ) MEF extracts in the presence or absence of mitochondria ( P5 ) isolated from Sendai virus-infected cells . ( C ) Ikbkg−/− ( Nemo−/− ) MEF cells stably expressing Flag-NEMOΔN were used to isolate endogenous NEMO–TBK1 complex , which was analyzed by immunoblotting . ( D ) Partially purified fraction from HeLa S100 that contains IRF3 stimulatory activity was further fractionated on Heparin-Sepharose , then each fraction was assayed for IRF3 dimerization in the presence or absence of the NEMO pull down ( PD ) as shown in ( C ) . The reactions also contained His8-E1 , Ubc5c , ubiquitin , ATP , and virus-activated mitochondria ( P5 ) . ( E ) Scheme for purification of an IRF3 activator . ( F ) Fractions from the last step ( monoQ ) shown in ( E ) were tested for their ability to stimulate IRF3 dimerization in the presence of NEMO-PD and virus-activated mitochondria ( top ) . Aliquots of the fractions were immunoblotted with a TRAF6 antibody ( bottom ) . ( G ) Recombinant TRAF6 activates IRF3 . Indicated amount of His-tagged TRAF6 purified from Sf9 cells was added into the IRF3 dimerization assay that contains E1 , UbcH5 , ubiquitin , 35S-IRF3 , NEMO-PD ( containing TBK1 complex ) , and mitochondria from Sendai virus-infected cells ( P5 ) . ( H ) Wild-type and Traf6−/− primary MEF cells were infected with Sendai virus for the indicated time , and phosphorylation of IRF3 and IκBα was analyzed by immunoblotting ( top ) . As a control , the same cells were treated with IL-1β , and the cell extract was analyzed by immunoblotting with an IκBα antibody ( bottom ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 004 Q-B contains multiple factors known to be important for virus-induced IRF3 activation , such as NEMO , TBK1 , ubiquitin , and E1 ( data not shown ) . Q-B also contains the IKK complex; however , IKKα/IKKβ double-deficient MEF cells activated IRF3 normally in response to infection by vesicular stomatitis virus ( VSV ) , an RNA virus ( Figure 1—figure supplement 1A ) . Moreover , GST-NEMO without its N-terminal IKK-binding site ( NEMOΔN ) rescued IRF3 activity in Nemo−/− ( Ikbkg−/− ) MEF cell extracts ( Figure 1—figure supplement 1B ) , indicating an IKK-independent role of NEMO in virus-induced IRF3 activation . NEMO has been reported to interact with TBK1 through TANK ( Zhao et al . , 2007 ) . Indeed , NEMOΔN pulled down endogenous TANK and TBK1 from Nemo−/− cell extracts ( NEMOΔN PD , Figure 1—figure supplement 1C ) . After NEMO depletion by a NEMO antibody , S100 lost its ability to support IRF3 dimerization in vitro , and the activity was restored by adding back NEMO PD , but not NEMO alone ( Figure 1C ) . This suggests that NEMO and the TBK1 complex function together in IRF3 activation . However , NEMOΔN PD does not fully replace Q-B in IRF3 activation in vitro even in the presence of ubiquitin and E1 ( data not shown ) , indicating that additional factor ( s ) might be required for IRF3 activation . We further fractioned Q-B on Heparin-Sepharose and tested the ability of individual fractions to support IRF3 dimerization in the presence or absence of NEMOΔN PD . In this assay , we replaced Q-A with purified Ubc5 and also included purified ubiquitin and E1 to avoid identifying these known factors . Several fractions from the Heparin column showed IRF3 stimulatory activity , which was dependent on NEMOΔN PD ( e . g . , fraction 14 in Figure 1—figure supplement 1D ) . Subsequently , five more steps of conventional chromatography were used to purify this activity ( Figure 1—figure supplement 1E ) . Fractions from the last monoQ column were subjected to silver staining and tandem mass spectrometry , which identified several proteins , including TRAF6 . Immunoblotting with a TRAF6 antibody confirmed that TRAF6 co-purified with the IRF3 dimerization activity ( Figure 1—figure supplement 1F ) . To determine whether TRAF6 is important for IRF3 activation in vitro , we performed reconstitution experiments using purified proteins and found that TRAF6 supported IRF3 activation in a manner that depended on MAVSΔTM and the ubiquitin system ( Figure 1D ) . Similarly , IRF3 activation by virus-activated mitochondria ( P5 ) was dependent on TRAF6 ( Figure 1—figure supplement 1G ) . Cytosolic extracts from Traf6−/− primary MEF cells were severely , albeit not completely , defective in supporting IRF3 dimerization and IκBα phosphorylation in vitro , and these defects were rescued by adding back wild-type TRAF6 ( Figure 1E , F ) . In contrast , TRAF6 RING mutant ( TRAF6-C70A ) , TRAF6 Zinc finger deletion ( TRAF6ΔZF ) , or TRAF6 with the TRAF-C domain replaced by a fragment of bacterial gyrase-B ( T6RZC ) ( Wang et al . , 2001 ) failed to rescue IRF3 activation in Traf6−/− cell extracts ( Figure 1E ) . These results suggest that both TRAF6 E3 ligase activity and its ability to interact with other proteins , that is , MAVS ( Seth et al . , 2005; Xu et al . , 2005 ) , are important for IRF3 activation in vitro . However , it has been shown that Traf6−/− cells exhibited normal virus-induced interferon production ( Seth et al . , 2005; Zeng et al . , 2009 ) . Consistent with these reports , Traf6−/− primary MEF cells supported IRF3 and IKK activation in response to Sendai virus infection , but were defective in activating IKK in response to IL-1β ( Figure 1—figure supplement 1H ) . Thus , it is possible that other TRAF proteins might also be involved in MAVS signaling . Indeed , crude extracts ( S5 ) from Traf2−/−/Traf5−/− ( Traf2/5 DKO ) MEF cells failed to support IRF3 dimerization or IκBα phosphorylation in vitro; both activities were restored by adding back wild-type TRAF2 or TRAF5 ( Figure 1G , H , I ) . Unlike TRAF6 , TRAF2 RING deletion ( TRAF2ΔR ) did not impair IRF3 dimerization in cytosolic extracts ( S5 ) from Traf2/5 DKO cells . In contrast , deletion of the TRAF-C domain from TRAF2 ( TRAF2ΔC ) abolished its activity . MEF cells lacking a single TRAF protein , TRAF2 , TRAF3 , or TRAF5 , were largely normal in activating IRF3 in vitro ( Figure 1G ) . Similar to Traf6−/− MEFs , Traf2−/− MEFs induced IFNβ and IL6 normally in response to Sendai virus infection ( Figure 2—figure supplement 1A , B ) . Moreover , in contrast to the profound defect of Traf2/5 DKO cell extracts in supporting IRF3 and IKK activation by MAVS , the DKO cells activated IRF3 and IKK and induced IFNβ normally after Sendai virus infection ( Figure 2A , B ) . Importantly , knockdown of TRAF6 expression by short hairpin RNA ( shTRAF6 ) in the DKO cells abolished IRF3 and IKK activation and IFNβ induction by Sendai virus , but did not impair STAT1 phosphorylation induced by IFNγ ( Figure 2A ) . The defects in IRF3 and IκBα phosphorylation in the DKO+shTRAF6 cells were rescued by expressing RNAi-resistant WT TRAF6 , but not the C70A mutant of TRAF6 . Similarly , VSV induction of several cytokines , including IFNβ , IL6 , IFNα , and CXCL10 , was abolished in the DKO+shTRAF6 cells but rescued by WT TRAF6 ( Figure 2C–E , Figure 2—figure supplement 1C , D ) . A TRAF6 mutant in which every lysine was substituted with an arginine ( T6-K0 ) , but not TRAF6-C70A , rescued the cytokine expression in the DKO+shTRAF6 cells . Thus , TRAF6 functions redundantly with TRAF2 and TRAF5 in a manner that depends on the E3 ligase activity of TRAF6 but not its ubiquitination at any lysine residue . 10 . 7554/eLife . 00785 . 005Figure 2 . TRAF6 functions redundantly with TRAF2 and TRAF5 to activate IRF3 in cells . ( A ) Depletion of TRAF6 in Traf2/5 DKO cells abolishes both IRF3 and NF-κB activation by virus . Traf2/5 DKO MEF cells stably expressing GFP ( as a control ) or an shRNA against TRAF6 were infected with Sendai virus for the indicated time followed by immunoblotting of the cell extracts with the indicated antibodies ( top ) . As a control , IFNγ induced STAT-1 phosphorylation was also analyzed by immunoblotting ( bottom ) . ( B ) Depletion of TRAF6 in Traf2/5 DKO cells abolishes IFNβ mRNA induction by virus . The cells described in ( A ) were treated with Sendai virus for the indicated time before total RNA was isolated . IFNβ mRNA level was analyzed by q-RT-PCR . ( C–E ) The catalytic activity of TRAF6 is required for antiviral immune responses . Traf2/5 DKO MEF cells stably expressing shRNA against TRAF6 ( DKO+shT6 ) and those in which endogenous TRAF6 was replaced with WT or RING mutant ( C70A ) Flag-TRAF6 were stimulated with Sendai Virus or VSV for the indicated time . Phosphorylation of IRF3 and IκBα was analyzed by immunoblotting . Total RNA was also isolated for the measurement of IFNβ and IL6 RNA by q-RT-PCR . T6-K0: a TRAF6 mutant in which all lysine residues were substituted with arginine . ( F–H ) The RING domain of TRAF2 is dispensable for its signaling functions . Traf2/5 DKO cells stably expressing WT or RING mutant ( C34A ) Flag-TRAF2 were stimulated with Sendai virus or VSV for the indicated time . Activation of IRF3 and phosphorylation of IκBα was analyzed by immunoblotting . Cytokine RNA levels were measured by q-RT-PCR . Unless indicated otherwise , error bars in this and other figures of this paper represent standard deviations of triplicate experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 00510 . 7554/eLife . 00785 . 006Figure 2—figure supplement 1 . TRAF2 , 5 , and 6 function redundantly to activate IRF3 in cells . ( A and B ) Wild-type , Traf6−/− , and Traf2−/− primary MEF cells were infected with Sendai virus for the indicated time . The RNA levels of IFNβ and IL-6 were measured by q-RT-PCR . ( C and D ) Traf2/5 DKO MEF cells stably expressing an shRNA against TRAF6 were reconstituted with WT or mutant TRAF6 as indicated . The cells were infected with VSV for the indicated time . Cytokine RNA levels were measured by q-RT-PCR . ( E–G ) Traf2/5 DKO MEF cells stably expressing an shRNA against TRAF3 were infected with VSV as indicated , and then RNA levels of IFNβ , IL6 , and TRAF3 were analyzed by q-RT-PCR . ( H ) Traf3−/− MEF cells stably expressing TRAF6 shRNA were infected with Sendai virus , followed by immunoblotting with phospho-IRF3 and TRAF6 antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 00610 . 7554/eLife . 00785 . 007Figure 2—figure supplement 2 . Multiple E3 ligases function redundantly to activate IRF3 in cells . ( A–D ) Wild-type , Traf6−/− , and Traf2/5 DKO MEF cells were treated with or without a SMAC mimetic compound ( SM ) for 1 hr before VSV infection . Cell extracts were analyzed by immunoblotting with the indicated antibodies ( A and B ) . The RNA levels of IFNβ and IL6 were measured by q-RT-PCR ( C and D ) . ( E and F ) Similar to Figure 2—figure supplement 1C , D , except that the cells were reconstituted with WT or C34A TRAF2 . ( G ) Cells described in ( E and F ) were stimulated with either TNF-α or IL-1β , followed by immunoblotting with a phospho-IκBα antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 007 Previous reports have suggested that TRAF3 and cIAPs function as ubiquitin E3 ligases that activate TBK1 and IRF3 ( Häcker et al . , 2006; Oganesyan et al . , 2006 ) . However , we have shown that in Traf3-deficient MEFs , RNA viruses can still activate IRF3 normally and induce IFNβ at a modestly reduced level ( Zeng et al . , 2009 ) . To determine if TRAF3 functions redundantly with other TRAF proteins , we knocked down TRAF3 in Traf2/Traf5 DKO MEFs and found that VSV induced IFNβ and IL6 normally in these cells ( Figure 2—figure supplement 1E–G ) . Similarly , knocking down TRAF6 in Traf3-deficient MEFs did not impair IRF3 activation by Sendai virus ( Figure 2—figure supplement 1H ) . To test the role of cIAPs in the MAVS pathway , we used a small molecule mimetic of SMAC , which is known to target the degradation of cIAP1 and cIAP2 ( Li et al . , 2004 ) . This treatment ( +SM ) did not impair the activation of IRF3 or the induction of IFNβ or IL6 by VSV in WT , Traf6−/− or Traf2−/−Traf5−/− MEFs ( Figure 2—figure supplement 2A–D ) . Taken together , these results suggest that TRAF3 and cIAPs are not essential for MAVS signaling , whereas TRAF6 functions redundantly with TRAF2 and TRAF5 to mediate MAVS signaling . To test if the E3 ligase activity of TRAF2 is important for MAVS signaling , we expressed WT or a RING domain mutant ( C34A ) of TRAF2 in the DKO+shTRAF6 cells . In contrast to a requirement of the E3 ligase activity of TRAF6 in MAVS signaling ( Figure 2C–E ) , both WT and C34A mutant of TRAF2 rescued virus activation of IRF3 as well as the induction of IFNβ , IL6 , IFNα , and CXCL10 in these cells ( Figure 2F–H , Figure 2—figure supplement 2E , F ) . WT and C34A TRAF2 also rescued IκBα phosphorylation induced by TNFα , but not IL-1β , in DKO+shTRAF6 cells ( Figure 2—figure supplement 2G ) . These results suggest that the E3 ligase activity of TRAF2 is either dispensable or redundant with that of another E3 ligase ( see below ) . Recent studies suggest that the linear ubiquitin E3 complex LUBAC plays a positive role in NF-κB activation by TNFα but a negative role in regulating the RIG-I pathway ( Tokunaga et al . , 2009; Inn et al . , 2011; Belgnaoui et al . , 2012 ) . To test the role of LUBAC in MAVS signaling , we used shRNA to knock down each individual subunit of the LUBAC complex , HOIP , HOIL-1 , and Sharpin , as well as MAVS or TRAF3 in wild-type MEF cells ( Figure 3A , B , Figure 3—figure supplement 1A , B ) . As expected , the knockdown of MAVS blocked IRF3 activation and IFNβ induction by VSV . The knockdown of HOIL-1 slightly enhanced IRF3 activation and IFNβ induction as compared to the control cells , consistent with its role as a negative regulator of RIG-I signaling ( Inn et al . , 2011 ) . However , the knockdown of HOIP , Sharpin , or TRAF3 did not appreciably affect IRF3 activation and modestly inhibited IFNβ induction by VSV ( Figure 3A , B ) . A previous report showed that IFNβ induction is enhanced in MEF cells derived from the mouse strain Sharpincpdm , which is deficient in Sharpin expression ( Belgnaoui et al . , 2012 ) . Contrary to this report , we found that loss of Sharpin slightly reduced IFNβ induction by VSV ( Figure 3C ) . At later time points , Sharpin deficiency partially reduced IFNβ induction by VSV and Sendai virus ( Figure 3—figure supplement 1C ) . Re-introducing Sharpin back into Sharpin-deficient MEFs did not inhibit or enhance the activation of IRF3 or IKK by Sendai virus ( Figure 3—figure supplement 1D ) . Consistent with our model that LUBAC is not a negative regulator of MAVS signaling , tetracycline-inducible knockdown of HOIP in the human osteosarcoma cell line U2OS modestly inhibited the activation of IRF3 and induction of IFNβ by VSV ( Figure 3—figure supplement 2A–C ) . 10 . 7554/eLife . 00785 . 008Figure 3 . LUBAC functions redundantly with TRAF2 to support MAVS signaling . ( A and B ) WT MEF cells stably expressing GFP or shRNA against HOIL-1 , Sharpin , HOIP , MAVS , or TRAF3 were infected with VSV for the indicated time . IRF3 dimerization and IκBα phosphorylation were analyzed by immunoblotting ( A ) . The levels of IFNβ RNA were measured by q-RT-PCR ( B ) . The efficiency of RNAi is shown in Figure 3—figure supplement 1A , B . ( C ) Primary MEF cells from heterozygous or Sharpincpdm mice were infected with VSV for the indicated time , and then IFNβ RNA levels were measured by q-RT-PCR . ( D ) Traf2/5 DKO MEF cells stably expressing an shRNA against TRAF6 were reconstituted with TRAF2 WT or ΔRING ( ΔR ) mutant ( lower panels ) . These cells , as well as WT MEF ( upper panel ) , were further depleted of HOIP ( lane 4–6 ) by lentiviral shRNA and then rescued with WT or the active site mutant ( CS ) of HOIP ( lanes 7–12 ) . In lanes 1–3 , a lentiviral vector expressing GFP was used as a control . The cells were infected with VSV for the indicated time , followed by measurement of IRF3 dimerization . ( E and F ) The cells described in ( D ) were analyzed for the expression of HOIP by q-RT-PCR ( E ) or immunoblotting with an HA antibody ( F ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 00810 . 7554/eLife . 00785 . 009Figure 3—figure supplement 1 . LUBAC is largely dispensable for MAVS signaling when it is depleted from wild-type cells . ( A and B ) Wild-type MEFs stably expressing shRNA against the indicated proteins were used for the experiments described in Figure 3A , B . The knockdown efficiency of each protein was analyzed by immunoblotting with the indicated antibody ( A ) . For mouse HOIP , for which no antibody was available , q-RT-PCR was used to measure its RNA levels ( B ) . ( C ) Primary MEFs from heterozygous or homozygous Sharpincpdm ( cpdm ) mice were infected with VSV or Sendai virus as indicated for 0–48 hr as indicated . IFNβ RNA levels were measured by q-RT-PCR . ( D ) Immortalized Sharpincpdm ( cpdm ) MEF cells stably expressing GFP or HA-Sharpin were infected with Sendai virus . IRF3 dimerization , IκBα phosphorylation ( left ) , and Sharpin expression ( right ) were analyzed by immunoblotting . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 00910 . 7554/eLife . 00785 . 010Figure 3—figure supplement 2 . LUBAC is required for MAVS signaling only when the functions of TRAF proteins are compromised . ( A–C ) U2OS cells stably expressing a tetracycline-inducible shRNA against HOIP were treated with tetracycline ( Tet; 1 μg/ml ) for 7 days , then infected with VSV for the indicated time , followed by immunoblotting of cell extracts with the indicated antibodies ( A ) . IFNβ RNA levels were measured by q-RT-PCR ( B and C ) . ( D ) Traf2/5 DKO MEF cells stably expressing shRNA against TRAF6 and HOIP ( or GFP ) were reconstituted with WT TRAF2 , TRAF6 , or ΔRING TRAF2 . The cells were infected with VSV for the indicated time , followed by immunoblotting with a phospho-IRF3 specific antibody . ( E ) Traf2/5 DKO MEF cells stably expressing the TRAF6 shRNA were reconstituted with TRAF2 WT or ΔRING mutant . These cells were then further depleted of HOIL-1 , Sharpin , or HOIP by lentiviral shRNA . After VSV infection , cell extracts were prepared for immunoblotting with the indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 010 To test if HOIP functions redundantly with TRAF2 , we knocked down HOIP in MEF cells lacking TRAF2 , TRAF5 , and TRAF6 ( Traf2/5 DKO + shTRAF6 ) , which were reconstituted with WT or ΔRING TRAF2 ( Figure 3D–F ) or WT TRAF6 ( Figure 3—figure supplement 2D ) . Interestingly , the depletion of HOIP abolished IRF3 activation by VSV in the cells expressing TRAF2-ΔRING , but not in those expressing WT TRAF2 or TRAF6 . The defect of IRF3 activation in the HOIP-depleted cells that expressed TRAF2-ΔRING was rescued by WT HOIP but not the catalytically inactive HOIP mutant , C693S/C696S ( Figure 3D ) . Similar to HOIP , knockdown of Sharpin , and to a lesser extent HOIL-1 , inhibited IRF3 and IKK activation in the Traf2/Traf5 DKO + shTRAF6 cells expressing TRAF2 ΔRING ( Figure 3—figure supplement 2E ) . Taken together , these results suggest that the E3 ligase activity of LUBAC can support MAVS signaling when the functions of the TRAF proteins are compromised . MAVS harbors binding motifs for TRAF2 , TRAF3 , TRAF5 , and TRAF6 ( Seth et al . , 2005; Xu et al . , 2005; Saha et al . , 2006; Paz et al . , 2011 ) ( Figure 4A ) . The motif PVQET ( 143–147 ) is known to bind TRAF2 and TRAF5 , and perhaps TRAF3 , whereas two motifs , PGENSE ( 153–158 ) and PEENEY ( 455–460 ) bind TRAF6 . The second TRAF6-binding motif has also been suggested to bind TRAF3 ( Paz et al . , 2011 ) . We showed previously that TRAF6 and TRAF2 shifted to high-molecular weight fractions together with MAVS polymers in response to Sendai virus infection ( Hou et al . , 2011 ) . Here , we examined whether mutations in MAVS that selectively disrupt its binding to specific TRAF proteins affect the downstream signaling . Mavs−/− MEFs stably expressing MAVS mutants were infected with Sendai virus followed by immunoblotting to examine IRF3 dimerization ( Figure 4B , C ) or q-RT-PCR to measure cytokine expression ( Figure 4D , E ) . Mutations of the TRAF2/3/5 ( Q145N ) or TRAF6 [E155D/E457D ( 2ED ) ] binding sites did not affect IRF3 activation by Sendai virus ( Figure 4B ) . In contrast , mutations of all known TRAF-binding sites ( QN2ED ) abolished IRF3 activation . Similarly , QN2ED , but not Q145N or 2ED , mutations in MAVS abolished IFNβ and IL-6 induction by Sendai virus . QN2ED MAVS also permitted higher levels of VSV replication ( Figure 4D–F ) . To confirm the specificity of the TRAF-binding motifs in MAVS , endogenous MAVS was knocked down by shRNA in Traf6−/− ( Figure 4G ) or Traf2/5 DKO ( Figure 4H ) MEFs , which were then rescued with different RNAi-resistant MAVS mutants . These cells were complemented with either Flag-TRAF6 ( Figure 4G ) or Flag-TRAF2 ( Figure 4H ) . Immunoprecipitation with Flag antibody revealed that the TRAF6 complex from the virus-infected cells contained WT and Q145N MAVS , but not 2ED or QN2ED MAVS ( Figure 4G ) . In contrast , the TRAF2 complex contained WT and 2ED MAVS , but not Q145N or QN2ED MAVS ( Figure 4H ) . Like WT MAVS , the QN2ED mutant formed high molecular weight aggregates after virus infection ( Figure 4—figure supplement 1A ) , suggesting that its inability to activate IRF3 is likely due to defective recruitment of TRAF proteins rather than a defect in polymerization . 10 . 7554/eLife . 00785 . 011Figure 4 . MAVS recruits multiple TRAF proteins to activate IRF3 upon virus infection . ( A ) The conserved binding motifs of TRAF2 , TRAF5 , and TRAF6 in human MAVS . Residues in red indicate mutation sites . ( B–E ) The TRAF-binding motifs on MAVS are essential for IRF3 and NF-κB activation by virus . Mavs−/− MEF cells reconstituted with WT or TRAF-binding mutant MAVS were infected with Sendai virus for the indicated time . IRF3 dimerization and the expression of different MAVS mutants were analyzed by immunoblotting . Cytokine RNA levels were measured by q-RT-PCR . 2ED: E155D/E457D; QN2ED: Q145N/E155D/E457D . ( F ) The TRAF-binding motifs on MAVS are important for restricting viral replication . MEF cells described in ( B ) were infected with VSV-ΔM51-GFP for the indicated time , and the GFP-positive cells were visualized under fluorescence microscope . ( G ) MAVS recruits TRAF6 after virus infection through TRAF6-binding motifs . Traf6−/− MEF cells were depleted of endogenous MAVS by shRNA , then reconstituted with Flag-TRAF6 and WT or mutant MAVS as indicated . The cells were infected with Sendai virus for the indicated time followed by immunoprecipitation of TRAF6 with a Flag antibody . The immunoprecipitates and whole cell lysates ( WCL ) were analyzed by immunoblotting with the indicated antibodies . ( H ) MAVS recruits TRAF2 upon virus infection through TRAF2 binding motif . Similar to ( G ) , except that Traf2/5 DKO MEF cells were reconstituted with Flag-TRAF2 and the binding between TRAF2 and different MAVS mutants was analyzed . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 01110 . 7554/eLife . 00785 . 012Figure 4—figure supplement 1 . MAVS recruits multiple TRAF proteins upon virus infection . ( A ) Mavs−/− MEF cells reconstituted with MAVS WT or QN2ED were infected with Sendai virus for the indicated time . Crude mitochondria ( P5 ) were isolated , solubilized in 1% DDM , and then separated by sucrose gradient ultracentrifugation . Fractions were analyzed by immunoblotting with a MAVS antibody . ( B ) MAVS ( Q145N ) only functions through TRAF6 . Traf6−/− MEF cells in which endogenous MAVS was knocked down by shRNA and replaced by wild-type or TRAF-binding mutant human MAVS were infected with Sendai virus as indicated . IFNβ RNA level was analyzed by q-RT-PCR . ( C ) Traf6−/− MEF cells stably expressing an shRNA against MAVS were reconstituted with MAVS Q145N and TRAF6 WT or C70A mutant . The cells were infected with Sendai virus to induce IFNβ RNA , which was measured by q-RT-PCR . ( D ) MAVS ( 2ED ) only functions through TRAF2 and TRAF5 . Traf2/5 DKO MEF cells with endogenous MAVS replaced by wild-type or TRAF-binding mutant human MAVS were infected with Sendai virus for the indicated time . IFNβ RNA levels were measured by q-RT-PCR . ( E ) Traf2/5 DKO MEF cells with endogenous MAVS replaced by MAVS ( 2ED ) were reconstituted with WT or RING mutant TRAF2 . The cells were infected with Sendai virus , and then IFNβ RNA induction was analyzed by q-RT-PCR . ( F and G ) Similar to ( E ) , except that Traf2−/− or Traf5−/− MEFs were used . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 01210 . 7554/eLife . 00785 . 013Figure 4—figure supplement 2 . MAVS recruits multiple TRAF proteins upon virus infection . ( A ) Traf6−/− or Traf3−/− MEF cells stably expressing an shRNA against MAVS were reconstituted with WT or TRAF-binding mutant human MAVS . These cells were infected with Sendai virus followed by the analysis of endogenous IRF3 dimerization . ( B ) A summary of results from the mutagenesis and complementation experiments that reveal the essential role of the distinct TRAF binding motifs in recruiting specific TRAF proteins . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 013 To further investigate the role of individual TRAF protein in MAVS signaling , we knocked down MAVS in MEF cells deficient in TRAF6 , TRAF2 , TRAF5 , or both TRAF2 and TRAF5 ( DKO ) , which were then rescued with RNAi-resistant WT or mutant MAVS ( Figure 4—supplement 1B–G ) . In Traf6−/− cells , Q145N but not 2ED MAVS was completely defective in inducing IFNβ , but this defect was rescued by ectopic expression of WT but not C70A TRAF6 ( Figure 4—figure supplement 1B , C ) . These results indicate that a functional TRAF6 was indispensable for signaling through the two TRAF6-binding sites ( i . e . , no other TRAF proteins , such as TRAF3 , could substitute the function of TRAF6 ) . In Traf2/5 DKO cells , 2ED but not Q145N MAVS was defective in inducing IFNβ , but this defect was rescued by WT TRAF2 and partially rescued by the C34A and RING-deleted mutant of TRAF2 ( TRAF2ΔRING ) ( Figure 4—figure supplement 1D , E ) . In Traf2−/− or Traf5−/− cells , only QN2ED MAVS failed to rescue IFNβ induction , suggesting that TRAF2 and TRAF5 could functionally substitute each other ( Figure 4—figure supplement 1F , G ) . Similar to WT cells ( Figure 4B ) , Traf3−/− cells could support IRF3 activation by WT , Q145N , and 2ED MAVS , but not QN2ED MAVS ( Figure 4—figure supplement 2A ) . In contrast , Traf6−/− cells could not support IRF3 activation by Q145N MAVS . Taken together ( Figure 4—figure supplement 2B ) , these results strongly support the conclusion that MAVS signals through its specific TRAF-binding motifs that recruit TRAF2 , TRAF5 , and TRAF6 , which then function in parallel to activate IRF3 and NF-κB . To test if MAVS polymerization is required for its binding to TRAF proteins , we mutated several conserved charged residues within MAVS CARD , including E26A , R64A , and R65A ( Figure 5A ) . Each of these mutations abrogated the ability of MAVS to induce an IFNβ luciferase ( IFNβ-Luc ) reporter gene stably integrated in HEK293T cells ( Figure 5B ) . WT or the mutant MAVS proteins were stably expressed in Mavs−/− MEFs , which were infected with Sendai virus followed by analysis of IRF3 dimerization . MAVS polymerization was analyzed by semidenaturing detergent agarose gel electrophoresis ( SDD-AGE; Figure 5C; Hou et al . , 2011 ) . Unlike wild-type MAVS , which activated IRF3 and formed SDS-resistant aggregates following viral infection , MAVS containing point mutations in the CARD domain failed to activate IRF3 or form aggregates . Similarly , in HEK293T cells in which endogenous MAVS was knocked down by shRNA and replaced by RNAi-resistant MAVS WT or CARD mutants , only WT MAVS activated IRF3 after Sendai virus infection and formed high molecular weight particles that sedimented to the bottom layers after sucrose gradient ultracentrifugation ( Figure 5D and Figure 5—figure supplement 1A ) . Importantly , each of the point mutations in MAVS CARD abolished the virus-induced interaction between TRAF6 and MAVS ( Figure 5E ) . The mutations that disrupt MAVS polymerization also abrogated its binding to TRAF2 and TRAF5 ( Figure 5—figure supplement 1B , C ) . These results strongly suggest that MAVS polymerization is important for the recruitment of TRAF6 , TRAF2 , TRAF5 , and likely other signaling molecules . 10 . 7554/eLife . 00785 . 014Figure 5 . Prion-like polymerization of MAVS is required for IRF3 activation and TRAF6 recruitment . ( A ) Sequence alignment of the CARD domain of MAVS from different species . Conserved residues mutated in this study are colored and shaded . ( B ) MAVS WT or CARD mutants were transfected into HEK293-IFNβ-luciferase reporter cells . Cells were lysed 24 hr later , followed by luciferase reporter assay . ( C ) Mavs−/− MEF cells reconstituted with Flag-MAVS WT or CARD mutants were infected with Sendai virus or mock treated for 12 hr , then mitochondrial extracts were separated by SDD-AGE ( top ) or SDS-PAGE ( middle ) followed by immunoblotting with a Flag antibody . Aliquots of the cytosolic extracts were separated by native gel electrophoresis ( bottom ) , followed by immunoblotting with an IRF3 antibody . ( D ) HEK293 cells in which endogenous MAVS was knocked down by shRNA and replaced by RNAi resistant MAVS WT or CARD mutants were infected with Sendai virus for the indicated time . Crude mitochondria were solubilized in a buffer containing 1% DDM and then subjected to sucrose gradient ultracentrifugation . Aliquots of the fractions were immunoblotted with a MAVS antibody . ( E ) MAVS CARD mutants defective in polymerization failed to recruit TRAF6 upon virus infection . Traf6−/− MEF cells were depleted of MAVS by shRNA and reconstituted with Flag-TRAF6 and WT or mutant MAVS as indicated . The cells were infected with Sendai virus for the indicated time , and TRAF6 was then immunoprecipitated with a Flag antibody . Co-immunoprecipitated MAVS was analyzed by immunoblotting . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 01410 . 7554/eLife . 00785 . 015Figure 5—figure supplement 1 . Mutations that disrupt MAVS polymerization abolish viral activation of IRF3 . ( A ) HEK293T cells stably expressing an shRNA against human MAVS were reconstituted with WT or mutant MAVS as indicated ( see also Figure 5D ) . The cells were infected with Sendai virus followed by analysis of endogenous IRF3 dimerization . Aliquots of the cell lysates were immunoblotted with a MAVS antibody . ( B and C ) Mavs−/− MEFs stably expressing Flag-TRAF2 ( B ) or Flag-TRAF5 ( C ) were infected with Sendai virus , then the Flag-tagged protein complexes were immunoprecipitated and analyzed by immunoblotting with the indicated antibodies . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 015 Because the catalytic activity of TRAF6 is important for IRF3 activation in the MAVS pathway , and both K63 and linear polyubiquitination have been suggested to be important for IKK activation ( Deng et al . , 2000; Wang et al . , 2001; Haas et al . , 2009; Tokunaga et al . , 2009; Zeng et al . , 2009 ) , we tested the effect of several ubiquitin mutants in IRF3 activation by MAVSΔTM in the in vitro assay ( Figure 6A ) . We used a tetracycline-inducible RNAi system to knock down endogenous ubiquitin in the human cell line U2OS ( Xu et al . , 2009 ) , and supplemented the cell extracts with different ubiquitin mutants , including K63-only , in which all lysine residues except K63 was substituted with arginine , His6-Ub , which cannot form linear polyubiquitin chains , and K63R . The strongest defect in IRF3 dimerization and IκBα phosphorylation was observed in cell extracts containing K63R , suggesting that K63 polyubiquitination is important for IRF3 and IKK activation by MAVS in vitro . 10 . 7554/eLife . 00785 . 016Figure 6 . Ubiquitination-dependent assembly of a MAVS signaling complex . ( A ) Lys63-linked polyubiquitination is important for MAVS-mediated IRF3 and IKK activation . U2OS cells stably integrated with tetracycline-inducible shRNA against ubiquitin genes were grown in the presence of tetracycline for 48 hr to deplete endogenous ubiquitin . The cytosolic extracts were then supplemented with WT or mutant ubiquitin ( 1 μg ) and 35S-IRF3 in the presence or absence of MAVSΔTM , followed by analyses of IRF3 dimerization and IκBα phosphorylation . K63-only: containing only one lysine at residue 63 of ubiquitin . Ubiquitin WT and mutants were analyzed by immunoblotting with an ubiquitin antibody . ( B ) Purified Flag-NEMO was incubated with HeLa S100 and His6-MAVSΔTM at the indicated temperatures in the presence or absence of vOTU , a viral deubiquitination enzyme . After the reaction , NEMO was immunoprecipitated with Flag antibody , and the co-precipitated proteins were detected with specific antibodies . ( C ) Similar to ( B ) , except that Flag-NEMO WT and UBD mutant were tested for their ability to interact with TRAF2 and MAVS . ( D ) A schematic diagram of TRAF2 protein . The ubiquitination sites identified by mass spectrometry are highlighted . ( E–G ) Traf2/5 DKO MEF cells stably expressing shRNA against TRAF6 were reconstituted with WT TRAF2 or a TRAF2 mutant containing arginine at each of the five ubiquitination sites ( K31/148/195/313/481R , 5KR ) . In some experiments , HOIP was further knocked down by shRNA as indicated . These cells were infected with VSV for the indicated time and cytokine RNA levels were analyzed by q-RT-PCR . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 01610 . 7554/eLife . 00785 . 017Figure 6—figure supplement 1 . SILAC experiments to identify ubiquitin-dependent NEMO-signaling complex both in vitro and in cells . ( A ) Design of the in vitro SILAC experiments . Cytosolic extracts from MEF cells cultured with heavy ( Lys8 , Arg10 ) or light ( Lys0 , Arg0 ) isotopes were incubated with Flag-NEMO protein and ATP in the presence ( A1 ) or absence ( A2 ) of His6-MAVSΔTM . In another set of experiments , the ‘heavy’ extracts were incubated with WT Flag-NEMO ( B1 ) , whereas the ‘light’ extracts were incubated with Flag-NEMO UBD mutant ( UBDm , B2 ) . Both samples were incubated with His6-MAVSΔTM . After incubation at 30°C for 1 hr , NEMO and its associated proteins were immunoprecipitated with a Flag antibody , mixed in pairs ( A1 and A2; B1 and B2 ) , and resolved by SDS-PAGE . The proteins were identified and analyzed by quantitative mass spectrometry . ( B ) Of the 187 proteins identified in both in vitro SILAC experiments as described in ( A ) , the ratios of ‘heavy’ to ‘light’ labeled proteins are plotted as indicated . Proteins known to function in the innate immunity pathways are highlighted in red . NEMO ( Cyan ) and Krt76 ( keratin , Green ) have low H/L ratios because these proteins were not labeled with the heavy isotopes . ( C ) Similar to ( A ) except that Ikbkg−/− ( Nemo−/− ) MEF cells reconstituted with Flag-NEMOΔN or Flag-NEMOΔN-UBDm were labeled with heavy or light isotopes and infected with VSV as indicated . ( D ) Of the 762 proteins identified in both cell-based SILAC experiments , the H/L ratios are plotted as indicated . Proteins that associate with WT NEMO in response to viral infection are shown in the upper right quadrant . ( E ) Venn diagram showing proteins with H/L ratios above 1 . 2 in both in vitro and cell-based SILAC experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 017 NEMO contains two ubiquitin binding domains , a NEMO–ubiquitin binding ( NUB; also known as UBAN ) domain and a C-terminal zinc finger domain ( Ea et al . , 2006; Wu et al . , 2006; Laplantine et al . , 2009 ) . We have shown previously that both of these domains are important for NEMO to mediate IRF3 activation in response to RNA virus infection ( Zeng et al . , 2009 ) . To identify ubiquitination target ( s ) that relays upstream signal to NEMO , we incubated Flag-NEMO with HeLa S100 in the presence of MAVSΔTM . MAVS and TRAF2 were found to co-immunoprecipitate ( Co-IP ) with Flag-NEMO after incubation at 30°C but not at 0°C ( Figure 6B ) . The MAVS-TRAF2-NEMO interaction was blocked when a deubiquitination enzyme , which contains the ovarian tumor type ( OTU ) domain of the Crimean Congo hemorrahagic fever virus , was included in the reaction mixture . Moreover , a NEMO mutant ( UBDm; Y308S/H413A/C417A ) , which contained point mutations in both ubiquitin-binding domains , failed to pull down the TRAF2-MAVS complex ( Figure 6C ) . These results suggest that MAVS and TRAF2 form a MAVS-induced ubiquitination-dependent signaling complex with NEMO in vitro . To further characterize the MAVS signaling complex , we preformed two sets of stable isotope labeling by amino acids in cell culture ( SILAC ) experiments to identify proteins that associate with NEMO in a manner that depends on MAVS signaling and NEMO–ubiquitin binding ( Figure 6—figure supplement 1A , C ) . In the first set ( Figure 6—figure supplement 1A ) , extracts from MEF cells labeled with ‘heavy’ isotopes ( K8R10 ) were stimulated with MAVSΔTM ( A1 and B1 ) in the presence of WT Flag-NEMO , whereas extracts containing ‘light’ labeled proteins ( K0R0 ) were either not stimulated with MAVSΔTM in the presence of WT Flag-NEMO ( A2 ) or stimulated with MAVSΔTM in the presence of ubiquitin-binding defective NEMO mutant ( Flag-NEMO-UBDm: Y308S/H413A/C417A; B2 ) . After immunoprecipitation with the Flag antibody , the precipitates from the ‘heavy’ and ‘light’ samples were combined , fractionated by SDS-PAGE and analyzed by nano liquid chromatography tandem mass spectrometry ( nanoLC-MS/MS ) . Among the proteins enriched in the heavy samples , we found that PLK1 , TRAF2 , and cIAP1 bound to NEMO in a manner that depended on MAVS and UBD of NEMO ( Figure 6—figure supplement 1B ) . PLK1 was previously shown as a negative regulator of MAVS ( Vitour et al . , 2009 ) . All three subunits of the LUBAC complex , HOIP , HOIL-1 , and Sharpin , bound to NEMO in a UBD-dependent manner , but this binding was largely independent of stimulation by MAVS . MAVS itself was not identified as a hit because it was not labeled with the heavy isotopes . In the second set of SILAC experiments ( Figure 6—figure supplement 1C ) , we sought to identify proteins that associate with NEMO in living cells infected with VSV . We used Nemo−/− MEFs reconstituted with Flag-NEMOΔN , which lacks the IKK-binding site but remained capable of stimulating TBK1 , or Flag-NEMOΔN-UBDm , which is defective in ubiquitin binding . The ‘heavy’ cells expressing Flag-NEMOΔN ( A1 and B1 ) and the ‘light’ cells expressing Flag-NEMOΔN-UBDm ( B2 ) were infected with VSV , whereas the ‘light’ cells expressing Flag-NEMOΔN were uninfected ( A2 ) . Extracts from these cells were subjected to immunoprecipitation followed by nanoLC-MS/MS . Among the proteins that associated with NEMO WT , but not NEMO UBDm , after VSV infection ( Figure 6—figure supplement 1D ) , several were known NEMO-interacting proteins , including TRAF6 , TRAF3 , TRAF2 , A20 , and MAVS . In addition , several proteins previously not known to associate with NEMO were identified . The role of these proteins in the MAVS pathway requires further investigation . A comparison of the NEMO interacting proteins identified by the in vitro and cell-based SILAC experiments revealed only two common proteins , MAVS and TRAF2 ( Figure 6—figure supplement 1E ) . At present , we do not know why some proteins such as TRAF6 , TRAF3 , and A20 interacted with NEMO only in the ‘in vivo’ but not the in vitro SILAC , whereas the LUBAC subunits behaved oppositely . One possibility is that these proteins interact with NEMO in a dynamic manner , which is recapitulated differently between the in vitro and ‘in vivo’ experiments . In any case , as TRAF2 was recruited to NEMO in a manner that depended on MAVS , VSV , and UBD of NEMO , we investigated whether ubiquitination of TRAF2 plays a role in MAVS signaling . We incubated Flag-tagged mouse TRAF2 with S100 from Traf2/5 DKO MEFs in the presence of MAVSΔTM , followed by Flag immunoprecipitation under denaturing conditions ( see ‘Materials and methods’ ) . Five ubiquitination sites on TRAF2 were detected by mass spectrometry , including K31 , K148 , K195 , K313 , and K481 as illustrated in Figure 6D . Among these lysines , K31 was identified as an autoubiquitination site of TRAF2 in the TNF pathway ( Li et al . , 2009 ) . We mutated these lysine residues to arginine in TRAF2 ( T2-5KR ) , and stably expressed the mutants as well as the WT TRAF2 in Traf2/5 DKO+shTRAF6 MEFs . Both WT and 5KR TRAF2 rescued the induction of IFNβ , IL6 , and CXCL10 by VSV ( Figure 6E–G ) . Further mutation of another lysine ( K38; to generate 6KR ) also did not impair the function of TRAF2 ( data not shown ) . Interestingly , knockdown of HOIP abolished the expression of these cytokines in cells expressing TRAF2-5KR , but not wild type TRAF2 . These results , together with the results shown in Figure 3 , suggest that ubiquitinated TRAF2 may function redundantly with another ubiquitination target of HOIP . NEMO has been reported to be ubiquitinated at lysine 285 and 309 , and this modification is important for NF-κB activation ( Abbott et al . , 2004; Tokunaga et al . , 2009 ) . We found that Nemo−/− MEF cells stably expressing NEMO K285/309R mutant produced IFNβ and IL6 after VSV infection and that the cytokine levels were similar to those produced by cells expressing wild-type NEMO ( Figure 7A–C ) . In contrast , the cells expressing NEMO-UBDm failed to induce the cytokines after VSV infection . To identify other potential ubiquitination sites on NEMO , we incubated endogenous amount of His6-Flag-NEMO with cytosolic extracts from Nemo−/− MEF cells in the presence of MAVSΔTM . NEMO was purified under denaturing conditions and analyzed by nanoLC-MS/MS ( see ‘Materials and methods’ ) , which detected signal-dependent ubiquitination of NEMO at K285 , K325 , K342 , and K344 ( Figure 7D ) . However , Nemo−/− MEF cells stably expressing NEMO 6KR ( K111/285/309/325/342/344R ) mutant still supported virus-induced cytokine production similar to those expressing wild-type NEMO ( Figure 7E , F ) . We also constructed a NEMO mutant in which 13 conserved lysine residues were substituted with arginine ( 13KR ) and found that this mutant could still support IRF3 dimerization in Nemo-deficient cell extracts in the presence of MAVSΔTM ( Figure 7G , H ) . These results suggest that NEMO ubiquitination per se is dispensable for IRF3 activation in the MAVS pathway . It is likely that ubiquitination of multiple targets as well as unanchored K63 polyubiquitin chains function cooperatively to recruit NEMO and other signaling molecules to the MAVS polymers and promote the activation of IKK and TBK1 ( Figure 8 and ‘Discussion’ ) . 10 . 7554/eLife . 00785 . 018Figure 7 . Mutation of NEMO ubiquitination sites does not impair viral induction of interferons . ( A–C ) Nemo−/− MEF cells stably expressing GFP , Flag-NEMO WT or mutants were infected with VSV for the indicated time . Cytokine RNA levels were measured by q-RT-PCR ( A and B ) . Expression of the NEMO proteins was analyzed by immunoblotting ( C ) . ( D ) A schematic diagram of human NEMO and the ubiquitination sites identified by mass spectrometry . ( E and F ) Nemo−/− MEF cells stably expressing GFP , Flag-NEMO WT , or 6KR ( K111/285/309/325/342/344R ) were infected with VSV for the indicated time . Cytokine RNA levels were analyzed by q-RT-PCR . ( G and H ) Nemo−/− MEF cell extracts ( S5 ) were supplemented with Flag-NEMO WT or 13KR ( K111/139/143/165/283/285/321/325/326/342/344/399R ) protein and incubated with 35S-IRF3 and His6-MAVSΔTM . IRF3 dimerization was detected by native gel electrophoresis followed by autoradiography ( G ) . Aliquots of the NEMO proteins were analyzed by immunoblotting ( H ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 01810 . 7554/eLife . 00785 . 019Figure 8 . A Model for MAVS-mediated IRF3 and NF-κB activation . Upon virus infection , MAVS undergoes prion-like polymerization to recruit and activate E3 ligases TRAF2 , TRAF5 , and TRAF6 ( possibly also LUBAC ) . These E3 ligases in turn synthesize polyubiquitin chains on TRAF2 and other proteins , resulting in the recruitment of NEMO through its ubiquitin-binding domains . NEMO then recruits IKK and TBK1 complexes to the MAVS polymer , where the kinases phosphorylate IκBα and IRF3 , respectively , leading to the induction of type-I interferons and other cytokines . DOI: http://dx . doi . org/10 . 7554/eLife . 00785 . 019
Although it is known that MAVS contains binding sites for several TRAF proteins ( Seth et al . , 2005; Xu et al . , 2005 ) , the role of these TRAF proteins in MAVS signaling has been enigmatic because cells lacking an individual TRAF protein , including TRAF2 , TRAF3 , TRAF5 , and TRAF6 , could still induce IFNβ normally in response to virus infection ( Seth et al . , 2005; Konno et al . , 2009; Zeng et al . , 2009 ) . In this report , we demonstrated that compound deletion of TRAF6 , TRAF2 , and TRAF5 completely abolished IFNβ induction by Sendai virus and VSV , suggesting that these proteins function redundantly in vivo to activate the downstream signaling cascades after they are recruited to the cognate binding sites on MAVS . Such redundant function of the TRAF proteins was not recapitulated in our cell free assay , in which IRF3 activation by MAVS was strongly inhibited in Traf6-deficient extracts , and largely abolished in the absence of TRAF2 and TRAF5 . It is possible that certain aspects of cell signaling in intact cells , such as the potential involvement of cellular structures ( e . g . , membrane organelles ) and transcriptional amplification , are not recapitulated in the in vitro system that employs soluble cytosolic extracts . Nevertheless , the in vitro assay recapitulates many aspects of MAVS signaling , such as its dependency on MAVS and NEMO . In the assay that involves endogenous MAVS , IRF3 activation occurs only when cytosolic extracts are incubated with mitochondria from virus-infected cells . Thus , the in vitro assays that we have established are valuable tools to dissect the biochemical mechanisms of MAVS signaling . Our finding that MAVS recruits multiple ubiquitin E3 ligases and other signaling proteins to activate IKK and TBK1 further underscores the importance of combining in vitro biochemical approach with cell-based assays to gain a mechanistic understanding of these complex signaling pathways . Currently , the prevailing view is that TRAF2 , TRAF5 , and TRAF6 mediate the activation of IKK and NF-κB , whereas TRAF3 is important for the activation of TBK1 and IRF3 . However , here we showed that TRAF2 , TRAF5 , and TRAF6 are essential for the activation of both IKK and TBK1 by MAVS , whereas TRAF3 is dispensable . Paz et al . have recently shown that the TRAF6-binding site ( 455-PEENEY-460 ) of MAVS also binds to TRAF3 and that this binding is important for MAVS signaling ( Paz et al . , 2011 ) . Another recent report confirmed that the PEENEY motif ( residues 455–460 ) of MAVS is important for IFNβ induction when an intermediate domain of MAVS is fused to an artificial oligomerization domain ( Takamatsu et al . , 2013 ) . However , this study did not directly investigate whether TRAF6 or TRAF3 is important for MAVS signaling . Our data showed that in cells expressing a MAVS mutant lacking the TRAF2/5 binding sites ( Q145N ) , activation of IRF3 and production of IFNβ were completely abolished in cells depleted of TRAF6 , indicating that other TRAF proteins , including TRAF3 , could not substitute the function of TRAF6 . It is possible that the role of TRAF3 in IRF3 activation depends on the cell types and specific signaling pathways . Alternatively , TRAF3 may regulate type-I interferon production through a mechanism that does not affect IRF3 activation . Our finding that TRAF2 , TRAF5 , and TRAF6 play an essential role in both NF-κB and IRF3 activation by MAVS is surprising because these TRAF proteins are also recruited to other receptors such as TNF receptor , TLRs , IL-1 receptor , and CD40 , which only stimulate NF-κB but not IRF3 . A possible explanation for these observations is that IRF3 activation requires certain adaptor proteins such as MAVS , STING , and TRIF . Indeed , we have recently shown that STING simultaneously binds to TBK1 and IRF3 , thereby specifying IRF3 phosphorylation by TBK1 ( Tanaka and Chen , 2012 ) . This mechanism explains why several innate immunity pathways , such as the TLR and IL-1 pathways , could lead to activation of TBK1 but not IRF3 ( Clark et al . , 2011 ) . Our model that MAVS signals through multiple TRAF proteins is further supported by mutagenesis experiments which showed that mutations of both TRAF2/5 and TRAF6 binding sites of MAVS , but not each alone , abolished IRF3 and IKK activation by virus infection . Importantly , MAVS binds to the TRAF proteins in a manner that depends on virus infection and MAVS polymerization . Mutations in the CARD domain of MAVS that disrupt its polymerization also abolish the recruitment of the TRAF proteins . These results explain why MAVS undergoes prion-like polymerization in response to viral infections . Presumably , each TRAF binding motif of MAVS has a low affinity for the TRAF proteins , but the polymerization of MAVS significantly increases the avidity of the interaction , resulting in the recruitment of the TRAF proteins . The recruitment of TRAFs to the MAVS polymers may lead to the oligomerization of TRAFs , which activates their ubiquitin E3 ligase activity ( Wang et al . , 2001; Sun et al . , 2004 ) . The E3 ligase activity of TRAF6 is likely important for IRF3 activation because mutation of the RING domain in TRAF6 abolishes its activity . Interestingly , the RING domain of TRAF2 is important for IRF3 activation only when the catalytic activity of HOIP is abrogated , suggesting that TRAF2 and HOIP may have redundant functions . HOIP is dispensable for IRF3 activation in cells expressing TRAF2 , TRAF5 , and TRAF6 . Conversely , the loss of TRAF2 , TRAF5 , and TRAF6 completely abolishes IRF3 activation in cells expressing other E3s such as LUBAC , cIAPs , and TRAF3 . These results indicate that TRAF2 , TRAF5 , and TRAF6 are the major E3s that mediate the activation of IKK and TBK1 downstream of MAVS . How do TRAF2 , TRAF5 , and TRAF6 activate IKK and TBK1 ? Both IKK and TBK1 employ NEMO as an essential regulatory subunit for their activation . NEMO contains two ubiquitin-binding domains , which are essential for the activation of IKK and TBK1 in response to virus infection . Although we could detect ubiquitination of NEMO at multiple lysines , mutations of these lysines to arginine did not impair the ability of NEMO to support IRF3 activation by MAVS . In fact , a NEMO mutant containing mutations at 13 conserved lysine residues ( 13KR ) could still activate IRF3 in the presence of MAVS ( Figure 7G , H ) . We also found that TRAF2 is recruited to NEMO in a manner that depends on ubiquitination and MAVS . TRAF2 is also ubiquitinated at multiple lysine residues . However , mutations of these residues to arginine did not impair its ability to induce cytokines in response to VSV infection unless HOIP was also depleted ( Figure 6E–G ) . Thus , HOIP may have an additional ubiquitination target that functions redundantly with TRAF2 ubiquitination to mediate IRF3 activation . An emerging theme from this work is that multiple ubiquitin E3s are recruited to MAVS , and each E3 targets ubiquitination of distinct proteins , which function cooperatively and redundantly to activate IKK and TBK1 . A similar model has recently been proposed for the regulation of DNA repair by sumoylation ( Psakhye and Jentsch , 2012 ) . It was shown that sumoylation of a group of DNA repair proteins ( ‘protein group’ ) , rather than a specific protein target , is responsible for the repair of DNA double-strand breaks .
Rabbit antibodies against human IRF3 , TRAF2 , TRAF3 , TRAF6 , NEMO , and mouse antibodies against human MAVS and ubiquitin were obtained from Santa Cruz Biotechnology , Dallas , TX; Rabbit antibody against human HOIP was from Abcam , Cambridge , MA; Flag antibody ( M2 ) , M2-conjugated agarose , anti-HA-conjugated agarose , and tubulin antibody were purchased from Sigma-Aldrich , St . Louis , MO; HA antibody was from Covance , Princeton , NJ; antibodies against pIRF3 Ser396 , pTBK1 Ser172 , pIκBα Ser32/36 , and pSTAT-1 Tyr701 were from Cell Signaling , Danvers , MA; mouse IRF3 antibody was from Invitrogen , Carlsbad , CA; pan cIAP antibody was from R&D Systems , Minneapolis , MN; TBK1 antibody was from IMGENEX Corp , San Diego , CA; TANK antibody was from BioVision , Milpitas , CA . The antibody against mouse Sharpin was generated by immunizing rabbits with the full-length mouse Sharpin produced in Sf9 cells . The antibody against mouse HOIL-1 was kindly provided by Dr Kazuhiro Iwai ( Kyoto University ) . Rabbit antibodies against human and mouse MAVS were generated as described before ( Seth et al . , 2005; Sun et al . , 2006 ) . For expression in mammalian cells , mouse cDNA encoding N-terminal Flag or HA tagged TRAF6 WT , TRAF6 C70A , TRAF6 K0 , TRAF2 WT , TRAF2 C34A , TRAF2ΔR ( Δ34-72 ) , TRAF2ΔC ( 1-359 ) , TRAF2 K31R , TRAF2 5KR ( K31/148/195/313/481R ) , TRAF2 6KR ( K31/38/148/195/313/481R ) , and TRAF5 WT were cloned into pcDNA3 . Human cDNA encoding N-terminal Flag or HA-tagged NEMO WT , K285/309R , 6KR ( K111/285/309/325/342/344R ) , UBDm ( Y308S/H413A/C417A ) , and ΔN ( 86-419 ) were cloned into pcDNA3 and pTY-EF1A-puroR-2a lenti-viral vectors . Human MAVS WT , QN ( Q145N ) , 2ED ( E155D and E457D ) , and QN2ED ( Q145N , E155D and E457D ) were cloned into pTY-EF1A-puroR-2a lenti-viral vector . Mouse HA-tagged HOIP WT and CS ( C693S/C696S ) mutant were cloned into pTY-EF1A-hygromycinR lentiviral vector ( see below ) . Mutants were constructed with the QuikChange Site-Directed Mutagenesis Kit ( Stratagene , La Jolla , CA ) . Flag or HA-tagged TRAF proteins were overexpressed in HEK293T cells , whereas Flag or HA-tagged NEMO proteins were overexpressed in Nemo−/− MEF cells . Proteins were purified with M2 or anti-HA agarose , followed by Flag or HA peptide elution . For expression in Escherichia coli , cDNA encoding N-terminal His6-Flag-tagged NEMO WT or K285/309R was inserted into pET23a; cDNA encoding N-terminal His6-tagged ubiquitin WT , K63R , K63 only , MAVSΔTM ( aa1–510 or 1–460 ) was also cloned into pET23a . Vectors encoding ubiquitin mutants were transformed and expressed in E . coli BL21 ( DE3 ) -pJY2 strain to prevent misincorporation of Lys residues . Other vectors were transformed and expressed in E . coli BL21 ( DE3 ) -pLysS strain . His6-MAVSΔTM ( 1–510 or 1–460 ) was purified in the presence of 4 M urea , whereas other His6-tagged proteins were purified under native condition , by nickel affinity chromatography . Ubc5c , His6-TRAF6 , His8-E1 , and His8-IRF3 were purified as described previously ( Zeng et al . , 2009; Hou et al . , 2011 ) . Nontagged ubiquitin and mutants were from Boston Biochem , Cambridge , MA . Sendai virus ( Cantell strain; Charles River Laboratories ) was used at a final concentration of 100 hemagglutinating unit/ml . VSV ( ΔM51 ) -GFP virus was from Dr John Bell ( University of Ottawa ) and propagated in Vero cells . All cells were cultured at 37°C in an atmosphere of 5% ( vol/vol ) CO2 . HEK 293T cells were cultured in Dulbecco’s modified Eagle’s medium ( DMEM ) supplemented with 10% ( vol/vol ) cosmic calf serum ( Hyclone , Thermo Fisher Scientific , Waltham , MA ) with penicillin ( 100 U/ml ) and streptomycin ( 100 μg/ml ) . Traf2/5 DKO MEF cells and derivatives were cultured in DMEM supplemented with 20% ( vol/vol ) fetal bovine serum ( Invitrogen ) and antibiotics . Other MEF cells were cultured in DMEM supplemented with 10% ( vol/vol ) fetal bovine serum ( Atlanta ) and antibiotics . Mavs−/− MEFs were immortalized from Mavs−/− mice generated in Sun et al . 2006 . Primary and immortalized Traf6−/− MEFs were generated from Traf6−/− mice provided by Dr Tak Mak ( University of Toronto ) . Traf2−/− MEFs , Traf5−/− MEFs and Traf2−/−/Traf5−/− MEFs were kindly provided by Dr Hiroyasu Nakano ( Juntendo University School of Medicine ) . Traf3−/− MEFs were gifts from Dr Genhong Cheng ( University of California , Los Angeles ) . Nemo-deficient MEFs and IKKα/IKKβ-deficient MEFs were kindly provided by Dr Inder Verma ( Salk Institute ) ; these cells are a complete null for the Ikbkg , Ikbkα and Ikbkβ loci , respectively . Sharpin-deficient MEFs were immortalized from Sharpincpdm mice purchased from the Jackson Laboratory ( Stock number: 007599 ) . Nemo−/− MEFs and MEFs stably expressing Flag-NEMOΔN were lysed in hypotonic buffer ( 10 mM Tris–HCl ( pH 7 . 5 ) , 10 mM KCl , 1 . 5 mM MgCl2 , and a protease inhibitor cocktail [Roche] ) . After centrifugation at 100 , 000×g for 30 min , the supernatants from both types of cells were mixed at a ratio of 5:1 , and the mixture was subjected to immunoprecipitation with M2 agarose at 4°C overnight . The agarose beads were washed three times with buffer B ( 10 mM Tris–HCl [pH 7 . 5] , 1 M NaCl , and 0 . 1% CHAPS ) , and the proteins were eluted with Flag peptide ( 0 . 2 mg/ml ) in buffer C ( 50 mM Tris–HCl [pH 7 . 5] , 0 . 1% CHAPS ) . The eluted proteins containing endogenous TBK1 and TANK from MEFs , designated as NEMOΔN PD were stored in buffer D ( 20 mM Tris–HCl [pH 7 . 5] , 50 mM NaCl , 10% glycerol , and 0 . 1% CHAPS ) after buffer exchange by repeated dilution and concentration . In vitro assays for IRF3 activation and IκBα phosphorylation were preformed as described previously ( Zeng et al . , 2009 ) , except that 20 ng of His6-MAVSΔTM was used to replace crude mitochondria ( P5 ) from virus-infected cells in some experiments . In the assays for purification of IRF3 activators , the reaction mixture ( 10 μl ) contained 100 ng His8-E1 , 50 ng Ubc5c , 5 μg ubiquitin , 20 ng His8-IRF3 , and 1 μl of NEMOΔN PD . To determine the NEMO–MAVS complex formation in vitro , reaction mixture ( 100 μl ) containing 20 mM HEPES-KOH ( pH 7 . 0 ) , 2 mM ATP , 5 mM MgCl2 , 200 ng His6-MAVSΔTM , 200 μg cytosolic extracts ( S5 or S100 ) , 200 ng Flag-NEMO was incubated at 30°C for 1 hr . Immunoprecipitation was then carried out using Flag antibody ( M2 ) agarose at 4°C overnight in the presence of 20 mM Tris–HCl ( PH 7 . 5 ) , 100 mM NaCl , 0 . 5% NP-40 , and the protease inhibitor cocktail ( Roche ) . The agarose beads were washed three times with lysis buffer A ( 20 mM Tris–HCl [PH 7 . 5] , 100 mM NaCl , 10% glycerol , and 0 . 5% NP-40 ) , and coprecipitated proteins detected by immunoblotting . Protein purification was carried out using AKTA-FPLC or ETTAN system ( GE Healthcare , Little Chalfont , UK ) at 4°C . Chromatographic columns and media were purchased from GE Healthcare unless indicated otherwise . HeLa S100 was prepared as described previously from 50 L of cells purchased from National Cell culture Center ( Deng et al . , 2000 ) . S100 was loaded onto a 60-ml Q-Sepharose column equilibrated with buffer Q-A ( 20 mM Tris–HCl [pH 7 . 5] , 10% glycerol , and 0 . 02% CHAPS ) , and eluted with 0 . 25 M NaCl . The eluate was diluted in buffer SP-A ( 20 mM HEPES-KOH [pH 6 . 5] , 10% glycerol , and 0 . 02% CHAPS ) and concentrated repeatedly to reduce the salt before loading to SP-Sepharose . The flow-through from the SP column was further fractionated on a Heparin-Sepharose column with a linear gradient of NaCl ( 0–300 mM ) in buffer SP-A , and active fractions eluted around 150 mM NaCl were pooled . After the salt was reduced by repeated dilution in buffer SP-A , the sample was fractionated on 2 ml ceramic hydroxyapatite ( CHT ) column ( Bio-Rad Laboratories Inc . , Hercules , CA ) with a linear gradient of KH2PO4–K2HPO4 ( 0–300 mM ) in buffer CHT-A ( 5 mM KH2PO4–K2HPO4 , pH 7 . 0 , 150 mM NaCl ) . Active fractions eluted around 150 mM PO4 were pooled and precipitated with 30% ammonium sulfate . The pellet was resuspended with buffer Q-A and fractioned on a 2 . 4-ml Superdex 200 PC 3 . 2/30 column in buffer Q-A containing 100 mM NaCl . Active fractions containing proteins with a size of ∼300 KD were loaded onto MonoQ-Sepharose and eluted with a linear gradient of NaCl ( 150–350 mM ) in buffer Q-A . The fractions were resolved by SDS-PAGE , silver stained , and identified by tandem mass spectrometry using LTQ-XL ( Thermo ) . The lentiviral shRNA vector , pTY-shRNA-EF1a-puroR-2a-GFP-Flag , was provided by Dr Yi Zhang ( Harvard Medical School ) . This vector was modified to pTY-shRNA-EF1a-hygroR-GFP and pTY-shRNA-EF1a-zeroR-GFP by replacing the puromycin resistant gene with hygromycin- and zeocin-resistant genes , respectively , such that multiple lentiviral vectors can be introduced to the same cell line . The original vector was also modified to pTY-shRNA-EF1a-GFP-IRES-puroR to circumvent the problem of incomplete cleavage by the 2A protease . The shRNA sequences were cloned into the vectors with U6 promoter . RNAi-resistant cDNA sequences were cloned into the vectors to replace GFP . Lentiviral infection and establishment of stable cells were described previously ( Tanaka and Chen , 2012 ) . The shRNA sequences are as follows ( only the sense strand is shown ) : mouse TRAF6 , 5′-GGATGATACATTACTAGTG-3′; mouse HOIP , 5′-GGCGCTCAGTGAAGTTTAA-3′; mouse MAVS , 5′-GATCAAGTGACTCGAGTTT-3′; human MAVS , 5′-GGAGAGAATTCAGAGCAAG-3′; mouse TRAF3 , 5′-GAATGAAAGTGTTGAGAAA-3′; mouse HOIL-1 , 5′-GGAGAAAGCCCGAGCTGTA-3′ ( 3′UTR ) ; mouse Sharpin , 5′-GCTACATACAAGCTAGTAA-3′ ( 3′UTR ) ; human HOIP , 5′-CCTAGAACCTGATCTTGCA-3′ . MAVS aggregation induced by Sendai virus infection was analyzed by sucrose gradient ultracentrifugation as previously described ( Hou et al . , 2011 ) . Briefly , crude mitochondria fraction was collected from the cells , loaded onto the top of 20–60% sucrose , and centrifuged at 170 , 000×g for 2 hr . Immunoblotting was used to analyze the distribution of proteins along the sucrose gradient . The formation of prion-like aggregates of MAVS was also analyzed by semidenaturing detergent agarose gel electrophoresis ( SDD-AGE; see Hou et al . , 2011 ) . To examine the recruitment of TRAF proteins to MAVS , cells were lysed with buffer containing 20 mM Tris–HCl ( pH 7 . 5 ) , 150 mM NaCl , 0 . 5% NP-40 , and protease inhibitor cocktail ( Roche , Basel , Switzerland ) . Flag antibody ( M2 ) agarose was added to the cell lysates to immunoprecipitate TRAF proteins at 4°C overnight . The agarose beads were washed for five times with lysis buffer before being boiled in SDS loading buffer . Co-immunoprecipitated MAVS was then analyzed by immunoblotting . Total RNA was isolated using TRIzol ( Invitrogen ) . 0 . 1 μg total RNA was reverse-transcribed into cDNA using iScript Kit ( Bio-Rad ) . The resulting cDNA served as the template for Quantitative-PCR analysis using iTaq Universal SYBR Green Supermix ( Bio-Rad ) and ViiTM7 Real-Time PCR System ( Applied Biosystems Inc . , Foster City , CA ) . IQ SYBR Green Supermix ( Bio-Rad ) and iQ5 real-time PCR detection system ( Bio-Rad ) were used for some of the experiments . Primers for specific genes are listed as follows: Mouse β-actin , 5′-TGACGTTGACATCCGTAAAGACC-3′ and 5′-AAGGGTGTAAAACGCAGCTCA-3′; Mouse IFNβ , 5′-CCCTATGGAGATGACGGAGA-3′ and 5′-CTGTCTGCTGGTGGAGTTCA-3′; Mouse IFNα , 5′-ATTTTGGATTCCCCTTGGAG-3′ and 5′-TATGTCCTCACAGCCAGCAG-3′; Mouse CxCl-10 , 5′-GGTCTGAGTGGGACTCAAGG-3′ and 5′-GTGGCAATGATCTCAACACG-3′; Mouse IL-6 . 5′-TCCATCCAGTTGCCTTCTTG-3′ and 5′-GGTCTGTTGGGAGTGGTATC-3′; Mouse TRAF3 , 5′-AGCAGCTGACTCTGGGACAT-3′ and 5′-CACCACACAGGGACAATCTG-3′; Mouse TRAF6 , 5′-GCCCAGGCTGTTCATAATGT-3′ and 5′-CGGATCTGATGGTCCTGTCT-3′; Mouse HOIP , 5′-TTATGCGAGACCCCAAGTTC-3′ and 5′-GCCTTGAGCCTGGTACTCTG-3′ . human IFNβ , 5′-ACTGCAACCTTTCGAAGCCTTT-3′ and 5′-TGGAGAAGCACAACAGGAGAGC-3′; human GAPDH , 5′-ATGACATCAAGAAGGTGGTG-3′ and 5′-CATACCAGGAAATGAGCTTG-3′ . To determine modification ( s ) on TRAF2 induced by MAVSΔTM in cell extracts , reaction mixture ( 10 ml ) containing 20 mM HEPES-KOH ( pH 7 . 0 ) , 2 mM ATP , 5 mM MgCl2 , 20 μg His6-MAVSΔTM , 20 mg Traf2/5 DKO cytosolic extracts ( S100 ) , and 20 μg Flag-TRAF2 was incubated at 30°C for 1 hr , followed by addition of 0 . 7% SDS to terminate the reaction . The reaction mixture was then diluted by adding 60 ml PBS , and immunoprecipitation was carried out using Flag antibody ( M2 ) agarose at 4°C overnight in the presence of 0 . 5% NP-40 . The agarose beads were washed three times with PBS containing 0 . 5% NP-40 , and the proteins were eluted with Flag peptide before SDS-PAGE and silver staining . To determine NEMO modification ( s ) induced by MAVSΔTM in cell extracts , reaction mixture ( 5 ml ) containing 20 mM HEPES-KOH ( pH 7 . 0 ) , 2 mM ATP , 5 mM MgCl2 , 10 μg His6-MAVSΔTM , 10 mg Nemo−/− cytosolic extracts ( S100 ) , and 10 μg His6-Flag-NEMO was incubated at 30°C for 1 hr , followed by addition of 1% SDS to terminate the reaction . The reaction mixture was then diluted by adding 45 ml PBS and immunoprecipitation was carried out using Flag antibody ( M2 ) agarose at 4°C overnight in the presence of 0 . 5% NP-40 . The agarose beads were washed three times with PBS and 0 . 5% NP-40 , and proteins were eluted with a buffer containing 20 mM Tris–HCl ( PH 8 . 0 ) , 100 mM NaCl , and 8M urea . The eluate was then incubated with Ni-NTA agarose at 4°C overnight . After washing three times with a buffer containing 20 mM Tris–HCl ( pH 8 . 0 ) , 100 mM NaCl , and 10 mM imidazole , the agarose beads were boiled in the SDS loading buffer before SDS-PAGE and silver staining . For both TRAF2 and NEMO , gel slices from each lane were excised and digested with trypsin in situ . Digested samples were subjected to mass spectrometry using Q Exactive ( Thermo Scientific ) , and raw data were analyzed by mascot search engine ( MATRIX SCIENCE ) . Wild-type MEF and Nemo−/− MEF expressing Flag-NEMOΔN WT or UBDm were cultured in SILAC-DMEM medium lacking lysine and arginine . The medium was supplemented with dialyzed FBS , penicillin , streptomycin , and amino acids l-lysine and l-arginine . The ‘light’ culture was supplemented with Lys0 ( 12C614N2 ) and Arg0 ( 12C614N4 ) , and the ‘heavy’ culture with Lys8 ( 13C615N2 ) and Arg10 ( 13C615N4 ) . All SILAC reagents were purchased from Pierce ( Thermo Scientific ) . The in vitro and cell-based SILAC experiments were preformed according to the outline described in Figure 6—figure supplement 1A , C , respectively . After SDS-PAGE and silver staining , 10 to 12 gel slices from each lane were excised and digested with trypsin in situ . Extracted peptides were fractionated on a homemade analytical column ( 75 μm ID , 100 mm in length ) packed with C18 resin ( 100 Å , 3 μm , MICHROM Bioresources ) using Dionex Ultimate 3000 nanoLC system ( Thermo Scientific ) . The column was coupled in-line to a Q Exactive mass spectrometer ( Thermo Scientific ) equipped with a nano-electrospray ion source , which was set at a spray voltage of 2 . 3 kV . Peptides were eluted with a 78 min gradient as follows: 2–30% B in 68 min , 30–35% B in 4 min , 35–40% B in 2 min , 40–60% B in 3 min , and 60–80% B in 1 min ( A = 0 . 1% formic acid; B = 100% acetonitrile in 0 . 1% formic acid ) . Full scan mass spectra were acquired from m/z 300 to 1500 with a resolution of 70 , 000 at m/z = 200 in the Orbitrap . MS/MS spectra ( resolution: 17 , 500 at m/z = 200 ) were acquired in a data-dependent mode , whereby the top 10 most abundant parent ions were subjected to further fragmentation by higher energy collision dissociation ( HCD ) . SILAC data were processed using MaxQuant computational platform ( Cox and Mann , 2008 ) version 1 . 3 . 0 . 5 which incorporates the Andromeda search engine ( Cox et al . , 2011 ) . Proteins were identified by searching the mouse UniProt database supplemented with frequently observed contaminants . The first search tolerance was set at 20 ppm , and main search deviation at 6 ppm . The required minimum peptide length was six amino acids . The false discovery rate ( FDR ) at both peptide and protein levels was set to 0 . 01 . SILAC quantification of each protein group was based on at least two ratio counts . | The innate immune system can detect and destroy viruses , bacteria and other pathogens that enter the human body . In particular , inside cells , viral RNA can bind to and activate a protein called RIG-I . This protein switches on another protein , called MAVS , which can activate other copies of itself . These MAVS molecules then aggregate together on the membrane of mitochondria and send a signal that leads to the production of small proteins , called cytokines , which stimulate an inflammatory response and ultimately neutralize the virus . Although many of the proteins that are activated by MAVS in the innate immunity signaling pathway have been identified , precisely how MAVS transmits this signal is unknown . Now , Liu et al . explore how this protein can propagate signals in the innate immune response by monitoring activation of the transcription factors IRF3 and NF-κB , which transcribe cytokine genes . Previous studies have suggested that a protein known as ubiquitin is needed to activate RIG-I , and that this protein collaborates with MAVS to signal through the innate immunity pathway . Liu et al . found that a group of proteins including TRAF2 , TRAF5 , TRAF6 and LUBAC relay the antiviral signal by binding to MAVS . These so-called ‘E3 ligases’ string ubiquitin together in chains called polyubiquitin , which is essential for activating signaling after , or downstream of , MAVS; however , the association of these E3 ligases with MAVS also requires that multiple copies of MAVS cluster together . MAVS , the TRAF proteins and LUBAC collectively recruit other innate immunity pathway proteins to activate IRF3 and NF-κB , and thus transcription of the genes that control the innate immunity response . Together , these results show the intricate interplay of proteins needed to eliminate viruses from the body . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"immunology",
"and",
"inflammation"
] | 2013 | MAVS recruits multiple ubiquitin E3 ligases to activate antiviral signaling cascades |
In mammals , helping is preferentially provided to members of one’s own group . Yet , it remains unclear how social experience shapes pro-social motivation . We found that rats helped trapped strangers by releasing them from a restrainer , just as they did cagemates . However , rats did not help strangers of a different strain , unless previously housed with the trapped rat . Moreover , pair-housing with one rat of a different strain prompted rats to help strangers of that strain , evidence that rats expand pro-social motivation from one individual to phenotypically similar others . To test if genetic relatedness alone can motivate helping , rats were fostered from birth with another strain and were not exposed to their own strain . As adults , fostered rats helped strangers of the fostering strain but not rats of their own strain . Thus , strain familiarity , even to one’s own strain , is required for the expression of pro-social behavior .
Pro-social behavior comprises actions that improve the well-being of others ( Eisenberg et al . , 1989 ) and is found widely in the animal world ( Owens and Owens , 1984; Wilkinson , 1984; Lee , 1987; Heinrich and Marzluff , 1995; Warneken and Tomasello , 2006; Nowbahari et al . , 2009; Yamamoto et al . , 2012; Baden et al . , 2013; Clay and de Waal , 2013; Hatchwell et al . , 2013 ) . Pro-sociality within a group promotes individual survival and reproduction ( Hamilton , 1984; Preston and de Waal , 2002; Decety and Svetlova , 2012 ) . On the other hand , acting pro-socially towards individuals from other groups may not be adaptive , as other groups often compete for valuable and limited resources . Thus animals can be motivated to act pro-socially or aggressively depending on the social context . In humans , pro-social behavior is modulated by the degree of affiliation and is extended preferentially towards in-group members and less often toward unaffiliated others ( Hornstein , 1978; Cialdini et al . , 1997; Preston and de Waal , 2002; Levine et al . , 2005; Sturmer et al . , 2006; Lamm et al . , 2010; Echols and Correll , 2012 ) . Yet humans can and often do act pro-socially towards strangers ( Batson et al . , 2005 ) . This human capacity to help unfamiliar individuals is often viewed as a cognitively complex behavior that depends on high cognitive capacities and cultural transmission ( Levine et al . , 2001 ) . Rodents have emerged as a valuable model system for social behavior and communication ( Decety and Svetlova , 2012; Mogil , 2012; Panksepp and Panksepp , 2013; Preston , 2013 ) . Rodents manifest emotional contagion ( Langford et al . , 2006; Chen et al . , 2009; Jeon et al . , 2010; Knapska et al . , 2010; Panksepp and Lahvis , 2011; Akyazi and Eraslan , 2014; Atsak et al . , 2011 ) , cooperation ( Rutte and Taborsky , 2007; Viana et al . , 2010; Tsoory et al . , 2012 ) , and helping ( Ben-Ami Bartal et al . , 2011; Church , 1959; Rice and Gainer , 1962 ) . We previously found that Sprague-Dawley ( SD ) rats learn to release SD cagemates trapped in a restrainer ( Ben-Ami Bartal et al . , 2011 ) . Here , we set out to investigate how previous social experience acquired during development and adulthood influences helping behavior . First , we tested whether rats will help unfamiliar individuals , strangers . Rats showed helping behavior equally towards cagemates and strangers , if the strangers were of their own strain . However , rats did not help strangers of an unfamiliar strain , suggesting that helping in rats may be innately biased towards the helper’s own strain . Yet , further experiments demonstrated that a short period of pair-housing with a rat of the unfamiliar strain was sufficient to motivate helping for that individual . Moreover , rats that had previously lived with a rat of a different strain were as motivated to help strangers of that strain as they were to help strangers of their own strain . This finding demonstrates that rats choose to help others depending on the social context , and extend pro-social motivation beyond individual identity , to groups defined by strain . Finally , we sought to determine if genetic relatedness is at all capable of influencing pro-social motivation towards an unfamiliar rat . We found that rats that were fostered from birth with another strain did not help strangers of their own strain as adults . Fostered rats only acted pro-socially towards rats of the foster strain , demonstrating that genetic relatedness alone is not capable of producing pro-social motivation . In each hour-long session , a free rat of the SD strain was placed in an arena containing another rat trapped inside a centrally located restrainer . ( A rodent stock , colloquially referred to as an outbred strain , is a colony of conspecifics derived from a small group of founder animals . Individuals are not genetically identical . Here we refer to stocks as strains . ) The door to the restrainer could only be opened from the outside and thus only by the free rat . Door-opening by the free rat led to the trapped rat’s release from the restrainer . If a free rat failed to open the door within 40 min , the experimenter opened it halfway , allowing the trapped rat to exit the restrainer . Only door-openings resulting from the free rat’s action , during the first 40 min of each session , were counted as such . Regardless of which rat opened the restrainer door and whether that occurred before or after halfway opening , both rats remained in the arena for the entire hour of every session . All rats were males and all free rats were from the SD strain . Sessions were repeated for 12 days .
To determine whether rats help unfamiliar individuals , free SD rats were placed with trapped SD rats that were either strangers ( n = 12 , SD stranger condition ) or cagemates ( n = 8 , SD cagemate condition ) in the helping behavior test described above . Cagemates were pair-housed for 2 weeks prior to the experiment . Strangers were SD males from a different cohort , born in the same week but on a different day as the test rats . On each day of testing , a different stranger , to whom the free rat had no prior exposure , was trapped in the restrainer . Most rats in both SD cagemate ( 6/8 , 75% ) and SD stranger ( 10/12 , 83% ) conditions acted pro-socially , learning to release the trapped rat , and becoming openers ( Figure 1; see Methods for opener definition ) . Rats in the two conditions were similarly active as measured by velocity ( two-tailed Student’s t test , p>0 . 05 ) , spent similar amount of time near the trapped rat ( two-tailed Student’s t test , p>0 . 05 ) , and began to open the door on around the same day ( 4 . 0 ± 1 . 1 days for trapped cagemates; 3 . 7 ± 0 . 8 for trapped strangers , mean ± SEM ) . Thus , rats were as motivated to help strangers as they were to help cagemates , showing that individual familiarity is not required for pro-social behavior in rats . 10 . 7554/eLife . 01385 . 003Figure 1 . Rats were as motivated to help strangers of the same strain ( SD stranger; right ) as they were to help cagemates ( SD cagemate; left ) . In both experimental conditions ( diagrammed at top ) , SD rats were housed with another SD rat . However in the cagemate condition , the cagemate served as the trapped rat ( ‘a’ ) whereas in the stranger condition , the trapped rat was a stranger ( ‘ ? ’ ) and not the cagemate ( ‘b’ ) . Across the days of testing , the median latency to door-opening ( bottom ) decreased for both conditions . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 003 The results above indicate that rats are motivated to help unfamiliar rats of their own strain . To determine if rats extend help to rats from a different strain , free SD rats were tested in the helping behavior test with trapped rats of the black-caped Long-Evans ( LE ) strain , which were either cagemates ( n = 7 , LE cagemate condition ) or strangers ( n = 16 , LE stranger condition ) . SD rats did not release LE strangers with only a minority becoming openers ( 4/16 , 25% openers; Figure 2; Video 1 ) . In contrast , most SD rats in the LE cagemate condition became openers ( 5/7 , 71%; χ2 , p=0 . 04 ) , learning to open the restrainer on average on day 3 . 8 ± 1 . 7 ( mean ± SEM , Figure 2 ) . Thus , rats help strangers of their own strain but not strangers of a different strain , suggestive of an in-strain bias for pro-social behavior . Moreover , short-term , paired-housing with a rat from a different strain is sufficient to motivate helping for that individual rat . 10 . 7554/eLife . 01385 . 004Figure 2 . Rats of the SD strain helped trapped rats of the LE strain only if they were familiar with an LE individual . Three experimental conditions are diagrammed at top . The free rat was always from the SD strain and was housed with a cagemate , denoted at top , from either the SD ( white ) or LE ( black and white ) strain . The free rat was then tested with an LE rat that was either the cagemate ( ‘a’ ) or a stranger ( ‘ ? ’ ) . Note that in the LE familiar condition ( right ) , rats had previously housed with an LE rat ( illustrated ) but were housed with an SD rat ( not illustrated ) at the time of testing . Across the days of testing , the median latency to door-opening ( bottom ) decreased for SD rats tested with trapped LE cagemates ( left ) , but not those tested with LE strangers ( middle ) . Like rats in the LE cagemate condtion , rats in the LE familiar condition ( right ) also became openers . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 00410 . 7554/eLife . 01385 . 005Video 1 . Rats help rats of a familiar strain . Rats that are familiar with at least one LE rat show interest in , open the restrainer door for , and rarely fight with trapped LE rats . Rats that are unfamiliar with LE rats show little interest in , do not help , and often fight with an LE rat . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 005 We then examined whether SD rats that were familiar with one LE individual would be motivated to help LE strangers . SD rats were pair-housed with an LE cagemate for 2 weeks , then re-housed with an SD rat , and a week later , tested in the helping behavior test with trapped LE strangers ( n = 12 , LE familiar condition; Figure 2 ) . In contrast to SD rats with no exposure to the LE strain , most SD rats familiar with one LE individual became openers for LE strangers ( 8/12 , 67%; χ2 , p=0 . 03 ) . They learned to open the door on average on day 4 . 0 ± 1 . 0 despite a lack of individual familiarity with the trapped LE rat ( Figure 2 ) . This experiment shows that social experience with one individual rat of a different strain is sufficient to motivate helping towards unfamiliar members of that strain . This provides further evidence that individual familiarity is not required for helping . Finally , to determine if helping behavior is at all influenced by the strain of the trapped rat ( SD , LE ) , or if strain bias in helping is entirely due to strain familiarity ( familiar , unfamiliar ) , SD rats were fostered and raised with LE rats from birth , in an environment that effectively prevented exposure to others of their own strain ( fostered conditions , Figure 3 ) . If rats are innately motivated to help rats of their own strain , then SD rats raised exclusively with LE rats ( fostered ) should act pro-socially towards other SDs . In contrast , if strain familiarity is the only determinant of pro-social behavior , these rats should not help their own kind . At 2 months of age , fostered rats were tested with trapped rats that were SD strangers ( n = 8 , fostered+SD ) or LE strangers ( n = 8 , fostered+LE ) . Fostered SD rats did not help trapped SD strangers ( Figure 4; Video 2 ) . Only 1 of 8 fostered SD rats ( 12 . 5% ) became an opener , establishing that rats are not innately motivated to help their own strain . In contrast , fostered rats did help LE strangers ( 5/8 , 62 . 5%; χ2 , p=0 . 04 ) as expected from the pro-social influence of strain familiarity . Thus , strain familiarity , even to one’s own strain , is required for the expression of helping behavior . 10 . 7554/eLife . 01385 . 006Figure 3 . Fostered SD rats were minimally exposed to other SD rats from birth . This diagram indicates how many LE pups ( black dots ) and SD pups ( red dots ) were present in each litter during postnatal ( P ) days 0–14 ( x-axis ) for rats in the fostered+LE ( A ) and fostered+SD ( B ) conditions ( see Figure 4 ) . On P0–P1 , two SD pups ( n = 32 ) were transferred into each LE litter ( n = 16 ) . If both pups survived one of the two SD pups was removed at P6 , leaving a single SD pup with an LE dam and LE littermates . As red dots indicate other SD pups , the number of red dots is representative of the amount of exposure each fostered SD rat had to other SD pups during their lives . Individual rats are arranged from the most exposure to SD rats ( upper left in each group ) to the least exposure ( bottom right ) . Following P11 , no fostered rat was exposed to other SD rats until testing . In two cases ( both in the fostered+SD condition illustrated in B ) , the SD pup in the LE litter died at a later date ( P11 , P12 ) . In these cases , an SD male who had been removed from an LE litter 5 or 7 days before was then re-added to the LE litter . Note that only one animal ( upper left in B ) was exposed to SD rats with his eyes open; this rat proved to be the only door-opener in the fostered+SD condition . Lines above the dots represent the strain of the dam ( red: SD; black: LE ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 00610 . 7554/eLife . 01385 . 007Figure 4 . Strain familiarity , even to one’s own strain , is required for the expression of helping behavior . Fostered SD rats were raised with LE rats from birth ( top diagram ) and were not exposed to or able to interact with other SD rats prior to testing . When fostered SD rats were adults , they were tested with trapped stranger rats ( ‘ ? ’ ) of either the SD ( left ) or LE ( right ) strain . Fostered SD rats did not help trapped SD strangers . In contrast , fostered SD rats helped trapped LE strangers . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 00710 . 7554/eLife . 01385 . 008Video 2 . Fostered rats do not help rats of their own strain . Normally rats show pro-social behavior toward rats of their own strain . However , SD rats raised exclusively with LE rats and in isolation from other SD rats did not show interest in or open for a trapped SD stranger . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 008 The finding that fostered rats did not help SD strangers indicates that a rat’s familiarity with himself is insufficient to motivate helping for individuals of his strain . Instead , social interaction with another rat , such as that occurring while two rats live together , is critical to shaping pro-social motivation . Moreover , since rats tested with strangers from unfamiliar strains did not become openers across the days of testing , the exposure and interactions afforded by testing sessions appear insufficient to produce pro-social motivation . Fostered rats were noticeably more anxious than naturally reared rats . In open-field testing conducted prior to the experiment , fostered rats spent more time in the corners of the arena than rats in other conditions ( 33 . 1 ± 2 . 7 s/min for fostered rats; 25 . 0 ± 1 . 9 s/min for naturally reared rats; files with open field data for SD cagemate and SD strangers were corrupted and therefore not included; two-tailed Student’s t test , p<0 . 02; Figure 5A ) . Moreover , fostered rats that opened the restrainer began to do so significantly later ( day 8 . 0 ± 1 . 1 for fostered rats; 4 . 5 ± 0 . 5 for all other rats; two-tailed Student’s t test , p<0 . 01; Figures 1–2 , 4B ) . Fostered rats were tested for an extra day , confirming that opening behavior persisted . These data suggest that fostering rats in a different-strain environment is associated with increased anxiety . A same-strain , cross-fostering control condition is needed to distinguish whether strain , fostering , or both drive this effect . 10 . 7554/eLife . 01385 . 009Figure 5 . ( A ) Fostered rats ( left of dashed line ) spent more time at the arena corners during open field testing than naturally reared animals ( right of dashed line ) . ( B ) The proportion of rats that opened increased across the testing sessions for rats tested with familiar strains ( open markers ) but not for those tested with unfamiliar strains ( filled markers ) . ( C ) No differences between rats tested with trapped rats from familiar strains ( white bars ) and those tested with rats from unfamiliar strains ( gray bars ) were observed in the average number of alarm calls or the amount of time spent at corners during open field testing . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 009 Across all conditions , the free rat’s familiarity with the strain of the trapped rat is the common determinant of whether helping behavior occurred ( Figure 6A ) . Most rats in familiar strain conditions became openers ( 33/47 , 74% ) whereas only 21% ( 5/24 ) of rats in unfamiliar strain conditions did ( χ2 , p<0 . 001 ) . Rats tested with a familiar strain experienced opening the restrainer door as rewarding , as they typically opened it on consecutive days ( Figure 6B–C ) . In contrast , rats tested with unfamiliar strains rarely opened on consecutive days , suggesting that opening was not rewarding for these animals . As open field performance and alarm calls were not different for familiar and unfamiliar strain conditions ( Figure 5C ) , it is unlikely that trait anxiety or the trapped rat’s distress accounts for the results . 10 . 7554/eLife . 01385 . 010Figure 6 . Rats experience helping rats of a familiar strain as rewarding . ( A ) The latency to door-opening decreased along the days of testing for rats tested with familiar strains ( open circles ) but not for those tested with unfamiliar strains ( filled squares ) . ( B ) The proportion of openings that were followed by a repeated opening on the next day of testing was higher for trapped rats from familiar strains ( left of dashed line ) than from unfamiliar strains ( right of dashed line ) . ( C ) Opening data from representative rats of each condition are illustrated . Fostered SD rats were tested for 13 days . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 010 When tested with familiar strains , helping behavior was demonstrated equally for strangers and cagemates . Yet , the underlying affective response might differ in these conditions . In support of this idea , rats displayed a different movement pattern when tested with a trapped cagemate or stranger . Rats tested with a trapped cagemate were significantly more active prior to door-opening than rats tested with a stranger ( MMA , p<0 . 05 , Figure 7A ) . However , there was no difference in the distance from the restrainer in these conditions ( Figure 7B ) . These results demonstrate that rats were equally motivated to help , but not equally aroused by , a trapped cagemate and stranger . 10 . 7554/eLife . 01385 . 011Figure 7 . Prior to door-opening , rats are equally motivated to help , but show more activity for trapped cagemates than for strangers from a familiar strain . ( A ) Across conditions and testing days , rats tested with cagemates ( white circles ) were more active in the period before door-opening than were rats tested with strangers from a familiar strain ( black circles ) . ( B ) The distance from the restrainer was not different for rats tested with cagemates ( white circles ) and strangers from a familiar strain ( black circles ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 011 Door-opening behavior overlapped greatly , but not completely , with strain familiarity . Across all conditions , rats who became openers demonstrated an affiliative behavioral pattern . Prior to door-opening , openers spent more time around the closed restrainer than did non-openers ( two-tailed Student’s t test , p<0 . 01; Figure 8A; Video 1 ) . Following the trapped rat’s exit from the restrainer , significantly less fights were observed for openers than for non-openers ( two-tailed Student’s t test , p<0 . 01; Figure 8A ) . Baseline anxiety level was assessed by the time spent in the corners of the arena during open-field testing . Only for openers , baseline anxiety was positively correlated with the first day of opening ( Figure 8B ) , suggesting that anxiety negatively impacts successful helping in the presence of pro-social motivation . Rats were less active before door-opening compared to after door-opening ( repeated measures ANOVA , main effect , p<0 . 001; Figure 8C ) . Pairwise comparisons found that this effect was due to significantly reduced activity in non-openers prior to door opening , which was not the case for openers ( pairwise comparisons , p<0 . 05; Figure 8C ) . This difference may result from either non-openers’ greater anxiety or reduced interest in the trapped rat . 10 . 7554/eLife . 01385 . 012Figure 8 . Openers showed an affiliative behavioral pattern . ( A ) Rats that became openers ( left ) spent more time around the closed restrainer ( open bars ) and fought less with the trapped rat ( gray bars ) than did non-openers ( right ) . ( B ) The first day of repeated opening was correlated with time spent at the corners during open-field testing for openers ( top ) but not for non-openers ( bottom ) . ( C ) Averaged across the 12 days of testing and regardless of strain familiarity , non-openers were less active when the restrainer was closed ( white bars ) than when it was open ( gray bars ) . DOI: http://dx . doi . org/10 . 7554/eLife . 01385 . 012
This study demonstrates that helping another rat , by releasing it from a restrainer , is flexibly applied to select others based on previous social experience . It is neither the individual identity nor the particular strain of the rat in need that motivates helping . Rather , it is the prior social experience of the free rat with any member of the trapped rat’s strain that determines the target group for helping . Thus , rats help another rat from a given strain ( same or different ) only if they have previously lived with a member of that strain . Rats pick up sensory cues from their cagemates that are retained and utilized to discriminate between strangers that share those cues and those that do not . This information then plays a key role in eliciting pro-social behavior , and could potentially be used for other behavioral decisions as well , when encountering an unfamiliar individual , effectively forming group categories . The experiments delineated above are not informative of the sensory modalities participating in the classification of strangers as similar or different to the cagemate . Previous research points to the importance of vision , olfaction and audition for affective communication between mice ( Langford et al . , 2006; Jeon et al . , 2010 ) . It is possible that helping behavior in rats similarly relies on visual , olfactory , auditory , and also tactile cues as the clear and perforated restrainer allows for full sensory communication between the free and trapped rats . The present study effectively demonstrates that strain identity is meaningless without social experience during development . Fostered rats raised without social interactions with their own strain were not motivated to help strangers of their strain , compelling evidence against an innate bias for pro-social behavior towards one’s own kind . This finding is congruent with studies showing that neither kinship nor perceived similarity is needed to motivate pro-social behavior in primates ( Batson et al . , 2005; Horner et al . , 2011; Baden et al . , 2013 ) . In nature , social animals typically live with related others , and thus a pro-social bias favoring familiar others would also favor genetically similar others . Yet , relying on social experience rather than genetic similarity for guiding pro-social behavior has an added value in that it allows animals to flexibly adapt to different circumstances ( Dugatkin , 2002 ) . Assigning an affective meaning to a group category following social experience would be an efficient mechanism for appropriately extending pro-social behavior towards unknown individuals belonging to that same group . While door-opening has an obvious pro-social outcome , a variety of motivations could contribute to this behavior . We have previously excluded the possibilities that door-opening depends on reward from either motor mastery or social contact ( Ben-Ami Bartal et al . , 2011 ) . In addition it is unlikely that rats were motivated to open the restrainer by aggressive or anti-social motivations ( Myer and White , 1965 ) because conflict between the free and trapped rats was minimal for opener pairs . As non-openers tested with an unfamiliar strain were not less active before door-opening than openers tested with a familiar strain ( Figure 8C ) , differences in general arousal cannot account for the observed results . Given the greater interest in the trapped rat shown by openers and the lack of aggressive interactions between opener pairs , the authors favor the interpretation that rats open the restrainer door in order to terminate the trapped rat’s distress . While we did not see any difference in helping behavior expressed toward strangers and cagemates , the underlying affective responses may have differed . Indeed , movement velocity was greater for rats tested with cagemates compared to those tested with strangers , indicative of different motivational states or magnitudes . Nonetheless , rats were sufficiently motivated to help strangers . This result is consistent with previous demonstrations of helping behavior expressed toward strangers in other animals , including humans ( Warneken and Tomasello , 2006; Batson et al . , 2007; Loggia et al . , 2008; Custance and Mayer , 2012; Tan and Hare , 2013 ) . While emotional contagion has been demonstrated for both individually familiar and unfamiliar mouse pairs ( Chen et al . , 2009; Jeon et al . , 2010 ) , it may be facilitated by individual familiarity ( Langford et al . , 2006; Jeon et al . , 2010 ) . Work in dogs and primates further supports a familiarity effect for emotional contagion ( Palagi et al . , 2009; Campbell and de Waal , 2011; Silva et al . , 2012 ) . Empathy , the capacity to share and recognize the emotional states of another ( Decety , 2011 ) , often motivates approach and pro-social behavior and caring in humans ( Eisenberg and Miller , 1987 ) . The empathic response involves activation of common neural networks for processing one’s own and another’s distress ( Decety et al . , 2012 ) . Some forms of empathy are primarily dependent on subcortical neural structures that are phylogenetically conserved across mammalian species ( Decety and Svetlova , 2012 ) . Emotional contagion , a fundamental component of empathy resulting when the emotion of one individual evokes a matching emotional state in another individual ( Preston and de Waal , 2002 ) , has been demonstrated in rodents ( Chen et al . , 2009; Langford et al . , 2006 ) . Thus , rodents , and mammals in general , may share a mechanism for mobilizing pro-social motivation in response to the distress of another individual . We would argue that a rodent form of empathy , as defined above , is the main motivation for the helping behavior observed in our studies . Moreover , the absence of door-opening for rats of unfamiliar strains likely reflects a reduced empathic arousal of the free rats in these conditions . Rats that are not familiar with a different strain do not act pro-socially towards others of that strain . The selectivity with which rats engage in helping behavior is further evidence that releasing a trapped rat is an intentional social behavior , congruent with an empathic drive to help some rats , but not others . Yet , social experience can modify this behavior . We conclude that through social interactions , rats form affective bonds that elicit empathy and motivate helping . This motivation to help is extended to strangers of familiar strains , showing that rats form groups based on social experience . As is the case for humans ( Xu et al . , 2009; Mathur et al . , 2010 ) , rats base the bias for pro-social behavior on group membership . And as for humans , a diverse social experience was effective in mitigating such bias ( Chiao et al . , 2008; Elfenbein and Ambady , 2003; Madsen et al . , 2007; Telzer et al . , 2013; Zuo and Han , 2013 ) . Moreover , we have demonstrated that for rats , genetic similarity does not influence pro-social motivation . Rather , groups are socially defined . Whereas social bias is hard to overcome through cognitive effort ( Dovidio et al . , 2002; Johnson et al . , 2002; Amodio et al . , 2004; Correll et al . , 2007 ) , our results support the existence of at least one mechanism for altering group membership that does not require complex cognition .
Sprague-Dawley ( SD ) and Long-Evans ( LE ) male rats ( Charles River , Portage , MI ) were used for all studies . Rats were 8–11 weeks old at the start of the experiment . Rats were housed in pairs with ad libitum access to chow and water in a 12:12 light-dark cycle . Animals were allowed 2–3 weeks to acclimate to the housing environment . Stranger rats were always housed either in a separate room or on a separate rack than the free rats with which they were to be tested . Stranger rats were also handled separately , such that there was never any contact between strangers and free rats before the start of testing . The fostered SD rats were bred in-house from pregnant females purchased from outside ( Charles River , Portage , MI; see more details below ) . Above every Plexiglas arena ( 50 × 50 cm , 32–60 cm high ) , a CCD color camera ( KT&C Co , Seoul , Korea ) was mounted . The cameras were connected to a video card ( Geovision , Irvine , CA ) in a dedicated PC . Sound recordings of ultrasonic ( 15–70 kHz ) vocalizations were recorded ( Avisoft Bioacoustics , Berlin , Germany ) through a single microphone in each testing room . Because recorded ultrasonic vocalizations could not be ascribed to individual rats and instead had to be assigned to each condition , all rats tested at one time in one room were always from the same condition . A Plexiglas rodent restrainer ( 25 by 8 . 75 by 7 . 5 cm , Harvard Apparatus , Holliston , MA ) was placed in the center of the arena . The body of the restrainer had several small slits and the back end had a large slit , allowing for olfactory and tactile communication between rats . At the other end , a customized door had two panes that were attached with three screws , and a pole ( 5 cm ) supporting two weights ( 25 g each ) . The weights were included in order to facilitate the door falling off to the side once the free rat pushed on the door . The door was designed to be opened only from the outside . The free rat could open the door from the top , from the side , or by pushing up on the door with its snout . Animals were habituated to the experimenters ( who were kept constant for each cohort of rats ) and the arenas prior to being tested . On day 1 , rats were transported to the testing room and left undisturbed in their home cages . On day 2 , rats were briefly handled and tested for ‘time-out’ ( see below ) . Starting with the second day of habituation , rats were weighed three times each week for the duration of the experiment; no animal lost weight during the experiment . On days 3–5 , rats were tested for time-out , marked , and handled for 5 min by each experimenter . Rats were then placed with their cagemate in the testing arenas for 30 min . All free rats and all cagemates were always placed in the same arena for habituation and testing . Rats that were used as strangers were placed in the same arena for habituation but in different arenas during testing . Habituation of free rats and strangers never occurred at the same time . After each habituation session , rats were returned to their home cages and to the housing room . All sessions were run during the rats’ light cycle between 0800 and 1730 . Order of testing was counterbalanced between sessions to control for effects of time of day on behavior . Time-out was measured as the latency from opening the homecage lid halfway to the time that the rat approached the front edge of the cage , reared up , and placed its paws on the ledge . This measurement was recorded 3–5 times for each rat in every cage during habituation . On the day following completion of habituation , rats were placed individually in an arena for 30 min and their activity recorded . Note that the arenas were the same as were used during habituation but that open field testing was the first time each rat had been in the arena alone . At the start of testing , the trapped rat was placed in the restrainer , the door closed , and the restrainer placed in the center of the arena . The free rat was then placed in the arena . If the free rat did not open the restrainer door within 40 min , the investigator opened the restrainer door halfway , to a 45° angle , greatly facilitating door-opening by either rat . Regardless of whether the door was opened before or after the 40 min mark , both rats always remained in the arena for the full hour-long session . After each session , the arena and restrainer were washed with 1% acetic acid followed by surface cleaner . Rats were tested once daily for 12 days . If a trapped rat succeeded in opening the door from inside the restrainer ( ∼30% of rats ) , the trapped rat was placed immediately back in the restrainer , and a Plexiglas blocker was inserted , preventing his access to the door . If the free rat subsequently opened the door , the blocker was removed , allowing the trapped rat to exit the restrainer . The blocker was then used for that trapped rat on all following test days . The free rat was a male SD rat in all conditions . The strain of the male trapped rat was either SD ( SD cagemate , SD stranger , fostered+SD ) or LE ( LE cagemate , LE stranger , LE familiar , fostered+LE ) as described further below . All rats used in all conditions were pair-housed . Ethovision tracking software ( Noldus Information Technology , Inc . Leesburg , VA ) was used to track the rats’ movements in the arena . To enable tracking both rats’ movements , free rats were colored red and trapped rats colored blue . The rat’s location was converted into x , y coordinates denoting the rat’s location at each frame at a rate of 7 . 5 FPS . These data were then used to calculate movement velocity and location in the arena ( time around the restrainer , time at arena corners ) . Freeware ( Jaywatcher V1 . 0 ) was used to manually code the rats’ interactions for 15 min following door-opening . If the door was opened less than 15 min before the end of session , only the remaining time in the arena was coded . Data were not analyzed for five rats ( four SD stranger , one LE cagemate ) due to technical problems . Coded behaviors included anogenital sniffing , pinning , wrestling , and boxing . Coding was performed by four judges . There was an 84% agreement between judges . Boxing was the behavior that was selected to represent fighting , and analyzed . One rat ( a non-opener in the SD cagemate condition ) boxed on 95 occasions , a total that was more than five standard deviations over the mean for all rats tested . Additionally , this rat boxed on 10 of the 12 testing days , more than two standard deviations over the mean . This rat was therefore considered an outlier and was excluded from the boxing analysis . Vocalizations during the first 40 minutes of the session were analyzed for days 1 , 3 , and 9 of testing using Avisoft SASLab Pro ( Berlin , Germany ) . Audio data was not available for rats in the SD cagemate and SD stranger conditions ( data were lost due to disk failure ) . Thus , audio data came from rats in the LE cagemate , LE stranger , LE familiar , and both fostered conditions . The mothering behavior of all dams was recorded at 5 min intervals for three hours ( an hour each , starting at 0900 , 1200 , and 1600 ) between P1 and P10 ( methods adapted from Champagne et al . , 2003 ) . The nursing behaviors recorded were ( 1 ) arched back nursing; ( 2 ) blanket nursing; or ( 3 ) passive nursing . In addition , we recorded if the dam was either ( 4 ) not in contact with pups; or ( 5 ) licking or grooming a pup . The proportion of time bins within each of the five categories was calculated for each dam on each day . The proportions of each behavior recorded were not different between rats in fostered+LE and fostered+SD conditions ( 1-way repeated measures ANOVAs , p>0 . 11 ) . Time to door-opening was calculated as the minute when the restrainer door was opened minus the start time . For rats that never opened , a cutoff time of 40 min was assigned . Rats that opened the restrainer on two sequential days , and did so at least three times were termed ‘openers’ . Thus , a rat that opened the restrainer 4 days in a row was considered an opener . In all cases except one , once this criterion was met , the rat continued to open the restrainer until the end of the experiment . In the one exception , the rat opened the restrainer on days 4–8 but not on days 9–11 . This rat ( in the LE familiar group ) was considered a non-opener . Three rats ( SD cagemate , LE cagemate , and fostered+LE conditions ) opened on the final 3 days of testing and did so at decreasing latencies . These rats were considered openers . The day on which opener rats learned to open the restrainer was defined as the first opening that was repeated the next day . Door-opening latencies , velocity , time in arena corners , and time around the restrainer were averaged per rat across all sessions and all days . ANOVA and two-tailed Student’s t tests were used to determine differences between groups . Fischer PLSD was used for all post-hoc analyses . A chi-square analysis was used to compare the proportions of openers and non-openers . In all cases , α <0 . 05 was used as criterion for significance . Door-opening latencies are displayed using the median since door-opening was not a normally distributed variable . Statistical comparisons were conducted using SPSS ( PASW 18 ) . | Humans help family members and friends under circumstances where they may not help strangers . However , they also help complete strangers through both direct actions , such as helping someone who has stumbled , and indirect actions , such as giving to charity . Ben-Ami Bartal et al . have now explored the biological basis of such socially selective helping by testing whether rats help strangers , and if so , under what circumstances . In the experiments a free rat was exposed to another rat trapped inside a plastic tube with an outward-facing door for 12 one-hour sessions . When tested with a cagemate trapped inside the tube , most free rats learned within a few days to release the trapped rat by opening the door . Ben-Ami Bartal et al . then exposed the free rats to strangers they had never met or seen before . Remarkably the rats consistently released the trapped stranger , acting toward strangers just as they had acted toward familiar cagemates . This result suggested that individual familiarity is not required for helping to occur . To test the limits of rat benevolence , Ben-Ami Bartal et al . tested free rats ( always white albino rats ) with trapped rats from a different outbred strain ( black-hooded rats ) . The rats helped cagemates of a different strain but not strangers of a different strain . These results could be explained by a requirement for strain familiarity or individual familiarity . To distinguish between these possibilities , albino rats were housed for 2 weeks with a rat of a different strain , and then re-housed with another albino rat before being tested with a trapped rat belonging to a different strain . Consistent with a requirement for strain but not individual familiarity , the free rats now helped stranger rats from the different , but now familiar , strain . To explore if there is any role for genetics or relatedness in socially selective helping , Ben-Ami Bartal et al . tested whether rats will help strangers of their own strain based on genetic relatedness alone . To do this albino pups were transferred to litters of a different strain on the day they were born , and never saw or interacted with another albino rat until testing . Remarkably , the albino rats helped strangers from the different strain that they were raised with , but they did not help strangers of their own strain because this strain was unfamiliar to them . The fact that the motivation to help other rats has its origins in social interactions rather than genetics provides the flexibility that is needed to navigate their way through social environments that often change unexpectedly . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2014 | Pro-social behavior in rats is modulated by social experience |
Insights into the conformational organization and dynamics of proteins complexes at membranes is essential for our mechanistic understanding of numerous key biological processes . Here , we introduce graphene-induced energy transfer ( GIET ) to probe axial orientation of arrested macromolecules at lipid monolayers . Based on a calibrated distance-dependent efficiency within a dynamic range of 25 nm , we analyzed the conformational organization of proteins and complexes involved in tethering and fusion at the lysosome-like yeast vacuole . We observed that the membrane-anchored Rab7-like GTPase Ypt7 shows conformational reorganization upon interactions with effector proteins . Ensemble and time-resolved single-molecule GIET experiments revealed that the HOPS tethering complex , when recruited via Ypt7 to membranes , is dynamically alternating between a ‘closed’ and an ‘open’ conformation , with the latter possibly interacting with incoming vesicles . Our work highlights GIET as a unique spectroscopic ruler to reveal the axial orientation and dynamics of macromolecular complexes at biological membranes with sub-nanometer resolution .
Numerous fundamental cellular processes such as energy conversion , signal transduction , and transport occur in the context of lipid membranes . In animals , approximately one-third of the proteome is localized at the various membranes of the cell , while more than 50% of the currently available pharmaceuticals target membrane proteins ( Santos et al . , 2017 ) . The mechanistic understanding of these proteins has been mainly addressed by techniques that arrest them in particular conformations , and take them out of their biological membrane context ( Fernandez-Leiro and Scheres , 2016; Kosol et al . , 2013 ) . Even more critical , we largely lack techniques to address the orientation and organization of large and highly flexible protein complexes at membranes involved in signaling or interorganellar communication and fusion . The structurally highly heterogeneous and dynamic systems are extremely difficult to tackle by traditional structural techniques that average ensembles . Single-molecule Förster resonance energy transfer ( smFRET ) has been used successfully to shed light on the structural heterogeneity of proteins lacking structural definition ( Roy et al . , 2008; Yang et al . , 2018; Zhu et al . , 2017 ) . However , with its limited dynamic range covering <10 nm as well as the need to introduce both donor and acceptor dyes with high fidelity and efficiency often obstructs its application . Membrane protein complexes , which often involve intrinsically disordered proteins or phase separation causing large-scale spatial re-arrangement , are therefore not amenable to smFRET . Single-molecule localization microscopy in turn is limited by axial localization precision of ~10 nm ( Gwosch et al . , 2020; Schmidt et al . , 2008; Shtengel et al . , 2009 ) . Thus , single-molecule-based techniques providing sufficient spatiotemporal resolution in the 10–20 nm regime , which is relevant for multi-protein complexes , are currently not available . Here , we introduce a novel approach for quantifying the axial organization and dynamics of large membrane protein complexes based on distance-dependent fluorescence quenching by graphene . This phenomenon is caused by radiation-less electromagnetic coupling of the excited fluorophore with graphene plasmons , which decays with the axial distance d by d−4 ( Swathi and Sebastian , 2009 ) . The atomic thickness of graphene ensures high optical transparency and an extremely high confinement of plasmons ( 106 times the diffraction limit ) ( Koppens et al . , 2011 ) . Compared to metal-induced energy transfer ( MIET ) ( Chizhik et al . , 2014 ) , with a dynamic range that extends over a distance of ~150 nm , graphene-induced energy transfer ( GIET ) occurs within a dynamic range of ~30 nm , thus covering the dimensions of large membrane protein complexes ( Ghosh et al . , 2019 ) . For exploiting GIET to probe axial organization of proteins in the context of membranes , we here developed a lipid monolayer assembly on graphene for site-specific protein capturing . We confirmed the theoretical distance-dependence of GIET within a dynamic range of 25 nm on graphene-supported lipid monolayers using DNA oligonucleotides as a nanoscale ruler . We successfully applied this approach for unraveling the axial organization and dynamics of large multi-protein complexes involved in vesicular transport and fusion , which is critically required for delivery of proteins and lipids in organellar homeostasis ( Bröcker et al . , 2010; Yu and Hughson , 2010 ) . We reconstituted the entire protein machinery required for tethering late endosomal vesicles , autophagosomes , and AP-3 vesicles to the vacuolar target membrane ( Bröcker et al . , 2010; Nickerson et al . , 2009 ) , including the Rab7-like Ypt7 , its guanine nucleotide exchange factor ( GEF ) Mon1-Ccz1 and the homotypic fusion and vacuole protein sorting ( HOPS ) complex ( Brett et al . , 2008; Nordmann et al . , 2010; Ostrowicz et al . , 2010 ) . In animal cells , HOPS is responsible for autophagy , the infectivity of Ebola virus , and linked to multiple diseases ( van der Beek et al . , 2019 ) . Based on ensemble and single-molecule GIET , we quantitatively unraveled the axial organization and dynamics of Ypt7 and its interacting HOPS complex . Our data reveal that HOPS adopts an upright orientation on membranes with characteristic axial dynamics . We thus introduce GIET as a powerful novel technique to uncover the nanoscale spatiotemporal architecture of extended multi-protein complexes at membranes .
To apply GIET to explore the structural and functional organization of protein complexes at membranes , we established lipid monolayer coating of graphene . Solid-supported graphene monolayers were prepared by transferring commercially available graphene sheets onto glass substrates . Coating of lipid on graphene was carried out by either liposome fusion or solution-assisted lipid deposition as reported previously ( Blaschke et al . , 2018; Lima et al . , 2016; Tabaei et al . , 2016 ) . For site-specific capturing of His-tagged proteins , tris-nitrilotriacetic acid ( trisNTA ) moieties were incorporated into the lipid monolayer . For this purpose , vesicles made from 1 , 2-dioleoyl-sn-glycero-3-phosphocholine ( DOPC ) containing 5% trisNTA conjugated with dioctadecyl amine ( trisNTA-DODA ) ( Beutel et al . , 2014; Lata et al . , 2005 ) were fused on freshly prepared graphene slides ( Figure 1A ) . Lipid monolayer formation , protein immobilization , and interactions were monitored in real-time using simultaneous total internal reflection fluorescence spectroscopy and reflectance interference ( TIRFS-RIF ) detection ( Gavutis et al . , 2005 ) . A mass signal of lipid deposition on graphene of 2 . 5 ng/mm² was observed after washing out excess vesicles ( Figure 1B ) . For trisNTA-functionalized lipid bilayer formation by vesicle fusion on a silica surface , which has been previously established for protein interaction analysis at membranes ( Beutel et al . , 2014; Gavutis et al . , 2005; Lata et al . , 2006 ) , a lipid deposition of 5 ng/mm² was observed ( Figure 1—figure supplement 1 ) . These results confirmed formation of a lipid monolayer on the hydrophobic graphene surface . Stable , Ni ( II ) ion specific immobilization of an anti-GFP nanobody fused to a C-terminal His-tag ( NB-H6 ) was observed on the graphene-supported lipid monolayer . Upon injection of monomeric enhanced green fluorescence protein ( mEGFP ) , we detected fast binding to the immobilized NB ( Figure 1B ) . The same binding assays were carried out with silica-supported lipid bilayers , yielding very similar binding kinetics for NB-H6 and mEGFP ( Figure 1—figure supplement 1 ) . Furthermore , quantitative removal of immobilized protein by imidazole was observed ( Figure 1B ) , which enabled for repeated immobilization . These experiments verified the functional integrity of His-tagged proteins captured onto graphene-supported lipid monolayers . The interaction with mEGFP was simultaneously quantified by TIRFS detection ( Figure 1C ) . Comparing the mEGFP fluorescence signal observed on graphene ( IG ) to that on the silica substrate ( I0 ) confirmed strong quenching by GIET with an efficiency of 83 . 6% for mEGFP bound to NB-H6 . Direct immobilization of H6-tagged mEGFP on graphene resulted into even stronger GIET ( Figure 1D ) with an efficiency of 92 . 4% . As the size of NB is ~2 nm based on the crystal structure ( Kirchhofer et al . , 2010 ) , these results highlight that GIET can reveal nanoscale distances of proteins immobilized on a graphene-supported lipid monolayer . To determine axial distances from the membrane , we calibrated the distance-dependent GIET efficiencies by using fluorescent-labeled DNA strands attached to the lipid monolayer . Four single strand DNA ( ssDNA ) with 20mer , 25mer , 35mer , and 50mer nucleotides were designed as the scaffold of nanorulers ( Figure 2A , DNA sequences in Supplementary file 1A ) . These ssDNAs share a common 20mer sequence at the 5´-end and cholesterol modification at the 3´-end for anchoring into the lipid monolayer ( Johnson-Buck et al . , 2014 ) ( anchor strand ) . To promote , by electrostatic repulsion , perpendicular orientation of oligonucleotides with respect to lipid layer , 5% negatively charged 1 , 2-di- ( 9Z-octadecenoyl ) -sn-glycero-3-phospho-L-serine ( DOPS ) was incorporated into the DOPC matrix . The probe strands have a common 20mer ssDNA sequence complementary to the anchor strand and were conjugated with 6-carboxyfluorescein ( FAM ) at either 3´- or 5´- end , respectively , for fluorescence read-out . By hybridization of the probe and anchor strands , eight distinct fluorophore distances from the monolayer surface were obtained , which are denoted as anchor strand-3´F or 5´F . For the 35mer and 50mer anchor strands , additional complementary 15mer and 30mer ssDNA were used as blockers , respectively , to obtain the fully length double strand DNA nanorulers ( Figure 2A ) . For the 25mer anchor strand , a 5mer unpaired region remained after hybridization with the probe strands , which may enhance the flexibility of this system . Using real-time , surface sensitive detection of TIRFS-RIF , we monitored formation of the lipid monolayer on graphene , integration of anchor strands , and hybridization with probe strands ( Figure 2—figure supplement 1 ) . For all probe-anchor hybridizations , detection of combination 20–5´F was not feasible , probably due to the steric hindrance upon inserting a FAM dye proximal to the lipid head groups . Instead , we used 2′ , 7′-difluorofluorescein-1 , 2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine ( OG488-DHPE ) , in which the dye molecule is directly linked to the lipid head group . In parallel , the same experiments were performed using silica-supported lipid bilayers . Simultaneous quantification of mass deposition and fluorescence intensity on both substrates ensured reliable quantification of GIET efficiencies ( Supplementary file 1B ) . Significantly reduced fluorescence intensities on graphene ( IG ) were observed compared to those on silica ( I0 ) . The IG/I0 ratios increased with the end-to-end distances of the seven different FAM positions on DNA nanorulers . However , the increase was significantly lower than the theoretical prediction from electromagnetic simulations , which assumed fully perpendicular orientation on top of the lipid monolayer coating ( Materials and methods: Electromagnetic simulation of GIET , and Figure 2—figure supplement 2 for sensitivity analysis ) . Since we have previously validated our simulations using well-defined silica adlayers on the graphene surface ( Ghosh et al . , 2019 ) , we wondered whether tilting of DNA strands on surface was the reason for the discrepancy ( Kabeláč et al . , 2012; Wong and Pettitt , 2004 ) . By assuming a tilting angle α between DNA and graphene , we calculated the vertical distance d by taking the 2 . 5 nm thickness of the lipid monolayer into account ( Attwood et al . , 2013; Blaschke et al . , 2018 ) ( Materials and methods: Validation of distance-dependent GIET by DNA nanorulers ) . Strikingly , the obtained correlation of IG/I0 vs d matches the predicted GIET curve well with a globally fitted tilting angle of ( 43 ± 1 ) ° ( Figure 2B ) , confirming the distance-dependent GIET on graphene-supported lipid monolayer . We also observed systematic deviations to higher experimental intensities at very short distances including OG488-DHPE , which may be ascribed to the fluorophores being attached via flexible linkers take a more distant position from the surface due repulsion from the negatively charged membrane surface . GIET efficiencies were furthermore quantified by time-correlated single photon counting ( TCSPC ) using confocal laser-scanning microscopy ( cLSM , Figure 2—figure supplement 3 ) . Lifetime histograms could be nicely fitted with mono-exponential decay functions in all cases ( Figure 2—figure supplement 4 ) , indicating a homogenous structural organization of the DNA rulers . On glass surfaces , constant fluorescence lifetimes of FAM were observed for DNA nanorulers , either 2 . 8 ns for 3´ labeling or 3 . 3 ns for 5´ labeling ( Figure 2—figure supplement 4 ) . The shorter lifetime at 3´-end could be attributed to quenching by the adjacent G-C base pair as reported previously ( Nazarenko et al . , 2002 ) . Ratios of fluorescence lifetimes on graphene ( τG ) to those on glass ( τ0 ) closely matched the corresponding intensity ratios , which further corroborated the simulated distance-dependent GIET efficiency ( Figure 2B , Supplementary file 1C ) . Results of both intensity and lifetime measurements showed a dynamic range within 3–25 nm corresponding to changes in GIET efficiency from 93–15% . A significant difference was detected for nanorulers between 35–5´F and 35–3´F , confirming that GIET is robust for detecting distance changes of ~1 . 2 nm between 6 nm and 11 nm . These consistent calibrations via fluorescent intensities and lifetime measurements corroborated the validity of our GIET model , with systematic deviations likely to be related to positional uncertainty caused by inherent flexibility of the nanoruler system . We therefore applied the theoretical GIET curve for subsequent quantification of distances from experimental IG/I0 . For the axial organization of mEGFP and mEGFP:NB complex tethered to the DOPC monolayer surface via trisNTA-DODA , distances 1 . 3 nm for H6-mEGFP directly tethered to the membrane ( IG/I0: 7 . 6% ) and 3 . 0 nm for mEGFP captured via NB-H6 ( IG/I0: 16 . 4% ) were estimated . The distance of 1 . 7 nm induced by the nanobody is in good agreement with the crystal structure of the GFP-NB complex ( Kirchhofer et al . , 2010 ) , highlighting the nanometer sensitivity of quantifying the axial position of proteins onto lipid monolayers by GIET . We next asked whether GIET is capable to measure distance changes larger than 10 nm . For this purpose , we chose the 35mer anchor strand with a fully extended length of 11 . 9 nm from 3´-end to 5´-end ( Figure 2C ) . Upon hybridizing with the 20mer 3´ FAM probe strand , a fluorescence lifetime τG of 0 . 24 ± 0 . 02 ns was detected ( Figure 2D ) . Comparing to τ0 of 3 . 04 ns on glass , the obtained τG/τ0 ratio of 8 . 0 ± 0 . 6% corresponds to a distance of 4 . 4 ± 0 . 2 nm from the graphene ( 1 . 9 nm from the lipid headgroups ) , suggesting that the hybridized probe-anchor DNA strand collapses onto the surface of the lipid monolayer due to the flexible , unpaired 15mer gap . To test the hypothesis , a 15mer ssDNA blocker strand was added , resulting into a ~ sixfold increase of the fluorescence lifetime ( 1 . 50 ± 0 . 03 ns , Figure 2D ) . The τG/τ0 ratio of 49 ± 1 . 0% corresponds to a distance of 10 . 9 ± 0 . 2 nm , that is a height of 8 . 4 nm above the lipid layer . Compared to the length of the 35mer DNA nanoruler ( 11 . 9 nm ) , the height corresponds to a tilting angle α = 45° , which is in good agreement with the tilting angle α obtained for the calibration . This observation not only supports the validation of the calibration , but also suggests loss in flexibility upon hybridization with the 15mer blocker strand forces the full duplex strand into an upstanding position ( Figure 2C ) . These results demonstrate the potential of GIET to quantify large-scale conformational rearrangements of biomolecules on membrane surfaces with nanometer precision . However , we did not observe lateral diffusion dynamics of anchored DNA strands as explored by fluorescence recovery after photobleaching ( FRAP , Figure 2—figure supplement 5 ) , thus ensuring that changes in fluorescence lifetime are not related to changes in the lateral organization within the lipid monolayer . We applied our new method to unravel the orientation and organization of proteins involved in vesicle tethering and fusion at the yeast vacuole . This process is initiated by the Rab7-like GTPase Ypt7 as a functional marker of the late endosomal membrane ( Figure 3A ) . Upon activation by its guanine nucleotide exchange factor ( GEF ) Mon1-Ccz1 on endosomes and vacuoles ( Nordmann et al . , 2010 ) , Ypt7-GTP recruits the 650 kDa heterohexameric HOPS tethering complex ( Brett et al . , 2008; Ostrowicz et al . , 2010 ) . This complex then tethers Ypt7-bearing membranes ( Hickey and Wickner , 2010; Ho and Stroupe , 2015; Lürick et al . , 2017; Orr et al . , 2015 ) , and catalyzes fusion of vacuoles or SNARE-carrying proteoliposomes ( Stroupe et al . , 2009; Wickner and Haas , 2000 ) . Our structural studies revealed that HOPS adopts a flexible tadpole-like conformation ( Bröcker et al . , 2012; Lürick et al . , 2017 ) . HOPS has four central subunits Vps11 , Vps18 , Vps16 , and Vps33 ( Rieder and Emr , 1997 ) , which are flanked by Vps39 and Vps41 as Rab-specific subunits at its tail and head , respectively ( Brett et al . , 2008; Bröcker et al . , 2012; Ostrowicz et al . , 2010 ) . However , the overall architecture and thus function of HOPS is controversial as two different structures have been observed by negative-stain electron microscopy ( Bröcker et al . , 2012; Chou et al . , 2016 ) . In its compact form , HOPS is about 30 nm in length and 10 nm in width ( Bröcker et al . , 2012 ) , whereas the more open conformation of a second study suggests an even longer particle ( Chou et al . , 2016 ) . Importantly , both structures were obtained from purified complexes in solution , bearing the question , how the functional complex is organized on membranes . To tackle this question by GIET , we reconstituted the entire tethering machinery on graphene-supported lipid monolayers using recombinantly expressed and purified components ( Figure 3B , C ) . Like all Rabs , Ypt7 has an N-terminal GTPase domain , followed by a hypervariable domain ( HVD ) , and a C-terminal prenyl group ( Goody et al . , 2017 ) . The HVD varies in length between different Rab proteins and is critical for functionality ( Li et al . , 2014 ) . Due to its composition , it has been assumed that the HVD adopts an extended conformation and would position the N-terminal GTPase domain of the Rab away from the membrane ( Burguete et al . , 2008 ) . However , this assumption has not been tested , neither alone nor in the presence of GEFs or effectors . To analyze the conformational organization of Ypt7 on membranes , we produced this protein N-terminally fused to mNeonGreen ( mNeon-Ypt7 ) . The purified protein was prenylated in vitro ( mNeon-pYpt7 ) in the presence of the chaperone GDI ( mNeon-pYpt7:GDI ) , and the complex was used for reconstitution into lipid layers ( Langemeyer et al . , 2018; Thomas and Fromme , 2016; Figure 3B , Figure 3—figure supplement 1 ) . In parallel , we generated a strain with mNeon-pYpt7 in vivo and observed full complementation ( Langemeyer et al . , 2020 ) . Full-length complexes of Mon1-Ccz1 and HOPS were purified to homogeneity and monodispersity ( Figure 3C ) , and we have previously shown that both function together with prenylated Ypt7 to promote fusion of proteoliposomes ( Langemeyer et al . , 2018 ) . To test membrane binding , we incubated mNeon-pYpt7:GDI with liposomes by destabilizing the interaction with GDI , and observed that Ypt7 recruitment required the prenyl anchor and GTP to prevent GDI-mediated extraction ( Figure 3D ) . We further analyzed the ability of activated pYpt7 to interact with HOPS , and thus incubated pYpt7-GTP carrying liposomes with purified HOPS ( Figure 3E ) . In agreement with previous work ( Lürick et al . , 2017; Orr et al . , 2015 ) , HOPS triggered liposome clustering in a pYpt7 and concentration-dependent manner ( Figure 3—figure supplement 2 ) . Efficient transfer of prenylated mNeon-pYpt7 to lipid monolayers ( on graphene ) and bilayers ( on glass ) and homogeneous distribution were confirmed by cLSM imaging ( Figure 3—figure supplement 3 ) . Binding of Mon1-Ccz1 to lipid mono- and bilayers ( Cabrera et al . , 2014; Poteryaev et al . , 2010 ) was confirmed by label-free RIF detection ( Figure 3—figure supplement 4A ) . Recruitment of HOPS to membrane-anchored mNeon-pYpt7-GTP resulted into significant changes of its mobility and the organization into dot-like structures ( Figure 3—figure supplement 5 , Video 1 , Video 2 ) . Such lateral heterogeneity of mNeon-pYpt7-GTP:HOPS was not observed for graphene-supported monolayers , in line with the lack of lateral mobility of this architecture . By quantifying the fluorescence lifetime of mNeonGreen on graphene and glass ( Figure 3—figure supplement 4B ) , the axial distances of active mNeon-pYpt7 from the membrane surface were determined based on the τG/τ0 ratio in the absence of additional effector proteins and in the presence of Mon1-Ccz1 and HOPS ( Figure 3G ) . pYpt7 was located surprisingly distal from the membrane in the absence of effector proteins with the mNeon fluorophore being located 4 . 5 ± 0 . 3 nm above the lipid monolayer surface . The site of mNeon fusion on Ypt7 is only ~1 nm away from the N-terminus of the HVD ( Wiegandt et al . , 2015 ) and therefore this observation suggests a rather extended conformation of the HVD . Upon interaction with Mon1-Ccz1 the distance of mNeon-pYpt7 decreased by 0 . 7 nm , while a further increase in the axial distance by 0 . 7 nm was observed in the presence of HOPS ( Figure 3F ) . The total of 30% distance difference suggests the large conformational changes of the HVD are functionally relevant for the interaction of Ypt7 with each complex . The HOPS complex binds Ypt7-GTP via its subunits Vps39 and Vps41 ( Brett et al . , 2008; Ostrowicz et al . , 2010 ) . Since these subunits have been mapped to opposite ends of the HOPS complex ( Bröcker et al . , 2012 ) , HOPS may be oriented differently on membranes , depending whether it binds Ypt7 via one or both sites . We therefore systematically mapped the axial arrangement of the HOPS complex bound to membrane-inserted pYpt7-GTP by GIET . Thus , we used HOPS variants with C-terminal yeast-enhanced GFP ( yEGFP ) on Vps39 , Vps11 , Vps18 , Vps16 , and Vps33 , respectively , and overexpressed and purified the complexes from yeast . Uncompromised complex assembly was verified by SDS-PAGE ( Figure 4A ) . yEGFP-tagged HOPS was specifically captured to lipid mono- and bilayers only after anchoring pYpt7-GTP ( Figure 4B ) . Significantly heterogeneous intensity distribution of yEGFP-tagged HOPS was observed on lipid bilayers but not on lipid monolayers on graphene ( Figure 4B , Figure 4—figure supplement 1 ) . These results suggest that HOPS clusters upon binding to diffusive membrane-anchored pYpt7-GTP , in line with the above-mentioned clustering and loss in mobility of mNeon-pYpt7-GTP upon binding to untagged HOPS ( Figure 3—figure supplement 5 ) . To determine the axial orientation of Ypt7-bound HOPS on membranes , GIET efficiencies were quantified by measuring fluorescence lifetimes on graphene ( τG ) and glass support ( τ0 ) . FLIM images of yEGFP-tagged HOPS bound to pYpt7-GTP showed homogeneous distribution on graphene , yielding a Gaussian-like distribution of fluorescence lifetimes ( Figure 4C , D ) . Interestingly , τG showed a broader distribution than τ0 , in line with some heterogeneity of the axial arrangement . The mean fluorescence lifetimes were used to estimate the average distance from the lipid head groups . For all yEGFP-fused subunits of HOPS , a significant decrease of τG compared to τ0 was found . The axial distances calculated for the different HOPS subunits are in line with an upright position of the complex bound to Ypt7 ( Figure 4E , F , Figure 4—figure supplement 2 ) : The two subunits at the putative HOPS tail , Vps39 ( 9 . 5 nm ) and Vps11 ( 8 . 5 nm ) , were found to be much closer to the membrane than those of the previously annotated head ( Vps16 at 12 . 0 nm , Vps18 at 11 . 4 nm , and Vps33 at 11 . 5 nm ) ( Figure 4G ) . These results suggest membrane anchoring of HOPS with membrane-anchored Ypt7-GTP occurs exclusively via interaction of Vps39 , in line with its higher affinity as compared to Vps41 ( Lürick et al . , 2017; Plemel et al . , 2011 ) . The ~8 nm distance of the head subunits Vps33 and Vps16 from Ypt7-GTP found by GIET analyses , however , is significantly lower than the ~20 nm head-to-tail distance ( between Vps33/16 and Vps39 ) observed in the EM structure of isolated HOPS ( Bröcker et al . , 2012 ) . These findings suggest that the HOPS complex adopts a more compact structure upon docking to membrane-anchored Ypt7-GTP . Negative-stain EM studies revealed that the HOPS complex can adopt different conformations in solution ( Bröcker et al . , 2012; Chou et al . , 2016 ) , which is in line with the broad distribution of GIET efficiencies observed in fluorescence lifetime histograms ( Figure 4D ) . To resolve such potential conformational heterogeneity of HOPS when bound to Ypt7 on membranes and possible transitions between these conformations , we turned to single-molecule imaging . The 97 . 7% transparency of graphene monolayer for visible light ( Nair et al . , 2008 ) allows single-molecule imaging on graphene by total internal reflection fluorescence ( TIRF ) microscopy with minimum loss of photons . For robust single-molecule GIET ( smGIET ) analysis , we used an anti-GFP nanobody site-specifically coupled to a photostable fluorophore ( Dy647NB ) at very low concentration ( 50 pM ) to sub-stochiometrically label yEGFP-tagged subunits in the HOPS complex at the head ( Vps33 ) and the tail ( Vps11 ) , respectively . Under these conditions , signals from specifically labeled individual HOPS complexes were detected by TIRF microscopy , as confirmed by photobleaching analyses on glass and graphene ( Figure 5A , B , Figure 5—figure supplement 1A ) . Analysis of ~100 single step bleaching fluorescence traces yielded IG/I0 of 0 . 76 ± 0 . 20 for HOPS Vps33 and 0 . 53 ± 0 . 14 for HOPS Vps11 ( Figure 5—figure supplement 1B , C ) . These mean values were in a good agreement with the ensemble lifetime measurements ( Supplementary file 2 ) . However , the large standard deviations corroborated significant structural heterogeneity in HOPS complexes that was already suggested by the broad fluorescence lifetime distribution observed in the ensemble experiments ( Figure 4D ) . Strikingly , single-molecule fluorescence of the HOPS complex bound to graphene-supported lipid monolayers showed pronounced fluctuations , which was not observed on glass ( Figure 5A , B , Video 3 ) . These fluctuations therefore indicate axial conformational dynamics of pYpt7-GTP bound HOPS on the membrane , which is monitored faithfully by distance sensitive smGIET . To quantify the conformational dynamics of the HOPS complex , we recorded time-lapse intensity traces of individual HOPS complexes on graphene and glass ( Figure 5C ) . HOPS Vps33-yEGFP was chosen based on its pronounced two lifetime populations observed on graphene in the ensemble experiments ( Figure 4E , Figure 4—figure supplement 2 ) . For pooled single-molecule intensities from traces of HOPS Vps33-yEGFP and HOPS Vps11-yEGFP , broad distributions were observed on graphene ( Figure 5D , E ) but not on glass ( Figure 5—figure supplement 2 ) . Analysis by Hidden Markov Modeling ( HMM ) ( McKinney et al . , 2006 ) identified three distinct states with ‘low’ ( L ) , ‘medium’ ( M ) , and ‘high’ ( H ) intensity , respectively , for both HOPS variants on graphene . The characteristic IG/I0 ratios , axial distances , and occupancies for each state are summarized in Table 1 . The similar changes of mean axial distance from L to M state ( ~2 nm ) in both HOPS Vps33-yEGFP and HOPS Vps11-yEGFP suggest a largely concerted axial movement of the entire HOPS complex . By contrast , the high discrepancy of distance changes during transitions from M to H ( Vps33: 4 . 4 nm; Vps11: 2 nm ) indicate a conformational re-organization of the HOPS complex itself , thus adopting an elongated conformation ( Figure 5F ) . In total , the conformational dynamic distance range on the membrane was ~7 nm for HOPS Vps33-yEGFP and 4 nm for HOPS Vps11-yEGFP , respectively . Furthermore , we obtained rate constants of the transition between L , M , and H states from HMM analysis for HOPS Vps33-yEGFP and HOPS Vps11-yEGFP ( Supplementary file 4 ) . Striking similarity of the transition probability densities were obtained for both labeled subunits , with the dominant transitions between L-M and M-H states ( Figure 5G , H ) . This observation supports correlated conformational dynamics of the L , M , and H states observed for HOPS Vps33-yEGFP and HOPS Vps11-yEGFP . Together , the time-resolved smGIET experiments reveal a 2-step conformational transition of the Ypt7-anchored HOPS complex from the ‘closed’ state L to the ‘open’ state H via an intermediate state M ( Figure 5I ) .
Quantitative insights into the structural organization and dynamics of protein complexes and large machineries at membranes is important to understand their function . Methods based on fluorescence-interference-contrast ( FLIC ) microscopy ( Braun and Fromherz , 1998 ) have been successfully employed to determine the axial organization of membrane proteins at ensemble level with at ~1 nm resolution ( Kiessling and Tamm , 2003; Kiessling et al . , 2018 ) . Resolving the axial conformational dynamics that involved in many membrane-associated processes , however , requires methods that provide sub-nanometer resolution with single-molecule sensitivity at sub-second timescale . To tackle this challenge , we have here introduced fluorescence quenching by GIET using graphene-supported lipid monolayer as a membrane model . Detailed calibration with DNA rulers confirmed excellent correlation of simulated and measured distance-GIET relationships between 3 nm and 15 nm , promising very good sensitivity up to a distance of ~25 nm from the graphene surface , thus substantially exceeding the ~10 nm limit of FRET . Next to rapid and robust ensemble measurements , we demonstrate highly sensitive single-molecule GIET detection yielding relative axial localization precision of 10% at video rate temporal resolution ( Materials and methods: Axial localization precision of smGIET ) . Thus , the spatial regime covered by GIET ideally fills the gap left by FRET and localization microscopy . Proof-of-concept experiments demonstrate the feasibility to reconstitute complex membrane-anchored multi-protein machineries onto graphene-supported lipid monolayers and to quantify their axial architecture and dynamics . We show that the HVD domain of Ypt7 adopts a strongly extended conformation , locating Ypt7 at a 4 . 5 nm distance above the membrane surface . This surprising observation could be explained by electrostatic repulsion of the negatively charged HVD . Significant changes in the axial positioning of Ypt7 were observed upon interaction with Mon1-Ccz1 and HOPS . Since Mon1-Ccz1 directly binds to the membrane ( Cabrera et al . , 2014 ) , its Ypt7 binding site appears to be located distally to the membrane and the HVD allows unhindered access . Importantly , ensemble and single-molecule GIET furthermore revealed the nanoscale organization and dynamics of the HOPS tethering complex upon recruitment to the membrane surface . While ensemble GIET measurements clearly identified the HOPS subunit Vps39 as the primary binding site to lipid-anchored Ypt7 , single-molecule GIET revealed rapid transition between three different conformational states with kinetics in the second to sub-second regime . Quantitative analysis of the kinetics suggests a 2-step transition from a ‘closed’ into an ‘open’ state involving concerted axial movement and conformational changes . Given the impaired diffusion in the graphene-supported lipid monolayer , however , the conformationally highly dynamic HOPS complex observed in our studies may represent an intermediate state prior to lateral HOPS clustering , which in turn could increase the efficiency of HOPS to tether incoming vesicles . By developing freely diffusive lipid architectures for graphene support , GIET will allow to dissect conformational organization and rearrangement of large protein complexes on membranes . Since GIET has negligible variation between spectrally different fluorophores ( Figure 2—figure supplement 2 ) , concerted conformational changes and axial movement can be potentially resolved by multicolor GIET . With graphene emerging as a routine support for transmission electron microscopy , correlative approaches using the same surface architecture can be envisioned .
Graphene monolayer was purchased from Graphenea , Spain ( Easy Transfer Monolayer , G/P-25–25 ) . Glass microscopy coverslips with thickness #1 . 5 and diameter of 24 mm were purchased from Carl Roth ( PK26 . 1 ) . 1 , 2-dioleoyl-sn-glycero-3-phosphocholine ( DOPC ) , 1 , 2-dioleoyl-sn-glycero-3-phosphoserine ( DOPS ) , 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine ( POPC ) , 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylethanolamine ( POPE ) , diacylglycerol ( DAG ) and DOGS-NTA were purchased from Avanti Polar Lipids , Alabama , USA . Phosphatidylinositol 3-phosphate ( PI-3-P ) was purchased from Echelon Biosciences Inc , Utah , USA . 2′ , 7′-difluoro-fluorescein conjugated with 1 , 2-dihexadecanoyl-sn-glycero-3-phosphoethanolamine ( OG488-DHPE ) was purchased from Thermo Fisher Scientific . ATTO488-1 , 2-dipalmitoyl-sn-glycero-3-phosphoethanolamine ( ATTO488-DPPE ) was obtained from ATTO-TEC GmbH , Siegen , Germany . DY647P1 maleimide was purchased from Dyomics GmbH , Jena , Germany . TrisNTA-DODA was synthesized as previously reported ( Beutel et al . , 2014 ) . DNA oligonucleotides were ordered from Integrated DNA Technologies , Inc The sequences and modifications are listed in Supplementary file 1A . Acetone was purchased from Merck ( Uvasol for spectroscopy , 100022 ) . Other reagents were purchased from Sigma Aldrich . Protein expression , purification and labeling mEGFP with an N-terminal hexahistidine tag ( H6-mEGFP ) was cloned in the plasmid pET21a . The protein was expressed in Escherichia coli BL21 ( DE 3 ) Rosetta cells , followed by purification of immobilized metal affinity chromatography ( IMAC ) and size exclusion chromatography ( SEC ) ( Wedeking et al . , 2015b ) . mEGFP without oligohistidine tag ( tagless mEGFP ) was expressed in E . coli . Lysed cells were heated to 80°C , centrifuged and the supernatant was purified by anion exchange column and SEC . Anti-GPF nanobody ‘enhancer’ ( Kirchhofer et al . , 2010 ) with a C-terminal hexahistidine tag ( NB-H6 ) was expressed in E . coli Rosetta as described previously ( Wedeking et al . , 2015a ) . For fluorescence labeling , a C-terminal cysteine was appended to the NB for conjugating with Dy647P1-malemide . The anti-GFP nanobody ‘enhancer’ fused to a C-terminal hexahistidine tag was cloned into pET21a ( pET21a-NB-H6 ) and expressed in E . coli BL21 ( DE 3 ) Rosetta cells . After cell lyses by sonication , NB-H6 was purified by immobilized metal affinity chromatography ( IMAC ) and size exclusion chromatography ( SEC ) . For fluorescence labeling , a short linker including an additional cysteine residue followed by an ybbR-tag , a PAS repeat sequence and a terminal hexahistidine tag ( a . a . : GSCGSGSKLDSLEFIASKLAPASPASPASPASPASLEHHHHHH ) was C-terminally fused to the NB . Expression and purification was performed as described for NB-H6 . The purified proteins were reacted with a twofold molar ratio of DY647P1 maleimide for 30 min and then purified by size exclusion chromatography . A degree of labeling close to 1 . 0 was obtained as quantified by UV-Vis spectroscopy . Endocytic proteins and protein complexes: ( 1 ) Rab GTPases and Gdi1 . pET24b-Ypt7 ( Cabrera et al . , 2014 ) , pET24d-GST-TEV-Ypt7 ( Lachmann et al . , 2012 ) , pET24d-GST-TEV-mNeon-Ypt7 and pGEX-6P-Gdi1 ( Thomas and Fromme , 2016 ) were transformed into competent E . coli BL21 ( DE 3 ) Rosetta cells . Rab GTPases and Gdi1 were purified as previously described with slight modifications ( Langemeyer et al . , 2018; Lürick et al . , 2017; Nordmann et al . , 2010 ) . The pET24b-Ypt7 , pET24d-GST-TEV-Ypt7 , pET24d-GST-TEV-mNeon-Ypt7 ( mNeon fused to Ypt7 via a GGSGx3 linker ) and pGEX-6P-Gdi1 were expressed in E . coli BL21 ( DE 3 ) Rosetta cells . Cells were grown in Luria broth ( LB ) medium containing the specific antibiotics until an OD600 of around 0 . 8 , before they were induced with 0 . 25 mM isopropyl-β-D-thiogalactoside ( IPTG ) for 18 hr at 16°C . For purification of the Rab GTPases , harvested cells were lysed by a Microfluidizer , Model M-110L ( Microfluidics , Newton , MA ) in 50 mM HEPES , pH 7 . 5 , 150 mM NaCl , 1 mM MgCl2 , 1 mM DTT , 1 mM phenylmethylsulfonyl fluoride ( PMSF ) , and 0 . 05-fold protease inhibitor cocktail . For purification of Gdi1 , lysis was performed in PBS supplemented with 5 mM β-mercaptoethanol and 1 mM PMSF . The lysate was centrifuged at 40 , 000 g for 30 min at 4°C , and the supernatant was incubated for 2 hr at 4°C with pre-equilibrated Ni-NTA agarose ( Macherey-Nagel , Germany ) for His-fused proteins or Glutathione Sepharose ( GSH ) 4B beads ( GE Healthcare ) for GST-fused proteins . Ni-NTA beads were washed with 50 ml lysis buffer lacking PMSF and protease inhibitor but supplemented with 10 mM imidazole . For elution , the imidazole concentration was increased to 300 mM . After elution , the buffer was exchanged via a PD10 column ( GE Healthcare ) containing no imidazole . GSH beads were extensively washed with 120 ml lysis buffer lacking PMSF and protease inhibitor cocktail , and proteins were eluted by cleaving the affinity-tag with TEV- or Precision-protease , respectively , for 2 hr at 16°C . ( 2 ) Rab GGTase and Rab Escort Protein . pET30-Mrs6 ( Pylypenko et al . , 2003 ) and pCDF-Duet-1-Bet2-Bet4 ( Thomas and Fromme , 2016 ) were transformed into E . coli BL21 ( DE 3 ) Rosetta cells . The Rab GGTase and the Rab escort protein were expressed and induced as Rab GTPases and Gdi1 . The Rab GGTase ( pCDF-Duet-1-Bet2-Bet4 ) and the Rab escort protein ( pET-Mrs6 ) were expressed and induced as described in 1 . 2 . Harvested cells were lysed in 50 mM Tris , pH 8 . 0 , 300 mM NaCl , 2 mM β-mercaptoethanol and 1 mM PMSF and centrifuged as described above . The supernatant was loaded on a pre-equilibrated Hi-Trap Ni-Sepharose column ( GE Healthcare ) . After extensive washing of the column with lysis buffer containing 30 mM imidazole but lacking PMSF , bound protein was eluted in a linear 30–300 mM imidazole gradient over 30 column volumes . Protein-containing fractions were pooled and dialyzed against buffer containing 50 mM HEPES , pH 7 . 5 , 150 mM NaCl , and 1 mM MgCl2 . The buffer was exchanged twice . ( 3 ) HOPS , yEGFP-fused HOPS and Mon1-Ccz1 . HOPS , yEGFP-fused HOPS and Mon1-Ccz1 were expressed in Saccharomyces cerevisiae and purified essentially as described before ( Lürick et al . , 2017 ) . For yEGFP-fused HOPS , yEGFP was fused to the C-terminus of Vps proteins via a linker of RTLNVDGSGAGAGAGAGAIL . yEGFP-fused HOPS and Mon1-Ccz1 complexes were expressed in S . cerevisiae . In short , cells were grown until an OD600 of around 8 . Harvested cells were resuspended in 50 mM HEPES , pH 7 . 4 , 300 mM NaCl , 1 . 5 mM MgCl2 , 1 mM DTT , 0 . 5 mM PMSF , 1x FY protease inhibitor mix ( Serva , Germany ) , and 10% glycerol . For purification of Mon1-Ccz1 , the salt concentration was decreased to 150 mM NaCl . Cell lysis in the presence of glass beads was conducted in a FastPrep-24 5G ( MP , Germany ) . After removal of the glass beads , the supernatant was centrifuged at 120 , 000 g for 1 hr at 4°C . Centrifuged lysate was incubated with pre-equilibrated immunoglobulin G ( IgG ) Sepharose ( GE Healthcare ) for 2 hr at 4°C . IgG beads were washed with 15 ml lysis buffer lacking PMSF and FY , and bound proteins were eluted by cleavage with TEV protease for 1 hr at 16°C . Strains and plasmids used in this study are listed in Supplementary file 3 . Rab–GDI complexes were obtained from prenylation reactions ( Thomas and Fromme , 2016 ) . 10 μM Rab GTPase pre-loaded with GDP was incubated with 9 μM GDI , 1 μM Rab escort protein ( REP ) Mrs6 , 1 μM geranylgeranyl transferase ( Bet2-Bet4 ) , and a sixfold excess of geranylgeranyl pyrophosphate in assay buffer ( 50 mM HEPES , pH 7 . 5 , 150 mM NaCl , 2 mM MgCl2 , 1 mM DTT ) for 1 hr at 37°C . Mrs6 and Bet2-Bet4 fused to a His6 tag were removed by subsequent incubation of the reaction with Ni-NTA Agarose ( Macherey-Nagel , Germany ) for 1 hr at 4°C . The stoichiometric complex containing the prenylated Rab and GDI was isolated from the supernatant by size exclusion chromatography . The functionality of prenylated Rabs was tested in membrane association and tethering assays as described before with modifications ( Lürick et al . , 2017 ) . For membrane association and tethering assays , lipid films were evaporated by a SpeedVac ( CHRIST , Germany ) and resuspended in HEPES-KOAc ( HK ) buffer ( 50 mM HEPES , pH 7 . 4 , and 120 mM KOAc ) . Liposomes as unilamellar vesicles were obtained by five freeze and thaw cycles in liquid nitrogen . For tethering assays , the 2 mM liposome suspension was extruded through polycarbonate filters ( 400 nm , 200 nm , and 100 nm pore size ) using a hand extruder ( Avanti Polar Lipids , Inc ) . To analyze the membrane association of mNeon-Ypt7 , 50 pmole mNeon-Ypt7 complexed with GDI was incubated with 50 nmole liposomes ( 62 mol % POPC , 18 mol % POPE , 10 mol % POPS , 8 mol % ergosterol , 1 mol % DAG , and 1 mol % PI-3-P ) in the presence or absence of GTP for 30 min at 27°C . Liposomes were sedimented by centrifugation for 20 min at 20 , 000 g . The fraction of membrane-bound mNeon-Ypt7 was determined by the fluorescent signal of the supernatant before and after sedimentation , which was quantified in a SpectraMax M3 fluorescence plate reader ( Molecular Devices , Germany ) . For analysis of HOPS-mediated tethering , liposomes containing 69 mol % POPC , 18 mol % POPE , 8 mol % ergosterol , 1 mol % DAG , 1 mol % ATTO488-DPPE , and 3 mol % DOGS-NTA or compensatory amounts of POPC were used . Ypt7 C-terminally fused to a His6 tag was loaded with GDP or GTP and afterwards targeted to liposomes via the lipid analogue DOGS-NTA ( Lürick et al . , 2017 ) . Liposomes lacking DOGS-NTA were directly loaded with prenylated Ypt7 . In this case , 50 pmole pYpt7:GDI complex was incubated with 50 nmole liposomes in the presence of GTP for 30 min at 27°C . For tethering reactions , 0 . 170 mM Ypt7-loaded liposomes were incubated with 50–350 nM HOPS complex or buffer , respectively , for 10 min at 27°C . Liposomal clusters were sedimented for 5 min at 1 , 000 g . The fraction of tethered liposomes in the pellet was determined on the basis of the ATTO488 fluorescent signal in the supernatant before and after sedimentation , using a SpectraMax M3 fluorescence plate reader . Silica-type substrates were used for coating with graphene monolayer . These include 1 × 1 cm2 silica-coated transducers for TIRFS-RIF detection and glass coverslips for microscopy imaging . Before coating , the substrates were cleaned by plasma cleaner ( Femto plasma system , Diener electronic GmbH/Germany ) . Coating of graphene monolayer on substrate was based on the manufacturer’s instruction . Briefly , the Easy Transfer Monolayer containing graphene monolayer with a thin protecting polymer film was cut into 0 . 5 × 0 . 5 cm2 pieces . The obtained piece was slowly emerged into MilliQ water to float on top of water . The polymer-protected graphene monolayer was fished out by a clean glass coverslip or TIRFS-RIF transducer from below , resulting in face-to-face contact of graphene monolayer with the substrate . The obtained substrate was left drying at room temperature for 30 min , followed by heating in a 150°C oven for 2 hr . The hot substrate was taken out and immediately stored under vacuum for 24 hr for cooling down . With the protecting polymer film , the obtained graphene-coated substrate could be stored at ambient conditions for weeks . Immediately before the experiments , the graphene-coated substrate was incubated in acetone for 1 hr , sequentially in isopropanol for another 1 hr , to remove the protecting polymer film on graphene . By blow-drying with N2 stream , the graphene monolayer-coated substrate was ready for use . For preparation of DOPC liposomes , 2 . 5 µmol DOPC was dissolved in chloroform in a 50 ml round bottom flask . For preparation of liposomes containing DOPC:DOPS ( 95:5 molar ratio ) , 2 . 4 µmol DOPC and 0 . 12 µmol DOPS dissolved in chloroform were mixed in a 50 ml round bottom flask . Similarly , a lipid mixture of 2 . 4 µmol DOPC and 0 . 125 µmol trisNTA-DODA was used for preparing liposomes of DOPC:trisNTA-DODA ( 95:5 molar ratio ) . Liposomes were prepared as small unilamellar vesicles ( SUVs ) by probe sonication as described before ( Beutel et al . , 2014 ) . To form solid-supported membranes on glass coverslips , 800 µl of liposome solution was added on top of a freshly cleaned coverslip . After incubation at room temperature for 20 min , the coverslip was rinsed with excess HBS to remove free vesicles . For binding His-tagged proteins on solid-supported membranes of DOPC:trisNTA-DODA , conditioning of trisNTA-DODA was carried out by loading with Ni2+ ions . The coverslip was rinsed with 250 mM EDTA and 200 mM imidazole , sequentially . After incubation with 10 mM NiCl2 for 10 min , the coverslip was washed by HBS buffer and 200 mM imidazole to remove possible non-specifically bound Ni2+ ions . A home-built set-up for simultaneous total internal reflection fluorescence spectroscopy ( TIRFS ) ( Gavutis et al . , 2006 ) and reflectance interference ( RIF ) detection ( Piehler and Schreiber , 2001; Schmitt et al . , 1997 ) has been described in detail before . For TIRFS-RIF detection of His-tagged protein binding , DOPC/trisNTA-DODA ( 95:5 molar ratio ) was injected into the flow chamber to form lipid bilayers on silica substrate , or lipid monolayer on graphene . Alternatively , solution-assisted lipid deposition ( SALD ) was used for forming lipid monolayer on graphene , in which ethanol was added to HBS to obtain a final mixture of HBS:EtOH ( 90:10 v/v ) . Conditioning of trisNTA-DODA by Ni2+ ion loading follows the same protocol as in vitro by using the flow-through system in the TIRFS-RIF setup . On the Ni2+-loaded trisNTA-DODA/DOPC lipid mono-/bilayers , 1 µM NB-H6 was injected , followed by injections of 100 nM tagless mEGFP , HBS buffer rinsing and imidazole washing for surface regeneration . For direct immobilization of H6-mEGFP , 1 µM H6-mEGFP was injected followed by imidazole washing . Mass signals of protein binding and fluorescence signals were recorded simultaneously in RIF channel and TIRFS channel , respectively . The ratio of fluorescence intensity IG/I0 was normalized to the amount of immobilized mEGFP according to Equation ( 1 ) . ( 1 ) IGI0 = ( m0mG ) ( IG_rawI0_raw ) where m0 and mG are the mass signals of mEGFP immobilized on silica and graphene , respectively , and I0_raw and IG_raw are the fluorescence intensities of mEGFP immobilized on silica and graphene , respectively . For DNA hybridization , liposomes with 250 µM DOPC/DOPS ( 95:5 molar ratio ) was injected to the surface of TIRFS-RIF transducer to form lipid layers on the solid support . Anchor strands with cholesterol modification were injected for immobilization on the surface . Sequentially , probe strand labeled with FAM was injected for hybridization with the anchor strand . All DNA concentrations were 1 µM . The running buffer was HBS buffer containing 5 mM Mg2+ for DNA hybridization ( HBS-Mg buffer , 20 mM HEPES , 150 mM NaCl , and 5 mM MgCl2 , pH 7 . 5 ) . Fluorescence ratio of IG/I0 was normalized to the amount of immobilized DNA according to Equation ( 1 ) . For calibration experiments using OG488-DHPE , liposomes containing 250 µM DOPC mixed with 0 . 1% molar ratio OG488-DHPE was used . IG/I0 was obtained by normalizing to the mass signals according to Equation ( 1 ) . Binding of Mon1-Ccz1 to lipid layers was probed as described before ( Cabrera et al . , 2014 ) . 250 µM DOPC was injected to the substrates for formation of lipid monolayers on graphene or lipid bilayers on silica , respectively , followed by injection of 100 nM Mon1-Ccz1 . For simulating the electrodynamic coupling of the excited fluorophore to graphene , the excited fluorophore was treated as an ideal electric dipole emitter and the graphene as a layer of matter with specific thickness and ( complex-valued ) bulk refractive index as described before ( Ghosh et al . , 2019 ) . Solving Maxwell’s equations of such a system leads to an expression for the emission power , S ( θ , d ) , of the electric dipole emitter as a function of dipole distance d and orientation ( described by the angle θ between the dipole axis and the vertical axis ) to the substrate . Considering the flexible linker and rapid rotation of the fluorophore in biomolecules , S ( θ , d ) was averaged over random orientations as a simplified S ( d ) . The relative fluorescence lifetime ( τG/τ0 ) was calculated as: ( 2 ) τGτ0 =S0⟮1−ϕ⟯S0+ϕS⟮d⟯where ϕ is the quantum yield ( QY ) , τ0 is the free-space lifetime in the absence of GIET , S0 is the free-space emission power of an ideal electric dipole emitter , S0=cnk04p2/3 , with c being the vacuum speed of light , k0is the wave vector in vacuum , n is the refractive index of water ( n = 1 . 33 ) , and p is the amplitude of the dipole moment vector . The τG/τ0 as a function of distance d was calculated for the following geometry: glass substrate ( n = 1 . 52 ) covered with single sheet graphene ( thickness = 0 . 34 nm , n = 2 . 76 + 1 . 40i for emission at 670 nm , or n = 2 . 68 + 1 . 21i for emission at 520 nm ) coated with lipid monolayer ( n = 1 . 44 , thickness = 2 . 5 nm ) , topped with water ( n = 1 . 33 ) . The emission maximum , QY , and τ0 of the fluorophores were: FAM ( 518 nm , 0 . 75 , 3 . 0 ns ) , EGFP ( 507 nm , 0 . 60 , 2 . 1 ns ) , mNeonGreen ( 517 nm , 0 . 80 , 2 . 8 ns ) , and Dy647P1 ( 667 nm , 0 . 27 , 1 . 3 ns ) . The emission maximum and QY were taken from the literature ( Mujumdar et al . , 1993; Shaner et al . , 2013; Zhang et al . , 2014 ) , values of τ0 were measured on glass-supported lipid bilayer in this work . Results of distance dependency and sensitivity analysis are summarized in Figure 2—figure supplement 2 . In the absence of static quenching , fluorescence intensity of fluorophore is proportional to its lifetime . Thus the relative fluorescence intensity follows the same theoretical prediction based on Equation ( 2 ) . To calibrate the distance-dependent GIET , the measured intensity ratios and lifetime ratios of DNA nanorulers need to be plotted vs the vertical distance of the FAM dye above graphene . The vertical distance d was calculated by considering tilting of the DNA strands on surface ( Wong and Pettitt , 2004 ) : ( 3 ) d=lDNA*sinα+lMLwhere lDNA is the end-to-end length of FAM dye on the hybridized probe strand to the 3´ end of the anchor strand . lDNA = 1 . 7 , 5 . 1 , 6 . 8 , 8 . 5 , 10 . 2 , 11 . 9 , and 17 . 0 nm . It is set as 0 nm for OG488-DHPE . α is the tilting angle between DNA and graphene surface . lML = 2 . 5 nm , is the thickness of lipid monolayer on graphene . The anchor DNA strands were modified with a cholesterol at 3´ end via a tri-ethylene glycol phosphate linker ( ‘Spacer 9’ of Integrated DNA Technologies , Inc ) . A full integration of the anchoring group into the phospholipid monolayer was assumed . Using the ‘lsqcurvefit’ function in Matlab , a global fitting of the intensity or lifetime ratios vs d to the simulated GIET curve of FAM yielded α of 42° ( intensity ) or 44° ( lifetime ) . Lifetime measurements were carried out on a confocal laser-scanning microscope ( FluoView 1000 , Olympus ) equipped with a FLIM/FCS upgrade kit from PicoQuant using a 60× ( NA 1 . 2 ) water-immersion objective ( UPLSAPO , Olympus ) . EGFP/mNeonGreen was excited either by the 488 nm line of an argon laser ( Olympus ) for cLSM/FRAP or by a picosecond pulsed 485 nm laser diode at 40 MHz repetition rate ( LDH-D-C-485 , Picoquant ) . Time-correlated single photon counting ( TCSPC ) was performed using the TCSPC module PicoHarp 300 ( PicoQuant together with a picosecond laser driver Sepia II ( PicoQuant ) and a single photon avalanche detector ( PicoQuant ) ) . Emission photons were filtered by a 500–550 nm bandpass filter ( BrightLine HC 525/50 , Semrock ) . TCSPC was acquired using point measurement or image mode ( i . e . FLIM mode ) . Acquisition time was more than 90 s to obtain >105 counts per sample for robust lifetime analyses . If not mentioned elsewhere , the TCSPC histograms were tail-fit with mono-exponential decay function using SymPhoTime64 integrated in the PicoQuant system . Only for mNeon-Ypt7 on graphene , the TCSPC histograms were fit with bi-exponential decay functions by fixing one component to the obtained average lifetime of mNeon-Ypt7 on glass ( <15% ) . 800 µl 250 µM DOPC/DOPS ( 95:5 molar ratio ) was added on top of graphene-coated glass coverslips . After incubation at room temperature for 5 min , the coverslips were washed for 5 times with 1 ml HBS-Mg buffer . The cholesterol-modified ssDNA anchor strands were added in the solution with a final concentration of 200 nM . After 5 min incubation , excess anchor strands were removed by washing for 5 times with HBS-Mg buffer . Probe strands of FAM-labeled ssDNA were added with a final concentration of 200 nM . After 5 min incubation , excess probe strands were removed by washing for 10 times with HBS buffer . For fluorescence lifetime measurements , a final concentration of 10 µM EDTA was added to the solution for preventing possible metal ion-mediated fluorescence quenching . TCSPC was acquired by point measurement mode . Lifetimes were obtained by monoexponential fitting of TCSPC curves . For FRAP experiments of DNA strands , a region of interest ( ROI ) was placed in areas with and without graphene , respectively . In FRAP experiments of DNA strands , a circular region of interest ( ROI ) with a radius of 10 μm was selected . Five pre-bleach images were recorded before the ROI was illuminated by 405 nm in 10 s . A background signal IBG was recorded from the center of the ROI immediately after photobleaching . Fluorescence recovery was monitored at a time resolution of 0 . 8 s for glass or 3 . 2 s for graphene . The fluorescence intensity in the ROI ( IROI ) was obtained by subtracting the background . A reference region outside the photobleached ROI was processed in the same way to obtain IREF . ROI intensity was normalized to the reference ( IROI/IREF ) for each time interval using Equation ( 4 ) : ( 4 ) IROIIREF= IROI_inside−IBGIROI_outside−IBGwhere IBG is the background signal , IROI_inside is the fluorescence intensity within the ROI , and IROI_outside is the intensity outside the photobleaching ROI . Plotting IROI/IREF vs time yields the time-lapse FRAP curve . Graphene-coated glass coverslips were incubated with 800 µl DOPC and washed as described above . Membrane association of 200 nM prenylated mNeon-Ypt7 or 150 nM prenylated Ypt7 complexed with GDI was conducted in the presence of 20 mM EDTA , pH 8 . 0 and 1 mM GTP for 30 min at 30°C . The loading reaction was stopped by addition of 25 mM MgCl2 and incubation for 10 min at room temperature . The coverslip was washed extensively with HBS ( >15 times ) to remove excess EDTA and Mg2+ ions . Where indicated , 100 nM Mon1-Ccz1 or HOPS were added to mNeon-pYpt7-GTP-loaded membranes for 5 min at room temperature . Before fluorescence lifetime determination , the unbound complexes in solution were removed by washing for 8 times with HBS . For loading of HOPS complexes containing yEGFP-fused subunits on membranes , 50 nM HOPS carrying one subunit ( Vps39 , Vps11 , Vps18 , Vps33 or Vps16 ) fused to yEGFP was added to pYpt7-GTP-loaded membranes for 5 min at room temperature . Before fluorescence lifetime determination , the coverslip was washed 5 times with HBS to remove unbound HOPS in solution . Fluorescent lifetimes of mNeon and yEGFP , respectively , were determined by TCSPC acquired in image mode ( FLIM ) . Fluorescence lifetime ratios were determined based on lifetimes obtained on graphene and glass , respectively , and the distance from the graphene surface d were calculated based on the calibration curve ( Supplementary file 2 ) . The distances from the monolayer surface h were calculated from d by subtracting the thickness of lipid monolayer ( 2 . 5 nm ) . For FRAP experiments of mNeon-pYpt7-GTP and mNeon-pYpt7-GTP in complex with Mon1-Ccz1 and HOPS , respectively , a ROI with a radius of 7 . 5 µm was illuminated with a 405 nm laser for 8 . 2 s . The fluorescence recovery was monitored for 72 s at a time resolution of 1 . 8 s ( excitation: 488 nm ) . The IROI/IREF for each time interval was obtained using Equation ( 4 ) as described above . All FRAP experiments were carried out at room temperature . Single-molecule imaging experiments were conducted by total internal reflection fluorescence ( TIRF ) microscopy with an inverted microscope ( Olympus IX71 ) equipped with a triple-line total internal reflection ( TIR ) illumination condenser ( Olympus ) and a back-illuminated electron multiplying ( EM ) CCD camera ( iXon DU897D , Andor Technology ) as described before ( Wilmes et al . , 2020 ) . A 150 × magnification objective ( UAPO 150×/NA 1 . 45 TIRFM , Olympus ) was used for TIR illumination resulting in an image pixel size of 107 nm . Image acquisition was performed at 30 frames per second using an exposure time of 32 ms per frame . All experiments were carried out at room temperature in presence of oxygen scavenger , that is 0 . 5 mg/ml glucose oxidase , 0 . 04 mg/ml catalase , 5% w/v glucose . 150 nM pYpt7-GTP was loaded onto glass-supported lipid bilayers and graphene-supported lipid monolayers for 30 min at 30°C . 50 nM HOPS with Vps33 or Vps11 fused to yEGFP was added in solution , respectively , followed by 10 times buffer washing . To ensure reliable single-molecule detection , 50 pM of dye-labeled NB was used to label the yEGFP-fused HOPS complex . The typical density of NB-labeled HOPS was ~0 . 2–1 molecule/µm2 . Photobleaching steps were counted manually from time-lapse single-molecule intensity traces as before ( Wedeking et al . , 2015b ) . The single-molecule intensity traces were obtained by using ImageJ software , in which a 5 × 5 pixel region of interest ( sROI ) was used for cropping individual molecules . More than 100 intensity traces were screened for counting the discrete steps for each sample ( Figure 5—figure supplement 1 ) . Single emitters in time-lapse TIRFM images were localized via a 2D Gaussian mask ( Arnauld et al . , 2008; Thompson et al . , 2002 ) . Intensity-time traces were built by grouping localizations within a 150 nm search radius of each other with a minimum observation time of 100 frames as described before ( Niewidok et al . , 2018 ) . In order to exclude crosstalk between traces only identified immobilization events with a minimum center-to-center distance of 500 nm apart were selected for further analysis . Individual traces were initially analyzed and evaluated for goodness-of-fit using the step transition and state identification ( STaSI ) algorithm ( Shuang et al . , 2014; Wedeking et al . , 2015b ) . The best fitting 33% of these traces were used to estimate optimal initial values ( initial state distribution , transition matrix and state emission functions ) for subsequent training of a Hidden Markov Model ( HMM ) with Gaussian state emissions ( Rabiner , 1989 ) . To the end , the most probable state sequence ( Viterbi path ) was calculated for each trace by assigning a state to each single-molecule localization . The relative state occupations were calculated as the number of localizations classified into the specific state with respect to all observations . State transition rates were calculated by estimating the dwell time of each state using a mono-exponential fit of identified state segments as well as the number of observed state transitions according to the method described before ( Yang et al . , 2018 ) . The state definition , occupancy and transition rates were summarized in Supplementary file 4 . The axial localization precision of smGIET was determined by comparing the root mean square deviation ( RMSD ) with the mean intensity of individual molecules . To avoid intensity fluctuations on graphene due to GIET , RMSD and mean intensities were quantified based on single-molecule detections on glass coverslips ( Figure 5—figure supplement 3 ) . For intensities in the range of 300–700 a . u . , the obtained ratios of RMSD to mean were very similar ( Supplementary file 5 ) . Based on the ratios , a relative error of 7 . 5 ± 0 . 5% was obtained for single-molecule detections . The axial localization precision depends on IG/I0 , which has a propagated relative error of 2 × 7 . 5% = 10 . 6% . Thus , the relative single-molecule axial localization precision was 10 . 6% of the corresponding distance . Given the sensitive range of GIET in 3–30 nm , the corresponding axial precision is in the range of 0 . 3–3 nm . All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials . Additional data related to this paper may be requested from the authors . | Proteins are part of the building blocks of life and are essential for structure , function and regulation of every cell , tissue and organ of the body . Proteins adopt different conformations to work efficiently within the various environments of a cell . They can also switch between shapes . One way to monitor how proteins change their shapes involves energy transfer . This approach can measure how close two proteins , or two parts of the same protein , are , by using dye labels that respond to each other when they are close together . For example , in a method called FRET , one dye label absorbs light and transfers the energy to the other label , which emits it as a different color of light . However , FRET only works over short distances ( less than 10nm apart or 1/100 , 000th of a millimeter ) , so it is not useful for larger proteins . Here , Füllbrunn , Li et al . developed a method called GIET that uses graphene to analyze the dynamic structures of proteins on membrane surfaces . Graphene is a type of carbon nanomaterial that can absorb energy from dye labels and could provide a way to study protein interactions over longer distances . Graphene was deposited on a glass surface where it was coated with single layer of membrane , which could then be used to capture specific proteins . The results showed that GIET worked over longer distances ( up to 30 nm ) than FRET and could be used to study proteins attached to the membrane around graphene . Füllbrunn , Li et al . used it to examine a specific complex of proteins called HOPS , which is linked to multiple diseases , including Ebola , measuring distances between the head or tail of HOPS and the membrane to understand protein shapes . This revealed that HOPS adopts an upright position on membranes and alternates between open and closed shapes . The study of Füllbrunn , Li et al . highlights the ability of GIET to address unanswered questions about the function of protein complexes on membrane surfaces and sheds new light on the structural dynamics of HOPS in living cells . As it allows protein interactions to be studied over much greater distances , GIET could be a powerful new tool for cell biology research . Moreover , graphene is also useful in electron microscopy and both approaches combined could achieve a detailed structural picture of proteins in action . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"structural",
"biology",
"and",
"molecular",
"biophysics",
"tools",
"and",
"resources"
] | 2021 | Nanoscopic anatomy of dynamic multi-protein complexes at membranes resolved by graphene-induced energy transfer |
While transcripts of neuronal mitochondrial genes are strongly suppressed in central nervous system inflammation , it is unknown whether this results in mitochondrial dysfunction and whether an increase of mitochondrial function can rescue neurodegeneration . Here , we show that predominantly genes of the electron transport chain are suppressed in inflamed mouse neurons , resulting in impaired mitochondrial complex IV activity . This was associated with post-translational inactivation of the transcriptional co-regulator proliferator-activated receptor gamma coactivator 1-alpha ( PGC-1α ) . In mice , neuronal overexpression of Ppargc1a , which encodes for PGC-1α , led to increased numbers of mitochondria , complex IV activity , and maximum respiratory capacity . Moreover , Ppargc1a-overexpressing neurons showed a higher mitochondrial membrane potential that related to an improved calcium buffering capacity . Accordingly , neuronal deletion of Ppargc1a aggravated neurodegeneration during experimental autoimmune encephalomyelitis , while neuronal overexpression of Ppargc1a ameliorated it . Our study provides systemic insights into mitochondrial dysfunction in neurons during inflammation and commends elevation of mitochondrial activity as a promising neuroprotective strategy .
Multiple sclerosis ( MS ) is a chronic inflammatory disease of the central nervous system ( CNS ) and the most frequent non-traumatic cause of neurological impairment during early adulthood . Neuronal loss occurs already from disease onset and correlates best with irreversible disability in MS patients ( Fisher et al . , 2008; Fisniku et al . , 2008; Tallantyre et al . , 2010 ) . While there has been substantial progress in the understanding and treatment of the immune response , the pathogenesis of concurrent neuronal damage is incompletely understood . Recently , two drugs , ocrelizumab and siponimod , have been licensed for the treatment of progressive MS patients; however , their mode of action relies on immune regulation ( Kappos et al . , 2018; Montalban et al . , 2017 ) . Since there are currently no therapies available to counteract the progression of neurodegeneration in MS patients ( Dendrou et al . , 2015; Feinstein et al . , 2015 ) and repurposing of drugs to directly target neurodegeneration in MS has been disappointing ( Chataway et al . , 2020 ) , a better understanding of the molecular processes that determine neuronal cell loss in MS is urgently needed . Neuronal loss in MS and its animal model , experimental autoimmune encephalomyelitis ( EAE ) , has been associated with enhanced production of reactive oxygen and nitrogen species by immune cells and with increased iron accumulation in the gray matter ( Stephenson et al . , 2014 ) . Both processes can lead to damage of neuronal mitochondria with subsequent metabolic failure ( Campbell et al . , 2011 ) . Moreover , disruption of neuronal ion homeostasis ( Craner et al . , 2004; Friese et al . , 2014 ) and aggregation of neuronal proteins ( Schattling et al . , 2019 ) consume high amounts of energy that might further drive neuroaxonal injury . Furthermore , excessive activation of calcium-dependent processes and neuronal calcium overload seems to be another important component of neuronal injury ( Friese et al . , 2007; Schattling et al . , 2012; Witte et al . , 2019 ) . While identifying druggable targets that specifically induce neuronal resilience has been extremely difficult due to insufficient insights into key modulators , mitochondria could serve as an important hub as they are pivotal for both energy production and calcium homeostasis ( Vafai and Mootha , 2012 ) . Mitochondria usually have a high calcium buffering capacity , which is driven by their negative membrane potential that is generated by the activity of oxidative phosphorylation ( Kann and Kovács , 2007; Zorova et al . , 2018 ) . However , overload of mitochondria with calcium as a consequence of CNS inflammation can result in inappropriate activation of the mitochondrial permeability transition pore ( PTP ) with subsequent mitochondrial swelling and cell death ( Giorgi et al . , 2018; Rizzuto et al . , 2012 ) , which is one of the neuropathological hallmarks in neurons during EAE ( Nikić et al . , 2011 ) . Inhibiting the mitochondrial matrix protein cyclophilin D , a regulator of the PTP can partially counteract this dysregulation ( Forte et al . , 2007 ) . Additionally , to counteract ion imbalance , a higher activity of ATP-dependent ion pumps would be required . However , postmortem studies of MS patients’ CNS tissue revealed a compromised neuronal ATP production by oxidative phosphorylation as decreased mitochondria complex IV activity in demyelinated axons and neurons was detected ( Campbell et al . , 2011; Mahad et al . , 2009 ) . Similarly , mitochondrial gene expression is suppressed in motor neurons of the spinal cord in EAE mice ( Schattling et al . , 2019 ) , further supporting a comprised mitochondrial function during CNS inflammation . Thus , excess neuroaxonal calcium , together with mitochondrial dysfunction , has been postulated to trigger neuroaxonal injury observed in MS and EAE . This would be predicted to lead to elevated levels of calcium within neuronal and axonal mitochondria , further perpetuating mitochondrial dysfunction and neuronal injury . However , it is currently unknown whether interventions that increase mitochondrial energy production in neurons can rescue neurodegeneration during CNS inflammation . Here , we discovered by an unsupervised survey of neuronal gene expression during CNS inflammation that genes involved in the electron transport chain ( ETC ) , especially in complex I and IV , are repressed in motor neurons . While this is not accompanied by a decrease of mitochondrial numbers in motor neuronal somata , we detected a decrease in neuronal mitochondrial complex IV activity . This implies a lower activity of neuronal mitochondria during EAE , and we were able to associate that with a post-translational inactivation of peroxisome proliferator-activated receptor gamma coactivator 1-alpha ( PGC-1α ) , one of the master regulators of mitochondrial numbers and function . Notably , neuronal overexpression of Ppargc1a , which encodes for PGC-1α , led to an increase of mitochondrial activity , especially in complex IV activity and substantially elevated calcium buffering capacity . Subjecting these mice to EAE resulted in a significantly better recovery from clinical disability in comparison to wild-type controls . Together , induction of neuronal mitochondrial activity commends as a promising therapeutic approach to counteract inflammation-induced neurodegeneration .
Since we have previously discovered that mitochondrial gene transcripts are markedly suppressed in motor neurons during CNS inflammation in the EAE model ( Schattling et al . , 2019 ) , we first investigated whether unifying biological processes are dysregulated . Therefore , we derived a gene expression signature of motor neurons in spinal cord during CNS inflammation ( Schattling et al . , 2019 ) and tested for gene set enrichment of biological process gene ontology ( GO ) terms . While the majority of terms was enriched during inflammation , only a confined group of 29 terms showed strong de-enrichment ( Figure 1A , dashed line ) and was dominated by terms representing mitochondrial function ( Figure 1B ) . A recurring theme was the suppression of the ETC ( Figure 1C ) . Suppression of gene transcripts that are involved in the ETC , which is the key mechanism of oxidative phosphorylation , can either result in reduced number of mitochondria or an impairment of mitochondrial function . Therefore , we quantified mitochondria content in the soma of motor neurons in EAE mice at the chronic stage in comparison to healthy control mice , but did not detect any difference in mitochondria area normalized to neuronal size ( Figure 1D , Figure 1—figure supplement 1A ) . However , we detected a tendency towards an increase in overall numbers of motor neuronal mitochondria ( Figure 1—figure supplement 1B ) that was based on an increased size of the neuronal cell body during EAE ( Figure 1E ) . By contrast , the size of mitochondria did not differ in motor neurons of healthy and EAE mice ( Figure 1—figure supplement 1C ) . Dismissing differences in mitochondrial content , we reasoned that downregulation of ETC transcripts could result in a compromised oxidative phosphorylation activity . By analyzing COX histochemistry that represents complex IV activity , we detected a significant decrease in absolute neuronal complex IV activity in the entire gray matter and ventral horn of acute EAE animals ( day 15 post immunization ) and in chronic EAE animals ( day 30 post immunization ) in comparison to healthy control mice ( Figure 1F ) . Normalization to HuC/HuD-positive neurons revealed that the reduced complex IV activity at day 30 was partly driven by neuronal loss , whereas at day 15 the reduced complex IV activity was independent of neuronal numbers ( Figure 1G , Figure 1—figure supplement 1D–E ) . Next we interrogated whether a unifying mechanism can explain the inflammation-induced downregulation of gene transcripts that are involved in ETC with consecutively impaired complex IV activity in neurons at the acute stage of EAE . Notably , we detected that many of the genes that drive the downregulation of the ETC theme are usually induced by the transcriptional coactivator Ppargc1a ( Lucas et al . , 2014; Puigserver et al . , 1998; Figure 2A ) . Consequently , we asked whether Ppargc1a could be the sought-after unifying factor that is functionally disturbed . Motor neuronal Ppargc1a expression itself was not altered in our previously published dataset comparing acute EAE to controls ( Figure 2B ) . After confirming this result by an additional translating ribosome affinity purification of motor neuronal transcripts followed by Ppargc1a qPCR ( Figure 2C ) and immunoblot of PGC-1α of the cervical spinal cord of EAE and healthy mice ( Figure 2D , E ) , we concluded that transcriptional or translational Ppargc1a/PGC-1α regulation cannot be the explanation of the observed ETC downregulation . Besides quantity , PGC-1α can be post-translationally inactivated by phosphorylation at serine 570 ( Li et al . , 2007; Xiong et al . , 2010 ) , thereby preventing the recruitment of PGC-1α to its regulated promoters , which we tested next . After validation of phosphorylation specificity of the pPGC-1αS570 antibody ( Figure 2—figure supplement 1 ) , we detected a significant increase in phosphorylated-serine570 PGC-1α in spinal cords of acute EAE mice in comparison to healthy control mice ( Figure 2E , F ) . Thus , we concluded that due to the phosphorylation of PGC-1α during acute stage of EAE that is associated with downregulation of its target gene transcripts and compromised mitochondrial function , Ppargc1a could be a potential target to rescue mitochondrial function in EAE-affected motor neurons . Next , we reasoned that elevating neuronal Ppargc1a could restore mitochondrial function and serve as a strategy to improve neuronal resilience in CNS inflammation . We utilized transgenic mice with a neuron-specific overexpression of Ppargc1a driven under the Thy1 promoter ( Thy1-Ppargc1a ) ( Mudò et al . , 2012 ) , which is expressed in mature neurons ( Figure 3—figure supplement 1A ) . We could not observe any differences in the health status of these mice . We made sure that elevated transgenic DNA copies of Ppargc1a in neurons ( Figure 3—figure supplement 1B ) resulted in elevated mRNA expression of Ppargc1a in cortex , hippocampus , and spinal cord as well as in cortical and hippocampal primary neurons of Thy1-Ppargc1a mice in comparison to wild-type mice ( Figure 3—figure supplement 1C–D ) . Additionally , we revealed that Thy1-Ppargc1a mice showed an increase of Ppargc1a-dependent transcriptional targets in primary neurons ( Figure 3—figure supplement 1E–F ) and cytochrome c oxidase subunit 4 ( Cox4i1 ) and citrate synthase ( Cs ) in CNS tissue ( Figure 3—figure supplement 1G–I ) . Increased mRNA level led to a corresponding increase in protein amount of PGC-1α in primary neurons ( Figure 3—figure supplement 1J ) and spinal cord , which we confirmed by staining for the transgenically fused FLAG-tag ( Figure 3—figure supplement 1K ) . Besides Cox4i1 , in primary hippocampal neurons , Ppargc1a overexpression induced 11 of the 17 Ppargc1a regulated ETC genes , which were downregulated in EAE ( Figure 3A ) . Accordingly , neuronal Ppargc1a overexpression resulted in increased mitochondrial content in primary neurons ( Figure 3B ) and CNS tissue ( Figure 3C ) . To determine the mitochondrial activity in these mice , we analyzed whether Ppargc1a-mediated Cox4i1 induction resulted in increased complex IV activity . We detected an elevated complex IV activity in the gray matter and ventral horn of Thy1-Ppargc1a mice , which was not due to a higher number of NeuN-positive neurons ( Figure 3D ) . Metabolically , primary neurons of Thy1-Ppargc1a mice showed an elevated oxygen consumption rate and maximum respiratory capacity ( Figure 3E , Figure 3—source data 1 ) , which was independent of neuronal numbers ( Figure 3—figure supplement 1L ) . In concordance with higher activity of oxidative phosphorylation , primary neurons of Thy1-Ppargc1a mice retained higher tetramethylrhodamin-ethylester ( TMRE ) fluorescence intensity than wild-type controls , representing a hyperpolarized mitochondrial membrane potential ( Figure 3F ) . As CNS inflammation results in substantial neuronal calcium influx ( Witte et al . , 2019 ) leading to neurodegeneration , which could potentially be buffered by mitochondria , we next asked whether increased neuronal mitochondrial activity and membrane potential in Thy1-Ppargc1a neurons could alleviate toxic calcium levels . By analyzing spontaneously active cortical primary neuronal cultures ( cPNC ) transduced with the virally encoded calcium indicator GCaMP6f ( Figure 4A ) , we recorded a significantly faster clearance of intracellular calcium concentrations and decreased amount of mean calcium in spontaneously active Thy1-Ppargc1a primary neurons in comparison to wild-type control neurons , whereas the amplitude and number of calcium transients did not differ ( Figure 4B , C ) . The results were confirmed with a chemical calcium indicator ( Figure 4D and Video 1 ) . As we assumed that inactivation of neuronal Ppargc1a during CNS inflammation aggravates mitochondrial dysfunction and spurs neuronal vulnerability , we first tested whether this hypothesis holds true in Ppargc1a-deficient neurons . We used Ppargc1aflx/flx × Eno2Cre+ mice to specifically delete Ppargc1a in neurons as Eno2 is widely expressed in neurons ( Figure 5—figure supplement 1A ) . Although Eno2 expression is also present by immune cells ( Heng et al . , 2008 ) , we could exclude Ppargc1a expression in immune cells ( Figure 5—figure supplement 1B , Heng et al . , 2008 ) , therefore securing that Ppargc1aflx/flx × Eno2Cre+ mice will not result in altered immune responses . Neuronal deletion of Ppargc1a resulted in decreased Ppargc1a mRNA levels ( Figure 5—figure supplement 1C ) and Ppargc1a -regulated downstream targets in hippocampus ( Figure 5—figure supplement 1D ) , cortex ( Figure 5—figure supplement 1E ) , and spinal cord ( Figure 5—figure supplement 1F ) . Consequently , neuron-specific Ppargc1a knockout mice ( Ppargc1aflx/flx × Eno2Cre+ ) showed a more severe EAE disease course ( Figure 5A , Figure 5—source data 1 ) and by using NeuN-staining an increased neuronal loss in the gray matter and ventral horn in comparison to Ppargc1aflx/flx control EAE mice ( Figure 5B ) . Since NeuN-staining can be compromised in the chronic phase of EAE , we corroborated our findings by using the neuronal marker HuC/D as an alternative staining of spinal cord neurons ( Figure 5—figure supplement 1G ) . Notably , Ppargc1a deletion in neurons did not result in different neuronal numbers in healthy mice by using NeuN staining ( Figure 5—figure supplement 1H ) . As we observed that neuronal Ppargc1a overexpression led to an increased oxidative phosphorylation and improved calcium buffering capacities , we next tested its translatability to the preclinical MS model . EAE induction in Thy1-Ppargc1a and wild-type littermate control mice resulted in a comparable maximum score in the acute phase of the disease . However , Thy1-Ppargc1a mice showed a significantly better recovery from clinical disability in comparison to wild-type controls ( Figure 5C , Figure 5—source data 2 ) . Notably , this protection is unlikely to be driven by an impaired immune response as numbers of infiltrating CD3+ T cells , activated IBA1+ microglia , and MAC-3+ macrophages were similar to WT mice ( Figure 5D ) . Moreover , neuronal overexpression of Ppargc1a resulted in a rescue of complex IV activity during acute EAE in the entire gray matter and ventral horn ( Figure 5E ) , leading to a significant reduction of neuronal loss in the gray matter and ventral horn and more intact axons at day 40 post immunization by using NeuN or HuC/HuD as neuronal marker ( Figure 5F , Figure 5—figure supplement 1I ) . Neuronal Ppargc1a overexpression did also not influence neuronal numbers in healthy mice by using NeuN staining ( Figure 5—figure supplement 1J ) . Thus , induction of neuronal mitochondrial activity represents an attractive neuroprotective strategy in CNS inflammation by compensating mitochondrial dysfunction .
Here , we show that neuronal oxidative phosphorylation is compromised during CNS inflammation , which directly contributes to neuronal vulnerability and can be counteracted by induction of neuronal mitochondrial activity . In an unsupervised survey , we discovered that expression of mitochondrial genes is massively suppressed in motor neurons during CNS inflammation ( Schattling et al . , 2019 ) and here in particular genes that participate in oxidative phosphorylation . This sparked our interest to explore upstream mechanisms that coordinate this mitochondrial shutdown and whether counter-regulation could rescue neuronal integrity . Our observation is in concordance with other reports that showed alterations of mitochondria in the brain of MS patients with compromised oxidative phosphorylation in demyelinated axons and neurons ( Campbell et al . , 2011; Mahad et al . , 2009 ) . This could potentially contribute to mitochondrial swelling and dysfunction of axons in EAE ( Nikić et al . , 2011; Sadeghian et al . , 2016 ) . While we could not detect a decrease in somatic neuronal mitochondrial numbers in the spinal cord during EAE , we discovered a compromised complex IV activity in the spinal cord gray matter . This compromised oxidative phosphorylation during CNS inflammation is likely coordinated by inactivation of PGC-1α , a transcriptional coactivator that acts as a master switch of mitochondrial function ( Mootha et al . , 2003; Puigserver et al . , 1998 ) . However , in contrast to MS patients , in which a marked decrease in cortical PGC-1α expression has been reported ( Witte et al . , 2013 ) , we could not detect a regulation of Ppargc1a in inflamed mouse motor neurons . By contrast , our data show that CNS inflammation results in a post-translational modification of PGC-1α ( Fernandez-Marcos and Auwerx , 2011 ) . Increased phosphorylation of serine 570 leads to an inactivation of PGC-1α , which could be mediated by increased AKT activity ( Li et al . , 2007 ) . Three different isoforms of AKT exist . Whereas AKT1 and AKT3 are mainly expressed in neurons , AKT2 is predominantly expressed in astrocytes ( Levenga et al . , 2017 ) . So far , the exact function of AKT in neurons is not clear , but AKT1 has been implicated in late long-term potentiation ( Levenga et al . , 2017 ) and AKT3 in neuronal growth ( Adams et al . , 2016; Rivière et al . , 2012 ) . AKT is activated by phosphatidylinositol 3-kinase ( PI3K ) , and the PI3K/AKT signaling pathway is known to reduce apoptosis and promote survival and proliferation ( Brunet et al . , 1999 ) and thereby counteracts neuronal cell death ( Peltier et al . , 2007 ) . As TNF-α activates the PI3K/AKT pathway ( Gu et al . , 2006; Li et al . , 2017; Osawa et al . , 2001 ) , upregulation of AKT could function as an immediate strategy of neurons against cell death during inflammation . As a trade-off , prolonged downregulation of mitochondrial gene transcripts will reduce ATP levels that could limit neuronal integrity during chronic inflammation . Thus , tuning Ppargc1a could act as a switch that modulates oxidative phosphorylation by regulating respective mitochondrial genes under challenging environmental conditions , such as CNS inflammation . This seems to be particularly relevant in tissues with a high energy demand , for example , muscle , liver , and CNS in which Ppargc1a is highly expressed . While astrocytes are mostly relying on glycolysis , neurons mainly generate their energy by oxidative phosphorylation with low glycolytic capacities ( Camandola and Mattson , 2017 ) . Therefore , mitochondrial damage and inactivation is particularly deleterious for neuronal metabolism . That Ppargc1a is important for neuronal health is supported by our observation of an aggravated EAE disease course in neuron-specific Ppargc1a-deficient mice . This is in accordance with the global Ppargc1a knockout mice that showed neurological symptoms such as myoclonus , dystonia , and limb clasping that was attributed to axonal striatal degeneration ( Lin et al . , 2004 ) , although we were not able to detect this phenotype in our neuron-specific Ppargc1a knockout mice . While a recent report using Ppargc1a overexpression under the Eno2 promoter led to an ameliorated EAE disease course ( Dang et al . , 2019 ) , we believe that this is mainly driven by suppressing the immune response and not by a neuron-intrinsic effect . As Eno2 is also expressed in immune cells and especially T cell function is influenced by altered mitochondrial activity ( Buck et al . , 2016 ) , in accordance with our notion , the authors reported a reduced immune cell infiltration , less demyelination , a later EAE onset , and an ameliorated EAE disease course ( Dang et al . , 2019 ) , which are typical features for an immunosuppressive phenotype . By contrast , our data show that specific neuronal overexpression of Ppargc1a elevated mitochondrial biogenesis and an improved activity of neuronal oxidative phosphorylation , particularly complex IV , which also resulted in a higher overall mitochondrial membrane potential . Notably , these alterations equipped mitochondria with an improved calcium buffering capacity . Besides MS , rise in neuronal intracellular calcium is one of the hallmarks of neurodegenerative diseases ( Mattson , 2007 ) . Several mechanisms are relevant that lead to intracellular neuronal Ca2+ accumulation in EAE ( Siffrin et al . , 2015 ) , among them upregulation of voltage-gated sodium channels , Na+-Ca2+ antiporters ( Craner et al . , 2004 ) , Ca2+ permeable ion channels ( Friese et al . , 2007; Schattling et al . , 2012 ) , and nanoscale ruptures of the axonal plasma membrane ( Witte et al . , 2019 ) . Therefore , the improved calcium buffering capacity in Thy1-Ppargc1a neurons could efficiently rescue calcium overload and neuronal demise , which was indicated by an increased neuronal survival during EAE . Calcium uptake by mitochondria holds the promise as a therapeutic strategy for several neurodegenerative diseases ( Lee et al . , 2018; Parone et al . , 2013 ) . As neuronal mitochondrial deficiency is also associated with other neurodegenerative diseases , for example , Alzheimer’s disease ( Pannaccione et al . , 2020 ) , induction of mitochondrial activity could serve as a unifying neuroprotective approach ( Murphy and Hartley , 2018 ) . Similarly , downstream targets of Ppargc1a are repressed in dopaminergic neurons of Parkinson's disease patients ( Zheng et al . , 2010 ) in which neuronal overexpression of Ppargc1a protects dopaminergic neurons in its mouse model ( Mudò et al . , 2012 ) . As we detected a higher neuronal resilience in Thy1-Ppargc1a mice during CNS inflammation , pharmacological increase of neuronal mitochondria and oxidative phosphorylation could be a promising neuroprotective strategy . Some drugs have already been described to induce Ppargc1a in different tissues ( Dumont et al . , 2012; Hofer et al . , 2014; Noe et al . , 2013; Ye et al . , 2012 ) ; however , we could not confirm neuronal induction of Ppargc1a by treating mice with bezafibrate or using transient receptor potential cation channel subfamily V member 4 ( Trpv4 ) knockout mice ( data not shown ) . Another possible drug candidate is sirtuin-1 that activates PGC-1α by deacetylation , and its neuronal overexpression leads to an ameliorated EAE disease course ( Nimmagadda et al . , 2013 ) . Besides extrinsic induction or activation of Ppargc1a , it can also be induced by intrinsic mechanisms such as cold exposure or exercise ( Lin et al . , 2005 ) . Exercise ameliorates the EAE disease course ( Klaren et al . , 2014; Rossi et al . , 2009 ) and shows several benefits in MS patients ( Motl et al . , 2017 ) ; however , the proof of a neuroprotective effect that is mediated by Ppargc1a induction is currently outstanding . Taken together , we provide evidence for a therapeutic potential of inducing mitochondrial activity in inflammation-induced neurodegeneration , supporting further studies that aim at finding drugs to target this pathway in neurons .
All mice ( C57BL/6J wild type [The Jackson Laboratory , Bar Harbor , USA] , Thy1-Flag-Ppargc1a [Thy1-Ppargc1a] mice on a C57BL/6J genetic background provided by D . Lindholm [Mudò et al . , 2012] , Eno2Cre+ mice [The Jackson Laboratory] , Ppargc1aflx/flx mice [The Jackson Laboratory] , and Chat-L10a-eGFP mice [Heiman et al . , 2008] ) were kept under specific pathogen-free conditions in the central animal facility of the University Medical Center Hamburg-Eppendorf , Hamburg , Germany . Neuron-specific knockout mice were generated by crossing Ppargc1aflx/flx with Eno2Cre+ mice . Mice were grouped housed in a facility with 55–65% humidity at 24 ± 2°C with a 12 hr light/dark cycle and had free access to food and water . Sex- and age-matched adult animals ( 10–20 weeks of age ) were used in all experiments . Wild-type mice were allocated to EAE or healthy group by cages . All procedures were carried out in accordance with the ARRIVE guidelines ( Kilkenny et al . , 2010 ) . Mice were anesthetized with isoflurane 1–2% v/v oxygen and immunized subcutaneously with 200 μg myelin oligodendrocyte glycoprotein 35–55 ( MOG35–55 ) peptide ( peptides and elephants ) in complete Freund’s adjuvant ( BD ) containing 4 mg/ml Mycobacterium tuberculosis ( BD , Franklin Lakes , USA ) . A 200 ng pertussis toxin ( Merck , Darmstadt , Germany ) was injected intravenously on the day of immunization and 48 hr later . Animals were scored daily for clinical signs by the following system: 0 , no clinical deficits; 1 , tail weakness; 2 , hind limb paresis; 3 , partial hind limb paralysis; 3 . 5 , full hind limb paralysis; 4 , full hind limb paralysis and forelimb paresis; and 5 , premorbid or dead . Animals reaching a clinical score ≥4 were sacrificed according to the regulations of the Animal Welfare Act . The experimenters were blinded to the genotype until the end of the experiment , including data analysis . Sex- and age-matched adult animals ( 8–12 weeks of age ) were used in all experiments . For Thy1-Ppargc1a mice and wild-type controls , three independent EAE experiments were conducted , and the data were pooled for final analysis . For Ppargc1aflx/flx × Eno2Cre+ EAE and Ppargc1aflx/flx controls , one EAE was conducted . For analysis of the disease course and weight , we only included animals that received a score ≥1 until day 15 and survived until the end of the experiment . Animals were either analyzed at acute stage of EAE ( day 12 to day 16 after immunization ) or chronic stage ( day 30 to day 40 after immunization ) . Gene signature of motor neurons during CNS inflammation was generated by ranking all expressed genes by the DESeq2-derived t statistics based on the comparison of healthy versus EAE motor neurons ( Schattling et al . , 2019 ) . Enrichment analysis was performed using the function ‘gseGO’ of the R package clusterProfiler ( Yu et al . , 2012 ) on biological process GO terms with at least 50 members . Gene sets with a Benjamini–Hochberg adjusted p value <0 . 05 were considered significant and ordered by their normalized enrichment score , with positive values indicating enrichment and negative values indicating de-enrichment . Core enrichment genes driving the de-enrichment of the term ‘GO:0022900 electron transport chain’ were extracted from clusterProfiler results and plotted as heatmap of gene expression values after variance stabilizing transformation . Plotting was performed with the R packages ggplot2 ( https://ggplot2 . tidyverse . org; Wickham , 2016 ) , clusterProfiler and tidyheatmaps ( Engler , 2020 ) . For the identification of Ppargc1a-induced downstream targets , a published microarray dataset of Ppargc1a-overexpressing SH-SY5Y cells was used ( Lucas et al . , 2014 ) . The expression matrix was downloaded via GEOquery ( GSE100341 ) and analyzed using limma . Genes with a false discovery rate ( FDR ) -adjusted p value <0 . 05 and a log2 fold change >1 in Ppargc1a overexpression were considered positively regulated Ppargc1a targets . Mice were anesthetized intraperitoneally with 100 µl solution ( 10 mg/ml esketamine hydrochloride [Pfizer , New York City , USA] , 1 . 6 mg/ml xylazine hydrochloride [Bayer , Leverkusen , Germany] dissolved in water ) per 10 g of body weight . For histopathology and immunohistochemistry , mice were perfused with 4% paraformaldehyde ( PFA ) , cervical spinal cord tissue was dissected , fixed for 30 min with 4% PFA , and then transferred to 30% sucrose in phosphate-buffered saline ( PBS ) at 4°C . Cervical spinal cord was used as it was shown that clinical scores correlate better with cervical than lumbar spinal cord lesions in EAE ( Fournier et al . , 2017 ) . We transversely cut midcervical spinal cord sections at 12 μm with a freezing microtome ( Leica Jung CM3000 ) and stored them in cryoprotective medium ( Jung ) at –80°C . For transmission electron microscopy , mice were perfused with 4% PFA , 1% glutaraldehyde ( GA ) in 0 . 1 M phosphate buffer , and spinal cords were dissected and fixed with 2% PFA , 2 . 5% GA in 0 . 1 M phosphate buffer . Antibodies directed against CD3 ( rabbit IgG , Abcam , Cambridge , UK; RRID: AB_443425 ) , Mac-3 ( rat IgG , BD Biosciences; RRID: AB_394780 ) , and Iba1 ( rabbit , Wako , Richmond , USA; RRID: AB_839504 ) were visualized by avidin-biotin technique with 3 , 3-diaminobenzidine ( DAB , Sigma , St . Louis , USA ) according to standard procedures of the UKE Mouse Pathology Facility using the ultraView Universal DAB Detection Kit ( Ventana , Oro Valley , USA ) . For complex IV/cytochrome c oxidase ( COX ) histochemistry , sections were incubated for 60 min at 37°C with COX reaction media ( diaminobenzidine tetrahydrochloride , cytochrome c and bovine catalase [all from Sigma] in 0 . 2 M phosphate buffer ) and embedded with Aquatex ( Merck ) . Images of tissue sections at ×200 magnification were scanned using Zeiss MIRAX MIDI Slide Scanner ( Carl Zeiss , MicroImaging GmbH , Germany ) . CD3+ , IBA1+ , and MAC-3+ infiltrating cells were analyzed with ImageJ software ( NIH , Bethesda , USA ) , with the same settings across all experimental groups . CD3+ cells were counted in entire spinal cord sections . For IBA1+ and MAC3+ quantification , we calculated the percentage of the positively stained area in the spinal cord . For quantification of complex IV activity , background was subtracted and mean gray value intensities of COX reactivity in complete gray matter or the ventral horn were measured by ImageJ . COX reactivity was normalized to corresponding neuronal count . Sections were incubated in blocking solution ( 10% normal donkey serum in PBS ) containing 0 . 05% Triton X-100 at room temperature for 45 min , incubated in 0 . 05% Triton X-100 containing a Fab Fragment anti-mouse IgG ( goat , 1:200; Jackson; RRID: AB_2338476 ) for 60 min in case of staining anti-mouse primary antibodies to prevent unspecific binding , and subsequently stained them overnight at 4°C with antibodies against the following structures: phosphorylated neurofilaments ( SMI 31 , mouse , 1:500; Covance , Princeton , USA; RRID: AB_10122491 ) , non-phosphorylated neurofilaments ( SMI 32 , mouse , 1:500; Covance; RRID: AB_2564642 ) , neuronal nuclei ( NeuN , chicken 1:500; Millipore , Burlington , USA; RRID: AB_11205760 ) , neuronal protein ( HuC/D , mouse 1:500; Thermo Fisher Scientific , Waltham , USA; RRID: AB_221448 ) , PGC-1α ( PGC-1α , rabbit , 1:100; Novus Biologicals , Littleton , USA; RRID: AB_1522118 ) , and FLAG sequence ( FLAG , mouse , 1:200; Sigma; RRID: AB_259529 ) . As secondary antibodies we used Alexa Fluor 488-coupled donkey antibodies recognizing chicken IgG ( 1:800 , Jackson; RRID: AB_2340375 ) , Cy3-coupled donkey antibodies recognizing mouse IgG ( 1:800 , Jackson; RRID: AB_2340816 ) , and Alexa Fluor 647-coupled donkey antibodies recognizing rabbit IgG ( 1:800 , Abcam; RRID: AB_2752244 ) . We analyzed the sections with a Zeiss LSM 700 confocal microscope . For quantification of neurons and axons , tile scans of each animal were taken . Numbers of neurons were manually counted in two defined areas ( 250 µm × 250 µm ) from the central gray matter or ventral horn . For absolute numbers of neurons in the ventral horn , all neurons in both ventral horns per animals were counted . Axons were quantified in three defined areas ( 90 µm × 90 µm ) with ImageJ software using fixed threshold intensity across experimental groups for each type of tissue examined . Cortical neurons ( DIV14 ) were fixed with 4% PFA for 30 min at room temperature , permeabilized with 0 . 05% Triton and blocked with 10% normal donkey serum in PBS . Cells were stained overnight at 4°C with antibodies directed against PGC-1α ( PGC-1α , rabbit , 1:200; Novus Biologicals; RRID: AB_1522118 ) and microtubule-associated protein 2 ( MAP2 , chicken , 1:2500; Abcam; RRID: AB_2138153 ) . As secondary antibodies we used Alexa Fluor 488-coupled donkey antibodies recognizing chicken IgG ( 1:800 , Jackson; RRID: AB_2340375 ) , Cy3-coupled donkey antibodies recognizing mouse IgG ( 1:800 , Jackson; RRID: AB_2340816 ) , and Alexa Fluor 647-coupled donkey antibodies recognizing rabbit IgG ( 1:800 , Abcam; RRID: AB_2752244 ) . Images were taken with a Zeiss LSM 700 confocal microscope . After dissection and fixation , spinal cords were cut in 1 mm transverse sections and post-fixed in 1% osmium tetroxide ( OsO4 ) for 1 hr . Hereafter , samples were dehydrated in serial dilutions of alcohol and embedded with Epon ( LX‐112 Resin , Ladd Research , USA ) . Next , ultrathin ( 50‐85 nm ) sections were collected on 200 mesh copper grids and counterstained with uranyl acetate and lead citrate . Images of individual motor neurons were taken at ×5000 magnification using a JEM 1010 transmission electron microscope ( JEM 1010 , Jeol USA ) . From each motor neuron , we quantified the surface area of neuronal cytoplasm and all cytoplasmic mitochondria using ImageJ ( NIH ) , which were used to calculate mitochondrial density per neuron . Primary hippocampal and cortical cultures were prepared from E16 . 5 embryos . Hippocampus and cortex were harvested , cut into smaller pieces , and incubated in 0 . 05% Trypsin-EDTA ( Gibco ) for 6 min at 37°C . Trypsination was stopped by DMEM-F12 containing 10% fetal calf serum ( FCS ) . Afterwards , tissue was dissociated in Hank's balanced salt solution ( HBSS ) and centrifuged for 2 min at 500 × g . The pellet was resuspended in primary growth medium ( PGM ) , and cells were plated at 5 × 104 per cm2 ( hippocampal neurons ) or 7 . 5 × 104 per cm2 ( cortical neurons ) on poly-d-lysine-coated cell culture plates . We maintained cells in PNGM ( Primary Neuron Growth Medium BulletKit , Lonza ) or neurobasal plus medium ( supplemented with B27 , penicillin , streptomycin and L-glutamine; Gibco ) at 37°C , 5% CO2 , and a relative humidity of 98% . To inhibit glial proliferation , we added cytarabine ( Sigma , 5 µM ) or floxuridine/uridine ( Sigma , 10 µM ) at 3–4 days in vitro ( DIV ) at 5 μM and maintained cultures for 14–21 days in vitro ( DIV14 or DIV21 ) . Cortical primary neurons were transduced with genetically encoded cytosolic calcium indicator GCaMP6f ( Addgene 100837 ) via adeno-associated virus ( AAV7 ) at day 7 in vitro . Calcium imaging was performed from cPNC from E16 . 5 embryos ( DIV15 ) . Alternatively , chemical calcium indicator Fluo4-AM ( Thermo Fisher , F14201 ) was used according to the manufacturer’s instructions . Briefly , cultures were incubated in optic buffer ( ddH2O supplemented with 10 mM glucose , 140 mM NaCl , 1 mM MgCl2 , 5 mM KCl , 20 mM HEPES , 2 mM CaCl2 , pH 7 . 3 adjusted with NaOH ) with 4 µM Fluo4-AM for 30 min at room temperature , washed and imaged in optic buffer . Live cell imaging was performed at day 15 in vitro with a Zeiss LSM 700 confocal microscope , 20× magnification ( objective ) and temporal resolution of 478 ms . Transduced cultures were placed to incubation chamber ( 33°C , 5% CO2 ) 20 min prior to imaging and then recorded . Ionomycin ( 8 µM ) was added at the end of each experiment for signal normalization and cultures were imaged for another 15 min to allow development of maximum signal intensity . Additional normalization with optic buffer without Ca2+ and with 50 µM BAPTA-AM ( Thermo Fisher ) was done for the experiments using Fluo4-AM as calcium indicator to determine absolute calcium concentration ( Yasuda et al . , 2004 ) . The 478 s of spontaneous Ca2+ transients before application of ionomycin were used for analysis . Time-series images were generated for further analysis . Manual image segmentation and generation of traces of the intensity from the single neuronal soma was performed in ImageJ . Normalization , calcium spikes , base line detection , and analysis of the calcium clearance rate were performed with the custom-made script written on Python 3 . 6 ( https://github . com/scriptcalcium/PGC1alpha; Rosenkranz , 2021; copy archived at swh:1:rev:af3c206a43f8b6e9fdbd7707c9b4601287d5968b ) ( Shaposhnykov , 2020 ) . Clearance time ( T ) , amplitude ( A ) , mean signal intensity ( Mean ) , and numbers of calcium transients = firing rate ( N ) were defined as follows:Ni=nif ( F ( n ) ≥F ( a ) , wherea∈[n−δ…n+δ] ) and ( F ( n ) >1+γ2θ+1∑j=n−θn+θF ( j ) ) , Firing rate ( Ni ) : N: calcium transient count; F: normalized signal intensity; i and n: indexes for defined frame ( n ) and cell ( i ) ; parameters δ , θ , and γ were set as 3 , 7 , and 0 . 25 for all analyzed cultures by manual assessment of effective pick detection >80%Ti=1ki∑j=1kiTj Clearance time ( Ti ) : clearance time for the ith cells with detected k transients , from maximum to half of the amplitudeAi=1ki∑j=1kiFpij Amplitude ( Ai ) : amplitude for the ith cell with detected k transients; Fpi – amplitude of individual transients of ith cellMeana , b=1b-a∑j=abF ( j ) Mean[a , b]: mean of recorded signal from frame a till frame b , F ( j ) – signal extracted from jth frame; parameters a and b were determined for each experiment individually for each experiment with a constant analyzed interval; last frame was recorded before ionomycin addition . Hippocampal neurons were seeded in a poly-d-lysine-coated XF 96-well cell culture microplate ( Seahorse Bioscience , Copenhagen , Denmark ) in triplicate at 5 × 105 cells per well in 1 ml neuronal growth medium and then incubated at 37°C in 5% CO2 . At DIV14 , media from neurons was removed , replaced by 180 µl of assay media ( Assay Media from Bioscience with 25 mM glucose , 1 mM sodium pyruvate , and 2 mM L-glutamine; pH 7 . 4 ) , and incubated in a CO2-free incubator at 37°C for 1 hr . Compounds ( 2 µM oligomycin , 1 µM FCCP , 0 . 5 µM rotenone/antimycin A , all in assay media ) were added into the appropriate ports of a hydrated sensor cartridge . For baseline measurement , control ports were left without compound addition . Cell plate and cartridge were then placed into the XFe96 Analyzer and results analyzed by WAVE Software . Hippocampal neurons were seeded in a poly-d-lysine-coated XF 96-well cell culture microplate ( Seahorse Bioscience , Copenhagen , Denmark ) in quadruplicate at 5 × 105 cells per well in 1 ml neuronal growth medium and then incubated at 37°C in 5% CO2 . At DIV14 , calcein-AM ( Sigma ) was added to the cultures ( 2 µM ) , incubated at 37°C in 5% CO2 for 30 min , washed in optic buffer ( ddH2O supplemented with 10 mM glucose , 140 mM NaCl , 1 mM MgCl2 , 5 mM KCl , 20 mM HEPES , 2 mM CaCl2 , pH 7 . 3 adjusted with NaOH ) at room temperature , and fluorescence intensity was measured according to the manufacturer’s instructions . Primary neurons were seeded on µ-dishes ( ibidi ) and TMRE assay ( Abcam ) was performed on DIV14 according to the manufacturer’s protocol . Cells were treated with 10 nM TMRE for 20 min in PGM , washed twice with HBSS w/o phenol red ( Gibco ) , and then equilibrated for 10 min in HBSS w/o phenol red and live imaged at 555 nm with a Zeiss LSM 700 confocal microscope . For internal control , 20 µM FCCP was added 10 min prior to TMRE . Images were taken with a Zeiss LSM 700 confocal microscope . Total fluorescence of five cells per biological replicate was analyzed . Chat-L10a-eGFP mice were anesthetized with ketamin/xylazin and perfused with 10 ml ice-cold dissection buffer ( 1 × HBSS , 2 . 5 mM HEPES-KOH pH 7 . 4 , 35 mM glucose , 4 mM NaHCO3 ) over 1 min . Cervical spinal cords were dissected in ice-cold dissection buffer containing 100 µg/ml cycloheximide , and three spinal cords were pooled for homogenization in lysis buffer ( 20 mM HEPES-KOH pH 7 . 4 , 150 mM KCl , 5 mM MgCl2 , 0 . 5 mM dithiothreitol , 100 µg/ml cycloheximide , 40 U/ml Rnasin , 20 U/ml Superasin ) using a glass homogenizer . Homogenates were centrifuged at 2000 × g for 10 min at 4°C to remove large cell debris , supernatant was transferred to a new tube , and NP-40 detergent solution ( Thermo Fisher Scientific ) and 1 , 2-diheptanoyl-sn-glycero-3-phosphocholine ( Avanti Polar Lipids ) were added to final concentrations of 1% and 30 mM , respectively . After 5 min incubation on ice , lysates were centrifuged at 20 , 000 × g for 10 min at 4°C . Ten percent of the supernatant was saved as input control , the remaining 90% were incubated with monoclonal GFP antibody ( Htz-GFP19C8 and Htz-GFP19F7; Memorial Sloan Kettering Cancer Center Monoclonal Antibody Core Facility ) -coated magnetic beads ( Streptavidin MyOne T1 Dynabeads , Invitrogen pre-coupled to biotinylated Protein L , Pierce ) with end-over-end rotation overnight at 4°C . Beads were subsequently collected on a magnetic rack , washed four times with high-salt wash buffer ( 20 mM HEPES , 350 mM KCl , 5 mM MgCl2 , 0 . 5 mM , 1% NP-40 , 0 . 5 mM dithiothreitol , 100 μg/ml cycloheximide ) , and immediately subjected to Trizol/chloroform-based RNA extraction ( Invitrogen ) . RNA was precipitated with sodium acetate and Glycoblue ( Ambion , Thermo Fisher Scientific ) in isopropanol overnight at –80°C , washed twice with 70% ethanol , resuspended in water , and further purified using the RNeasy Micro Kit ( Qiagen , Venlo , The Netherlands ) with on-column DNaseI digestion . For higher RNA yields , all steps were carried out in non-stick microfuge tubes ( Ambion ) . Mice were anesthetized and killed by inhalation of CO2 , and spleens were excised with sterile instruments and collected in ice-cold PBS . Pooled single-cell suspensions from spleen were prepared by homogenization through a 40 μm cell strainer . Cells were pelleted by centrifugation ( 300 × g , 10 min , 4°C ) , and lysis of splenic erythrocytes was initiated by RBC lysis buffer ( 0 . 15 M NH4Cl , 10 mM KHCO3 , 0 . 1 mM Na2EDTA , pH = 7 . 4 ) for 2 . 5 min at 4°C and stopped with MACS buffer ( PBS , 0 . 5% bovine serum albumin ( BSA ) , 2 mM EDTA ) . For fluorescence-activated cell sorting , cells were washed with PBS and stained with Fixable Viability Stain 700 ( BD Biosciences ) in PBS for 20 min at 4°C in the dark to exclude dead cells . Cell suspension was washed in PBS . Cells were stained with CD45-APC/Cy7 ( clone 30F11 , BioLegend; RRID: AB_312981 ) , CD3-BV605 ( clone 17A2 , BioLegend; RRID: AB_2562039 ) , CD4-FITC ( clone GK 1 . 5 , BioLegend; RRID: AB_312691 ) , CD8-Pacific Blue ( clone 53-6 . 7 , BioLegend; RRID: AB_493425 ) , and CD19-PE/Cy7 ( clone 1D3 , BD; RRID:AB_10894021 ) or with CD45-APC/Cy7 ( clone 30F11 , BioLegend; RRID: AB_312981 ) , CD3-BV605 ( clone 17A2 , BioLegend; RRID: AB_2562039 ) , CD19-BV605 ( clone 6D5 , BioLegend; RRID: AB_11203538 ) , Ly6G-BV711 ( clone 1A8 , BD; RRID:AB_2738520 ) , F4/80-BV421 ( clone T45-2342 , BD; RRID: AB_2734779 ) , NK1 . 1- PE/Cy7 ( clone PK136 , BioLegend; RRID: AB_389364 ) , and CD11b-FITC ( clone M1/70 , BioLegend; RRID: AB_312789 ) in FACS buffer ( PBS , 0 . 5% BSA , 0 . 02% NaN3 ) supplemented with Fc-Block ( True Stain anti-mouse CD16/32 , clone 93; BioLegend; RRID: AB_1574975 ) for 30 min at 4°C in the dark , washed with PBS , and filtered and resuspended in PBS with 10 µM EDTA . CD4+ T cells ( CD45+CD3+CD4+ ) , CD8+ T cells ( CD45+CD3+CD8+ ) , B cells ( CD45+CD19+ ) , and macrophages ( CD45+CD3-CD19- Ly6G-NK1 . 1-CD11b+ F4/80+ ) were sorted into collection tubes coated with FCS and filled with complete RPMI 1640 medium ( 1% penicillin , streptomycin , 0 . 1% 2-ME ) with 20% FCS using a FACSAria device ( BD Biosciences ) . Purity of sorted populations was routinely above 95% . Then cells were pelleted by centrifugation at 300 × g for 10 min at 4°C , dry frozen in liquid nitrogen , and stored at –80°C until RNA isolation . For isolation of microglia , mice were intracardially perfused with ice-cold PBS immediately after killing by inhalation of CO2 to remove blood from intracranial vessels . Brain and spinal cord were excised with sterile instruments , and mechanically dissected and incubated with agitation in RPMI medium 1640 ( PAN-Biotech ) containing 1 mg/ml collagenase A ( Roche ) and 0 . 1 mg/ml DNaseI ( Merck ) for 60 min at 37°C . Tissue was triturated through a 40 μm cell strainer and washed with PBS ( 300 × g , 10 min , 4°C ) . Homogenized tissue was resuspended in 30% isotonic Percoll ( GE Healthcare , Chicago , USA ) and underlaid with 78% isotonic Percoll . After gradient centrifugation ( 1500 × g , 30 min , 4°C ) , CNS-immune cells were recovered from the gradient interphase and washed twice in ice-cold PBS . For fluorescence-activated cell sorting , cell suspensions were stained with Fixable Viability Stain 700 , followed by staining of surface antigens with CD45-APC/Cy7 ( clone 30F11 , Biolegend; RRID: AB_312981 ) , and CD11b-PerCPCy5 . 5 ( clone M1/70 , Biolegend; RRID: AB_312981 ) and Fc-Block as described above . Microglia ( CD45med CD11b+ ) were sorted and processed for RNA isolation as described above . RNA was purified using the RNeasy Mini Kit ( Qiagen ) and reverse transcribed to cDNA with the RevertAid H Minus First Strand cDNA Synthesis Kit ( Thermo Fisher ) according to the manufacturer’s instructions . Gene expression was analyzed by quantitative real-time PCR performed in an ABI Prism 7900 HT Fast Real-Time PCR System ( Applied Biosystems , Waltham , USA ) using TaqMan Gene Expression Assays ( Thermo Fisher ) for Ppargc1a ( Pgc-1α; Mm00464452_m1 ) , Alas ( Mm01235914_m1 ) , Cox4i1 ( Mm01250094_m1 ) , Cs ( Mm00466043_m1 ) , Nrf1 ( Mm01135606_m1 ) , Tfam ( Mm00447485_m1 ) , and Tbp ( Mm01277042_m1 ) . Gene expression was calculated as 2−ΔCT relative to Tbp as endogenous control . DNA was purified using the DNeasy Blood and Tissue Kit ( Qiagen ) according to the manufacturer’s instructions . To destroy RNA , samples were treated with RNAse ( Qiagen ) . Mitochondrial DNA copy numbers relative to diploid chromosomal DNA content were analyzed by quantitative real-time PCR performed in an ABI Prism 7900 HT Fast Real-Time PCR System ( Applied Biosystems ) using TaqMan Gene Expression Assays ( Thermo Fisher ) for Cox2I ( Mm03294838_g1 ) and β-actin ( Mm_00607939_s1 ) . MtDNA copy numbers were quantified as 2−ΔCT relative to β-actin . DNA was lysed with 50 µl QuickExtract DNA Extraction Solution ( Lucigen , Middleton , USA ) . Copy numbers were analyzed by quantitative real-time PCR performed in an ABI Prism 7900 HT Fast Real-Time PCR System ( Applied Biosystems ) using TaqMan Copy Number Assays ( Thermo Fisher ) for Ppargc1a ( Pgc-1α , Mm00164544_cn , FAM ) and Tfrc ( TaqMan Copy Number reference assay for mouse , VIC ) in a duplex PCR . Ppargc1a copy numbers were quantified as 2−ΔCT relative to Tfrc . After dissection , 15 mg of mouse spinal cord were lysed in 500 μl RIPA buffer supplemented with 50× protease inhibitor ( complete , Sigma-Aldrich ) and 2% 50x phosphatase ( PhosSTOP , Sigma Aldrich ) , homogenized and incubated at 4°C for 30 min on a rotating wheel ( GLW Storing Systems GmbH ) , and subsequently centrifuged at 3200 × g for 5 min at 4°C . Respective total protein amount of supernatant was determined with bicinchoninic acid protein assay . Only for validation of antibodies lambda protein phosphatase ( New England BioLabs Inc , Ipswich , USA ) was added to the samples . Samples were denatured by boiling 12 μg protein in a mixture of 15 μl 4× NuPAGE sample buffer ( novex ) , 6 μl 10× Bolt sample reducing agent ( novex ) , and RIPA buffer in a total volume of 60 μl for 5 min at 95°C . Samples were then loaded onto a Bolt 4–12% Bis-Tris Plus Gel ( Invitrogen ) in a Mini Gel Tank ( Life Technologies , Carlsbad , USA ) . Chambers were filled with 1× Bolt MOPS SDS running buffer . Proteins were separated at 165 V for 1 hr . Spectra Multicolor High Range Protein Ladder ( Thermo Fisher ) was used as a protein standard . Proteins were then transferred to a nitrocellulose membrane at 10 V for 1 hr . For immunodetection , membrane was first blocked in 5% BSA in 1× TBS-T for 1 hr at room temperature and then incubated overnight at 4°C on a rotation wheel in 5% BSA in 1× TBS-T containing the primary antibodies against the following proteins: PGC-1α ( rabbit 1:2000; Novus Biologicals; RRID: AB_1522118 ) , phosphorylated PGC-1αS570 ( rabbit 1:1000; R and D Systems , Minneapolis , USA; RRID: AB_10890391 ) , and β-actin ( rabbit 1:1000; Cell Signaling Technology , Cambridge , UK; RRID: AB_330288 ) . The membrane was washed thrice with 1× TBS-T for 5 min each and subsequently incubated for 1 hr at RT on a shaker in 5% BSA in 1× TBS-T supplemented with a horseradish-peroxidase conjugated secondary antibody ( goat 1:15000; LI-COR Biosciences , Lincoln , USA; RRID: AB_2721264 ) . The membrane was washed again three times in 1× TBS-T at RT for 5 min . Bound antibodies were detected by chemiluminescence using the WesternSure ECL Substrate ( LI-COR Biosciences ) using the CCD camera-based ImageQuant LAS 4000 Mini system ( GE Healthcare ) . No statistical methods were used to predetermine sample sizes; our sample sizes are similar to those reported in previous publications ( Schattling et al . , 2019; Schattling et al . , 2012 ) . Experimental data were analyzed using Prism 8 software ( GraphPad ) and are presented as mean values ± s . e . m . Statistical analyses were performed using the appropriate test indicated in the figure legends . Outlier identification was first performed via Grubb’s test with p=0 . 05 . Shapiro–Wilk test was used to analyze normality . In normally distributed data , differences between two experimental groups were determined by unpaired , two-tailed Student's t-test; in non-normally distributed data , we tested differences between two experimental groups by unpaired , two-tailed Mann–Whitney test . Differences between three experimental groups were determined by multiple comparisons test following one-way ANOVA . Significant results are indicated by asterisk: *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , ****p<0 . 0001 . | Multiple sclerosis is a life-long neurological condition that typically begins when people are in their twenties or thirties . Symptoms vary between individuals , and within a single individual over time , but can include difficulties with vision , balance , movement and thinking . These occur because the immune system of people with multiple sclerosis attacks the brain and spinal cord . This immune assault damages neurons and can eventually cause them to die . But exactly how this happens is unclear , and there are no drugs available that can prevent it . One idea is that the immune attack in multiple sclerosis damages neurons by disrupting structures inside them called mitochondria . These cellular ‘organs’ , or organelles , produce the energy that all cells need to function correctly . If the mitochondria fail to generate enough energy , the cells can die . And because neurons are very active cells with high energy demands , they are particularly vulnerable to the effects of mitochondrial damage . By studying a mouse version of multiple sclerosis , Rosenkranz et al . now show that mitochondria in the neurons of affected animals are less active than those of healthy control mice . This is because the genes inside mitochondria that enable the organelles to produce energy are less active in the multiple sclerosis mice . Most of these genes that determine mitochondrial activity and energy production are under the control of a single master gene called PGC-1alpha . Rosenkranz et al . showed that boosting the activity of this gene — by introducing extra copies of it into neurons — increases mitochondrial activity in mice . It also makes the animals more resistant to the effects of multiple sclerosis . Boosting the activity of mitochondria in neurons could thus be a worthwhile therapeutic strategy to investigate for multiple sclerosis . Future studies should examine whether drugs that activate PGC-1alpha , for example , could help prevent neuronal death and the resulting symptoms of multiple sclerosis . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"immunology",
"and",
"inflammation"
] | 2021 | Enhancing mitochondrial activity in neurons protects against neurodegeneration in a mouse model of multiple sclerosis |
Many organisms spanning from bacteria to mammals orient to the earth's magnetic field . For a few animals , central neurons responsive to earth-strength magnetic fields have been identified; however , magnetosensory neurons have yet to be identified in any animal . We show that the nematode Caenorhabditis elegans orients to the earth's magnetic field during vertical burrowing migrations . Well-fed worms migrated up , while starved worms migrated down . Populations isolated from around the world , migrated at angles to the magnetic vector that would optimize vertical translation in their native soil , with northern- and southern-hemisphere worms displaying opposite migratory preferences . Magnetic orientation and vertical migrations required the TAX-4 cyclic nucleotide-gated ion channel in the AFD sensory neuron pair . Calcium imaging showed that these neurons respond to magnetic fields even without synaptic input . C . elegans may have adapted magnetic orientation to simplify their vertical burrowing migration by reducing the orientation task from three dimensions to one .
Many organisms such as birds , butterflies and turtles use the magnetic field of the earth ( geomagnetic field ) to navigate across the globe ( Johnsen and Lohmann , 2005; Guerra et al . , 2014 ) . Many animals migrate horizontally by preferentially using either the horizontal ( e . g . , salmon , Quinn et al . , 1981 ) , or the vertical component of the earth's field ( e . g . , turtles , Light et al . , 1993 ) . By contrast , magnetotactic bacteria use the geomagnetic field to migrate roughly vertically by following the magnetic dip line ( Blakemore , 1975 ) . Across hemispheres , magnetotactic bacteria reverse their polarity-seeking preference , thus conserving the adaptiveness of the response ( Blakemore et al . , 1980 ) . Although much is known about magnetosensation in bacteria , the cellular and molecular basis for magnetosensation in animals is gaining in understanding . Recent progress has been made identifying central neurons that respond to magnetic fields ( e . g . , Wu and Dickman , 2012 ) . Moreover , advancements have also been made identifying candidate magnetosensory transduction mechanisms ( e . g . , Gegear et al . , 2010; Lauwers et al . , 2013 ) . Despite this progress , no magnetosensory neurons have been identified in any animal ( Edelman et al . , 2015 ) . Understanding how animals detect and use magnetic fields will allow us to better predict the behavior of magnetosensitive organisms , and will aid the study of how natural and artificial magnetic fields affect living systems ( Engels et al . , 2014 ) . We show for the first time that the soil nematode Caenorhabditis elegans orients to earth-strength magnetic fields . This ability is required for vertical burrowing migrations directionally influenced by their satiation state . The direction and strength of the behavioral response to magnetic fields of wild-type strains isolated around the world correlated with their native magnetic field's inclination , and with the amplitude of the field's vertical ( but not its horizontal ) component . The AFD sensory neurons respond to earth-strength magnetic fields as observed by calcium imaging , and are necessary for magnetic orientation , and for vertical migrations . Expression of the cyclic nucleotide-gated ion channel , TAX-4 , in AFD neurons is necessary for worms to engage in vertical migrations , to orient to artificial magnetic fields , and for the AFD neurons to activate in response to an earth-strength magnetic stimuli .
While much is known about C . elegans crawling on agar surfaces , in the wild worms likely spend most of their time burrowing through their substrate . After 50 years of C . elegans research , studies looking at their burrowing behavior have only recently begun ( Kwon et al . , 2013; Beron et al . , 2015 ) . Because worms are known to orient to a variety of sensory stimuli that vary with depth in their native soil niches ( Braakhekke et al . , 2013 ) , we hypothesized that burrowing worms engage in vertical migrations like magnetotactic bacteria . To test this , we placed worms in the center of 20-cm long , agar-filled cylinders . Three layers of aluminum foil and a Faraday cage blocked light and electric fields respectively from penetrating the cylinders . Pipettes were then aligned horizontally in the ‘north-south’ or the ‘east-west’ directions , or vertically in the ‘up-down’ direction in the absence of artificial magnetic fields ( Figure 1A ) . Directional preference during burrowing was quantified with a burrowing index computed as the difference between the number of worms reaching either side divided by the total number of worms reaching both sides . We found that when starved , the wild-type lab strain , N2 , originally from Bristol , England ( Dougherty and Calhoun , 1948 ) preferentially migrated down in vertically oriented cylinders , but did not show a burrowing preference when cylinders were arranged horizontally ( Figure 1B ) . 10 . 7554/eLife . 07493 . 003Figure 1 . C . elegans engages in vertical migrations whose direction depends on satiation state . ( A ) To determine if C . elegans engaged in burrowing migrations we injected worms into agar-filled pipettes aligned horizontally ( east-west and north-south ) , or vertically ( up-down ) . Alternatively , we disrupted the local magnetic field around vertical pipettes ( where magnetic north is down ) , by reversing the local field polarity ( thus making magnetic north up ) with a magnetic coil system . ( B ) Only worms in pipettes aligned vertically displayed burrowing bias , preferentially migrating down unless the local magnetic field polarity was reversed ( red bar ) with the help of a magnetic coil system ( C ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 00310 . 7554/eLife . 07493 . 004Figure 1—figure supplement 1 . Merritt coil system for 3D control of magnetic fields . To expose worms to controlled and homogeneous earth-strength magnetic fields we constructed a triple magnetic Merritt coil system ( Merritt et al . , 1983 ) . ( A ) Each system creates a magnetic field along the x ( i ) , y ( ii ) , and z ( iii ) directions and consists of four 1-m2 squares , each arranged orthogonal to the other two . The system generates magnetic and electric fields . To prevent electric fields from affecting our experiments we built a Faraday cage around the experimental volume ( iv ) . Dedicated DC power supplies for each coil ( v ) allowed us to control the orientation and the magnitude of the net magnetic field within the coil system . Assay plates ( vi ) were then placed inside the coil system for testing . We empirically calibrated the field within the coil system with the aid of a milligausmeter ( vii ) from AlphaLab Inc . ( Utah , USA ) . ( B ) In each magnetic coil system experiment , the north direction of the imposed magnetic field is signified by the 0° on the top of the circular plot . Directly beneath this , and inside the circular plot , the strain's genetic background or geographic origin is indicated . The solid circular histograms represent the heading of the tested populations in a circle where the radius equals 10% of the entire population . Well-fed animals are represented by the black contour , while starved worms are represented by the grey contour . Circular plots had 18 bins ( 20° each ) . Similarly , the black and grey arrows represent the mean heading vector for the well-fed and starved populations respectively . The length of the vector is 0 if the population of animals migrated at random , and it is 1 if all animals migrate to a single point . The brown and green dashed curves indicate the heading that would result in ( respectively ) downward or upward translation at the original isolation site of each strain . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 004 Most animals determine the up and down direction by sensing the gravitational field of the earth ( e . g . , protozoans: Roberts , 2010; crustaceans: Cohen , 1955; vertebrates: Popper and Lu , 2000 ) . Alternatively , magnetotactic bacteria have been shown to use the earth's magnetic field to migrate up or down within the water column ( Blakemore , 1975 ) . To help distinguish magnetotactic vs gravitactic mechanisms we built a magnetic coil system capable of producing homogeneous magnetic fields of any desired 3D orientation ( Figure 1—figure supplement 1 ) . Our magnetic coil system is comprised of three independently-powered , orthogonal Merritt coil systems that allow the generation of magnetic fields of up to 3× earth strength ( Figure 1—figure supplement 1 , Merritt et al . , 1983 ) . Within the 1-m3 coil system , a smaller 20-cm2 Faraday cage , made with copper fabric , protects the test volume from the electric field that is concomitantly created alongside the magnetic field . Though often ignored in studies on animal magnetic orientation , this precaution was necessary because C . elegans and other animals exhibit strong behavioral responses to electric fields ( Gabel et al . , 2007; Manière et al . , 2011 ) . We repeated our vertical burrowing assay with an artificial magnetic field of earth-strength oriented opposite to the local earth's magnetic field ( i . e . , magnetic north pointing up rather than the natural orientation where magnetic north points down , Figure 1C ) . Under these conditions , we expected that worms responding to gravitational cues would continue to migrate down , while worms responding to magnetic cues would reverse their direction and now migrate up . Consistent with magnetic stimuli dictating vertical migration we found that starved N2 strain worms reversed their burrowing behavior and migrated up ( Figure 1B red bar ) . Our above results indicated that C . elegans might be able to detect and orient to magnetic fields of earth strength . To further investigate how worms respond to magnetic fields we placed them at the center of an agar plate and in turn placed this at the center of our 1-m3 Merritt coil system ( Figure 1—figure supplement 1; Figure 2A , B ) . We placed an anesthetic ( NaN3 ) around the circumference of the plate , which allowed us to immobilize and tally worms after they arrived at the plate's periphery . To determine if the coil system generated an unwanted temperature gradient , we measured temperatures across the assay plate in response to a magnetic field ( Figure 2—figure supplement 1A , B ) . Temperature gradients between the assay's start position ( at the center of the plate ) and the finish position ( at its edge ) were negligible across time , and did not vary significantly whether we imposed a magnetic field of one earth strength , or if we cancelled out the earth's field by imposing a field of equal strength but opposite direction ( two-way repeated measures ANOVA , N = 5 , p = 0 . 123 ) . 10 . 7554/eLife . 07493 . 005Figure 2 . Preferred magnetotaxis orientation to a spatially uniform , earth-strength field depends on satiation state , and local field properties . ( A ) The core of the earth generates a magnetic field that bisects the ground at different angles across the planet . The vertical component is strongest at the poles and weakest near the equator . ( B ) We constructed a Faraday cage within three orthogonal magnetic coil systems to test the response of worms under earth-like magnetic conditions . Circular histograms show the average percentage of worms migrating in each of 18 20-degree-wide headings . The radius of each circle represents 10% of the tested population and the mean heading vector arrow would be as long as the radius if every worm converged on one heading , and have zero length if worms distributed randomly around the plate ( see Figure 1—figure supplement 1B for explanation of circular plots ) . When the magnetic field around them was cancelled C . elegans migrated randomly ( C ) . ( D ) Worms migrated at an angle to an imposed field when its amplitude was half maximum earth strength ( 0 . 325 Gauss ) . ( E ) When exposed to a field equaling maximum earth strength ( 0 . 625 Gauss ) , worms showed stronger orientation . ( F ) Starving the worms for 30 min resulted in animals migrating in the opposite direction to their heading while fed . ( G ) Burrowing worms mirrored the magnetic coil results with fed worms preferentially burrowing up while starved worms preferentially migrating down . For the standard lab strain N2 ( native to Bristol , England ) the virtual up and down direction is represented in the circular plots by a green and brown dashed arch and was not found to vary significantly from the mean heading angle of fed and starved worms respectively . For the magnetic coil assays migration along the imposed field would translate animals towards the 0°/N signs . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 00510 . 7554/eLife . 07493 . 006Figure 2—figure supplement 1 . Testing the presence of temperature gradients . To determine if the artificial magnetic fields introduced unwanted temperature gradients in our assay we used high-accuracy thermometers capable of measuring 1/100 of a degree Celcius . ( A ) We recorded the temperature inside the coil system at the edge of the assay plate ( where worms were tallied ) , and at the center of the assay plate ( where worms began the experiment ) . We took temperature measurements under two magnetic regiments: when the earth's magnetic field was actively cancelled out inside the cage ( 0 . 000 Gauss , blue ) , and when we created an artificial magnetic field of earth strength inside the cage ( 0 . 650 Gauss , red ) . ( B ) The temperature difference between the center and the edge of the plate was reported every 5 min for 30 min before powering the cage on; every 10 min for an hour while the cage was on; and every 5 min for 30 min after powering down the cage . A two-way repeated measures ANOVA failed to reveal significant differences between both treatments ( p = 0 . 123 ) . ( C ) We measured the temperature difference between the end points of our magnet assays . Two temperature probes were placed at the target zones of magnet assay plates in the absence of a test magnet ( blue ) , or when a magnet was present above one of the two test areas ( red ) . ( D ) We report the difference between both temperature probes every 10 min for 1 hr . A two-way repeated measures ANOVA failed to find a significant difference between the two experimental conditions ( p = 0 . 559 ) . ( E ) To empirically confirm that both probes were accurately calibrated we placed them inside a beaker containing 1 l of dH2O and compared their readings between experiments . ( F ) Throughout our experiments both probes remained in agreement within 1/100th of a degree Celsius . In all experiments the two probes were positioned 5 cm apart . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 006 First , as a control , we asked how worms respond when the earth's magnetic field is cancelled . We accomplished this within the test volume of the magnetic coil system using a magnetic field of equal strength and orientation to the field of the earth , but with opposite direction . Worms in this regiment experienced a net magnetic field of 0 . 000 Gauss in three dimensions . Under this condition animals migrated randomly , distributing evenly around the circumference of the assay plate ( Figure 2C ) . We next produced a homogeneous magnetic field of 0 . 325 Gauss ( corresponding to half of earth's maximum field intensity ) directed across the assay plate . Worms assayed this way showed a biased distribution directed ∼120° to the imposed magnetic vector ( Figure 2D ) . Increasing the field strength to match the earth's maximum field strength ( 0 . 650 Gauss ) resulted in worms migrating approximately the same direction ( 132° ) to the imposed vector ( Figure 2E ) . Surprisingly , if worms were allowed to starve for just 30 min , they reversed their migratory distribution by ∼180° , now migrating at 305° relative to the field vector ( Figure 2F ) . These results demonstrate that C . elegans does not migrate simply toward magnetic north like magnetotactic bacteria ( Frankel et al . , 2006 ) ; rather , they display a preference to migrate at particular angles relative to magnetic north that depend on feeding state . Similar plasticity for opposite migration preferences in C . elegans has been documented for other sensory modalities ( e . g . Bretscher et al . , 2008; Russell et al . , 2014 ) . Do these seemingly arbitrary migratory angles serve a relevant purpose in the worm's soil niche ? As mentioned earlier , the standard wild-type C . elegans lab strain ( N2 ) was originally isolated in Bristol , England and cryogenically preserved there for distribution and study around the world . We turned to available geomagnetic data from NOAA to determine whether these migratory angles related to the earth's magnetic field in England ( Maus et al . , 2009 ) . In Bristol , the earth's magnetic vector enters the ground ( north pointing down ) at approximately 66° of inclination ( Figure 2G ) . Thus , in Bristol , to optimally orient upward , a worm would need to migrate 156° to the magnetic field penetrating the earth ( green arc in Figure 2E , G ) ; to optimally orient downward , a worm would need to migrate 336° to the magnetic field ( brown arc in Figure 2F , G ) . To determine whether the preferred migratory angles of starved and well-fed worms in our magnetic coil system assay matched these directions we performed a V test ( Batschelet , 1981 ) . We found that the mean heading of fed worms ( 132° , N = 1268 animals ) did not differ significantly from the upward direction for England ( 156 . 3° ) . Likewise , the mean heading of starved worms ( 304 . 6° , N = 1079 animals ) did not differ significantly from the downward direction for England ( 336 . 3° ) . Please refer to Supplementary file 1A through 1e for descriptive and analytical statistics for all the data presented in this study . These results predicted that worms use the earth's magnetic field to migrate at angles to the vector that would translate them up if they are fed , or down if they are starved . To test this hypothesis , we placed well-fed or starved worms in vertically arranged agar-filled pipettes away from artificial magnetic and electric fields as before . We found that , consistent with this idea , starved worms preferentially migrated down while well-fed worms migrated up ( Figure 2G ) . These results are parsimonious with C . elegans directing its vertical burrowing behavior by using the earth's geomagnetic field . Soil nematodes feed on bacteria associated with rotting fruit on the soil surface ( Félix and Braendle , 2010 ) and on root rhizobacteria deep in the soil ( Horiuchi et al . , 2005 ) . Vertical migrations could be associated with travel between these distinct food sources . Like magnetotactic bacteria , C . elegans has been isolated across the world . Magnetotactic bacteria from different hemispheres migrate in opposite directions to the field vector . Bacteria that inhabit the northern hemisphere ( where magnetic north points down ) are termed north-seeking magnetotactic bacteria , while those inhabiting the southern hemisphere ( where magnetic south points down ) are termed south-seeking magnetotactic bacteria ( Frankel et al . , 2006 ) . The distribution of different wild-type C . elegans isolates from around the world with distinct magnetic environments affords us a valuable opportunity to investigate how animals in magnetically distinct environments respond to magnetic fields . We therefore repeated our magnetic coil system and burrowing assays with wild-type C . elegans worms isolated from Adelaide ( Australia ) where the magnetic field of the earth is similar in strength and angle to that in England but differs in the key respect of having the opposite polarity ( Figure 3A ) . Unlike British worms , we found that Australian worms placed in a plate within our magnetic coil system migrated to an earth-strength field at 302 . 5° if fed , and 117 . 4° if starved . While oriented oppositely in preference from the angles displayed by the British N2 strain , these angles were similar in that they would also result in upward translation in Australia for fed animals and downward translation for starved ones ( Figure 3B , C respectively ) . To test if this response to an imposed artificial magnetic field reflected the migratory burrowing preference of worm in a natural magnetic field , we compared the burrowing behavior of Australian worms to that of the British strain . Paralleling our magnetic coil system results , we found that in our lab ( located in Texas , USA ) Australian worms migrated down when well-fed , and they migrated up when starved in the burrowing assay ( Figure 3D ) . Identical results were found for Hawaiian worms which migrated at angles that would optimally orient them up and down when satiated and starved in Hawaiian ( Supplementary file 1B ) . Overall , these results suggest that unlike magnetotactic bacteria , which follow magnetic field lines during their migrations , worms migrate at angles to the imposed field that would result in optimal vertical translation in their native locations . 10 . 7554/eLife . 07493 . 007Figure 3 . Magnetic orientation varies with satiation state and local field properties . To investigate if worms from distinct locations around the world displayed different magnetic orientations we tested C . elegans isolated from Adelaide ( Australia ) where the magnetic field is similar to that of the lab strain ( Bristol , England ) in strength and inclination but opposite in polarity ( A ) . Worms from Australia showed a magnetotactic response reversed from the British strain . Plots for well-fed ( B ) and starved ( C ) worms are shown and the local angle relative to the up and down direction are shown as green and brown dashed arches respectively . For each population , the radius of the circle represents 10% of the animals . The histograms show the percent of the worms that migrated in each of 18 20-degree headings . The mean heading vector shows the average direction of the animals and is equal to zero if all animals migrated randomly , and to the circle radius if all animals migrate on a single heading . ( D ) We compared the burrowing preference of fed and starved British and Australian worms placed in the local ( Texas ) magnetic field and found that consistent with our magnetic cage experiments both strains migrated in opposite directions . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 007 The results of our magnetic coil and burrowing experiments suggested that worms use the local magnetic field to guide vertical migrations . Unfortunately these experiments are limited to a few assays at the time , preventing their use in larger-scale behavioral screens . To mitigate this shortcoming , we developed a new assay using strong rare-earth magnets to quickly assess the ability of different strains to respond to imposed magnetic fields ( Figure 4A ) . This assay allowed us to run many assays at the same time . Briefly , worms were placed at the center of an assay plate and allowed to migrate freely ( Figure 4A , and ‘Materials and methods’ ) . A magnet was then placed above one of two equidistant ‘goal’ areas . Magnetotactic performance was quantified with a magnetotaxis index computed as the difference between the number of worms reaching either goal divided by the number of worms reaching both goals . We found that when no magnet was present , worms distributed evenly between these two goals . However , if the magnet was present , worms preferentially migrated toward it ( Figure 4—figure supplement 1 , Supplementary file 1A ) . To ensure the presence of the magnet did not introduce an unwanted thermal gradient , we recorded the temperature difference between goals in the presence and absence of a magnet and found that the two treatments did not significantly differ from each other ( Figure 2—figure supplement 1C , D ) . We used this assay to compare the ability of different strains to detect and migrate in a biased way in the presence of strong magnetic field . We first turned our attention to many wild C . elegans strains isolated from different locations across the world . 10 . 7554/eLife . 07493 . 008Figure 4 . Magnetotactic variability between wild C . elegans isolates result from differences in local magnetic field properties . ( A ) We developed a novel assay to rapidly assess the ability of worms to detect and orient to magnetic fields . Worms placed at the center of a test plate were allowed to migrate freely toward or away from a magnet . The number of animals by the magnet M , or by a control area C were compared and used to calculate a magnetotaxis index: MI = ( M − C ) / ( M + C ) . Wild-type C . elegans have been isolated across the planet at locations with diverse local magnetic fields . ( B ) Earth's magnetic field inclination map plotted from data obtained from NOAA ( Maus et al . , 2009 ) showing the isolation location for twelve wild-type strains of C . elegans used in this study ( circles ) . ( C ) The ability of these wild isolates to magnetotax in our magnet assay strongly correlated with the inclination of the magnetic field at their origin . We used the white ten isolates to compute the correlation between these variables . This correlation was able to predict the magnetotaxis index of an additional strain obtained from California ( red circle ) . ( D ) Map of the vertical component of the earth's magnetic field ( Maus et al . , 2009 ) . ( E ) Performance in the magnet assay was even more correlated with the vertical component of the earth's magnetic field . However , the horizontal component of the magnetic field ( F ) showed no correlation with the magnetotaxis index of the wild isolates . The blue circle represents the lab strain ( N2 ) from England . All assays conducted at location indicated by the lone star . All values reported are means . Error bars represent S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 00810 . 7554/eLife . 07493 . 009Figure 4—figure supplement 1 . A new assay for testing magnetotactic ability . ( A ) We developed a convenient assay able to determine the ability of worm populations to detect and orient to magnetic fields . Worms were placed in the center of an agar plate . A 1 . 5 μl drop of anesthetic ( NaN3 ) was placed at the center of two test areas equidistant from the start , and a magnet was then centered above one of the two test areas . We calculated the magnetotaxis index as: Magnetotaxis Index = ( M − C ) / ( M + C ) . Where M is the number of worms found immobilized by the test area at the magnet , and C is the number of worms immobilized by the control test area . ( B ) If no magnet was present , worms distributed evenly between the two test areas . If a magnet was introduced above one of the areas , about two thirds of the worms preferentially migrated to the magnet test area . We repeated the experiment in assay plates wrapped in several layers of aluminum foil and observed that migration towards the magnet did not require light . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 009 The magnetic field of the earth varies greatly around the world ( Maus et al . , 2009 ) . If C . elegans uses the magnetic field for vertical migrations , what happens near the equator where the vertical component of the earth's field is weakest ? The global heterogeneity in field characteristics made us wonder if selection pressure for magnetosensation ability may drop off nearest the equator where the vertical component of the magnetic field is at its weakest . To test this , we used our magnet assay ( Figure 4A ) on wild-type populations isolated from ten locations across the planet where the local magnetic field varies in inclination and vertical strength ( Figure 4B , D , Maus et al . , 2009 ) . We found that the ability of different worm populations to orient to an artificial magnetic field was strongly correlated with the inclination ( Figure 4B , C ) and vertical strength ( Figure 4D , E ) of the magnetic field at their native sites . The horizontal component of the field , however , was a poor predictor of this magnetotactic ability ( Figure 4F ) . The strong correlation between magnetotaxis performance , local field inclination , and vertical strength allowed us to successfully predict the magnetotaxis index of an additional wild-type isolate from California , USA ( Figure 4 red circle in panels B-E , Supplementary file 1C ) . Wild-type isolates from equatorial locations where the magnitude of the vertical component was close ( or below ) 0 . 2 Gauss , were either unable or barely able to magnetotax ( Supplementary file 1A ) . These results are consistent with local adaptations to global magnetic field variations , and could perhaps be used to model how other species may respond to temporal field variations ( such as magnetic polar drift or field reversals , Cox et al . , 1964 ) . We conclude from these results that like many animals ( Johnsen and Lohmann , 2005 ) C . elegans can use the magnetic field's polarity and inclination to guide its migrations . Having determined that C . elegans orients to magnetic fields , we turned our magnet assay to next investigate the cellular and molecular underpinnings of this fascinating behavior . To investigate the neuromolecular substrates for magnetosensation , we tested mutants with deficiencies in a variety of previously characterized sensory pathways . Mutants with severe defects in some sensory modalities displayed normal or nearly normal magnetic orientation ( Figure 5 ) . These included worms deficient in the touch-form of mechanosensation ( mec-10 , Arnadottir et al . , 2011 ) , light detection ( lite-1 , Edwards et al . , 2008 ) , taste ( che-1 , Uchida et al . , 2003 ) , and oxygen sensation ( gcy-33 , Zimmer et al . , 2009 ) . However , we also found mutants that were significantly impaired in magnetotaxis . This group comprised worms with mutations in genes co-expressed in a single sensory neuron pair called AFD , first implicated in thermosensation ( Mori , 1999 ) . These included two independent mutant alleles of ttx-1 , important for AFD differentiation , and the triple mutant lacking guanylyl cyclases , gcy-23 , gcy-8 , and gcy-18 , which together are critical for AFD function . Furthermore , we identified a set of transduction mutants that failed to perform magnetic orientation . These included two independent mutant alleles of each tax-4 and tax-2 genes . These encode subunits of a cGMP-gated ion channel already implicated in sensory transduction in many sensory neurons , including AFD ( Komatsu et al . , 1996 ) . 10 . 7554/eLife . 07493 . 010Figure 5 . Magnetotaxis requires intact AFD sensory neurons . We used our magnet assay to test a large number of sensory mutants . Mutations that impair the mechano- ( mec-10 ) , light- ( lite-1 ) , oxygen- ( gcy-33 ) , and taste- ( che-1 ) sensory pathways spared magnetotaxis , while mutations in genes specifically required for AFD sensory neurons ( ttx-1 and gcy-23 , -8 , -18 ) abolished magnetotaxis . Mutations that impair the cGMP-gated ion channel TAX-4/TAX-2 that are expressed in the AFD sensory neurons ( and other cells ) similarly prevented magnetotaxis . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 010 To test the requirement of the AFD neuron pair in magnetosensation , we genetically ablated them via cell-specific expression of a transgene for a human cell-death caspase . One advantage of this technique is that only a fraction of individual worms will inherit the artificial chromosome carrying the transgene . This allowed us to compare the performance of sister worms grown and tested together under identical conditions and only differing in having or not said transgene . After each assay , individual worms with genetically ablated neurons were distinguished from their unaffected sisters by the co-expression of a fluorescent transgene reporter . We found that worms lacking the AFD sensory neurons failed to orient to an artificial magnetic field , while their unaffected sisters oriented normally ( Figure 6A , Supplementary file 1A ) . This could not be explained by non-specific defects , because these worms could move and orient normally to olfactory stimuli ( Figure 6—figure supplement 1 ) . Similarly ablating nearby sensory neuron pairs ASE and AWC had no effect on magnetotaxis . The sensory ending of the AFD neurons consists of dozens of villi arranged anterior-to-posterior ( in an antenna-like formation ) imbedded inside glial cells ( Perkins et al . , 1986; Doroquez et al . , 2014 ) . Genetic ablation of the glia surrounding these structures , results in worms with viable AFD neurons but lacking villi ( Bacaj et al . , 2008 ) . These worms were unable to orient to artificial magnetic fields ( Figure 6A ) . This supports the idea that the villi may be the site of magneto-transduction ( and/or that the glia themselves contribute to this sense ) . Taken together , our results demonstrate that the AFD sensory neurons are required for magnetotaxis . 10 . 7554/eLife . 07493 . 011Figure 6 . Geomagnetotaxis requires the TAX-4/TAX-2 cGMP-gated ion channel in the AFD sensory neurons . ( A ) Genetic ablation of the AFD neurons ( or their sensory villi via ablation of amphid glial cells ) prevented magnetotaxis . However , ablation of adjacent sensory neurons ( ASE and AWC ) did not impair this behavior . ( B ) Genetic rescue of the cGMP-gated ion channel TAX-4 via cDNA specifically in the AFD neurons , or via genomic DNA in additional tax-4-expressing neurons was sufficient to restore magnetotactic ability ( white bars ) compared to their tax-4-mutant background controls ( grey bars ) . However , rescue of tax-4 expressing neurons that excluded the AFD neurons failed to restore magnetotactic behavior . We retested some of the mutants impaired in the magnet assay in the magnetic coil system under earth-like fields . Impairment of the AFD neurons by mutations in the ttx-1 ( C ) , tax-2 ( D ) , or tax-4 ( E ) genes resulted in worms that failed to orient to magnetic fields of earth strength ( 0 . 625 Gauss ) . Migration along the imposed field would translate animals towards the 0°/N mark . ( F ) Genetic manipulations that impaired ( or rescued ) magnetotaxis had a similar effect on geomagnetotaxis of vertically burrowing worms . Starved British worms lacking the tax-4 gene failed to burrow down . However , control sister worms with the tax-4 gene rescued specifically in the AFD neurons ( AFD+ others- ) were able to burrow down . Conversely , starved British worms lacking the AFD neurons ( AFD dead ) failed to migrate down , while control sister worms ( AFD alive ) migrated down . Ablation of AFD in Australian worms similarly abolished geomagnetotaxis . * p < 0 . 05 , ** p < 0 . 001 . All values reported are means , and error bars represent S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 01110 . 7554/eLife . 07493 . 012Figure 6—figure supplement 1 . Genetic ablation of AFD does not impair chemotaxis . ( A ) Genetic ablation of the AFD neurons did not impair the ability of worms to move , or orient to the chemical attractant diacetyl compared to control sister worms that did not carry a cell-death transgene ( ICE ) . Comparison between the AFD neurons of animals expressing GFP ( B ) , or GFP and ICE ( C ) , revealed that in animals expressing the cell-death gene the AFD neuron is impaired and shows many of the typical signs of neurodegeneration ( e . g . , circular soma , beaded and fragmented processes ) . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 012 Many sensory neurons in C . elegans require the TAX-4 cGMP-channel for sensory transduction ( Komatsu et al . , 1996 ) . To determine if TAX-4 function in the AFD neurons was sufficient for magnetic orientation , we selectively rescued expression of tax-4 in the AFDs neurons in a tax-4 mutant background . Specific rescue of TAX-4 in AFD neurons was sufficient to partially restore the ability of tax-4 ( null ) mutant worms to orient to an artificial magnetic field ( Figure 6B , AFD+ others- ) . To investigate the possibility that TAX-4 may also mediate magnetic orientation through additional neurons we further rescued TAX-4 in all tax-4-expressing neurons by using its endogenous promoter and regulatory elements . However , this did not result in an increased rescue ( Figure 6B , AFD+ others+ ) . To test if tax-4 contributed to magnetotaxis through any other neuron asides from AFD , we tested tax-4 mutants where this gene was rescued in all tax-4-expressing neurons except for the AFD neurons ( gift from Dr R Baumeister ) . While rescuing tax-4 in all but AFD neurons resulted in a rescue of the ability of these animals to orient to chemical stimuli ( Supplementary file 1A ) , these animals remained unable to orient to magnetic fields ( Figure 6B , AFD- others+ ) . These results support the hypothesis that the cGMP-gated ion channel TAX-4 plays an important role in the AFD sensory neurons for orientation to magnetic fields . To confirm the relevance of these findings in a more natural magnetic assay , we retested selected mutants under earth-like fields ( in our coil system ) and found similar results ( Figure 6C–E ) . The results above were obtained for worms orienting to artificial magnetic fields . To determine if these results generalized to the ability of worms to engage in vertical migrations we tested selected strains in our vertical burrowing assay ( without artificial magnetic field ) . Consistent with our observations in the magnet and in the magnetic coil assays , tax-4 mutant worms did not show preferential vertical migration unless the gene was selectively rescued in the AFD neurons ( Figure 6F ) . Starved British and Australian wild-type isolates lacking the AFD neurons similarly failed to engage in biased vertical migrations , although their sisters not carrying the transgene ( used to kill AFD ) remained able to migrate down or up respectively ( Figure 6F ) . To determine whether the AFD neurons are directly responsive to magnetic fields , we measured the fluorescence of a genetically encoded calcium indicator , GCaMP3 , in fully immobilized worms ( Figure 7A , and Figure 7—figure supplement 1A ) . After recording baseline activity ( Figure 7B ) , we exposed mechanically immobilized worms to an 8-s , 65-Gauss ( 100× earth ) rotating ( 2 Hz ) magnetic stimulus ( see ‘Materials and methods’ for details ) . We observed a transient increase in the average brightness of the AFD neurons ( Figure 7C ) . Successive stimuli consistently produced a reduced response ( Figure 7D ) . We observed a similar response when the magnetic stimulus was decreased to 10× and 1× earth stimuli ( 6 . 5 and 0 . 65 Gauss respectively , Figure 7E , F ) . To help determine if the AFD neurons themselves are magnetosensitive , and not just synaptically downstream from ‘real’ magnetoreceptive neuron ( s ) , we measured AFD calcium responses in worms impaired in rapid and dense-core synaptic transmission ( unc-13 and unc-31 mutant strains , Ahmed et al . , 1992; Ann et al . , 1997 ) . In the absence of chemical synaptic or neuromodulatory inputs , the AFD neurons continued to respond to magnetic fields ( Figure 7G , H ) . Qualitatively similar results ( but higher in amplitude ) were observed for worms that were partially restrained ( Figure 7—figure supplement 1B ) . Magnetic-induced calcium responses in AFD were not observed in a tax-4 mutant background , suggesting that this requires Ca2+ entering the TAX-4 cGMP-gated ion channel ( Figure 7I , Figure 7—figure supplement 1B ) . Responses were also not observed in an adjacent sensory neuron pair AWC ( Figure 7J and Figure 7—figure supplement 1B ) . To quantitatively compare the magnetosensory response of AFD for different conditions and mutant backgrounds , we plotted the average GCamp3 . 0 intensity during the final 4 s of the magnetic stimulus relative to a 4-s baseline before presentation of the stimulus ( for the no-stimulus control we used the same time window as for the other recordings ) . We found that the change in brightness was significantly greater than control for all test conditions except in the case of tax-4 mutant background ( Figure 7K ) . Our imaging results provide physiological evidence that the AFD sensory neurons respond to magnetosensory stimuli relevant to geomagnetic orientation . 10 . 7554/eLife . 07493 . 013Figure 7 . The AFD sensory neurons respond to magnetic stimuli . ( A ) Calcium activity indicator GCaMP3 in the AFD neurons . ( B ) In the absence of a magnetic stimulus the soma of AFD neurons rests at baseline . Exposing restrained worms to a sinusoidal 65 Gauss ( 100× earth strength ) magnetic stimulus caused the soma of the AFD neurons to transiently increase brightness by 2% above baseline in response to the first stimulus ( C ) , and ∼1% in response to subsequent stimuli ( D ) . The AFD neurons responded when the magnetic stimuli was reduced to 6 . 5 ( E ) and 0 . 65 Gauss ( F , earth strength ) . The AFD magnetic response remained even in synaptic mutants ( G: unc-13 and H: unc-31 ) that render these cells synaptically isolated from other neurons . ( I ) Animals lacking a functional copy of the tax-4 gene did not show an increase in brightness in response to a magnetic stimulus . ( J ) A 65 Gauss stimulus failed to elicit a response in neighboring sensory neuron AWC . ( K ) The average soma brightness for the final 4 s prior to stimulus , and the final 4 s of the stimulus were compared . While the ‘no-stimulus’ , the ‘tax-4’ , and the ‘AWC’ conditions resulted in no significant brightness change , all other test conditions produced a significant increase in AFD brightness above baseline . Change in relative fluorescence key for panels B–J depicted in B with the exception of panel H which has its own key . N = 11 for B–D; 6 for E–H; 14 for I; and 7 for K . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 01310 . 7554/eLife . 07493 . 014Figure 7—figure supplement 1 . Measuring AFD calcium activity in partially and fully restrained worms . ( A ) Worm-immobilization chip for high-resolution fluorescence microscopy . The two-level device consists of a valve layer ( pink ) sitting above a flow layer where the worms reside ( grey ) . Animals enter the immobilization chamber via the worm input as fluid flow is directed to the fluid output . Small channels across the outer edge of the immobilization chamber permit fluid flow to pass but block the passage of the worms ( left ) . As the flow pushes the worms against the outer edge of the chamber the valve layer is pressurized to fully immobilize the worms ( right ) . A magnified view of a single animal pressed against the small channels along the outer edge of the immobilization chamber is shown during immobilization . ( B ) Alternatively , we partially restrained worms on an agar pad while measuring the brightness of the AFD ( or AWC ) sensory neurons before , during , and after exposure to a 60-Gauss magnetic stimulus . Images were taken only when the AFD soma was stationary . While consistent with our immobilized-worm experiments in sign ( Figure 7 ) , the amplitude of the responses were about 10 times larger in partially restrained animals . ( C ) Expression of GCaMP3 in AFD neurons did not impair the worm's ability to orient to magnetic fields . * p < 0 . 05 , ** p < 0 . 001 . All values reported are means , and error bars represent S . E . M . DOI: http://dx . doi . org/10 . 7554/eLife . 07493 . 014
There are many possible ways in which the AFD neurons may play a role in magnetic orientation . The magnetosensory AFD neurons also respond to temperature ( Mori , 1999 ) , CO2 ( Bretscher et al . , 2008; 2011 ) , and moisture ( Russell et al . , 2014 ) gradients in a satiation-dependent manner . All of these parameters vary with depth in the soil ( Jassal et al . , 2005 ) supporting the role of the AFD neurons in vertical burrowing migrations . Because C . elegans performs magnetotaxis in darkness ( assays were wrapped in 3-layers of foil ) , it is possible that it detects fields with nano-scale compasses made of biological magnetic material previously described in C . elegans ( Cranfield et al . , 2004 ) , rather than by a light-dependent mechanism , although this study did not investigate this possibility . However , based on our behavioral , mutant , transgenic , and physiological analyses we hypothesize magnetic particles , perhaps such as those found in magnetotactic bacteria ( Frankel et al . , 2006 ) and previously reported in C . elegans , may ( either directly or indirectly ) be associated with the anterior and posterior-directed microvilli of the AFD neurons . Magnetic stimulation of these structures could lead to activation of unspecified guanylyl cyclases ( such as GCY-8 , GCY-18 , or GCY-23 ) , in turn activating the TAX-4 channel and resulting in Na+ and Ca2+ influx and cell depolarization . These findings could represent an intriguing lead into the putative magnetotransduction mechanism of the AFD neurons . It will be intriguing to investigate whether the diverse range of other magnetotactic animals employ magnetosensitive neurons with analogous morphology and transduction mechanisms as the AFD neurons . The magnetic field of the earth provides reliable directional and positional information to organisms capable of its detection . Aside from magnetotactic bacteria , magnetic orientation has been largely observed in animals that migrate horizontally ( Johnsen and Lohmann , 2005 ) . Our finding that vertical migrations by an animal may also be guided by this sensory modality opens a new niche for the study of magnetic navigation . Magnetotactic bacteria passively migrate along field lines: with south-seeking bacteria swimming down , and north-seeking bacteria swimming up in the southern and northern hemispheres respectively ( Frankel et al . , 2006 ) . Unlike bacteria , C . elegans does not follow the field vector but rather migrates at an angle that appears to maximize its vertical translation . The difference in migration angles and in response amplitude between wild-type isolates from around the world suggests that this sensory modality is under considerable selective pressure and will be the subject of future studies . Our findings that the direction of vertical migrations could be reversed by an imposed magnetic field and that wild-type populations of worms from opposite hemispheres displayed opposite vertical migration preference strongly suggests that C . elegans relies on the geomagnetic field rather than gravity . Many organisms deduce their vertical orientation by using the earth's gravitational field ( gravitaxis ) . Studies on paramecia suggest that the relative density of the organism against its media is instrumental in gravitaxis ( Kuroda and Kamiya , 1989 ) . For terrestrial and marine animals , the relative density of their media ( air and water respectively ) is largely constant . However , for nearly buoyant worms imbedded in a soil matrix , the relative density of their surrounding media is highly variable and may preclude reliance on gravitaxis . Many magnetotactic bacteria use a mechanism known as ‘polar magneto-aerotaxis’ where these cells preferentially migrate up or down in chemically stratified water or sediment columns using an single sensory pathway to integrate magnetotaxis and aerotaxis ( Blakemore et al . , 1980; Frankel et al . , 1997; Popp et al . , 2014 ) . Like C . elegans , polar magneto-aerotactic bacteria from different hemispheres have adapted their polarity preference to match their native environment ( Blakemore , 1975 ) . It appears that the role of magnetotaxis in C . elegans , as in bacteria , may be to increase the efficiency of taxis to other sensory cues by reducing a search problem from three dimensions to one . Unlike bacteria which migrate along the dip line at an angle relative to the vertical direction , however , C . elegans appears to align motion at an angle to the magnetic field that would enable a more vertical trajectory in its native environment . Our finding that the strains from England , Australia and Hawaii each displayed a preferred magnetic orientation preference that matched the geomagnetic field orientation at their source of isolation rather than the one at the experimental site ( Texas , USA ) suggests a genetic encoding of magnetic orientation preference . From previous work ( Bretscher et al . , 2008; Russell et al . , 2014 ) , it is clear that satiation affects the sign of the response to sensory stimuli that the AFD neurons respond to . However , our experiments did not provide evidence to answer why fed worms migrate up and starved worms migrate down . One possibility concerns the vertical stratification of food sources . C . elegans eats bacteria growing on rotting fruit on the soil's surface ( Félix and Braendle , 2010 ) , and also root rhizobacteria deep in the soil ( Horiuchi et al . , 2005 ) . Vertical migrations may direct travel between these segregated food sources following the marginal value theorem ( Carnov , 1976; Milward et al . , 2011 ) . Rotting fruit on the surface represents an extremely rich , but transient , food supply . By contrast , rhizobacteria represent a low-quality but stable source of food . Surface populations likely grow exponentially until they exhaust their resources . For a starved worm on the surface , burrowing down would be adaptive because it leads to rhizobacteria in the plant roots . Rhizobacteria , however , represent a lower quality food source . From here , fed worms may venture to emerge in search of better and more plentiful food . Worms on poor diets have been shown to be more likely to abandon the relative safety of their food patch in an attempt to find a higher quality source ( Shtonda and Avery , 2006 ) . For these worms , an adaptive locomotor strategy would be to burrow up in search of higher quality food sources , as burrowing down would not be likely to result in finding a higher quality food patch . Future experimental studies will distinguish between these and other possibilities by mimicking specific soil conditions . Many animal species ( including other nematodes , Prot , 1980 ) engage in vertical soil migrations ( Price and Benham , 1977 ) . Therefore magnetic orientation may be more widespread than previously believed . While the scale and nature of magnetosensation make it challenging to study in large animals with extensive ranges , the small size , genetic tractability , and research amenability of C . elegans make it an optimal model to begin to unlock potentially conserved cellular-molecular mechanisms by which animals detect and orient to the magnetic field of the earth .
All worms were raised and tested at The University of Texas at Austin , Texas , USA ( 30° 20′ N 97° 45′ W ) between 2011 and 2014 . The local characteristics of the magnetic field of the earth during the duration of the experiments were as follows: Declination 4°35′ to 4°13′ ( East ) ; Inclination 59°19′ to 59°12′ ( Down ) ; Horizontal Intensity 0 . 245 to 0 . 244 Gauss; Vertical Intensity 0 . 413 to 0 . 410 Gauss ( Down ) ; Total Intensity 0 . 480 to 0 . 477 Gauss ( Maus et al . , 2009 ) . We conducted over 1200 assays ( >61 , 000 worms ) , averaging ∼48 worms per assay . All behavioral assays were conducted with experimenter blind to genotype of the worms assayed . Because of the multimodal properties of the AFD neurons , assays controlled for many physiological and environmental aspects prior to testing . To ensure worms were in comparable physiological states , all assays were performed on ( never starved ) day-1 adult hermaphrodite C . elegans . Worms were never allowed to overpopulate their plates . To minimize physiological changes due to unsealing of test plates ( e . g . , altering the O2/CO2 ratio ) , worms were tested within 20 min of unsealing their incubation plates . To test worms in comparable satiation states , worms tested under the ‘fed’ status were assayed within 10 min of being extracted from their bacterial lawn . To test worms in the ‘starved’ state , we allowed worms to remain suspended in liquid Nematode Growth Media ( NGM ) for 30 min prior to beginning their run . Incubation temperature was between 19–21°C in standard NGM agar plates seeded with Escherichia coli ( OP50 ) lawns ( Brenner , 1974 ) . Artificial magnetic fields were removed from the vicinity of the worms , and the local field surrounding the worms was determined to be of earth strength and direction with a DC Milligauss Meter Model MGM magnetometer ( AlphaLab , Utah , USA ) . To minimize novel background mutations , all strains were tested within 3 months of thawing from cryopreserved stocks , with additional re-thaws of fresh samples at 3-month intervals . We used the GATEWAY system to generate transgenes for transgenic strains ( Hartley et al . , 2000 ) . Fluorescent reporters were used to identify transgenic individuals ( Pmyo-2::mCherry , Pmyo-3::mCherry , or Punc-122::GFP; see Supplementary file 1E for specifics ) . We used the AFD-specific promoter ( Pgcy-8 ) to target the AFD neurons ( Inada et al . , 2006 ) . To genetically ablate these neurons , we constructed plasmids containing the human caspase gene ICE ( gift from V Maricq ) and transformed N2 wild-type worms to generate strains JPS264 vxEx264[Pgcy-8::ICE] . Identical results were found for two independently derived strains , JPS265 and JPS271 , in a N2 background . AFD neurons were also killed in the Australian wild-type isolate AB1 to generate strain JPS545 and JPS546 . The ability of this transgene to kill the AFD neurons was assessed by comparing GFP expression in the AFD neurons of worms carrying the ICE construct , with that of worms not carrying it ( Figure 6—figure supplement 1 ) . To measure intracellular calcium levels in AFD neurons in vivo we expressed GCaMP3 ( Tian et al . , 2009; gift from L Looger ) vxEx316[Pgcy-8::GCaMP3] in lite-1 ( ce314 ) worms to generate the strain JPS316 . Identical results were found for three independently derived strains , JPS275 , JPS294 , and JPS315 . All of these strains were capable of magnetosensation . We rescued tax-4 specifically in AFD by constructing plasmid with a wild-type copy of the tax-4 cDNA ( gift from Dr Ikue Mori ) expressed in AFDs as described above to generate the strain JPS458 tax-4 ( ks11 ) vxEx458[Pgcy-8::tax-4 ( + ) ] . Identical results were found for the independently derived strain , JPS459 . We also rescued with fosmid VRM069cE04 containing genomic region of tax-4 with its promoter , UTR and endogenous regulatory elements vxEx458[VRM069cE04] with strain JPS458 . unc-13 or unc-31 mutants were crossed with JPS316 males and F2 worms were selected that exhibited GCaMP3 fluorescence and an uncoordinated phenotype to generate strains JPS496 and JPS495 respectively . For details on the construction of additional neuronal ablation strains please refer to the Extended Data section in Russell et al . ( 2014 ) . To determine if worms could sense and respond to magnetic fields , we picked 50 never-starved ( day-1 ) adults from an OP50 bacterial lawn and into a 1-μl drop of liquid NGM . We used the latter to clean the worms off bacteria , and to transfer worms to the center of a 1-day old , 10-cm diameter , chemotaxis-agar assay plate . Equidistant from the worms , we drew 3 . 5-cm circles on either side and placed 1-μl drops of 1-M NaN3 at the center of these circles to immobilize and count any worm that reached the area ( Figure 4A ) . A N42 Neodymium 3 . 5-cm diameter magnet ( K&J Magnetics Inc . , Pennsylvania , USA ) was placed above one of the circles so that the assay plate was now traversed by a vertical magnetic field gradient that became stronger toward the magnet and weaker away from it . ( Note that more commonly found weaker strength magnets produced qualitatively the same behavioral results . ) Worms were released from the liquid NGM droplet by wicking excess liquid with filter paper . The total manipulation time ( from bacterial lawn to the beginning of the assay ) was kept under 10 min to avoid inadvertently starving the worms . Worms released from the liquid NGM became able to freely migrate around the plate . After 30 min we counted the number of worms NaN3-paralyzed in each circle and calculated the magnetic orientation index ( MI ) as: MI = ( M − C ) / ( M + C ) . Where M is the number of worms paralyzed within the magnet's circle , C is the number of worms paralyzed within the control circle . We repeated the test a minimum of 10 times for each population ( please refer to Supplementary file 1A , C–E for the number of assays and worms used in each experiment ) . The absence of artificial magnetic fields and temperature gradients were empirically determined before each assay with DC Milligauss Meter Model MGM magnetometer ( AlphaLab , Utah , USA ) and two high accuracy Fisher thermometers accurate to 1/100 of a degree ( Figure 2—figure supplement 1E , F ) . Assays were run over multiple days and across a range of temperatures ( 19–21° Celsius ) . To ensure that unaccounted gradients in the room did not affect the assays , we ran multiple assays in parallel . We arranged assay plates so that their magnetic gradients were not aligned with one another or with the magnetic field of the earth . In this configuration , the magnetic field on the plate surface ranged between 40 and 2900 Gauss ( Figure 4A ) . To test if worms could perform magnetotaxis in the dark , we wrapped assay plates in heavy-duty aluminum foil at least three layers thick . All burrowing experiments were conducted similarly with the pipettes wrapped in multiple layers of aluminum foil . All magnetic coil system experiments were conducted in the dark . Additionally , burrowing and magnetic coil system experiments were conducted within an opaque Faraday cage consisting of copper mesh . In order to test worms under earth-like homogeneous magnetic fields we constructed a triple Merritt coil system ( Merritt et al . , 1983 ) of 1 m3 in volume capable of generating a homogeneous magnetic field in the central 20 cm3 of the space . A 22 cm3 , copper fabric , Faraday cage around the test volume prevented electric fields from interfering with our assays . Each of the three coil systems was orthogonal to the other two ( Figure 1—figure supplement 1 ) and was independently powered by Maxtra Adjustable 30V 5A DC power supplies . We used a DC Milligauss Meter Model MGM from Alphalab Inc . ( Utah , USA ) to measure the magnetic field inside the coil system before and after each experiment . Before the start of each experiment , the system was used to neutralize the magnetic field of the earth within the coil system by creating a field of equal strength and opposite orientation . A single 10-cm diameter , agar-filled plate ( Ward , 1973 ) , with ∼50 worms in its center , was placed in the center of the coil system ( Figure 1—figure supplement 1 ) . To immobilize and count the worms that reached the plate's edge , we placed a 10-μl ring of 0 . 1-M NaN3 anesthetic in the agar around circumference of the plate . We next closed the Faraday cage and allowed the worms to migrate freely within the plate with a homogeneous earth-strength ( 0 . 325 and 0 . 650 Gauss ) magnetic field aligned with the plane of the assay plate . Alternatively , we allowed worms to migrate in plates when the effective magnetic field inside the coil system was 0 . 000 Gauss ( earth-neutralized ) . After an hour , the angle at which each worm had migrated with respect to the imposed field vector was recorded . To ensure that the worms were responding to the generated magnetic field , and not to some unknown gradient in the room , all assays were run in darkness with the direction of the imposed magnetic field , and also the orientation of the magnetic coil system was varied between trials with respect to the room and the earth's magnetic north . All experiments were conducted blind with experimenter unaware of strain genotype . In addition to magnetic and electric fields , the wires of coil system also produce a small degree of heat . To test for the presence of unwanted temperature gradients we used two high accuracy ( 0 . 01°C ) thermometers ( Fisher Scientific , New Hampshire ) to record the temperature at the center of the plates ( where the worms begin the assay ) and at the edge of the plates ( where they complete the assay ) . Please refer to Supplementary file 1B for a summary of sample and population sizes , and for a statistical description ( and comparisons ) of each dataset in the magnetic coil system assays . We filled 5-ml plastic pipettes with 3% chemotaxis agar and cut and sealed the ends with Parafilm to minimize the formation of gaseous/humidity gradients similar to our previous study ( Beron et al . , 2015 ) . We made three equidistant holes 10 cm apart in the pipettes . We injected 50 never-starved ( day-1 ) adults into the center hole , and 1 . 5 μl 1-M of NaN3 into the end holes to immobilize and easily count the worms that reached either side . Worms were first picked from their incubation plates into a 1-μl liquid NMG solution and transferred within 5 min into the center of the pipettes . Care was taken to ensure that worms were injected into solid agar ( rather than remaining suspended in liquid solution once injected ) . The genotype of the strains was kept blind during the prep and running of the assay . The holes in the pipettes were then sealed with Parafilm . We wrapped the pipettes in aluminum foil multiple times ( >3 ) , to maintain the assays in complete darkness . Pipettes were aligned horizontally , in either the north-south or the east-west direction , or vertically in the up-down direction . Pipettes were placed inside a Faraday envelope made with copper cloth to prevent electric fields from interfering with the assays . The assays were allowed to run overnight and the worms immobilized at either end of the pipette were counted . The burrowing index ( BI ) was calculated as: BI = ( A − B ) / ( A + B ) . Where A and B are the number of worms on opposite ends of the assay pipette . To assess if the surrounding magnetic field could disrupt burrowing behavior , we arranged burrowing pipettes vertically inside the magnetic coil system . We next generated a magnetic field of earth strength that had the opposite inclination ( magnetic north up ) to the local magnetic field ( magnetic north down ) . Worms were allowed to burrow and their burrowing index was calculated as described above . To determine if the presence of an artificial magnetic field in the Merrit Coil System produced a temperature gradient in the assay plate we used a Fisher High Accuracy thermometer sensitive to 0 . 01°C ( Figure 2—figure supplement 1 ) . We placed one probe on the center of the plate ( where worms begin the assay ) and one probe on the edge of the plate ( where worms normally end the assay ) . After closing the Faraday cage , and with the coil system off , we measured the temperature on each probe in 5 min intervals for 20 min . At this point we powered the cage on to produce either a horizontal magnetic field of 0 . 650 Gauss ( 1× earth ) , or with a magnetic field able to cancel out the earth's own magnetic field inside the cage ( field cancelled ) . We continued to record the temperature of the probes every 10 min for 1 hr ( the duration of a typical coil system experiment ) . We next powered off the magnetic cage and once again recorded the temperature every 5 min for an additional 20 min ( Figure 2—figure supplement A , B ) . To determine if the magnets placed above the plates in our magnet assays produced a temperature gradient we placed temperature probes on the surface of the agar above the test and control positions of agar plates as indicated in the magnet assay procedure above . We compared the two temperature readings every 10 min for 1 hr Figure 2—figure supplement C , D ) . To determine if temperature gradients were created by the presence of a magnet we carried out these experiments both in the presence and in the absence of a magnet . Between temperature experiments , the thermal probes were immersed in a beaker containing 1 l of dH2O at the same distance they had during the experiments to determine if their reading differed from one another ( Figure 2—figure supplement E , F ) . Worms carrying extrachromosomal arrays with transgenes do not pass this construct to all their offspring . This permitted us to blindly test clones that are identical in their genetics and in their upbringing , only differing from one another in having ( or not ) the extrachromosomal array . In these experiments , a mixed population of worms ( both carrying and not carrying the array ) was tested together . After the assay , we used a fluorescent co-injection marker linked to the transgene of interest to identify and count the number of worms belonging to each population separately . We next calculated the magnetic orientation ( or burrowing ) index for each subpopulation of worms . This ensured that the comparisons were made between genetically identical populations that had grown under identical conditions , their only difference being whether or not they carried the extrachromosomal array . Expression of GCaMP3 . 0 did not interfere with ability to perform magnetotaxis to an artificial magnet ( Figure 7—figure supplement 1C ) . Day-1 adult wild-type ( and mutant ) worms expressing the calcium-activity reporter GCaMP3 in neurons of interest were loaded with liquid NGM into a microfluidic chip ( Figure 7—figure supplement 1A ) . Worms were immobilized and imaged in an Olympus BX51 scope at ×60 magnification . Images were sampled at 3 . 5–8 Hz using a CoolSNAP ES camera run by Windwiew 32 . Each run lasted 50 s and begun with a 12 . 5 s of baseline followed by the presentation of a 6-s sinusoidal magnetic stimulus , and a 31 . 5-s recovery period . Consecutive recordings were made 3–5 min apart . A N42 Neodymium 3 . 5-cm diameter magnet ( K&J Magnetics Inc . , Pennsylvania , USA ) was used to deliver the magnetic stimulus . The intensity of the stimulus was calibrated with a DC Milligauss Meter Model MGM magnetometer ( AlphaLab , Utah , USA ) . The putative role of the magnetic sensor in the worm is to detect the direction of the magnetic field . It therefore follows that the cell must have an optimal stimulation angle between its sensor and the surrounding field . Because we could not infer what this optimal alignment angle may be , we decided to rotate the stimulus vector throughout 360° to ensure that each worm was stimulated with its presumably optimal angle . We did this by rotating the magnet along the xy plane followed by the xz plane at a rate of 2 Hz . A series of TIFF files was exported into ImageJ where the brightness of the cell body was measured across each photo series . The average soma brightness accounting for bleaching was calculated as previously described ( Kerr and Schafer , 2006 ) . Worms expressing GCaMP3 were incubated as described above and placed along with 3-μl liquid NGM on a 10% agar pad on a microscope slide with the coverslip pressed down to inhibit swimming , but permit slow crawling . Neurons were only imaged when the cells remained in the focal plane for the duration of the experiment . We placed the slide in an upright Olympus BX53 microscope equipped with and Retiga 2000R Fast 1394 camera ( Q-Imaging , BC , Canada ) . Worms were illuminated with a Series 1200 UV light source ( X-cite , Cincinnati , Ohio ) . We took a series of four pictures before , during , and after exposing the setup , for 10 s , to a 60 Gauss magnetic field generated by a Neodymium magnet . The Tiff files of the images were exported as 8-bit files . We used ImagePro 6 . 0 ( MediaCybernetics , Rockville , MD ) to measure the brightness of the soma in each picture series . We averaged the eight images before and after magnet exposure and compared this value to the average soma brightness of the four images taken during magnet exposure . We reported the percent change in brightness of the test condition to the average of before and after ( Figure 7—figure supplement 1B ) . All plots were graphed using SigmaPlot 12 ( Aspire Software ) and Matlab R2013b ( Mathworks ) . Multiple plates were assembled in CorelDRAW X6 ( Corel ) . All bars correspond to means , and variation is given as SEM throughout . Linear statistical analyses were performed using SigmaPlot 12 ( Aspire Software ) . Comparisons between different experimental groups were performed by planned , two-tailed paired or unpaired t-tests to compare different groups that were normally distributed . Differences between non-normally distributed groups ( or groups that failed the test of equal variance ) were evaluated using the Mann–Whitney Ranked Sum Test , and two way repeated measures ANOVA ( Temperature experiments ) . Correlations between parameters were determined using linear regressions and assessed using Pearson product–moment correlation coefficients . Circular statistical analyses ( descriptive and comparative ) were performed using a circular statistics toolbox for Matlab 2013b ( Berens , 2009 ) . We tested the significance of mean directions using Rayleigh tests , and for difference between vertical ‘up’ or ‘down’ direction and the mean direction of the population using V tests ( Batschelet , 1981 ) . Throughout this study , p values were reported using the convention: * p < 0 . 05 , ** p < 0 . 001 . Please refer to Supplementary file 1A–E for descriptive , and Supplementary file 1A , B for comparative statistics . | The Earth has a magnetic field that protects the planet from the harmful effects of cosmic rays , which is generated by the movement of the layer of molten metal that surrounds the planet's solid inner core . The orientation of the magnetic field relative to the Earth's surface varies around the globe , and is like the pattern adopted by iron filings around a bar magnet . Many organisms , from bacteria to birds such as the Arctic Tern and mammals such as wolves , are able to exploit this variation in the Earth's magnetic field to help them navigate . In bacteria , this ability has been linked to the possession of tiny magnetic particles that align with the Earth's magnetic field lines . However , it is not clear how animals are able use the magnetic field to navigate . Vidal-Gadea et al . have now obtained clues to this process from a surprising source , namely a soil-dwelling nematode worm called Caenorhabiditis elegans . Worms that were placed inside a gelatin-filled cylinder preferred to burrow downward , but exposure to an artificial magnetic field that in effect reversed the Earth's magnetic field caused them to burrow upward . By contrast , worms that had been well-fed burrowed in the opposite direction . When worms were instead allowed to crawl across a flat surface with a parallel magnetic field , a map of the Earth's magnetic field revealed that the worms , which were originally from England but which were tested in the US , moved at an angle that would correspond—in England—to burrowing downwards when they were hungry and upwards when were not . Consistent with this , worms from Australia crawled in the opposite direction to English worms , setting off at an angle that would also be equivalent—in Australia—to downwards when hungry and upwards when full . Identical results were found for Hawaiian worms . Worms from countries on the Equator were less sensitive to magnetic fields than their northern and southern counterparts; which suggests that the response is genetically encoded . And indeed , a survey of mutant worms with defects in different neurons uncovered a pair of sensory neurons called ‘AFD neurons’ that serve the purpose of a compass , and allow the animals to navigate using geomagnetism . It remains unclear exactly how AFD neurons detect the Earth's magnetic field , and it is also not clear why hungry worms should burrow downwards while satisfied worms burrow upwards . One possibility is that hungry worms move downwards to feed on abundant but nutrient-poor bacteria growing on plant roots , whereas sated worms can afford to risk moving to the surface in search of more desirable , but less reliable , food sources . Future work could set out to test these explanations . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2015 | Magnetosensitive neurons mediate geomagnetic orientation in Caenorhabditis elegans |
Familiarity discrimination has a significant impact on the pattern of food intake across species . However , the mechanism by which the recognition memory controls feeding is unclear . Here , we show that the nematode Caenorhabditis elegans forms a memory of particular foods after experience and displays behavioral plasticity , increasing the feeding response when they subsequently recognize the familiar food . We found that recognition of familiar food activates the pair of ADF chemosensory neurons , which subsequently increase serotonin release . The released serotonin activates the feeding response mainly by acting humorally and directly activates SER-7 , a type 7 serotonin receptor , in MC motor neurons in the feeding organ . Our data suggest that worms sense the taste and/or smell of novel bacteria , which overrides the stimulatory effect of familiar bacteria on feeding by suppressing the activity of ADF or its upstream neurons . Our study provides insight into the mechanism by which familiarity discrimination alters behavior .
Wholesome food is essential for survival and animals have developed nervous systems that guide food intake . The nervous system senses diverse extrinsic and intrinsic cues , integrates the information and activates muscle movements that are required for food intake . The nervous system also stores past food experiences , which change the pattern of food intake . Dissection of the neural pathways that control food intake is not only a key to stop the epidemic of obesity and eating disorders , but may also provide insight into fundamental problems in neuroscience such as sensory perception and learning and memory . Recognition is the ability to identify and to judge a recently encountered item as having been presented previously ( Brown and Aggleton , 2001 ) . In response to previously encountered stimuli , this ability allows knowledge gained from prior experience to guide animals to respond with an altered output that is beneficial for their survival . Recognition is classified into two types: recollection and familiarity discrimination . Recollection is knowledge of prior occurrence with vivid contextual details . In contrast , familiarity discrimination is mere sensation of prior occurrence and thus does not accompany episodic memory ( Brown and Aggleton , 2001 ) . Accumulated studies show that mere exposure to particular food alters subsequent consumption of the food in many different species ( Pliner et al . , 1993; Wang and Provenza , 1996; Diaz-Cenzano and Chotro , 2010; Morin-Audebrand et al . , 2012 ) , suggesting that feeding regulation by familiarity discrimination is conserved across species . Some species including humans consume familiar food more actively than novel food ( Diaz-Cenzano and Chotro , 2010 ) , probably to avoid possible pathogens . In contrast , other species consume familiar food less actively than novel food ( Wang and Provenza , 1996 ) , probably to assure balanced nutrition intake . Despite extensive studies of recognition ( Brown and Aggleton , 2001; Barker et al . , 2006; Seoane et al . , 2009; Uslaner et al . , in press ) and subsequent behavioral plasticity ( Kandel and Schwartz , 1982; Kravitz , 1988 ) , the mechanisms by which familiarity discrimination alters food intake are still poorly understood . Its genetic tractability and simple anatomy make the bacteria-eating roundworm C . elegans ( Schafer , 2005 ) an attractive model system for study of the process . Although it is unknown if familiarity discrimination alters food intake in C . elegans , the following observations support the possibility: The nervous system in C . elegans senses various aspects of food , such as the efficiency with which it supports growth ( Shtonda and Avery , 2006 ) and its pathogenicity ( Zhang et al . , 2005 ) , and triggers behavioral plasticity . The nervous system can also form memories of various olfactory or gustatory cues ( Bargmann , 2006 ) , which is likely to be crucial for the recognition of familiar food . Here , we show that C . elegans discriminates familiar food from novel food and selectively increases feeding in response to familiar food . Using the behavioral pattern that we identified , we uncover the mechanism by which familiarity discrimination increases the feeding response .
To test if familiarity discrimination alters feeding in C . elegans , we tested if exposure to a particular bacterium alters its subsequent consumption . For this assay , we trained wild-type animals to develop familiarity either to Escherichia coli HB101 ( H ) or to Pseudomonas DA1878 ( D ) ( also called B7 in our previous study , Avery and Shtonda , 2003 ) by exposing the animals to one or the other bacterium from the first larval stage ( L1 ) until adulthood ( Figure 1A ) . HB101 and DA1878 are benign bacteria that support the growth of C . elegans at a similar rate ( Avery and Shtonda , 2003 ) . Once the animals reached adulthood , we compared feeding rates on previously experienced bacteria ( HH and DD groups ) to feeding rates on novel bacteria ( DH and HD groups ) ( Figure 1A; see ‘Feeding assay’ and ‘Statistical analysis and data presentation’ in ‘Materials and methods’ for details ) . We found that the feeding rates of the animals on familiar bacteria were significantly higher than the rates on novel bacteria , regardless of bacterial type ( Figure 1B , C ) . The increased feeding rates on familiar food compared to the rates on novel food persisted 7–8 hr after the training was over , supporting that worms discriminate familiar food from novel food and that recognition of familiar bacteria increases feeding ( Figure 1D , E ) . Consistent with this , the familiar food-induced increase in feeding was not affected by the worm's nutritional status ( Figure 1B–E ) . We then tested if the behavior is selective for the two tested bacteria by repeating the experiment using other benign bacterial strains , Enterobacteria JU54 and Pseudomonas PA14 pstP ( Tan et al . , 1999 ) ( Figure 2A ) . Consistent with the idea that mere exposure to any benign bacterium increases subsequent consumption of that bacterium , the feeding rates on familiar bacteria were higher than the rates on novel bacteria , regardless of bacteria type ( Figure 2B–E ) . Finally , we considered the possibility that cultivation on two different bacteria caused the differences in feeding rates on familiar food and novel food by affecting development . If the feeding differences are caused by developmental differences , not familiarity discrimination , an exposure to particular bacteria during adulthood is not expected to cause an increased feeding response to the bacteria . However , 9 hr of exposure to a particular bacterium during adulthood was sufficient to induce a preferential response to that bacterium ( Figure 3D , E ) . In contrast to 9 hr , 6-hr exposure to the training food failed to increase subsequent consumption ( Figure 3B , C ) , suggesting that the behavioral plasticity was determined by the duration of exposure to the training food . These data imply that C . elegans forms a recognition memory of particular bacteria after experience , which allows the worms to discriminate familiar bacteria from novel bacteria , and that the recognition of familiar bacteria increases the feeding response . 10 . 7554/eLife . 00329 . 003Figure 1 . Recognition of familiar food increases feeding response in C . elegans . The memory of familiar food lasts for at least 7 hr . ( A ) Experimental design for the feeding assay . The periods during which animals were exposed to HB101 , DA1878 , and starvation are shown in blue , red , and white , respectively . Each condition is coded by two letters representing the training and test food in order . H and D represent HB101 and DA1878 , respectively . ( B ) – ( C ) Feeding rates of wild-type worms on HB101 ( B ) and DA1878 ( C ) just after training the animals on one or the other bacterium . ( D ) – ( E ) Feeding rates of wild-type worms on HB101 ( D ) and DA1878 ( E ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Data shown as mean ± SEM , ***p<0 . 001 , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays for each group ) is shown in parenthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00310 . 7554/eLife . 00329 . 004Figure 2 . Recognition of familiar food increases feeding response in C . elegans . The memory of familiar food lasts for at least 7 hr . ( A ) Experimental design for the feeding assay . The periods during which animals were exposed to HB101 , JU54 or PA14 pstP , and starvation are denoted blue , red , and white , respectively . Each condition is coded by two letters representing the training and test foods in order . H , J and P represent HB101 , JU54 and PA14 pstP , respectively . ( B ) – ( C ) Feeding rates of wild-type worms on HB101 ( B ) and JU54 ( C ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . ( D ) – ( E ) Feeding rates of wild-type worms on HB101 ( D ) and PA14 pstP ( E ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Data shown as mean ± SEM , ***p<0 . 001 , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00410 . 7554/eLife . 00329 . 005Figure 3 . A 9-hr exposure , but not a 6-hr exposure , to particular bacteria during adulthood increases its subsequent consumption . ( A ) Experimental design for the feeding assay . Coding is as in Figure 1A . Each condition is coded by three letters representing the cultivation , training and test food in order . ( B ) – ( E ) feeding rates of wild-type worms on HB101 ( B and D ) And DA1878 ( C and E ) After a 7- to 8-hr interval from training the animals on one or the other bacterium . Data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 , one-way ANOVA , post hoc Tukey test . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown in parenthesis . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 005 The neurotransmitter serotonin increases feeding in C . elegans ( Croll , 1975; Avery and Horvitz , 1990 ) and serotonin has long been suggested to be a food signal in C . elegans ( Horvitz et al . , 1982 ) . However , the serotonin effect on feeding was tested only on familiar bacteria; thus it is unknown if the serotonin feeding signaling is activated by novel bacteria . We thus hypothesized that recognition of familiar bacteria might increase the feeding response by activating serotonin signaling . To test our hypothesis , we first tested if a tph-1 null mutation suppresses the difference in the feeding rates between animals on familiar bacteria and novel bacteria . tph-1 encodes a tryptophan hydroxylase required for serotonin biosynthesis ( Sze et al . , 2000 ) . We found no difference in the feeding rates between the tph-1 null mutant animals on familiar bacteria and novel bacteria ( Figures 4A , B and 5 ) . Furthermore , exogenous serotonin treatment increased the feeding rates on novel food to rates comparable to those on familiar food without affecting the rates on familiar food ( Figure 4C , D ) . These data suggest that recognition of familiar bacteria indeed increased feeding rate by activating serotonin signaling . To find out which serotonin receptors mediate the serotonin action on feeding , we compared the feeding rates between wild-type and serotonin receptor null mutants in presence of serotonin . The C . elegans genome encodes four serotonin-activated G protein coupled receptors , SER-1 , SER-4 , SER-5 and SER-7 , and a serotonin-gated Cl− channel MOD-1 . Among the tested mutants , only the ser-7 null mutant failed to activate feeding in presence of serotonin , confirming the previous report that serotonin activates feeding via a type 7 serotonin receptor SER-7 ( Hobson et al . , 2006; Song and Avery , 2012 ) ( Figure 6 ) . Consistent with the idea that serotonin increases feeding via SER-7 in response to familiar food , the feeding rates of ser-7 null mutants on familiar food were substantially decreased compared to wild type ( Figure 7A , B ) . Furthermore , no differences were found in the feeding rates among the tph-1 single null mutant , the ser-7 single null mutant and the tph-1; ser-7 double null mutant on familiar bacteria ( Figure 8A ) . Since SER-7 is the major receptor to mediate the serotonin action , we initially expected no difference between feeding rates of the ser-7 null mutant on familiar food and novel food , as for the tph-1 null mutant . When the ser-7 null mutant animals were tested on DA1878 , the feeding rates of the animals did not differ on familiar and novel food ( Figures 4F and 7B , D ) . However , when the animals were tested on HB101 , the feeding rate of the animals was lower on familiar test food than on novel test food ( Figures 4E and 7A , C ) . To resolve the unexpected observation , we examined feeding rates in wild type and the ser-7 null mutant animals in the presence or absence of serotonin . Serotonin suppressed feeding in the ser-7 null mutant ( Figure 4G ) , suggesting that serotonin suppresses as well as activates feeding . We also found that null mutations in ser-4 and mod-1 completely relieved the suppression of feeding by serotonin in the ser-7 mutant ( Figure 4G ) , suggesting that serotonin suppressed feeding by acting on SER-4 and MOD-1 . In support of the idea that the inhibitory serotonin signal via SER-4 and MOD-1 is active only on familiar food , the feeding rates of the ser-4; mod-1; ser-7 triple null mutant were greater than the rates of the ser-7 single null mutant on familiar food ( Figures 4E , F and 7C , D ) whereas the feeding rates of the two mutant animals were not different on novel food . Interestingly , the amount of suppression by the inhibitory serotonin signal on familiar food , that is the difference between feeding rates of the ser-4; mod-1; ser-7 mutant and the ser-7 null mutant animals on familiar food , was greater on HB101 than DA1878 ( Figures 4E , F and 7F ) . Although it is not clear why the activities of the inhibitory serotonin signal on the two bacteria are different , it explains the unexpected observation seen in the feeding rate differences between the ser-7 null mutant animals on familiar food and novel food in Figure 4E and F . We next tested whether the inhibitory serotonin signal is essential for recognition of familiar food or its regulation of feeding by comparing the feeding rates of the ser-4; mod-1 double null mutant animals on familiar food and novel food . Consistent with the idea that SER-7 is the major serotonin receptor , the double mutant , like wild type , showed increased feeding responses on familiar food compared to novel food ( Figure 9 ) . 10 . 7554/eLife . 00329 . 006Figure 4 . Recognition of familiar food increases feeding response by activating serotonin signaling via SER-7 . SER-4 and MOD-1 are putative inhibitory receptors . ( A ) – ( B ) Familiarity of food does not alter the feeding rates in tph-1 ( mg280 ) . ( C ) – ( D ) Exogenous serotonin treatment selectively increases feeding rates of wild type worms on novel food to the level of the worms on familiar food . The average values of the feeding rates presented in ( C ) are 197 . 7 ± 4 . 6 , 174 . 3 ± 5 . 8 , 191 . 2 ± 4 . 9 and 190 . 7 ± 3 . 4 in order . The average values of the feeding rates presented in ( D ) are 189 . 0 ± 7 . 9 , 158 . 4 ± 9 . 2 , 184 . 2 ± 3 . 3 and 181 . 2 ± 4 . 6 in order . ( E ) – ( F ) ser-7 ( tm1325 ) is defective in increasing feeding response to familiar food compared to novel food . Familiarity of food does not alter the feeding rates in ser-4 ( ok512 ) ; mod-1 ( ok103 ) ; ser-7 ( tm1325 ) . The average values of the feeding rates presented in ( E ) are 189 . 7 ± 2 . 6 , 225 . 5 ± 3 . 6 , 238 . 3 ± 2 . 1 and 231 . 3 ± 3 . 4 in order . The average values of the feeding rates presented in ( F ) are 178 . 8 ± 5 . 5 , 188 . 4 ± 4 . 2 , 196 . 3 ± 3 . 2 and 191 . 9 ± 3 . 1 in order . ( G ) Serotonin controls feeding positively via SER-7 and negatively via SER-4 and MOD-1 . The feeding rate of the ser-4; mod-1; ser-7 triple null mutant is not altered by serotonin treatment . These assays were conducted on 3- to 5-hr-old L1 larvae , which pumped much more slowly than the adults used in other measurements . The average values of the feeding rates presented in ( G ) are 29 . 1 ± 4 . 0 , 73 . 6 ± 3 . 3 , 29 . 8 ± 4 . 2 , 12 . 2 ± 1 . 8 , 42 . 7 ± 11 . 3 , 26 . 3 ± 4 . 3 , 45 . 9 ± 8 . 0 and 86 . 8 ± 11 . 5 in order . ( H ) Serotonin signaling via SER-7 that activates the feeding response is more active on familiar food than novel food . The y axis indicates the difference in the feeding rates between wild-type and ser-7 ( tm1325 ) animals . Each value corresponds to the difference in the feeding rates between wild-type and the ser-7 null mutant presented in Figure 7A and B . Data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001; for Figure 4A and B , unpaired t-test and Mann–Whitney U test ( two-tailed ) , for Figure 4C–G , one-way ANOVA , post hoc Tukey test and for Figure 4H , Student's t test ( two-tailed; see ‘Detailed data analysis’ in ‘Materials and methods’ ) . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown in parentheses or at the bottom of each bar . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00610 . 7554/eLife . 00329 . 007Figure 5 . Serotonin is required to increase feeding in response to familiar food . Feeding rates of tph-1 ( mg280 ) on HB101 ( A ) and JU54 ( B ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00710 . 7554/eLife . 00329 . 008Figure 6 . Feeding rates of wild type and five serotonin receptor null mutants in presence of serotonin . Among five serotonin receptor null mutants , only ser-7 ( tm1325 ) failed to activate feeding in response to serotonin . A null mutation in ser-4 also decreased the feeding rate in presence of serotonin but the effect was relatively small . These assays were conducted on 3- to 5-hr-old L1 larvae , which pumped much more slowly than the adults used in other measurements . The average values of the feeding rates presented in this figure are 73 . 6 ± 3 . 3 , 63 . 5 ± 6 . 2 , 52 . 1 ± 5 . 4 , 82 . 3 ± 10 . 5 , 12 . 2 ± 1 . 8 and 90 . 9 ± 11 . 6 in order . *p<0 . 05 , ***p<0 . 001; one-way ANOVA , post hoc Tukey test . The number of animals tested ( n ≥ 2 independent assays per each group ) is shown at the bottom of the bar . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00810 . 7554/eLife . 00329 . 009Figure 7 . Feeding rates of wild-type , ser-7 single and ser-4; mod-1; ser-7 triple null mutant on HB101 and DA1878 and model of feeding regulation by serotonin . ( A ) – ( B ) Feeding rates of wild-type ( + ) and ser-7 ( tm1325 ) on HB101 ( A ) and DA1878 ( B ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Wild-type worms feed more actively on familiar food than novel food . On novel food the feeding rate of wild-type is slightly higher than that of the ser-7 null mutant . The difference may be due to constitutive activity of SER-7; that is , SER-7 is active to some extent even in absence of its ligand , serotonin ( Hobson et al . , 2003 ) . ( C ) – ( D ) Feeding rates of ser-4 ( ok512 ) ; mod-1 ( ok103 ) ; ser-7 ( tm1325 ) and ser-7 ( tm1325 ) on HB101 ( C ) and DA1878 ( D ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . ser-7 ( tm1325 ) is defective in increasing the feeding response to familiar food compared to novel food . Familiarity of food does not alter feeding rates in ser-4 ( ok512 ) ; mod-1 ( ok103 ) ; ser-7 ( tm1325 ) . Like the positive SER-7-mediated signal , the inhibitory SER-4- and MOD-1-mediated serotonin signaling is more active on familiar food than novel food , but it decreases the feeding rate . ( E ) A simple linear model explaining how different serotonin receptors might contribute to the regulation of pumping on familiar food and on novel food . There are three effects: B: Basal activity of SER-7 , S: Serotonin-stimulated activity of SER-7 , and −I: Serotonin-stimulated activity of inhibitory serotonin receptors SER-4 and MOD-1 . The net effect of serotonin on wild-type ( + ) pumping is S + B − I; the net effect on pumping in a mutant lacking SER-7 is −I . While it is presented as an aid to thinking about the results , none of the results presented in the paper depend on this model . Figure 4H , in particular , is a direct measurement of the effect of SER-7 under differing conditions , calculated as the difference in feeding rates between wild-type ( + ) and the ser-7 null mutant worms . A change in this number suggests the action of serotonin via SER-7 . We use this as the measure of serotonin action via SER-7 because it is model-independent and robust . ( F ) Serotonin signaling via SER-4 and MOD-1 that suppresses the feeding response on familiar food is more active on HB101 than DA1878 . The y axis indicates the difference in the feeding rates between ser-4 ( ok512 ) ; mod-1 ( ok103 ) ; ser-7 ( tm1325 ) and ser-7 ( tm1325 ) animals . Each value corresponds to the difference in the feeding rates between the triple null mutant and the ser-7null mutant presented in ( C ) and ( D ) . For ( A–D ) and ( F ) , data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , *p<0 . 05 , ***p<0 . 001; for ( A–D ) , one-way ANOVA , post hoc Tukey test and for ( F ) , Student's t test . The number of animals tested ( n ≥ 3 independent assays per group ) is shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 00910 . 7554/eLife . 00329 . 010Figure 8 . Serotonin from ADF activates feeding in response to familiar food mainly by directly activating SER-7 in MC pharyngeal motor neurons . Active SER-7 in MC ( and possibly in M4 ) acts mainly via cholinergic transmission from MC to the pharyngeal muscles . ( A–B ) tph-1 expression in ADF , but not in NSM , restores feeding rate in the tph-1 null mutant . The rescue effect is suppressed by loss of ser-7 , but not by loss of mod-5 . No difference was found in feeding rates between the tph-1 single null mutant , the ser-7 single null mutant and the tph-1; ser-7 double null mutant . The average values of the feeding rates presented in ( A ) are 266 . 1 ± 3 . 0 , 207 . 7 ± 1 . 9 , 206 . 6 ± 4 . 2 , 277 . 2 ± 6 . 0 , 261 . 1 ± 5 . 3 , 216 . 4 ± 4 . 0 , 267 . 3 ± 3 . 1 , 217 . 5 ± 2 . 8 , 204 . 1 ± 3 . 0 and 216 . 9 ± 1 . 8 in order . ( C–D ) ADF-minus animals , but not NSM-minus animals , feed significantly less in response to familiar food . ( E ) Expression of ser-7 cDNA driven either by the flp-2 promoter or by the flp-21 promoter ( MC , M4 , and other neurons ) but not by the ser-7b promoter ( M4 only ) fully restored the feeding rate in the ser-7 null mutant in response to serotonin . The rescue effect was suppressed by blocking cholinergic transmission from MC to the pharyngeal muscles . #Pharyngeal pumping rate was lower in the eat-2; ser-7 double null mutant than the eat-2 single null mutant ( p<0 . 001 ) and the ser-7 single null mutant ( p=0 . 002 ) . The difference suggests that acetylcholine marginally activates pumping in an EAT-2-independent manner and that there is residual acetylcholine release in absence of SER-7 in response to serotonin . No difference in feeding rates was found between the eat-2; ser-7 mutant expressing pflp-21::gfp and the mutant expressing pflp-21::ser-7 cDNA . The average values of the feeding rates presented in ( E ) are 73 . 6 ± 3 . 3 , 12 . 2 ± 1 . 8 , 39 . 3 ± 7 . 1 , 51 . 2 ± 8 . 9 , 117 . 8 ± 13 . 7 , 107 . 7 ± 10 . 4 , 132 . 6 ± 10 . 6 , 16 . 9 ± 3 . 1 , 2 . 9 ± 1 . 0 , 8 . 3 ± 4 . 1 and 15 . 6 ± 1 . 4 in order . ( F ) Expression of ser-7 cDNA driven either by the flp-2 promoter or by the flp-21 promoter fully restored the feeding rate in the ser-7 null mutant in response to familiar food . Expression of ser-7 cDNA in M4 ( and occasionally in M2 ) driven by the ser-7b promoter also increased the feeding rate , but the effect was relatively small . The average values of the feeding rates presented in ( F ) are 189 . 7 ± 2 . 6 , 194 . 9 ± 3 . 4 , 212 . 0 ± 5 . 1 , 245 . 1 ± 4 . 7 , 240 . 7 ± 4 . 6 and 253 . 6 ± 2 . 5 in order . Data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , *p<0 . 05 , ***p<0 . 001; for ( A–B ) and ( E–F ) , one-way ANOVA , post hoc Tukey test , for ( C–D ) , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown on each bar . ‘0’ and ‘wt’ in this figure indicate absence of transgene and wild type , respectively . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 01010 . 7554/eLife . 00329 . 011Figure 9 . SER-4 and MOD-1 are not essential for discriminating familiar food from novel food . Feeding rates of ser-4 ( ok512 ) ; mod-1 ( ok103 ) on HB101 ( A ) and DA1878 ( B ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Like wild type worms , ser-4 ( ok512 ) ; mod-1 ( ok103 ) show increased feeding response on familiar food compared to novel food . Data shown as mean ± SEM , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays per group ) is shown in parentheses . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 011 To test if serotonin feeding signaling via SER-7 indeed gets activated by recognition of familiar food , we compared the differences between feeding rates of wild-type and the ser-7 null mutant animals ( the SER-7 effect ) on familiar food with the differences on novel food . Any feeding rate difference between wild-type and the ser-7 mutant animals indicates active serotonin signaling via SER-7 because ser-7 specifically affects serotonergic signaling , with some contribution from basal activity of SER-7 in the absence of serotonin ( Hobson et al . , 2003 ) ( Figure 7E ) . If serotonin signaling is equally active on familiar food and novel food , we expect the SER-7 effect to be similar on familiar food and novel food . However , on familiar food the SER-7 effect was far greater than on novel food ( Figure 4H; see ‘Detailed data analysis’ in ‘Materials and methods’ ) , suggesting that serotonin signaling via SER-7 is indeed more active on familiar food than novel food . We concluded that recognition of familiar food increases the feeding response mainly by activating serotonin signaling via SER-7 . To gain insight into how serotonin signals familiar bacteria , we asked which serotonergic neurons regulated the feeding response . Serotonin is detected in five types of neurons in C . elegans hermaphrodites: NSM , ADF , HSN , RIH and AIM ( Sze et al . , 2000 ) . RIH and AIM obtain serotonin by taking up extracellular serotonin ( Jafari et al . , 2011 ) . HSN is also unlikely to be necessary for the behavioral plasticity because feeding rates of males , which do not have HSN , were also greater on familiar food than the rates on novel food ( Figure 10 ) . We therefore hypothesized that either NSM or ADF uses serotonin to control feeding . The NSM neurons are a pair of secretory neurons located in the pharynx , whereas the ADF neurons are a pair of chemosensory neurons located outside the pharynx ( Sze et al . , 2000 ) that have been suggested to sense bacteria ( Bargmann and Horvitz , 1991 ) . We asked if serotonin either in ADF or in NSM regulates the feeding response by expressing tph-1 cDNA in the tph-1 null mutant using either the srh-142 promoter or the ceh-2 promoter . The srh-142 promoter drives expression specifically in ADF and the ceh-2 promoter drives expression in NSM and three additional neurons ( Liang et al . , 2006 ) . We found that restoring serotonin synthesis in ADF , but not in NSM , rescued the feeding response in the tph-1 mutant ( Figure 8A , B ) , suggesting that ADF regulates feeding in response to familiar bacteria . Laser killing of ADF , but not NSM , also decreased feeding on familiar food ( Figure 8C , D ) —the difference in feeding rates between ADF-minus and mock-operated animals ( 49 . 8 ± 7 . 4; feeding rates of ADF-minus and mock-operated animals were 199 ± 7 . 7 and 249 ± 3 . 5 , respectively ) was comparable to the difference in the feeding rates between tph-1 and wild-type animals ( 58 . 4 ± 3 . 5; feeding rates of tph-1 and wild-type animals were 207 . 7 ± 1 . 9 and 266 . 1 ± 3 . 0 , respectively ) , further supporting the idea that ADF regulates feeding in response to familiar bacteria . 10 . 7554/eLife . 00329 . 012Figure 10 . Male worms discriminate familiar food from novel food . ( A–B ) Feeding rates of wild-type male worms on HB101 ( A ) and DA1878 ( B ) after a 7- to 8-hr interval from training the animals on one or the other bacterium . Data shown as mean ± SEM , unpaired t-test and Mann–Whitney U test ( two-tailed ) . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown in parentheses above each bar . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 012 In tph-1; Ex[ADF::tph-1 ( + ) ] animals , in which serotonin synthesis activity was restored only in ADF , serotonin was detected in other serotonergic neurons in addition to ADF ( Figure 11B; see ‘Immunohistochemistry’ in ‘Materials and methods’ for details ) . This suggests that serotonin synthesized by ADF might act in either of two possible ways: it could activate SER-7 directly , or it could be taken up and subsequently released by other serotonergic neurons . To distinguish between these possibilities , we compared the feeding rates of tph-1; Ex[ADF::tph-1 ( + ) ] animals with or without mod-5 . MOD-5 is a serotonin transporter required to take up extracellular serotonin into some serotonergic neurons ( Ranganathan et al . , 2001; Jafari et al . , 2011 ) . In mod-5; tph-1; Ex[ADF::tph-1 ( + ) ] , serotonin was detected only in ADF ( Figure 11C ) , suggesting that mod-5 loss blocks serotonin uptake into other cells . If serotonin synthesized by ADF acts only through other serotonergic neurons , mod-5 loss should substantially decrease the feeding rate of tph-1; Ex[ADF::tph-1 ( + ) ] . However , we found that ADF could activate feeding as effectively in the absence of mod-5 as in its presence ( Figure 8A ) , suggesting that serotonin from ADF directly activates SER-7 . 10 . 7554/eLife . 00329 . 013Figure 11 . Recognition of familiar food may increase serotonin release from ADF . ( A ) Schematic of experimental design for anti-serotonin staining . Coding is as in Figure 1A . ( B–C ) serotonin immunoreactivity in tph-1; Is[ptph-1::gfp]; Ex[ADF::tph-1 ( + ) ::gfp] ( B ) and in mod-5; tph-1; Is[ptph-1::gfp]; Ex[ADF::tph-1 ( + ) ::gfp] , the paired control animals defective in serotonin uptake ( C ) . The Is[ptph-1::gfp] allows the identification of NSM and ADF by GFP expression . Filled arrowheads and open arrowheads indicate ADFs and serotonin-uptaking cells , respectively . The serotonin signals not marked by arrowheads are neuronal processes . ( D ) Increase in the average number of serotonin-positive serotonin-uptaking cells during the 1 hr refeeding on familiar or novel food in tph-1; Is[ptph-1::gfp]; Ex[ADF::tph-1 ( + ) ::gfp] and in mod-5; tph-1; Is[ptph-1::gfp]; Ex[ADF::tph-1 ( + ) ::gfp] ( see ‘Immunohistochemistry’ , ‘Quantification of serotonin positive neurons’ and ‘Detailed data analysis’ in ‘Materials and methods’ for details ) . The baseline for each measurement is the average number of serotonin positive AIMs and RIH after starvation . The baseline was 2 . 06 ± 0 . 07 in animals trained on HB101 and 1 . 99 ± 0 . 05 in animals trained on DA1878 . The number of animals examined ( n = 3 independent assays per each group ) is shown under each bar . Data shown as mean ± SEM , Student's t test ( see ‘Detailed data analysis’ in ‘Materials and methods’ ) . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 013 Next , we confirmed that ADF regulates feeding through SER-7 . A ser-7 null mutation suppressed the rescue effect of restoring serotonin in ADF in the tph-1 null mutant ( Figure 8A ) . To understand how serotonin increases feeding at a neural circuit level , we asked where SER-7 acts . SER-7 is expressed mostly in pharyngeal neurons ( Hobson et al . , 2006 ) , which regulate the motions of pharyngeal muscles ( Avery and Horvitz , 1989 ) . Among the pharyngeal neurons , MC is particularly interesting because it is essential for normal fast feeding on bacteria ( Avery and Horvitz , 1989 ) , and SER-7 was suggested to activate MC ( Hobson et al . , 2006; Song and Avery , 2012 ) . To ask if SER-7 acts in MC , we expressed SER-7 in the ser-7 null mutant using the flp-21 and the flp-2 promoters . The flp-21 and the flp-2 promoters drive expression in several neurons , and the expression patterns of the two promoters overlap only in MC and M4 ( Kim and Li , 2004 ) . We found that both pflp-21::SER-7 and pflp-2::SER-7 fully rescued the feeding rate in the ser-7 mutant in response to familiar food as well as serotonin ( Figure 8E , F; Part of the data in Figure 8E , F were reported previously in Song and Avery , 2012 and were re-analyzed and presented here . ) . In contrast , expression of SER-7 in M4 and occasionally in M2 using the ser-7b promoter failed to alter the pumping rate and had only a small effect on pumping in the ser-7 null mutant in response to serotonin and familiar food , respectively ( Figure 8E , F ) , suggesting that SER-7 in MC activates pharyngeal pumping . The failure of rescue is unlikely to be due to insufficient expression because expression of the rescue construct significantly activated isthmus peristalsis , the other feeding motion in C . elegans , which is controlled by M4 ( Song and Avery , 2012 ) . To test whether SER-7 indeed acts through MC , we used an eat-2 null mutation to test if blocking neurotransmission from MC suppresses the rescue effect of pflp-21::SER-7 in the ser-7 mutant . eat-2 encodes a nicotinic acetylcholine receptor subunit specifically localized in the pharyngeal muscles postsynaptic to MC ( McKay et al . , 2004 ) . Thus , an eat-2 null mutation selectively blocks cholinergic transmission from MC to the pharyngeal muscles . We found that the eat-2 null mutation suppressed the rescue effect of pflp-21::ser-7 ( + ) in response to serotonin ( Figure 8E ) , supporting our hypothesis that SER-7 in MC increases the feeding response . In summary , we conclude that serotonin released from extrapharyngeal ADF increased feeding in response to familiar bacteria mainly by activating SER-7 in MC directly , which in turn activates cholinergic transmission from MC to the pharyngeal muscles . We next asked why ADF increases feeding response on familiar but not novel bacteria . A simple explanation is that only familiar bacteria can activate ADF , and this activation causes increased serotonin release . We tested the hypothesis first by asking if ADF is more active on familiar bacteria than novel bacteria by directly measuring the response of the ADF neurons to novel and familiar bacteria using ratiometric calcium imaging . For this , we imaged the ADF neurons using the genetically-encoded calcium sensor Cameleon YC3 . 60 . Changes in intracellular calcium concentration were reported as changes in the ratio of fluorescence emission at distinct wavelengths ( ΔR/R ) ( Nagai et al . , 2004 ) . Slow yet substantial increases in the activity of ADF neurons , reported as an increase in ΔR/R , were observed in response to familiar food ( Figure 12A , B ) . In contrast , only marginal changes in the activity were observed in response to novel food ( Figure 12A , B ) . To further quantify the response observed in ADF , we measured the averaged signed area under ΔR/R for each experimental group ( see ‘Ca2+ imaging’ in ‘Materials and methods’ for details ) . Consistent with our hypothesis , the increases in ADF activity were greater in response to familiar food than novel food ( Figure 12C ) . The activities in response to novel food were not different from baseline ( data not shown ) . These data indicate that familiar food , but not novel food , activates ADF neurons . 10 . 7554/eLife . 00329 . 014Figure 12 . Familiar bacteria , not novel bacteria , activate ADF serotonergic neurons . ( A–B ) . Average calcium transients in response to familiar ( HH and DD groups ) or novel ( DH and HD groups ) food . Traces represent the average percentage change from baseline over time of the fluorescence emission ratio of the ratiometric calcium sensor Cameleon YC3 . 60 . Dashed line at t = 0 represents the time at which the stimulus is delivered . The number of individual recordings is indicated in parenthesis next to each group . ( C ) Average response to familiar or novel food . The bars represent the average signed area below ΔR/R between t = 0 and t = 120 s . Data shown as mean ± SEM , *p < 0 . 05; one-way ANOVA , post hoc Tukey test . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 014 We then asked if ADF releases more serotonin in response to familiar bacteria than novel bacteria . Since direct measurement of serotonin release from ADF in response to food is challenging , we developed a method to detect released serotonin indirectly by its uptake into other serotonergic neurons . As mentioned above , in tph-1; Ex[ADF::tph-1 ( + ) ] animals , serotonin signal is detected in other cells in addition to ADF ( Figure 11B; see ‘Immunohistochemistry’ in ‘Materials and methods’ for details ) . Since no serotonin was detected in the tph-1 null mutant animals ( Figure 13 ) and since the ADFs are the only cells capable of synthesizing serotonin in tph-1; Ex[ADF::tph-1 ( + ) ] , all the serotonin in these animals must have been synthesized in ADF , and its appearance in other serotonergic neurons must have occurred after release from ADF and uptake into the other neurons . In confirmation of this hypothesis , in mod-5; tph-1; Ex[ADF::tph-1 ( + ) ] , serotonin is detected only in ADF ( Figure 11C ) . Thus , the presence of serotonin in cells other than ADF in tph-1; Ex[ADF::tph-1 ( + ) ] animals is an indication of serotonin release from ADF . 10 . 7554/eLife . 00329 . 015Figure 13 . Serotonin immunoreactivity in tph-1 single and mod-5; tph-1 double null mutants . ( A–B ) . No serotonin signal was detected in the tph-1 ( mg280 ) ;Is[ptph-1::gfp] ( A ) and in the mod-5 ( n3314 ) ;tph-1 ( mg280 ) ;Is[ptph-1::gfp] ( B ) mutant animals . Filled arrowheads indicate ADFs . The Is[ptph-1::gfp] allows the identification of NSM and ADF by GFP expression . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 015 We thus tested if ADF releases more serotonin in response to familiar bacteria than novel bacteria by comparing the numbers of serotonin positive serotonin-uptaking cells in tph-1; Ex[ADF::tph-1 ( + ) ] animals on familiar bacteria and novel bacteria . As for the feeding assay , we trained the animals on HB101 or DA1878 and tested them on HB101 or DA1878 after a 7 hr interval ( Figure 11A; see ‘Immunohistochemistry’ in ‘Materials and methods’ for details ) . Consistent with our hypothesis , the increase in the number of serotonin positive serotonin-uptaking cells during the 1 hr incubation on familiar food was greater than the increase on novel food ( Figure 11D; see ‘Quantification of serotonin positive neurons’ and ‘Detailed data analysis’ in ‘Materials and methods’ for details ) . There are several other possible explanations for the greater increase on familiar food than novel food that we cannot exclude , such as increased efficiency of serotonin uptake in serotonin-uptaking cells or increased serotonin release from NSM on familiar food compared to novel food . However , considering that an increase in the efficiency of serotonin uptake is likely to result in a decrease in the level of extracellular serotonin that can activate SER-7 in MC and that restoring serotonin in NSM in the tph-1 null mutant or killing NSM in wild type did not affect the feeding response on familiar food , it is more likely that the greater increase is due to increased serotonin release from ADF . In conclusion , only familiar bacteria activate ADFs , which increases serotonin release from the neurons and subsequently activates the feeding response . Worms may recognize familiar bacteria by taste , smell or texture . To get insight into the mechanism , we tested if worms recognize familiar bacteria by their taste or smell . For this , we examined feeding rates of worms on bacteria mixed with LB broth or with medium conditioned by one of the bacteria ( Figure 14A ) . The conditioned media do not contain any bacterial particles , thus , if the media alter the feeding responses to familiar food or novel food , it suggests that gustatory or olfactory cues in the media were sensed by worms and affected discrimination of familiar food from novel food . We found that the media from bacteria that are familiar to the tested worms did not alter the feeding responses to the novel bacteria ( Figure 14B , C ) . In contrast , the media from novel bacteria decreased the feeding rates of worms on familiar food ( HH and DD groups ) to a level comparable to those of worms on novel food ( DH and HD groups ) ( Figure 14B , C ) . These results , together with the fact that the conditioned media did not affect feeding rates on familiar food or novel food of the same bacterial type as those that conditioned the media ( Figure 14B , C ) , indicate that taste and/or smell of novel bacteria overrides the stimulatory effect of familiar bacteria and suppresses feeding activation . In conclusion , worms sense taste and/or smell of novel bacteria , which negatively regulates the feeding response . 10 . 7554/eLife . 00329 . 016Figure 14 . Taste and/or smell of novel bacteria override the stimulatory effect of familiar bacteria on feeding . ( A ) Experimental design for the feeding assay . Coding is as in Figure 1A . Each condition is coded by three symbols . The first and the second letter above the bar represent training and test food in order . The third symbol , below the bar , represents the conditioned media that was mixed with the test food . LB , Hsup and Dsup represent LB broth and conditioned media filtered from cultures of HB101 and DA1878 , respectively . ( B–C ) Conditioned media from novel bacteria override the stimulatory effect of familiar bacteria on feeding . The average values of the feeding rates presented in ( B ) are 218 . 5 ± 5 . 0 , 197 . 5 ± 7 . 3 , 214 . 6 ± 3 . 9 , 197 . 9 ± 5 . 4 , 195 . 8 ± 4 . 7 , 201 . 6 ± 5 . 2 in order . The average values of the feeding rates presented in ( C ) are 199 . 4 ± 5 . 6 , 169 . 2 ± 6 . 3 , 197 . 2 ± 4 . 7 , 175 . 8 ± 3 . 8 , 174 . 0 ± 5 . 1 and 172 . 8 ± 6 . 7 in order . Data shown as mean ± SEM , n . s . , not significant ( p≥0 . 05 ) , *p<0 . 05 , **p<0 . 01; one-way ANOVA , post hoc Tukey test . The number of animals tested ( n ≥ 3 independent assays per each group ) is shown on each bar . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 016
Given that NSM is a prominent reservoir of serotonin in the pharynx , and that NSM is implicated in regulating the enhanced slowing response on food ( Sawin et al . , 2000 ) , it is surprising that serotonin in NSM does not affect feeding in presence of familiar food . One plausible explanation is that NSM releases little or no serotonin , which is insufficient to activate MC neurons in the pharynx in response to familiar food . This explanation does not contradict previously reported NSM function in the enhanced slowing response ( Sawin et al . , 2000 ) because serotonin from NSM has only a small effect on the behavior ( Sawin et al . , 2000; Zhang et al . , 2005 ) . An endocrine serotonin signal from ADF , not a local serotonin signal from NSM , may have been employed for the feeding activation on familiar food to systemically control multiple behaviors and physiological adaptations . To test this possibility , it would be informative to study whether familiarity of food affects behaviors ( Horvitz et al . , 1982; Avery and Horvitz , 1990; Sawin et al . , 2000; Sze et al . , 2000 ) and various aspects of physiology ( Liang et al . , 2006; Petrascheck et al . , 2007; Srinivasan et al . , 2008 ) that are controlled by serotonin in presence of food ( e . g . , the systemic suppression of stress response that requires serotonin from ADF; Ranganathan et al . , 2001 ) . Further studies will be helpful to understand how recognition of familiar food contributes to survival in C . elegans . How then are ADF and the downstream serotonin feeding signal controlled to increase feeding on particular bacteria after experience ? Our results that conditioned media from novel bacteria override the stimulatory effect of familiar bacteria and suppress the feeding response ( Figure 14 ) and that ADF is active only on familiar bacteria ( Figure 12 ) indicate that perception of the smell and/or the taste of novel bacteria suppresses feeding activation on novel food by inhibiting the activity of ADF or its upstream neurons ( Figure 15 ) . Given that familiar food substantially increased ADF activity compared to the baseline ( Figure 12A , B ) , at least two neural pathways should act antagonistically in controlling the activity of ADF or its upstream neurons . The simplest model would be that ADF or its upstream neurons are positively regulated by perception of food and negatively regulated by perception of olfactory and gustatory cues of novel bacteria ( Figure 15 ) . Our study predicts that novel bacteria dictate activation of ADF and the subsequent serotonin-dependent feeding response in the natural habitat where more than one bacterial type are likely to grow mixed together . Preliminary data show that C . elegans remembers two bacterial types at least for 24 hr ( data not shown ) , supporting the possibility that C . elegans accumulates past food experience and uses them for feeding regulation in its natural habitat . The following observations suggest that ADF releases serotonin and increases the feeding response when worms encounter familiar bacteria: ( 1 ) Serotonin from ADF increases the feeding response ( Figure 8A ) ; ( 2 ) ADF is activated selectively by familiar bacteria within 1 min ( Figure 12A , B ) ; ( 3 ) ADF releases more serotonin in response to familiar bacteria than novel bacteria ( Figure 11D ) . Interestingly , serotonin transmission from ADF was also shown to be critical for the learned aversion to pathogenic bacteria ( Zhang et al . , 2005; Ha et al . , 2010 ) , which is opposite in direction to the appetitive change in feeding behavior that we describe here . For the aversive learning , it is not yet clear when the serotonin signal from ADF acts . It will be interesting to understand how serotonin signaling from ADF and the physiological context are integrated to produce seemingly opposite experience-dependent behaviors . Further studies to understand regulation of the two seemingly opposite behaviors at the neural circuit level will also help us understand how the C . elegans nervous system differentially encodes , maintains and retrieves the appetitive and aversive memories of bacteria .
Many questions remain to be answered to fully understand the mechanism underlying recognition of familiar bacteria in C . elegans . How do worms sense different bacteria ? What changes in the nervous system underlie the process of becoming familiar to particular bacteria during experience ? Further quests to explore these unanswered questions may deepen our understanding of sensory information processing and familiarity discrimination .
Except when stated otherwise , C . elegans was cultured at 19°C as described by ( Brenner , 1974 ) . Except in Figure 10 , all worms used were hermaphrodites . The following mutant alleles were used: mod-1 ( ok103 ) V , mod-5 ( n3314 ) I , ser-4 ( ok512 ) III , ser-7 ( tm1325 ) X , tph-1 ( mg280 ) II . In the main text only the gene name is shown . The wild-type strain was N2 ( Brenner , 1974 ) , and the mutant strains used were DA2100: ser-7 ( tm1325 ) X , MT15434: tph-1 ( mg280 ) II , MT9772: mod-5 ( n3314 ) I , DA2289: tph-1 ( mg280 ) II; kyEx947[pceh-2::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , DA2290: tph-1 ( mg280 ) II; kyEx949[psrh-142::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , DA2293: tph-1 ( mg280 ) II; ser-7 ( tm1325 ) X , DA2294: tph-1 ( mg280 ) II; ser-7 ( tm1325 ) X; kyEx949[psrh-142::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , DA2295: mod-5 ( n3314 ) I; tph-1 ( mg280 ) II , DA2296: mod-5 ( n3314 ) I; tph-1 ( mg280 ) II; kyEx949[psrh-142::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , DA2301: ser-7 ( tm1325 ) X; nyIs80[pflp-21::gfp ( + ) ] , DA2297: ser-7 ( tm1325 ) X; adEx2297[pser-7::ser-7 ( + ) pflp-21::gfp] , DA2298: ser-7 ( tm1325 ) X; adIs2298[pflp-21::ser-7 ( + ) pflp-21::gfp ( + ) ] , DA2299: mod-5 ( n3314 ) I; tph-1 ( mg280 ) II; yzIs71[ptph-1::gfp rol-6 ( su1006 ) ]; kyEx949[psrh-142::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , DA2300: tph-1 ( mg280 ) II; yzIs71[ptph-1::gfp rol-6 ( su1006 ) ]; kyEx949[psrh-142::tph-1 ( + ) ::gfp punc-122::gfp ( + ) ] , GR1333: yzIs71 [tph-1::gfp , rol-6 ( su1006 ) ] V , OT180: ser-4 ( ok512 ) ; mod-1 ( ok103 ) V; ser-7 ( tm1325 ) X , DA2445: ser-7 ( tm1325 ) X; adEx2245[pflp-2::ser-7 ( + ) pflp-21::gfp] , XL188: ntIs16[ptph-1::yc3 . 60 lin-15 ( + ) ] . Pseudomonas PA14 pstP and Enterobacteria JU54 were kind gifts from Dr . Fred Ausubel and Dr . Gary Ruvkun , respectively . Developmentally synchronized L1 larvae were cultured until adulthood ( for 54 hr at 19°C ) on training food . For Figure 1B , C , individual animals were transferred to test food for measuring feeding rates after removing bacteria by letting them crawl on unseeded plates for ∼1 min . Removal of bacteria was checked by the absence of traces of bacteria on the track . For Figures 1D , E , 4A–F , 5 , 7 , 9 , 10 , and 11 , individual animals were transferred to test food for measuring feeding rates after being starved for 7–8 hr on unseeded NGM plate at room temperature ( RT , 23°C ) after the training . We starved the animals for the following reasons: First , the 7–8 hr of starvation synchronized the nutritional status of worms . Second , the 7- to 8-hr gap between the test and the last exposure to the training bacteria allows us to test if C . elegans can actually remember the familiar bacteria . For Figure 3 , the animals that were cultured on HB101 or DA1878 for 54 hr at 19°C were incubated on training food or control food for either 6 or 9 hr at 19°C . Then , individual animals were transferred to test food for measuring feeding rates after 7–8 hr of starvation on unseeded NGM plates at room temperature ( RT ) . For Figure 8A , D , F , the animals that were cultured on HB101 for 54 hr at 19°C ( Figure 8A , B , F ) or until adulthood ( Figure 8C , D ) , were transferred to HB101 for measuring feeding rates after 7–8 hr of starvation on unseeded NGM plates at room temperature ( RT ) . Feeding rates of individual worms were quantified by counting pharyngeal contractions 2–5 min after the transfer to test food at room temperature ( about 23°C ) . Feeding motions of individual animals were observed with a Zeiss Stemi SV11 Apo microscope . The feeding rate of each animal ( pumps per min ) was calculated by averaging the three measures from each animal ( pumps per 30 s ) and subsequently by multiplying by 2 . For each experiment , preparation of worms and reagents and feeding assays were performed in the same way in a designated place , mostly using the same batch of reagents . In contrast , conditions for feeding assays for different experiments varied in several ways , for instance , temperature and amount of test bacteria . As a result , we got consistent range of pumping rates within experiments , but not among different experiments . For all figures except Figure 4A , B , 5 , and 8 , test food were prepared by seeding 10 μl of bacterial culture in LB ( OD = 5 . 0 ) on new NGM plates and incubating the seeded plates at RT for a defined amount of time ( 5 . 5 hr for HB101 , JU54 and PA14 pstP and 7 hr for DA1878 ) . For Figure 4A , B , 5 , and 8A–D , test food was prepared in the same way except that 100 μl of bacterial culture was seeded . The variation in food preparation was necessary because tph-1 mutants would not stay on food that was prepared from 10 μl of the culture . For conditioned media in Figure 14 , bacterial suspensions of OD = 5 . 0 were prepared by collecting bacterial pellets from overnight cultures and resuspending the pellets in LB after rinsing once . After incubating the bacterial suspensions at 37°C for 12 hr while shaking , supernatants were obtained by filtering each bacterial using a 0 . 2-μm microfilter ( Nalgene , 190-2520 ) . 200 μl of the supernatant from each bacterial culture was added to each 4 . 5 ml NGM in a 35-mm plate 6 . 5 hr prior to the feeding assay and air-dried for 1 hr . Test food was then prepared by seeding 10 μl of bacterial culture in LB ( OD = 5 . 0 ) on each plate as for other feeding assays . The ser-7b promoter ( 2 . 2kb ) for expression in M4 and M2 ( Hobson et al . , 2003 ) was cloned by PCR ( pser-7 F: 5′- CAAACAGGTAGACAATGTTGTAAACTGTGA -3′ and pser-7 R: 5′- TTCACCCCTCAGGCTGTG -3′] from N2 genomic DNA . 1 . 3 kb ser-7 cDNA ( Hobson et al . , 2003 ) was cloned by PCR ( SER-7 cDNA F primer , 5′-CCCGGGATGGCCCGTGCAGTC-3′ and SER-7 cDNA R primer 5′-CCCGGGCTAGACGTCACTTGGTTCGT-3′] from a cDNA pool that was reverse transcribed from N2 mRNA extracts . The flp-21 promoter and the flp-2 promoter were kind gifts from Dr . Chris Li ( Kim and Li , 2004 ) . The ser-7b promoter and the flp-21 promoter were cloned into HindIII-BamHI digested pPD96 . 52 ( Addgene plasmid 1608 ) . ser-7 cDNA was cloned into the EcoRI site of the two vectors containing each of the promoters . The pflp-2::ser-7 ( + ) rescue construct was generated by the PCR-fusion method ( Hobert , 2002 ) using the following primers: pflp-2 A primer , 5′- TCTGTGTTCACTCTACCAGGAACTTTTCTCACTTTTTAATACATATTTTCATGAAC -3′ , pflp-2 A′ primer , 5′- TCTGTGTTCACTCTACCAGGA -3′ , pflp-2 B primer , 5′-GAGATATGTTGACTGCACGGGCCATGGTTTGCGACAATTGGTTTGGCAACG -3′ , SER-7 cDNA C primer , 5′- ATGGCCCGTGCAGTC-3′ , pPD9575 3′UTR D primer , 5′- GGAAACAGT TATGTTTGGTATATTGGG -3′ . To generate DA2297 , DA2298 and ser-7; Ex[pflp-2::ser-7 ( + ) pflp-21::gfp] , germline transformation was performed in ser-7 ( tm1325 ) with pser-7b::ser-7 ( + ) ( 100 ng/μl ) , pflp-21::ser-7 ( + ) ( 50 ng/μl ) or pflp-2::ser-7 ( + ) ( 75 ng/μl ) , along with pflp-21::gfp ( 50 ng/μl ) as an injection marker . For DA2298 , the extrachromosomal array was integrated into the chromosome by gamma irradiation ( 6 krad ) . The integration line was outcrossed five times against DA2100 . The ceh-2 promoter and the srh-142 promoter were kind gifts from Dr . Cori Bargmann . The ceh-2 promoter spans 1 . 5 kb on chromosome I from AAGCTTAAATCTTATCAGAC to TTCTAATATTCGGAGTGAAA and the srh-142 promoter spans 4 kb on chromosome V from TAGATTCATGTACTTGGCTC to TTTTTGCCAATATGAGTTGT . Laser ablation of ADFs and NSMs was performed by a modified procedure ( Avery and Horvitz , 1987 ) . We destroyed both ADFs or NSMs in newly hatched GR1333 larvae ( 0–4 hr old ) using a MicroPoint laser ablation system ( Andor Technology USA , South Windsor , CT ) . For laser ablation , we mounted the larvae on 2% agarose pads containing 10 mM sodium azide . To minimize the variation caused by sodium azide , we retrieved all the mock-operated and the putative ADF-ablated animals from the agarose pad after the same incubation time ( 12 min ) . Mock-operated groups were treated in the same way except that the laser was not fired . The retrieved animals were cultured on HB101 until adulthood . After 7 hr of starvation , feeding assays were performed as described in ‘Materials and methods’ . Successful ablation of ADFs was confirmed by absence of the ADF or NSM GFP signals . Only data from animals in which both ADFs or NSMs were specifically destroyed are included . One-day-old adults were placed in a T-shaped microfluidic chamber ( Figure 16 ) , with the tip of their nose exposed to constantly flowing LB broth . After 30–50 s , the solution was switched to LB broth containing either DA1878 or HB101 bacteria ( OD = 10 . 0 ) . ADF neurons were visualized through a Zeiss plan-apochromat 63X , 1 . 4 NA oil immersion objective . Excitation light ( 436/10 nm ) was provided by an X-cite 120 illuminator ( Lumen Dynamics Group Inc , Ontario , Canada ) . For ratiometric imaging , images in cyan ( 480/15 nm ) and yellow ( 535/20 nm ) wavelength bands were simultaneously acquired by the calcium imaging camera ( Hamamatsu; ORCA-AG , Bridgewater , NJ ) by means of a beam splitter ( Optical Insights; OI-DV-FC , Tucson , AZ ) . The camera was controlled by Micromanager ( Edelstein et al . , 2010 ) . Frames were acquired at 2–10 Hz with a 8 × 8 binning . The resulting images were analyzed off-line using custom analysis routines written in Igor Pro ( Wavemetrics , Lake Oswego , OR ) . Briefly , fluorescence intensity I in the 480 and 535 nm wavelength images was measured in a circular region of interest ( ROI ) centered on the neuron . Background fluorescence I' was measured in a second ROI surrounding the first one . The raw emission ratio was computed as R = ( I535 − I'535 ) / ( I480 − I'480 ) − 0 . 65 , where the latter term corrects for 480-nm channel bleed-through into the 535-nm channel . This raw emission ratio was corrected for photobleaching and normalized by fitting a single exponential function to the emission ratio trace and dividing the latter by the fitted function; thus all ratio changes were expressed in terms of changes in fluorescence , ΔR/R . 10 . 7554/eLife . 00329 . 018Figure 16 . The schematic of the experimental setup for calcium imaging . The microfluidic chamber consisted of a T-shaped channel formed in PDMS bonded to a glass coverslip . The main branch of the channel was connected to an inlet and outlet , allowing LB or bacterial solutions to flow through . Switching from LB to the bacterial solutions was achieved via an upstream valve . A smaller branch orthogonal to the first one contained the worm . This tapered channel allowed immobilization of the animal while exposing only the tip of the nose to the flowing LB or bacterial solutions . DOI: http://dx . doi . org/10 . 7554/eLife . 00329 . 018 To establish the time course of the ADF response to bacteria , the change in emission ratio ΔR/R was averaged in 10 s bins for each animal , and the data then averaged across animals . The data were further quantified by measuring the signed area under ΔR/R for each animal during the first 120 s of the stimulus , and averaging across animals . Samples of DA2299 and DA2300 were prepared as for the feeding assay . Just after the 7-hr starvation we divided each group into three equal subgroups and fixed one . The remaining two groups were separately refed on either DA1878 or HB101 and fixed after 1 hr . This assay could not be done immediately after training because most of the serotonin-uptaking cells were serotonin positive in DA2300 as in wild type worms . The background was too high to detect any increase in serotonin release . Immunohistochemistry was performed using a protocol from Curtis Loer ( http://home . sandiego . edu/∼cloer/loerlab/anti5htlong . html ) with the following antibodies: anti-serotonin rabbit IGG: S5545 ( Sigma-Aldrich , Inc ) , Anti-GFP chicken IGG: GFP-1020 ( Aves Labs , Inc ) , ALEXA FLOUR 488 goat anti-chicken IGG: A11039 ( Invitrogen Corporation ) , Cy-3 conjugated donkey anti-rabbit IGG: 711-165-152 ( Jackson ImmunoResearch ) . Failure of the co-immunostaining against serotonin and against GFP was 0% . The GFP signal was used to identify serotonergic neurons ( ptph-1::gfp ) and to find transgenics that express tph-1 cDNA in ADF ( psrh-142::tph-1::gfp ) . Due to Is[ptph-1::gfp] in DA2299 and DA2300 , GFP signal was found in NSM , ADF , HSN and sporadically in RIH and AIM in all animals . Although ptph-1::gfp expresses GFP in ADF , we could still recognize the transgenic animals that express Ex[psrh-142::tph-1::gfp] because GFP signal in ADF is much stronger in those transgenics . The strong GFP signal in ADF was perfectly correlated with serotonin signal . In DA2300 , serotonin signal was found in four different classes of serotonergic neurons ( ADF , NSM , RIH and AIM ) . In DA2299 , the control strain that is defective in serotonin uptake , serotonin signal was found only in ADF . Images were obtained with a Zeiss LSM510-meta confocal microscope using a 40× oil-immersion objective . To calculate the increase in the average number of serotonin-positive serotonin-uptaking cells during 1 hr of refeeding for each group ( Figure 11D ) , we first blindly counted the number of serotonin positive AIMs and RIH from each animal and calculated the average number for each group . ADF , NSM and HSN were not included because serotonin was detected in all ADFs and NSMs even before refeeding and in none of HSNs even after refeeding . To minimize variation , only the animals expressing TPH-1 in both ADFs ( as indicated by the presence of the strong GFP signal from psrh-142::tph-1::gfp ) were considered for counting the number of serotonin positive AIMs and RIH . Then , we subtracted the baselines from each familiar and novel food group as follows: ( Serotonin positive cell ) HD/HH = ( Average number of serotonin positive AIMs and RIH ) HD/HH − ( Average number of serotonin positive AIMs and RIH ) H . ( Serotonin positive cells ) DH/DD = ( Average number of serotonin positive AIMs and RIH ) DH/DD − ( Average number of serotonin positive AIMs and RIH ) D . For data presentation , we combined the values from three independent experiments ( see ‘Detailed data analysis’ in ‘Materials and methods’ for details ) . To examine serotonin effects on feeding rate in absence of bacteria , feeding rates were quantified from 3- to5-hr-old L1 larvae that had never been exposed to bacteria . The feeding assay was performed with L1 larvae because it was easier to examine feeding responses of developmentally synchronized worms that are free from bacteria in large numbers . The strategy is particularly useful for the developmentally retarded mutants that carry eat-2 ( ad465 ) , which makes comparisons of the feeding rates in adults rather difficult . We confirmed that the effects of single null mutations of the serotonin receptors in response to serotonin are consistent in L1 and adults ( data not shown ) and thus , it is likely that our observations made in L1 larvae are still valid in adults . After collecting embryos by egg preparation , we incubated them on unseeded NGM plates for 2 hr . Newly hatched L1 larvae ( 0- to 2-hr-old ) were then transferred to unseeded NGM plates and incubated for 3 hr . 15 min after mounting the larvae ( 3- to 5-hr-old ) on 2% agarose pads containing 20 mM serotonin ( H7752 , Sigma-Aldrich , Inc ) in M9 , the feeding motions of each larva were observed using a Zeiss Axiophot microscope with a 63× objective . 2-min videos were taken from each larva with a Hitachi kP-160 CCD camera and digitized using Adobe Premiere v6 . 5 for quantification of feeding rates . Each experiment continued for 1 hr . Feeding rates shown in Figures 4G and 8E were calculated by averaging two measurements per animal ( pumps per 55 s ) . Except the data that are analysed by one-way ANOVA or by Student t test , data were statistically analysed by both the unpaired t-test and the Mann–Whitney U test ( two-tailed ) . The two tests produced the same conclusions for all data analyses . For data presentation , the more conservative p value was selected . Only the familiar food group and the novel food group that were tested on the same test food were compared since the feeding rates on HB101 were significantly higher than the rate on DA1878 ( in non-starved wild-type worms [p=0 . 006] and in 7–8 hr starved wild-type worms [p=0 . 003] ) , on JU54 ( p=0 . 003 ) and since the feeding rate on PA14 pstP was significantly higher than the rate on HB101 ( p=0 . 001 ) . The ser-7 effect on the feeding rate ( shown in Figure 4H ) for each food condition was calculated by subtracting the averaged feeding rate of the ser-7 mutant from the rate of wild-type worms that were tested under each food condition . Student's t-test ( two-tailed ) was used to compare the ser-7 effects on feeding rate between HH and DH groups , and between DD and HD groups ( see ‘Detailed data analysis’ in ‘Materials and methods’ for details ) . GraphPad Prism ( version 5 . 0 ) was used for statistical analysis . The effects of inhibitory serotonin signal on feeding rates in HH and DD groups ( shown in Figure 7F ) were calculated and compared as for the ser-7 effect using the feeding rates of the ser-7 single mutant and the ser-4; mod-1; ser-7 triple mutant . To compare the increase in the numbers of serotonin positive cells during 1 hr of refeeding on familiar food with the increase on novel food , we tested the data shown in Figure 11D using Fisher's method ( Fisher , 1954 ) for combining the results of several independent tests bearing upon the same overall hypothesis . We first compared the difference in the increase in the number of serotonin positive cells between HH and DH groups , and between DD and HD groups in each experiment using Student's t test ( two-tailed ) . p values from three independent experiments were then combined using Fisher's method and tested by χ2 test ( see ‘Detailed data analysis’ in ‘Materials and methods’ for details ) . For clarity , some results are presented in more than one panel . The feeding rates of wild type and the ser-7 null mutant in presence of serotonin are shown in Figures 4G and 6 . The feeding rates of the ser-7 null mutant on familiar food and novel food are presented in Figures 4E , F and 7A–D . The feeding rates of the ser-4; mod-1; ser-7 triple mutant on familiar food and novel food are presented in Figures 4E , F and 7C , D . In Figure 4H , comparison of the differences in the feeding rates between wild-type ( N2 ) and the ser-7 null mutant animals on familiar food with the differences on novel food using Student t-test . STEP 1 . Calculation of the parameters ( means and standard error of the means of the differences in the feeding rates between N2 and ser-7 under HH , DH , DD and HD ) for the comparisons using Student t-testMean ( N2HH–ser-7HH ) = Mean ( N2HH ) – Mean ( ser-7HH ) Standard error of the mean ( N2HH–ser-7HH ) = [Var ( N2HH ) /n ( N2HH ) + Var ( ser-7HH ) /n ( ser-7HH ) ]1/2Mean ( N2DH–ser-7DH ) = Mean ( N2DH ) – Mean ( ser-7DH ) Standard error of the mean ( N2DH–ser-7DH ) = [Var ( N2DH ) /n ( N2DH ) + Var ( ser-7DH ) /n ( ser-7DH ) ]1/2Mean ( N2DD–ser-7DD ) = Mean ( N2DD ) – Mean ( ser-7DD ) Standard error of the mean ( N2DD–ser-7DD ) = [Var ( N2DD ) /n ( N2DD ) + Var ( ser-7DD ) /n ( ser-7DD ) ]1/2Mean ( N2HD–ser-7HD ) = Mean ( N2HD ) – Mean ( ser-7HD ) Standard error of the mean ( N2HD–ser-7HD ) = [Var ( N2HD ) /n ( N2HD ) + Var ( ser-7HD ) /n ( ser-7HD ) ]1/2 Mean ( XY ) is the mean of the feeding rate of animals of genotype X under Y condition; Var ( XY ) is the variance of the feeding rate of animals of genotype X under Y condition; n ( XY ) is number of animals of genotype X that were tested under Y condition . STEP 2 . Comparison of the differences in the feeding rates between wild-type ( N2 ) and the ser-7 null mutant animals on familiar food with the differences on novel food using Student t-test A . Comparison between HH and DHt = [{Mean ( N2HH ) – Mean ( ser-7HH ) } – {Mean ( N2DH ) – Mean ( ser-7DH ) }][Var ( N2HH ) /n ( N2HH ) + Var ( ser-7HH ) /n ( ser-7HH ) + Var ( N2DH ) /n ( N2DH ) + Var ( ser-7DH ) /n ( ser-7DH ) ]1/2 p<0 . 001 B . Comparison between DD and HDt = [{Mean ( N2DD ) – Mean ( ser-7DD ) } – {Mean ( N2HD ) – Mean ( ser-7HD ) }][Var ( N2DD ) /n ( N2DD ) + Var ( ser-7DD ) /n ( ser-7DD ) + Var ( N2HD ) /n ( N2HD ) + Var ( ser-7HD ) /n ( ser-7HD ) ]1/2 p<0 . 001 These statistical analyses concluded that the difference in the feeding rates between wild-type ( N2 ) and ser-7 is greater on familiar food than the difference on novel food , suggesting that serotonin signaling via SER-7 is more active on familiar food than novel food . In Figure 7F , comparison of the differences in the feeding rates between the ser-4; mod-1; ser-7 ( OT180 ) and the ser-7 null mutant animals on familiar food using Student t-test was done in the same way . t = [{Mean ( OT180HH ) – Mean ( ser-7HH ) } – {Mean ( OT180DD ) – Mean ( ser-7DD ) }] [Var ( OT180HH ) /n ( OT180HH ) + Var ( ser-7HH ) /n ( ser-7HH ) + Var ( OT180DD ) /n ( OT180DD ) + Var ( ser-7DD ) /n ( ser-7DD ) ]1/2 p<0 . 05 In Figure 11D , comparison of the numbers of serotonin positive serotonin-uptaking cells in tph-1; Ex[ADF::tph-1 ( + ) ] animals on familiar bacteria and novel bacteria . STEP 1 . Comparison of the increase in the average number of serotonin positive serotonin-uptaking cells in the tph-1;Is[ptph-1::gfp];Ex[ADF::tph-1 ( + ) ::gfp] animals that were refed on familiar food with the increase in the animals that were refed on novel food using Student t-test A . Calculation of the parameters ( means and standard error of the means for HH-H , DH-D , DD-D and HD-H ) for the comparisons using Student t-test ( The subtraction was for isolating the increase in the number during the 1 hr of refeeding . ) Mean ( HH-H ) = Mean ( HH ) - Mean ( H ) Standard error of the mean ( HH-H ) = [Var ( HH ) /nHH + Var ( H ) /nH]1/2Mean ( DH-D ) = Mean ( DH ) - Mean ( D ) Standard error of the mean ( DH-D ) = [Var ( DH ) /nDH + Var ( D ) /nD]1/2Mean ( DD-D ) = Mean ( DD ) - Mean ( D ) Standard error of the mean ( DD-D ) = [Var ( DD ) /nDD + Var ( D ) /nD]1/2Mean ( HD-H ) = Mean ( HD ) - Mean ( H ) Standard error of the mean ( HD-H ) = [Var ( HD ) /nHD + Var ( H ) /nH]1/2 Mean ( X ) is mean of number of serotonin-positive serotonin-uptaking cells in group X; Var ( X ) is variance of number of serotonin-positive serotonin-uptaking cells in group X; nX is sample number of group X . B . Comparison between HH-H and DH-D in each experimentt = [{Mean ( HH ) – Mean ( H ) } – {Mean ( DH ) – Mean ( D ) }][Var ( HH ) /nHH + Var ( H ) /nH + Var ( DH ) /nDH + Var ( D ) /nD]1/2 Experiment 1: p=0 . 239 Experiment 2: p=0 . 380 Experiment 3: p=0 . 007 C . Comparison between DD-D and HD-H in each experimentt = [{Mean ( DD ) – Mean ( D ) } – {Mean ( HD ) – Mean ( H ) }][Var ( DD ) /nDD + Var ( D ) /nD + Var ( HD ) /nHD + Var ( H ) /nH]1/2 Experiment 1: p=0 . 155 Experiment 2: p=0 . 005 Experiment 3: p=0 . 540 STEP 2 . Comparison of the increases in number of serotonin-positive serotonin-uptaking neurons ( HH-H vs DH-D and DD-D vs HD-H ) using χ2test after combining the data from three independent experiments using Fisher's method . χ2=−2∑i=1kloge ( pi ) , where pi is the p value for the ith hypothesis test . When the p-values tend to be small , the test statistic χ2 will be large , which suggests that the null hypotheses are not true for every test . When all the null hypotheses are true , and the pi ( or their corresponding test statistics ) are independent , χ2 has a distribution with 2k degrees of freedom , where k is the number of tests being combined . This fact can be used to determine the p value for χ2 . A . Comparison between HH and DH: χ2=14 . 63 ( degree of freedom=6 ) p=0 . 023 B . Comparison between DD and HD: χ2=15 . 40 ( degree of freedom=6 ) p=0 . 017 These statistical analyses conclude that the increases in the average number of serotonin positive serotonin-uptaking neurons of HH and DD groups are greater than the increases of DH and HD groups , respectively . | Many species , including our own , show a preference for familiar foods over novel ones . This behavior probably evolved to reduce the risk of consuming items that turn out to be poisonous , but the mechanisms that underlie a preference for familiar foods are largely unknown . The nematode worm , C . elegans , is a useful organism in which to study such processes . Having only around 1000 cells and a simple anatomy , C . elegans is an attractive model system for studying molecular biology , and was the first multicellular organism to have its genome fully sequenced . C . elegans feeds on bacteria , which it detects using a pair of sensory cells called ADF neurons , which sense extrinsic cues . When the ADF neurons detect bacteria , they release the transmitter serotonin . Serotonin stimulates the worm to consume the bacteria by pumping them into the pharynx , its feeding organ , and then transporting them to its intestine after crushing them . Now , Song et al . have demonstrated that C . elegans consumes familiar bacteria more rapidly than it does novel ones , and have identified the molecular mechanism behind this behavior . They found that familiar bacteria stimulated the release of serotonin from the ADF cells: this activated SER-7 receptors on a specific type of motor neuron in the pharynx and this , in turn , triggered the worms' feeding response . Novel bacteria , on the other hand , failed to either activate ADF or to trigger feeding . Moreover , when Song et al . offered the worms familiar bacteria in medium that had previously contained novel bacteria , the residual traces of the novel bacteria stopped the worms from responding to familiar food . Further research is needed to determine whether the mechanisms that underpin the more active consumption of familiar food by C . elegans can also explain the preference for familiar foods shown by other species . A better understanding of the mechanisms by which different foods elicit feeding could also offer important insights into factors that contribute to obesity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Conclusion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2013 | Recognition of familiar food activates feeding via an endocrine serotonin signal in Caenorhabditis elegans |
Sepsis is a systemic inflammatory response to infection , accounting for the most common cause of death in intensive care units . Here , we report that peripheral administration of the hypothalamic neuropeptide orexin improves the survival of mice with lipopolysaccharide ( LPS ) induced endotoxin shock , a well-studied septic shock model . The effect is accompanied by a suppression of excessive cytokine production and an increase of catecholamines and corticosterone . We found that peripherally administered orexin penetrates the blood-brain barrier under endotoxin shock , and that central administration of orexin also suppresses the cytokine production and improves the survival , indicating orexin’s direct action in the central nervous system ( CNS ) . Orexin helps restore body temperature and potentiates cardiovascular function in LPS-injected mice . Pleiotropic modulation of inflammatory response by orexin through the CNS may constitute a novel therapeutic approach for septic shock .
Systemic inflammatory response syndrome induced by infection , or sepsis ( Bone et al . , 1992; Dellinger et al . , 2013 ) , can lead to a life-threatening medical emergency requiring intensive care ( Martin et al . , 2003 ) . Septic shock is defined as cardiovascular dysfunction triggered by sepsis , representing the most severe stage of the illness ( Angus and van der Poll , 2013 ) . Lipopolysaccharide ( LPS ) , a major cell wall component of Gram negative bacteria , plays a central role in sepsis as the endotoxin inducing a systemic inflammatory response , and LPS-induced endotoxin shock is one of the several well-studied animal models of septic shock . Recent advances have started to reveal the highly complex pathophysiology of sepsis ( Cohen , 2002 ) . However , researchers have failed to translate the advances in understanding the pathophysiology of sepsis into effective new therapies , and the mortality of sepsis still remains high . To improve the outcome of patients with sepsis , new therapeutic strategies and agents are essential . Recent studies have revealed the regulation and integration of inflammatory responses by the central nervous system ( CNS ) through the neuroendocrine and autonomic nervous systems ( Tracey , 2002 ) . For example , vagal afferents activated by endotoxin and cytokines in sepsis stimulate the hypothalamic-pituitary-adrenal axis and exert anti-inflammatory effects through the release of glucocorticoids ( Tracey , 2002 ) . A cholinergic anti-inflammatory pathway has also been reported , in which the activation of efferent vagus nerves suppresses systemic inflammatory responses ( Borovikova et al . , 2000; Wang et al . , 2004 ) . Acetylcholine attenuates cytokine production from LPS-activated macrophages in the spleen through the nicotinic acetylcholine receptor ( Wang et al . , 2003 ) . Vagus nerve stimulation also leads to the activation of the splenic nerve and release of norepinephrine in the spleen ( Vida et al . , 2011 ) . Norepinephrine inhibits cytokine production in the spleen and suppresses systemic inflammation in experimental sepsis through β2-adnenoceptors on lymphocytes ( Vida et al . , 2011 ) . Thus , the anti-inflammatory effects of the sympathetic and parasympathetic nervous systems seem to be synergistic . The hypothalamic neuropeptide orexin , which plays a crucial role in controlling sleep/wakefulness ( Sakurai , 2007 ) , regulates the hypothalamo-pituitary-adrenal axis by activating the paraventricular nucleus ( PVN ) ( Kuru et al . , 2000 ) and integrates autonomic functions by interacting with brainstem centers ( Zheng et al . , 2005 ) . Recent reports show that intracerebroventricular ( ICV ) administration of orexin modulates heart rate and body temperature , and increases the level of adrenocorticotropic hormone ( ACTH ) in a murine sepsis model induced by cecal ligation and puncture ( Deutschman et al . , 2013 ) . ICV orexin also partially increases locomotor activity suppressed by a low dose of LPS in rats ( Grossberg et al . , 2011 ) . However , no information is available as to whether orexin actually improves the survival and/or suppresses the systemic inflammation . Moreover , CNS administration of therapeutic agents in human patients may often be unfeasible; clinical applications of orexin in humans require a tactic to deliver orexin into the brain . Systemic inflammation , the hallmark of sepsis , enhances the permeability of the blood-brain barrier ( BBB ) in rodents ( Kowal et al . , 2004; Xaio et al . , 2001 ) and humans ( Ballabh et al . , 2004 ) ; some peptides and proteins , including insulin , albumin ( Xaio et al . , 2001 ) , and antibodies ( Kowal et al . , 2004 ) , can enter the brain under the condition of systemic inflammation . In this study , we deliver orexin into the brain by taking advantage of the enhanced BBB permeability under the condition of systemic inflammation , rescuing mice with endotoxin shock by targeting the CNS .
Orexin cannot normally penetrate the BBB ( Fujiki et al . , 2003 ) . Surprisingly , however , we found that the survival of mice with endotoxin shock was dramatically improved ( Figure 1A ) by subcutaneous ( SC ) infusion of orexin-A ( OXA; 1 mg/mouse/24 hr ) starting 30 min before a lethal dose of LPS injection ( 10 mg/kg IP , which kills 70–90% of mice ) . No effect was seen in Hcrtr1-/-;Hcrtr2-/- mice ( abbreviated as OXRKO mice ) ( Figure 1B ) . Furthermore , by SC administration of OXA ( 2 mg/mouse/24 hr ) starting 30 min after LPS injection , the survival rate and duration were also significantly improved ( Figure 1C ) . Although orexin was administered for the first 24 hr only , its life-supporting effect manifested over the next several days . 10 . 7554/eLife . 21055 . 003Figure 1 . Effects of peripherally administered orexin-A ( OXA ) on survival in mice with endotoxin shock . ( A ) Kaplan-Meier survival curves of wild-type mice subcutaneously ( SC ) administered with saline or orexin-A ( OXA; 1 mg/mouse/24 hr ) 30 min before lipopolysaccharide ( LPS; 10 mg/kg ) injection ( each group n = 10 ) . ( B ) Kaplan-Meier survival curves of Hcrtr1-/-;Hcrtr2-/- ( OXRKO ) mice SC-administered with saline or OXA ( 1 mg/mouse/24 hr ) 30 min before LPS injection ( each group n = 7 ) . ( C ) Kaplan-Meier survival curves of wild-type mice SC-administered with saline or OXA ( 2 mg/mouse/24 hr ) 30 min after LPS injection ( each group n = 10 ) . ( D , E ) Effect of OXA treatment on the levels of catecholamines ( D ) in the serum and corticosterone ( E ) in the plasma from LPS-injected mice , compared to saline treatment ( each group n = 3–5 , *p<0 . 05 ) . Data are presented as mean±s . e . m . Statistical significance assessed by Mantel Cox log-rank test ( A–C ) and unpaired t-test ( D , E ) . Data are replicated in at least three independent experiments . n . s: not significant . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 00310 . 7554/eLife . 21055 . 004Figure 1—figure supplement 1 . Serum catecholamines ( A ) and plasma corticosterone ( B ) levels at 22 hr after LPS injection . Norepinephrine level in the serum decreased and corticosterone level in the plasma increased at 22 hr after LPS injection ( 22H-LPS ) , compared to saline injection ( 22H-Sal ) ( each group n = 3–5 , *p<0 . 05 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 004 LPS induces excessive cytokine production systemically , leading to further amplification of the inflammatory response ( Beutler and Rietschel , 2003 ) . We measured the levels of 32 cytokines in serum from LPS-injected mice . Whereas LPS elevated the levels of most cytokines , SC infusion of OXA ( 1 mg/mouse/24 hr ) significantly inhibited the increase of a number of cytokines at 4 hr and 22 hr after LPS injection ( Figure 2 ) . Simultaneously , OXA ameliorated hypothermia induced by LPS at 4 hr and 22 hr after LPS injection ( Figure 2—figure supplement 1 ) . Particularly , in the early stage of endotoxin shock , the levels of cytokines produced by LPS-activated macrophages ( e . g . , tumor necrosis factor-α ( TNF-α ) , macrophage inflammatory protein-1α ( MIP-1α , also termed CCL3 ) , and MIP-1β [CCL4] ) were significantly decreased by OXA treatment . In the later stage , the levels of many more cytokines produced by macrophages and/or lymphocytes were significantly decreased , including the levels of interferon-γ ( IFN-γ ) and interleukin-17 ( IL-17 ) , of which the decrease by OXA treatment was particularly robust . In healthy mice injected with saline instead of LPS , peripherally administered OXA did not affect the cytokine levels in the peripheral blood ( Figure 2 ) . To examine the inflammatory state of the CNS , we measured the levels of the same set of 32 cytokines in homogenized brains of LPS-injected mice . LPS increased the levels of most cytokines in the brain , and peripherally administered OXA significantly inhibited the increase of certain cytokines in the brain at 4 hr and/or 22 hr after LPS injection , particularly the microglia-derived chemokines such as MIP-1α/1β ( Figure 2—figure supplement 2 ) . These findings suggest that orexin may modify the course of disease in endotoxin shock by suppressing excessive cytokine production . 10 . 7554/eLife . 21055 . 005Figure 2 . Effects of peripherally administered OXA on cytokine production in mice with endotoxin shock . Effects of OXA treatment on the levels of 32 cytokines in serum at 4 hr and 22 hr after injection of LPS ( LPS-OXA ) or saline ( Sal-OXA ) , compared to saline treatment ( LPS-Sal , Sal-Sal ) ( each group n = 8 , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . OXA ( 1 mg/mouse/24 hr ) started to be SC-administered at 30 min before LPS injection . Statistical significance assessed by 2-way ANOVA coupled with the Bonferroni’s test . Data are replicated in at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 00510 . 7554/eLife . 21055 . 006Figure 2—figure supplement 1 . Effects of SC-infused OXA on body temperature at 4 hr or at 22 hr after LPS injection . The changes of body temperature at 4 hr or 22 hr after LPS injection by SC administration of OXA ( 1 mg/mouse/24 hr , LPS-OXA ) , compared to saline administration ( LPS-Sal ) ( each group n = 8 , ***p<0 . 001 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 00610 . 7554/eLife . 21055 . 007Figure 2—figure supplement 2 . Effects of SC-infused OXA on cytokine production in brain at 4 hr or at 22 hr after LPS injection . LPS increased most cytokines in all samples assayed . SC administration of OXA ( 1 mg/mouse/24 hr , LPS-OXA ) for 4 hr or 22 hr decreased several cytokine levels in the brain from LPS-injected mice , compared to saline SC administration ( LPS-Sal ) ( each group n = 8 , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 00710 . 7554/eLife . 21055 . 008Figure 2—figure supplement 3 . Effects of OXA on expressions of IL-6 ( A ) , TNF-α ( B ) , IL-17 ( C ) , and IFN-γ ( D ) mRNA in cultured peritoneal macrophages from wild-type and OXRKO mice . Hypoxanthine phosphoribosyl transferase-1 ( HPRT ) was used as an internal control ( each group n = 6 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 008 The excessive cytokine production in sepsis is primarily attributed to macrophages activated by LPS ( Beutler and Rietschel , 2003; Moore et al . , 1976 ) . Our findings suggest that orexin may regulate LPS-induced macrophage activation either by acting through the CNS or by acting directly on macrophages . To examine the latter possibility , we prepared peritoneal macrophages from wild-type or OXRKO mice , and then activated cultured macrophages by LPS in the absence or presence of OXA ( 1 nM and 100 nM ) . Levels of IL-17 , IFN-γ , IL-6 , and TNF-α mRNA in cultured macrophages from both wild-type and OXRKO mice were increased after LPS addition . We found no evidence for orexin’s direct action on macrophages , however , since the induction of these mRNAs was not affected in the presence of OXA ( 1 nM and 100 nM ) ( Figure 2—figure supplement 3 ) . Catecholamines and corticosterone could be the possible immune modulators mediating the anti-inflammatory effects of orexin through the CNS ( Tracey , 2002 ) . Norepinephrine levels in the serum from LPS-injected mice were significantly decreased , compared to healthy mice ( Figure 1—figure supplement 1A ) . OXA increased serum epinephrine and partially inhibited the decrease of norepinephrine levels in LPS-injected mice ( Figure 1D ) . Corticosterone levels in the plasma from LPS-injected mice were significantly increased , compared to healthy mice ( Figure 1—figure supplement 1B ) . This is consistent with the report that LPS activates PVN ( Elmquist et al . , 1996 ) . OXA further augmented the increase of corticosterone levels in LPS-injected mice ( Figure 1E ) . These findings suggest that orexin simultaneously modulates a multitude of peripheral effectors to suppress the systemic inflammatory response . In rodents ( Blanqué et al . , 1996 ) and humans ( Kushimoto et al . , 2013 ) , hypothermia is regarded as an important indicator for the severity of inflammation in sepsis . We investigated the correlation between survival rate and body temperature in mice with endotoxin shock . Core body temperature of orexin-administered mice with endotoxin shock remained higher than those of saline-administered mice with endotoxin shock ( Figure 3A ) . The mice that survived through endotoxin shock had significantly higher body temperatures at 24 hr post-LPS injection , as compared with non-survivors ( Figure 3B ) . A cut-off value of 28 . 5°C gave an optimal receiver operating characteristic ( ROC ) curve , predicting survival with highest sensitivity and specificity ( Figure 3C ) . This suggests that body temperature at 24 hr after LPS injection can be a good predictor or marker of eventual survival in our mice with endotoxin shock . This does not necessarily mean that maintenance of body temperature is a sole cause of the improved survival in endotoxin shock . Indeed , we found that the survival of mice with endotoxin shock was not improved by simply warming up the animal on a heat pad to the same degree as the orexin-treated mice ( Figure 3—figure supplement 1 ) , although it is reported that the survival is improved by dramatically elevating body temperature through external heating to 42°C to activate the heat shock protein pathway ( Chu et al . , 1997 ) . 10 . 7554/eLife . 21055 . 009Figure 3 . Effects of OXA on body temperature and heart rate in mice with endotoxin shock . ( A ) The changes in body temperature of LPS-injected mice treated with OXA ( 1 mg/mouse/24 hr ) or saline ( each group n = 10 ) . ( B ) The correlation between the survival and body temperature in mice with endotoxin shock ( Survivor n = 12 , Non-survivor n = 8 , ***p<0 . 001 ) . ( C ) Receiver operating characteristic ( ROC ) curve between survival and body temperature in mice with endotoxin shock . AUC: area under the curve . ( D , E ) Transient effects of bolus ICV-injected OXA ( Wild-type_OXA ) or saline ( Wild-type_Sal ) on body temperature ( D ) and heart rate ( E ) in LPS-injected wild-type mice ( each group n = 4 , *p<0 . 05 ) . ( F , H ) Transient effects of bolus IP-injected OXA on body temperature ( F ) and heart rate ( H ) in mice with endotoxin shock . ( G , I ) IP injection of OXA ( Wild-type_OXA ) but not saline ( Wild-type_Sal ) increased body temperature ( G ) and heart rate ( I ) transiently in LPS-injected wild-type mice , but not in LPS-injected OXRKO mice ( OXRKO_OXA ) ( each group n = 4 , **p<0 . 01 , ***p<0 . 001 ) . Statistical significance assessed by unpaired t-test ( B ) , 2-way ANOVA ( D , E ) , and 1-way ANOVA coupled with the Bonferroni’s test ( G , I ) . Data are replicated in at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 00910 . 7554/eLife . 21055 . 010Figure 3—figure supplement 1 . Effects of whole-body warming with a heat pad on body temperature ( upper ) and survival rate ( lower ) in mice with endotoxin shock . Warming elevated body temperature at 24 hr after LPS injection , but did not improve the survival of mice with endotoxin shock , compared to room temperature exposure ( each group n = 4 , *p<0 . 05 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 01010 . 7554/eLife . 21055 . 011Figure 3—figure supplement 2 . Transient effects of IP-injected OXA on body temperature and heart rate in healthy wild-type mice injected with saline ( A , B , C , D ) or in Hcrtr1-/- ( OX1RKO ) and Hcrtr2-/- ( OX2RKO ) mice with endotoxin shock ( E , F , G , H ) . IP injection of OXA had no effects on body temperature and heart rate in healthy wild-type mice . In OX2RKO mice , transient effects of OXA on body temperature and heart rate in endotoxin shock were canceled . In OX1RKO mice , the transient effect of OXA on body temperature in endotoxin shock was significantly potentiated . ( each group n = 4 , *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 011 Assessment of cardiovascular function ( e . g . , heart rate and blood pressure ) is also important for the evaluation of the severity of endotoxin shock . We investigated the effects of transiently administered OXA on body temperature and heart rate in mice with endotoxin shock . Body temperature and heart rate of mice ICV-injected with OXA remained significantly higher for 3 hr , compared to saline-injected mice ( Figure 3D , E ) . Peripheral ( IP ) injection of OXA ( 0 . 1 mg/mouse ) also increased body temperature ( Figure 3F , G ) and heart rate ( Figure 3H , I ) transiently in wild-type mice but not OXRKO mice with endotoxin shock , indicating that this transient effects of orexin are mediated by orexin receptors . In contrast , IP injection of OXA had no effects on body temperature and heart rate in healthy wild-type mice ( Figure 3—figure supplement 2A–D ) . Peripherally administered orexin does not efficiently penetrate the BBB under healthy conditions ( Fujiki et al . , 2003 ) . However , we found that the uptake of [125I]OXA by the brain after intraperitoneal ( IP ) administration was significantly increased in mice with endotoxin shock , compared to healthy mice ( Figure 4A , B ) . These results indicate that peripherally administered orexin can penetrate the BBB under endotoxin shock , likely due to the dysfunction of BBB induced by systemic inflammation . 10 . 7554/eLife . 21055 . 012Figure 4 . Direct action of OXA on the CNS in mice with endotoxin shock . ( A ) [125I]OXA-autoradiography of 1 mm coronal brain sections from control ( saline ) and endotoxin shock ( LPS ) mice . The brains were removed without perfusion at 2 hr after intraperitoneal ( IP ) administration of [125I]OXA , and were fixed in 4% PFA overnight . The sections were exposed to imaging plates for five days , and then scanned by BAS-2500 ( Fuji Film ) . ( B ) The levels of radioactivity in the whole brain from LPS- or saline-injected mice at 2 hr after IP administration of [125I]OXA ( each group n = 4 , ***p<0 . 001 ) . ( C ) Kaplan-Meier survival curves of wild-type mice intracerebroventricularly ( ICV , upper; saline: n = 8 , OXA: n = 7 ) or subcutaneously ( SC , lower; each group n = 7 ) administered with saline or OXA ( 0 . 3 mg/mouse/24 hr ) before LPS injection . ( D ) Effects of ICV-administered OXA on the levels of MIP-1α , MIP-1β , TNF-α and IL-6 in serum at 4 hr after LPS injection ( LPS-OXA ) , compared to saline treatment ( LPS-Sal ) ( each group n = 8 , *p<0 . 05 ) . OXA administration ( 0 . 1 mg/mouse/4 hr ) started 30 min before LPS injection . Data are presented as mean±s . e . m . Statistical significance assessed by unpaired t-test ( B ) , Mantel Cox log-rank test ( C ) , and 2-way ANOVA coupled with the Bonferroni’s test ( D ) . Data are replicated in at least three independent experiments . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 01210 . 7554/eLife . 21055 . 013Figure 4—figure supplement 1 . Effects of ICV-infused OXA on body temperature at 4 hr after LPS injection . The changes of body temperature at 4 hr after LPS injection by ICV administration of OXA ( 0 . 3 mg/mouse/24 hr , LPS-OXA ) , compared to saline administration ( LPS-Sal ) ( each group n = 8 , **p<0 . 01 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 013 We then found that central ( ICV ) but not peripheral ( SC ) administration of OXA at a lower dose ( 0 . 3 mg/mouse/24 hr ) starting 30 min before LPS injection improves the survival ( Figure 4C ) , further suggesting a central action of orexin . Here again , although orexin was administered for the first 24 hr only , its life-supporting effects manifested over the next several days . We also investigated the early-phase effects of centrally administered OXA on the levels of relevant cytokines/chemokines ( see Figure 2 ) . ICV administration of OXA ( 0 . 1 mg/mouse/4 hr ) significantly suppressed the increase in the level of MIP-1α and showed a tendency to suppress the increase in the levels of MIP-1β and TNF-α in the peripheral blood at 4 hr after LPS injection ( Figure 4D ) . ICV administered OXA simultaneously ameliorated hypothermia induced by LPS ( Figure 4—figure supplement 1 ) . These findings suggest that orexin modifies the course of inflammatory processes in endotoxin shock by acting directly on the CNS . To clarify the central targets of orexin’s life-supporting effects in mice with endotoxin shock , we then focused on the transient thermogenic effect of orexin as a surrogate index for the survival . Raphe nuclei , containing serotonergic neurons , exist in the pons/medulla and midbrain . Medullary raphe nuclei are known as one of the important sites of action for orexin’s thermogenic effect ( Tupone et al . , 2011 ) . In wild-type mice with endotoxin shock , IP administration of OXA ( 0 . 1 mg/mouse ) activated serotonergic neurons in the medullary raphe pallidus and raphe magnus ( RPa/RMg ) , as assessed by double immunostaining with anti-Fos and anti-5HT antibodies ( Figure 5A ) . There was a significant increase in the percentage of Fos-positive serotonergic neurons in RPa/RMg after OXA administration , compared to saline administration ( Figure 5B ) . This Fos induction was not seen in healthy wild-type mice ( Figure 5B ) or in OXRKO mice with endotoxin shock ( Figure 5C ) . Midbrain dorsal raphe nucleus ( the dorsal raphe nucleus , dorsal part and ventrolateral part; DRD/DRVL ) is reported to be a thermosensor activated by elevation of body temperature ( Hale et al . , 2011 ) . In wild-type mice with endotoxin shock , there was also a significant increase of Fos activity in serotonergic neurons in DRD/DRVL by OXA administration ( Figure 5—figure supplement 1A–C ) . We systematically surveyed other known orexin targets in the CNS , and detected no statistically significant changes in Fos activities upon OXA administration under endotoxin shock in these areas ( Figure 5—figure supplement 1D–H ) . Neurons in locus coeruleus ( LC ) , rostral ventrolateral medulla ( RVLM , a sympathetic nucleus ) , nucleus tractus solitaries ( NTS , a parasympathetic sensory nucleus ) , and dorsal motor nucleus of the vagus ( DMNX , a parasympathetic motor nucleus ) are known to be activated after LPS injection ( Elmquist et al . , 1996 ) . Stimulation of LC by corticotropin-releasing hormone is reported to exert an immunosuppressive effect on the peripheral lymphocytes ( Rassnick et al . , 1994 ) . We confirmed that these neurons were activated after LPS injection ( Figure 5—figure supplement 2A ) , independently of orexin signaling ( Figure 5—figure supplement 2B ) . We concluded that , in mice with endotoxin shock , peripheral administration of orexin activated the serotonergic thermoregulatory system in an orexin receptor-dependent manner . 10 . 7554/eLife . 21055 . 014Figure 5 . The hypothetical schema of multiple pathways by which OXA may improve survival in mice with LPS-induced endotoxin shock ( see also main text ) . ( A ) IP-administered OXA ( LPS-OXA , right ) but not saline ( LPS-Sal , left ) activated serotonergic neurons in the raphe pallidus nucleus and raphe magnus nucleus ( RPa/RMg ) in LPS-injected wild-type mice by immunohistochemistry using anti-c-fos and anti-5HT antibodies . Black arrows indicate double positive cells and white arrows indicate 5HT-single positive cells . ( B , C ) IP-administration of OXA significantly activated serotonergic neurons in RPa/RMg of LPS-injected wild-type mice , compared to saline administration ( each group n = 4 , *p<0 . 05 ) ( B ) . There was no significant difference in the activities of serotonergic neurons in RPa/RMg between saline and OXA administration in healthy wild-type mice ( B ) or in OXRKO mice with endotoxin shock ( C ) . Statistical significance assessed by unpaired t-test . Scale bar: 100 μm . Data are replicated in at least three independent experiments . ( D ) LPS activates macrophages via toll-like receptor 4 and produces TNF-α , IL-6 , and other cytokines . These cytokines activate NTS and DMNX through sensory vagus nerves ( Tracey , 2002 ) . Activated motor vagus nerves regulate the inflammatory response and decrease heart rate and body temperature . Orexin sustains heart rate and body temperature through sympathetic nervous system by activating medullary raphe , and also regulates the inflammatory response by central actions . Orexin thus improves the survival through a multitude of pathways , including neuroendocrine and autonomic nervous systems . NTS: nucleus tractus solitaries . DMNX: dorsal motor nucleus of the vagus . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 01410 . 7554/eLife . 21055 . 015Figure 5—figure supplement 1 . Systematic survey of orexin’s targets in the CNS of mice with endotoxin shock . ( A ) Typical photographs of double staining for c-fos and 5HT in DRD/DRVL from LPS-injected mice treated with saline ( LPS-Sal , left ) or OXA ( LPS-OXA , right ) . ( B , C ) The percentage of c-fos positive serotonergic neurons in DRD/DRVL from LPS-injected mice treated with saline or OXA , form saline-injected normal mice treated with saline or OXA ( B ) and from LPS-injected OXRKO mice treated with saline or OXA ( C ) . ( D , E ) The percentage of c-fos positive serotonergic neurons in DRV/DRI ( D ) and in MnR ( E ) from LPS-injected mice treated with saline or OXA and from normal mice treated with saline or OXA . ( F ) The percentage of c-fos positive orexin neurons in LH from LPS-injected mice treated with saline or OXA and from normal mice treated with saline or OXA . ( G , H ) The percentage of c-fos positive noradrenergic neurons in LC ( G ) and histaminergic neurons in TMN ( H ) from LPS-injected mice treated with saline or OXA and from normal mice treated with saline or OXA . ( B–H each group n = 4 , **p<0 . 01 ) . DRD: dorsal raphe nucleus , dorsal part; DRVL: dorsal raphe nucleus , ventrolateral part , DRV: dorsal raphe nucleus , ventral part , DRI: dorsal raphe nucleus , interfascicular part , LH: lateral hypothalamus , LC: locus coeruleus , TMN: tuberomammillary nucleus . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 01510 . 7554/eLife . 21055 . 016Figure 5—figure supplement 2 . Activation of neurons in the LC , RVLM , NTS and DMNX at 3 hr after LPS injection in wild-type ( A ) or OXRKO ( B ) mice . Typical photographs of Fos staining in LC ( left ) , RVLM ( middle ) , NTS and DMNX ( right ) from wild-type mice or OXRKO mice injected with saline or LPS . LC: locus coeruleus . RVLM: rostral ventrolateral medulla . NTS: nucleus tractus solitaries . DMNX: dorsal motor nucleus of the vagus . DOI: http://dx . doi . org/10 . 7554/eLife . 21055 . 016
Septic shock is a life-threatening condition with no effective treatment reported so far ( Peake et al . , 2014; Mouncey et al . , 2015; Yealy et al . , 2014; Ranieri et al . , 2012 ) . We found that the survival of mice with endotoxin shock is improved by central administration of orexin . Moreover , surprisingly , peripheral injection of orexin also improved survival , because orexin could penetrate into the CNS due to BBB dysregulation under the condition of systemic inflammation . Our findings show that orexin could be a new therapeutic agent for septic shock; we propose here an innovative approach to target the CNS for systemic inflammation , employing an unexpected method to deliver neuropeptides into the brain by taking advantage of the pathophysiology of sepsis . In mice with endotoxin shock , peripherally administered orexin robustly decreased the levels of IL-17 and IFN-γ in peripheral blood along with other pro-inflammatory cytokines such as IL-6 and TNF-α . Inhibiting either IL-17 or IFN-γ by neutralizing antibodies within 24 hr after cecal ligation and puncture ( CLP ) is reported to improve survival ( Flierl et al . , 2008; Márquez-Velasco et al . , 2011 ) . Although these antibodies are administered in the early period only , their life-supporting effects manifest over the next several days , similar to the effects of orexin in the present study . IL-17 produced during sepsis is implicated in the ‘cytokine storm , ’ in which the inflammatory response is amplified , increasing the lethality of septic shock ( Flierl et al . , 2008 ) . IL-17 synergistically potentiates LPS-induced activation of peritoneal macrophages in vitro and increases the levels of IL-6 , TNF-α , and IL-1β . In our study , orexin did not directly inhibit the increase of cytokine production in LPS-activated peritoneal macrophages , although TNF-α production in LPS-stimulated microglia , the primary immune cells in the CNS , is reported to be decreased by OXA pretreatment ( Xiong et al . , 2013 ) . Together with the findings that centrally administered orexin was effective at a lower dose in improving the survival of mice with endotoxin shock and in suppressing MIP-1α levels in their peripheral blood , our results indicate that the anti-inflammatory effects of orexin are through its action in the CNS rather than in the peripheral tissues . We found that orexin increases blood corticosterone levels of mice with endotoxin shock , which is consistent with the report that ICV administration of orexin increases the expression of ACTH in CLP-induced septic shock model ( Deutschman et al . , 2013 ) . The blood corticosterone level is increased in mice with LPS-induced lung injury ( Landgraf et al . , 2014 ) , and even a slight upregulation of corticosterone by exogenous leptin may improve LPS-induced lung injury ( Landgraf et al . , 2014 ) . Therefore , the increase of corticosterone levels in endotoxin shock may play a role in the anti-inflammatory effects of orexin , although further studies with glucocorticoid agonists/antagonists are necessary to substantiate this possibility . Indeed , glucocorticoids are sometimes used for human patients with sepsis and effective in some cases ( Dellinger et al . , 2013 ) . We also found that orexin increased the levels of epinephrine and norepinephrine in the blood of mice with endotoxin shock: Further studies with adrenergic antagonists would fully qualify the role of these neuroendocrine changes in the observed improvements in survival . The increases of catecholamines might not only stabilize hemodynamics by potentiating cardiovascular function , but also modulate inflammation through adrenoceptors ( Spengler et al . , 1990; Vida et al . , 2011 ) . However , although catecholamines are used to stabilize hemodynamics in human patients with sepsis ( Dellinger et al . , 2013 ) , their effects are often limited because vascular responsiveness to catecholamines is reduced during sepsis ( Parker and Parrillo , 1983 ) . Also , adrenergic modulation of inflammation is complicated . The adrenergic anti-inflammatory effects are mostly mediated by β2-adrenoceptors ( Vida et al . , 2011 ) , while stimulation of α2-adrenoceptors induces pro-inflammatory effects ( Spengler et al . , 1990 ) . Moreover , systemically administered catecholamines act on all tissues without coordination . In contrast , catecholamines endogenously released under the CNS command are localized in the specific target tissues and their actions on each target are integrated by the CNS . This may be why the life-supporting effects of orexin administration on sepsis may be longer lasting and more effective than those of catecholamine administration alone . Indeed , in the rat septic shock model , administration of norepinephrine for 24 hr improves the survival rate by only 20% ( Li et al . , 2009 ) . In contrast , administration of orexin for 24 hr showed 70–80% and 40% improvements in the survival rate for prophylactic and post-hoc protocols , respectively , in our study ( Figures 1 and 3 ) . Furthermore , it is reported that the treatment with both catecholamines and glucocorticoids synergistically attenuates the production of pro-inflammatory cytokines and IL-17 during sepsis ( Bosmann et al . , 2013 ) . There has been no report showing the role of the raphe nuclei in septic shock . Raphe nuclei are reported to be an orexin-regulated region for brown adipose tissue thermogenesis and heart rate modulation ( Tupone et al . , 2011 ) . Taken together , our findings show that orexin improves the survival of mice with endotoxin shock through a multitude of pathways , including the neuroendocrine and autonomic nervous system , by targeting the CNS ( hypothesized in Figure 5D ) . The precise mechanism through which CNS orexin modulates inflammatory responses merits further investigation . The present study suggests that a coordinated intervention through the CNS against the complex pathophysiology of systemic inflammation can be a novel therapeutic approach for septic shock . While we therapeutically exploited the leakiness of BBB under sepsis in this study , BBB dysfunction may also permit cytokines and inflammatory cells to freely enter into the CNS , which may constitute an additional element of pathophysiology of sepsis , such as ‘sepsis-associated encephalopathy’ ( Widmann and Heneka , 2014 ) . The failure of traditional therapy ( e . g . , catecholamines ) in septic shock has been commonly attributed to defects in peripheral response mechanisms . Our study suggests that it may be the CNS regulatory machinery that is dysfunctional , which could be partially restored by orexin .
Eight- to 10-week-old male mice from four genotypes ( Hcrtr1-/-;Hcrtr2-/- , Hcrtr1-/- , Hcrtr2-/- , and wild-type , on a C57BL/6J background , generated by crosses between homozygous mice ) were randomly assigned to experimental groups . Experimental mice were individually housed and kept on a 12 hr:12 hr light:dark schedule at an ambient temperature of 23 ± 1°C under specific pathogen-free conditions . Lipopolysaccharide ( LPS , Lipopolysaccharides from Escherichia coli 055:B5 , Sigma , St . Louis , MO ) was dissolved in saline ( Otsuka , Japan ) and LPS solution was adjusted to 1 mg/ml . Human orexin-A ( Peptide Institute , Japan ) was dissolved in saline and adjusted for each experiment . Phosphate Buffered Saline ( PBS ) was made from Phosphate Buffered Saline Powder ( 0 . 01 mol/l , pH 7 . 2–7 . 4 , Wako , Japan ) and Phosphate Buffer ( PB ) was made from Phosphate Buffered Powder ( 1/15 mol/l , pH 7 . 4 , Wako ) . Paraformaldehyde ( PFA , Nacalai , Japan ) was dissolved in PBS and adjusted to 4% ( w/w ) . To induce the endotoxin shock , mice were intraperitoneally ( IP ) injected with LPS ( 10 mg/kg ) at ZT10 . 5 . To investigate the central effect of orexin , mice were implanted with a guide cannula into the left lateral ventricle . To monitor body temperature and heart rate of mice continuously every ten minutes , the mice were subcutaneously implanted with a telemetry probe ( TA11TA-F10 , Data Science International , St . Paul , MN ) . Mice were recovered for one week after operation . To investigate the prophylactic effect of peripherally administered orexin on survival in mice with endotoxin shock , mice were implanted subcutaneously with an osmotic pump ( Alzet , Durect , Cupertino , CA ) containing 1 mg orexin-A dissolved in 200 µl saline ( or containing 200 µl saline as control ) under anesthesia before LPS injection . Subcutaneous infusion started at a rate of 8 µl/hr for 24 hr at the same time of implantation of the pump . To investigate the post-hoc therapeutic effect of peripherally administered orexin on survival , mice were implanted subcutaneously with a programmable microinfusion pump ( iPRECIO , SMP300 , Primetech , Japan ) containing 1 mg orexin-A dissolved in 100 µl saline ( or containing 100 µl saline as a control ) under anesthesia at 30 min before LPS injection . These pumps are wirelessly controlled by an external scheduler . The infusion schedule was programmed with iPRECIO Management Software ver 1 . 1 . At 30 min after LPS injection , subcutaneous infusion started at a rate of 8 µl/hr for 24 hr . The next day iPRECIO was refilled with orexin A ( 1 mg orexin-A dissolved in 100 µl saline ) or 100 µl saline . To investigate the effect of centrally administered orexin on survival , mice were intracerebroventriculerly ( ICV ) administered with 0 . 06 mg orexin-A dissolved in 3 µl saline ( or 3 µl saline as a control ) in 5 min and implanted subcutaneously with an osmotic pump containing 200 µl saline as transfusion ( at a rate of 8 µl/hr ) under anesthesia at 30 min before LPS injection . At the same time of LPS injection , ICV infusion was started at the rate of 0 . 5 µl /hr ( orexin-A concentration was 0 . 02 mg/µl ) . To investigate the effect of warming up to the same degree as orexin-treated mice on survival , mice were warmed on a heat pad at 30°C for 24 hr at the same time of LPS injection . In all the survival experiments , body temperature was monitored for seven days after LPS injection . We assessed survival as primary endpoint , and body temperature at 24 hr after LPS injection as secondary endpoint . Observation was stopped in seven days or by death . Relationship between survival and body temperature was expressed by the receiver operating characteristic ( ROC ) curve , plotting true positive rate ( Y-axis; sensitivity ) against the false positive rate ( X-axis; one minus specificity ) at various body temperature values . The ROC curve was assessed by area under the curve ( AUC ) , and an optimal cut-off values was determined , which gave maximum AUC . The sample size was calculated on the following assumptions; α is 0 . 05 , power is 0 . 8 , the usual survival rate in control group at two days is 0 . 2 and the expected survival rate in the treatment group at seven days is 0 . 6–0 . 9 . All the experimental mice were used as data samples without inclusion/exclusion . Monitoring body temperature , LPS- or saline-injected mice were IP-administered with [125I]orexin-A at 10 ( Borovikova et al . , 2000 ) cpm . At 2 hr after [125I]orexin-A administration , we removed the whole brain and the levels of radioactivity were determined in a γ-counter . The brains were then fixed in 4% PFA overnight . One-mm coronal sections were exposed to imaging plates ( Bas-SR 2040 , Fuji Film , Japan ) for five days and then scanned by BAS-2500 ( Fuji Film ) . Mouse peritoneal macrophages were harvested five days after IP injection of 1 . 5 ml thioglycolate ( 2 . 4 g/100 ml; BD Biosciences , Franklin Lakes , NJ ) by peritoneal lavage with PBS and the purity of cell suspension was >95% macrophages . Macrophages ( 5 × 10−6 per experimental condition ) were allowed to adhere to tissue culture plates overnight . Cells were incubated in 10% FBS/DMEM containing LPS ( 50 ng/ml ) in the absence or presence of orexin A ( 10−9 , 10−7 M ) for 6 hr at 37°C . Then culture medium was sampled and total RNA was prepared by using RNeasy ( Qiagen ) . The expression levels of IL-6 , IL-17 , IFN-γ , and TNF-α were measured by QPCR using the following primers: 5’-CACAGAGGATACCACTCCCAACA-3’ and 5’-TCCACGATTTCCCAGAGAACA-3’ for IL-6; 5’-GAAGGCCCTCAGACTACCTCAA 3’ and 5’-TCATGTGGTGGTCCAGCTTTC-3’ for IL-17; 5’-CGCCTATCTTCGGGATGAATC 3’ and 5’-CCAACCGATACTCCATGAAAATG-3’ for IFN-γ; 5’-GGCCTCCCTCTCATCAGTTC-3’ and 5’-GACAAGGTACAACCCATCGGC-3’ for TNF-α . To investigate transient effects of peripherally administered orexin , mice were bolus IP-administered with 0 . 1 mg ( 30 nmol ) orexin-A dissolved in 100 µl saline ( or with 100 µl saline as a control ) at 13 . 5 hr after LPS or saline injection . To investigate transient effects of centrally administered orexin , mice were ICV-administered with 0 . 01 mg ( 3 nmol ) orexin-A dissolved in 6 μl saline ( or with 6 μl saline as a control ) in 6 min at 13 . 5 hr after LPS injection . Body temperature and heart rate were continuously measured for 6 hr after orexin treatment . Wild-type and OXRKO mice were bolus IP-injected with 0 . 1 mg ( 30 nmol ) orexin-A dissolved in 100 µl saline ( or with 100 µl saline as a control ) at 13 . 5 hr after LPS or saline injection . At 1 . 5 hr after orexin treatment , mice were anesthetized with pentobarbital sodium ( 50 mg/kg , IP ) and were perfused with PBS followed by 4% PFA . Brains were removed from the skull and were post fixed with 4% PFA overnight and then immersed in 30% sucrose dissolved in PB until they had sunk . After freezing of the brains in Optical Cutting Temperature Compound ( O . C . T . Compound , Sakura Finetek , Torrance , CA ) , the brains were cut serially at 40 µm by the cryostat ( Leica , Germany ) . All sections were incubated with rabbit anti-Fos antibody ( 1:10 , 000; RRID: AB_2314421 ) overnight at 4°C . After incubating with primary antibodies , sections were incubated in biotinylated anti-rabbit IgG antibody ( 1:200; RRID: AB_2313606 ) for 1 . 5 hr and were treated with avidin–biotin complex ( ABC kit , Vector Labs . , Burlingame , CA ) for 30 min at room temperature . The immunoreactive product was visualized in a solution of 3 , 3’-diaminobenzidine ( DAB , Vector Labs . ) with nickel and H2O2 , to label the nuclei of the Fos-positive cells in dark black . For identifying the neurochemical feature of the neurons , after staining for Fos , the sections in the raphe nuclei were incubated with goat anti-5HT antibody ( 1:5 , 000; RRID: AB_572262 ) , the ones in the lateral hypothalamus ( LH ) were incubated with goat anti-orexin-A antibody ( 1:50; RRID: AB_653610 ) , and the ones in the tuberomammillary nucleus ( TMN ) were incubated with rabbit anti-histidine decarboxylase ( HDC ) antibody ( 1:1 , 000; RRID: AB_1002154 ) overnight at 4°C . After incubating with primary antibodies , sections were incubated secondly with each biotinylated antibody ( anti-goat IgG antibody ( 1:200; RRID: AB_2336123 ) and anti-rabbit IgG antibody ( 1:200; RRID: AB_2313606 ) ) for 1 . 5 hr , and ABC solution for 30 min at room temperature . The immunoreactive product was visualized in a solution of DAB without nickel intensification to label the cytoplasm of the cells with the neurochemical feature in light brown . Finally , sections were mounted on slides , air dried , dehydrated and cover-slipped with mounting medium for microscope preparation . Cell counts were conducted by three investigators blind to the assignment of the treatment group by using an LSM700 microscope ( Zeiss , Germany ) . Cytoplasm-labeled cells and double-labeled cells were counted in the sections of each region . We assessed the ratio of double-labeled cells to cytoplasm-labeled cells plus double-labeled cells . Survival rates were expressed by a Kaplan-Meier curve and comparisons of survival curves were performed with a Mantel Cox log-rank test using PRISM Ver . 5 . 0 ( RRID: SCR_002798 ) . All values are expressed as means ± s . e . m . To compare two groups , data were analyzed with unpaired student t-test . To compare more than two groups , data were analyzed with a 1-way ANOVA , and individual group means were then compared with a Bonferroni’s test . To compare the time-elapsed data , the data were analyzed with a 2-way ANOVA , and individual group means were then compared with a Bonferroni’s test . Differences were considered significant when p<0 . 05 . | The body has a range of defenses to fight infection , which play a crucial role in keeping us healthy . However , sometimes the response to infection may damage the body’s own tissues and organs , leading to a life-threatening condition called sepsis . In the most severe stage of sepsis – known as septic shock – blood pressure drops to dangerously low levels and the individual often dies . There is currently no effective therapy for septic shock . Recent studies have revealed how the brain regulates immune responses via chemical signals and nerve impulses . A molecule called orexin is made in the brain and regulates the activity of a group of neurons that control sleep . Orexin can also alter heart rate and body temperature in rats , which suggests that it may have potential to be developed as a treatment for septic shock . To test this idea , Ogawa , Irukayama-Tomobe et al . injected orexin under the skin of mice with septic shock . The experiments show that the injected orexin is able to enter the brain , where it helps the mice to survive and recover from septic shock by restoring normal body temperature and boosting heart rate . Further experiments suggest that orexin is likely to regulate immune responses through multiple signaling pathways in the brain . The next step following on from this work is to find out the precise mechanism through which orexin regulates the responses of the immune system . This orexin treatment strategy should also be tested on primates with septic shock before planning any clinical trials in humans . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience",
"immunology",
"and",
"inflammation"
] | 2016 | Peripherally administered orexin improves survival of mice with endotoxin shock |
The RNA polymerase II largest subunit ( Rpb1 ) contains a unique C-terminal domain ( CTD ) that plays multiple roles during transcription . The CTD is composed of consensus Y1S2P3T4S5P6S7 repeats , in which Ser , Thr and Tyr residues can all be phosphorylated . Here we report analysis of CTD Tyr1 using genetically tractable chicken DT40 cells . Cells expressing an Rpb1 derivative with all Tyr residues mutated to Phe ( Rpb1-Y1F ) were inviable . Remarkably , Rpb1-Y1F was unstable , degraded to a CTD-less form; however stability , but not cell viability , was fully rescued by restoration of a single C-terminal Tyr ( Rpb1-25F+Y ) . Cytoplasmic and nucleoplasmic Rpb1 was phosphorylated exclusively on Tyr1 , and phosphorylation specifically of Tyr1 prevented CTD degradation by the proteasome in vitro . Tyr1 phosphorylation was also detected on chromatin-associated , hyperphosphorylated Rpb1 , consistent with a role in transcription . Indeed , we detected accumulation of upstream antisense ( ua ) RNAs in Rpb1-25F+Y cells , indicating a role for Tyr1 in uaRNA expression .
RNA polymerase II ( RNAP II ) is a multisubunit enzyme responsible in eukaryotes for transcription of all mRNAs and many non-coding RNAs . Rpb1 , the largest subunit , contains a unique C-terminal domain ( CTD ) , composed of up to 52 heptad repeats with the consensus sequence Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7 ( YSPTSPS ) . The CTD performs multiple functions during transcription ( Buratowski , 2009; Munoz et al . , 2010; Egloff et al . , 2012; Hsin and Manley , 2012; Heidemann et al . , 2013 ) , most of which are dependent on phosphorylation of specific CTD residues . For example , Ser5 phosphorylation ( Ser5-P ) promotes recruitment of capping enzyme ( Fabrega et al . , 2003 ) and Ser2-P can be important for 3′ mRNA processing ( Kim et al . , 2004 ) . Ser7-P and Thr4-P also function in 3′ processing , of snRNAs ( Egloff et al . , 2007 ) and histone mRNAs ( Hsin et al . , 2011 ) , respectively , with the later also functioning in transcription elongation ( Hintermair et al . , 2012 ) . Tyr1 can be phosphorylated in mammals ( by c-Abl; Baskaran et al . , 1993 ) and yeast ( Mayer et al . , 2012 ) , where it may prevent premature recruitment of termination factors . However , the function ( s ) of Tyr1 and Tyr1-P in metazoans are unknown .
We previously utilized chicken DT40 cells to study properties of the Rpb1 CTD . We showed that an Rpb1 derivative containing a CTD with 26 YSPTSPS repeats ( Rpb1-26r ) plus the ten C-terminal non-consensus residues , important for stability ( Chapman et al . , 2005 ) , confers cell viability , while a comparable derivative with all Thr4 residues changed to Val was inviable ( Hsin et al . , 2011 ) . To investigate the functions of Tyr1 , we constructed a plasmid encoding a Flag-tagged Rpb1 derivative , Rpb1-Y1F , identical to Rpb1-26r but with all Tyr1 residues replaced by Phe , and expressed this in Rpb1 conditional knock-out cells ( DT40-Rpb1; Hsin et al . , 2011 ) . Tyr1 was suggested to be essential for viability in S . cerevisiae ( West and Corden , 1995 ) , but not in S . pombe ( Schwer and Shuman , 2011 ) . To determine whether Tyr1 is required for growth in vertebrate cells , DT40-Rpb1 cells were transfected with the Rpb1-Y1F vector , and tetracycline ( tet ) was added to turn off wild-type Rpb1 expression . Rpb1-Y1F was unable to complement Rpb1 , whereas Rpb1-26r fully restored viability ( Figure 1—figure supplement 1A ) . We next established cell lines stably expressing Rpb1-Y1F to analyze how the Y1F mutation affects Rpb1 function . Cells expressing Rpb1-Y1F ( Y1F ) stopped growing around 24 hr in medium containing tet ( Figure 1A ) . To examine whether the inviability of Y1F cells might result from different Rpb1 levels , we analyzed several independent Y1F cell lines by Western blot ( WB ) with anti-FLAG antibodies ( Abs ) . Rpb1-Y1F levels were indeed significantly reduced compared to Rpb1-26r ( Figure 1B ) . Importantly , accumulation of a lower molecular weight form ( indicated by * ) was observed in all Y1F cell lines . This corresponds to a derivative likely precisely lacking the CTD , as it migrated slightly more rapidly than an Rpb1 derivative containing six heptads ( Figure 1B ) . 10 . 7554/eLife . 02112 . 003Figure 1 . Growth properties of Rpb1 cell lines . ( A ) Cells were cultured in medium containing 1 µg/ml tetracycline ( tet ) . Control cells , 26r , were split on day 2 . Average cell counts from two independent experiments were plotted . ( B ) Cells were treated with tet for 24 hr . Whole-cell lysates from 26r , Y1F cells , and cells ( 6r ) expressing an Rpb1 with 6 YSPTSPS repeats were analyzed by western blotting . Flag-tagged Rpb1 proteins were detected using Flag antibody . Partially degraded CTD-less Rpb1 is indicated by an asterisk ( * ) . Full-length and degraded Rpb1 isoforms were quantified using ImageJ , and % degradation is displayed . nd , Degraded Rpb1 not detected . ( C ) Cell lysates from four independent 25F+Y and four independent Y1F cell lines were analyzed as in ( B ) . Asterisk indicates partially degraded Rpb1 . The lower molecular weight species in the 26r sample is of unknown identity and was not observed reproducibly , but was included in the quantitation . ( D ) Growth curves of two independent 25F+Y cell lines and 26r cells were plotted as in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 00310 . 7554/eLife . 02112 . 004Figure 1—figure supplement 1 . Complementation test and western blotting analysis in cells with Y1F mutations . ( A ) Complementation test was performed as described and results were shown . ( B ) Western blotting analysis of 20F+Y cells . Cells ( 20F+6Y ) expressing an Rpb1 with the last 6 Y1F repeats replaced with normal YSPTSPS repeats were treated with tet for 24 hr . Rpb1 proteins were detected using Flag antibody . ( C ) Whole-cell lysates prepared from cells expressing the indicated Rpb1 derivative and treated with tet for 24 hr were analyzed by western blotting . Western blots were quantified using ImageJ , and % degradation from 3–4 independent replicates is presented . Error bars denote standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 004 To begin to investigate the basis for Rpb1-Y1F instability , we determined how many Tyr1 residues were necessary to restore stability . We first analyzed an Rpb1-Y1F derivative ( 20F+6Y ) in which the F residues in the C-terminal six heptads were reverted to Y , and found that this derivative was completely stable ( Figure 1—figure supplement 1B ) , although cells expressing Rpb1-20F+6Y remained inviable ( Figure 1—figure supplement 1A ) . Next , we analyzed an Rpb1-Y1F derivative in which only a single F , in the C terminal-most heptad , was changed back to Y ( Rpb1-25F+Y ) . Strikingly , this single Tyr residue was sufficient to prevent Rpb1 degradation , as the truncated isoform , which we denote Rpb1-b , was absent , and Rpb1-25F+Y levels were comparable to Rpb1-26r in multiple 25F+Y cell lines ( Figure 1C; quantitation of the amount of degraded Rpb1 observed in multiple experiments is shown in Figure 1—figure supplement 1C ) . However , despite the restoration of Rpb1 stability , 25F+Y cells remained inviable ( Figure 1D ) . We next set out to determine how Tyr1 residues stabilize Rpb1 . A first question was whether Rpb1 is indeed Tyr1-phosphorylated in DT40 cells . To address this , we utilized an anti-phospho-Tyr1 Ab ( Mayer et al . , 2012 ) to examine Tyr1 phosphorylation ( Tyr1-P ) of Rpb1-25F+Y and Rpb1-26r by WB; both proteins were indeed Tyr1-phosphorylated ( Figure 2A ) . We next investigated where in cells the Rpb1-b isoform accumulates . We analyzed cytoplasmic , nuclear and chromatin-bound fractions from 26r and Y1F cells by WB with an N-terminal Rpb1 Ab ( N20 ) . Rpb1-b ( indicated by * ) was detected in all three fractions from Y1F cells , but barely or not at all in the 26r fractions ( Figure 2B ) . The relative ( and absolute ) Rpb1-b levels were lowest in the cytoplasm , while Rpb1-b was essentially the only form on Y1F chromatin . As anticipated , Rpb1-b was not detected in 25F+Y cell fractions ( Figure 2—figure supplement 1A ) . We next determined whether Tyr1-P could also be detected on Rpb1 in all three fractions , in this case using extracts from wild-type DT40 ( Figure 2C ) and human HEK293 ( Figure 2—figure supplement 1B ) cells . Robust Tyr1-P was indeed detected in all three fractions in both cell types . Notably , in both cytoplasm and nucleoplasm , Tyr1-P was observed only on hypophosphorylated Rpb1 ( the lower band ) , while it was found primarily on the hyperphosphorylated isoform on chromatin . This suggests both that CTD phosphorylation is limited to Tyr1 in the cytoplasm and nucleoplasm and that Tyr1-P is present on hyperphosphorylated RNAP II found on active genes . We also examined phosphorylation on Ser 2 , 5 and 7 and Thr4 ( Figure 2C , Figure 2—figure supplement 1B ) . All these modifications were nearly undetectable in cytoplasmic and nuclear fractions , present almost exclusively on chromatin-associated , hyperphosphorylated Rpb1 . Together , our data show that Tyr1 , and only Tyr1 , is phosphorylated before RNAP II engages in transcription , and support the idea that Tyr1-P functions in stabilizing the CTD when RNAP II is not transcribing , and perhaps also plays a role during transcription . Consistent with this , Tyr1-P was detected on Rpb1 immunoprecipitated by Abs recognizing Ser5-P and Ser2-P ( Figure 2—figure supplement 1C ) . 10 . 7554/eLife . 02112 . 005Figure 2 . Rpb1 Tyr1 phosphorylation is found in all cell fractions . ( A ) Flag-tagged Rpb1 proteins were immunoprecipitated from cells treated with tet for 24 hr , and analyzed using western blotting . Phosphorylation on Tyr1 ( Tyr1-P ) was detected by the 3D12 antibody . ( B ) 26r and Y1F cells were treated with tet for 24 hr , subcellular fractionation was performed , and cytoplasmic , nuclear , and chromatin fractions were analyzed by western blotting . U2AF65 ( a nuclear protein ) , and chromatin bound histone H3 protein served as controls for subcellular fractionation . The asterisk ( * ) indicates partially degraded Rpb1 . ( C ) Wild-type DT40 cells were subjected to subcellular fractionation . The localization of Rpb1 phosphorylated on Tyr1 , Ser 2 , 5 and 7 and Thr4 was determined using antibodies as described in ‘Materials and methods’ . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 00510 . 7554/eLife . 02112 . 006Figure 2—figure supplement 1 . Western blotting analysis . ( A ) Cells were treated with tet for 24 hr , and then subjective to subcellular fractionation . Rpb1 localization was determined by western blotting . Nuclear protein U2AF65 , and chromatin-bound histone H3 served as controls for fractionation . Asterisk ( * ) indicates the degraded Rpb1 fragment . ( B ) Subcellular fractionation assay was performed in HEK293 cells . The localization of Rpb1 phosphorylated on Tyr1 , Ser 2 , 5 , 7 , and Thr4 was determined using antibodies as described in ‘Materials and methods’ . ( C ) Rpb1 from DT40 cell lysates were immunoprecipitated using antibodies recognizing phosphoserine 5 ( 3E8 ) or phosphoserine 2 ( 3E10 ) . The association of tyrosine phosphorylation with phosphoserine 5 or phosphoserine 2 was determined by western blotting using the 3D12 antibody . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 006 We next wished to determine how the CTD is degraded , and whether Tyr1-P indeed plays a role . Since one Tyr residue in the final heptad confers stability , it is unlikely that endoproteolytic cleavage occurred between the CTD and the Rpb1 body . One possibility is that the CTD is degraded by the proteasome , which has been shown to associate both with RNAP II and with active genes ( Gillette et al . , 2004 ) . Additionally , certain naturally unstructured proteins can be degraded by the proteasome in a ubiquitin-independent manner ( Sheaff et al . , 2000; Tofaris et al . , 2001 ) . Since the CTD is considered a structure-less domain ( Meinhart et al . , 2005 ) , we hypothesized that the CTD is a natural proteasome substrate , and that Tyr1-P prevents its proteasomal degradation . To test this directly , we performed in vitro proteasome assays using a GST-CTD substrate ( containing full-length wt CTD ) and purified 20S proteasomes , and detected products by WB . Using an anti-GST Ab , the amount of full-length GST-CTD was greatly diminished and a ladder-like pattern of low molecular weight bands was detected , indicating that the proteasome degraded the GST-CTD protein from the C-terminus ( Figure 3A , lane 1 and 2 ) . Consistent with this , the low molecular weight products were undetectable by WB using an anti-CTD Ab ( 8WG16; Figure 3B ) . Notably , the 20S proteasome used was in a latent status with a closed gate and minimal enzymatic activity ( Groll et al . , 2000; Forster and Hill , 2003 ) . Thus , GST-CTD , like for example the unstructured protein p21 ( Forster and Hill , 2003 ) , was capable of activating the 20S proteasome . Low concentrations of SDS render the proteasome gate disordered , leading to proteasome activation ( Groll et al . , 2000; Forster and Hill , 2003 ) . Indeed , addition of 0 . 01% SDS to reaction mixtures increased CTD degradation ( Figure 3A , lane 3 and 4 ) . In contrast , the proteasome inhibitor MG132 inhibited degradation ( Figure 3A , lane 5 and 6 ) . 10 . 7554/eLife . 02112 . 007Figure 3 . Tyr1 phosphorylation functions in CTD stability . ( A ) In vitro 20S proteasome assay . 200 nM purified GST-CTD was incubated with 5 nM bovine 20S proteasome , and the reaction was carried with or without 0 . 01% SDS . MG132 was used to inhibit the proteasome , and reaction with 2 . 5% ethanol ( ETOH ) served as control . Western blotting was performed using antibody against GST , and the CTD ( B ) . Position of 25 and 75 kD molecular weight markers are indicated . ( C ) 40 nM GST-CTD phosphorylated by recombinant c-Abl or by purified CDK7 complex was incubated with 2 nM 20S proteasome with or without 0 . 01% SDS for 1 hr , and reactions were analyzed by western blotting . ( D ) Y1F cells grown in the absence of tet were treated with 50 nM epoxomicin or 5 µM MG132 for indicated time . DMSO treatment served as control . Cell lysates were analyzed by western blotting . Partially degraded Rpb1 is indicated . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 00710 . 7554/eLife . 02112 . 008Figure 3—figure supplement 1 . Western blotting analysis and quantification . ( A ) GST-CTD proteins were phosphorylated by Abl tyrosine kinase for indicated time in vitro . Reactions were analyzed by western blotting with indicated antibodies . ( B ) Y1F cells grown in the absence of tet were treated with 50 nM epoxomicin or 5 µM MG132 for 4 hr . Cell lysates were analyzed by western blotting as in Figure 3D . The levels of full-length Rpb1 were quantified using ImageJ , and ratios of full-length Rpb1 to degraded Rpb1 ( IIB ) and actin were presented . N = 3 . Error bars display standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 008 We next asked whether phosphorylation of GST-CTD affects its stability in the proteasome assay . For this , we first used a recombinant c-Abl derivative to phosphorylate GST-CTD . This resulted in conversion of a fraction of the GST-CTD to a low-mobility , Tyr1-P isoform , although the majority remained unphosphorylated ( Figure 3C , lane 1 , Figure 3—figure supplement 1A ) , consistent with the processive phosphorylation by c-Abl observed previously ( Duyster et al . , 1995 ) . We then performed the proteasome assay described above using c-Abl-phosphorylated GST-CTD ( Figure 3C ) . Strikingly , the Tyr1 hyperphosphorylated GST-CTD ( top panel , upper band , and lower panel ) was resistant to degradation ( lane 2 ) , while the remaining unphosphorylated GST-CTD ( top panel , bottom band ) was degraded . Addition of 0 . 01% SDS again promoted degradation of unphosphorylated GST-CTD , but the Tyr1-P isoform remained resistant ( lane 3 ) . Significantly , GST-CTD phosphorylated by the Ser5/Ser7 kinase CDK7 , which converted essentially all of the substrate to the hyperphosphorylated form , was not protected from degradation ( lanes 4–6 ) , indicating a specific role of Tyr1-P in preventing proteasomal degradation . We next investigated whether the proteasome functions in Rpb1-Y1F degradation in vivo . We added the proteasomal inhibitors epoxomicin and MG132 to Y1F cells , and measured intact Rpb1 Y1F levels by WB ( Figure 3D ) . Both inhibitors led to approximately twofold elevated levels of Rpb1-Y1F ( see Figure 3—figure supplemental 1B for quantification ) . Although considerable truncated Rpb1-Y1F remained , this likely reflects incomplete proteasomal inhibition and/or accumulation of the truncated form prior to addition of the inhibitors . In any event , our data provide evidence that Tyr1 , and specifically Tyr1-P , prevents proteasomal degradation of the CTD in vitro and in vivo . We next wished to determine the genome-wide impact of the 25F+Y mutation on transcript levels . Using 3′READS ( Hoque et al . , 2013 ) , a deep sequencing method to quantitate poly ( A ) + RNAs , we analyzed 25F+Y and 26r cells , as well as S2A , S5A and T4V cells ( all of which , like 25F+Y , are inviable; Hsin et al . , 2011; Hsin et al . , 2014 ) for comparison . Cells were treated with tet for 24 hr , and a total of ∼5 million reads mapping to 3′ regions of genes were generated for each cell type ( Figure 4—figure supplement 1 ) . Reads were classified into sense RNAs and upstream antisense ( ua ) RNAs ( Figure 4A ) . uaRNAs were defined as transcripts that did not overlap any known protein-coding genes and used a poly ( A ) site within 2 kb from the TSS ( Figure 4—figure supplement 2 ) . Unexpectedly , the number of genes with upregulated uaRNAs was significantly greater than the number of genes with downregulated uaRNAs , by ∼16-fold ( p=10−21 . 5 ) , in 25F+Y cells ( Figure 4B ) . S2A and S5A cells showed similar trends but to much lesser extents , fourfold ( p=10−6 . 7 ) and 5 . 6-fold ( p=10−9 . 0 ) , respectively , while T4V cells in fact showed a trend in the opposite direction ( Figure 4B ) . Using RT-qPCR , we validated several of the uaRNAs ( Figure 4C ) . uaRNAs associated with the ARGLU1 , METTL14 , SH3BP5 and WEE1 genes were upregulated about twofold in two independent 25F+Y cell lines , consistent with results from RNA-seq ( Figure 4—figure supplement 3A ) . Levels of RPLP1- and CCNB2-associated uaRNAs were indistinguishable in 26r and 25F+Y cells by both methods . 10 . 7554/eLife . 02112 . 009Figure 4 . Tyr1 functions in expression of upstream antisense transcripts . ( A ) Schematic of sense RNA and upstream antisense ( ua ) RNAs analyzed . ( B ) Regulation of uaRNA expression . Cells treated with tet for 24 hr were processed for 3′READS RNA-seq analysis . The number of genes with upregulated ( UP ) and downregulated ( DN ) uaRNAs ( compared to sense RNA expression ) are shown . Their ratio ( UP/DN ) and p-value ( Chi-squared test ) are shown . ( C ) RT-qPCR was performed to measure uaRNA levels associated with select genes detected in the RNA-Seq analysis . Fold relative to 26r cells is plotted . N = 3 . ( D ) Rpb1 levels on sense and antisense genes were determined by ChIP using primers as indicated . The number in parenthesis next to the examined gene indicates the distance between the amplicon and the TSS ( minus sign denotes upstream ) . N = 3 . Error bars display standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 00910 . 7554/eLife . 02112 . 010Figure 4—figure supplement 1 . RNA-Seq analysis of Rpb1 cell lines . RNA from cells treated with tet for 24 hr were processed for deep sequencing as described . S2A cells express an Rpb1 derivative with 26 YAPTSPS repeats , whereas S5A cells express an Rpb1 with 28 YSPTAPS repeats . The number of reads mapped to polyA sites for each cell type are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 01010 . 7554/eLife . 02112 . 011Figure 4—figure supplement 2 . uaRNA polyA site analysis . ( A ) The distribution of uaRNA polyA sites . The number of polyA sites in generated reads was counted and plotted against their distance from TSS . ( B ) The nucleotide profiles of polyA sites in ( A ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 01110 . 7554/eLife . 02112 . 012Figure 4—figure supplement 3 . uaRNA expression analysis in 25F+Y and 26r cells . ( A ) Expression difference of uaRNAs vs sense strand RNAs for 25F+Y and 26r cells . Each dot is a gene with uaRNA expression detected . Genes with significant difference in expression of uaRNAs vs sense strand RNAs ( p<0 . 05 , Fisher's exact test ) were highlighted . The five genes validated in Figure 4C were marked in the plot . ( B ) Expression of antisense poly ( A ) + RNA near the transcription start site ( TSS ) in 26r and 25F+Y cells . Reads per million ( RPM ) per base for poly ( A ) sites were shown ( y-axis ) . All genes with upstream antisense ( ua ) RNAs detected in either 26r or 25F+Y cells were used for plotting . The curves were smoothened by the ‘lowess’ function . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 01210 . 7554/eLife . 02112 . 013Figure 4—figure supplement 4 . Exosome subunit levels . Cells were treated with tet for 24 hr . The levels of exosome subunits Exosc 3 ( Rrp40 ) , Exosc 9 ( Rrp45 ) , Exosc 10 ( Rrp6 ) , and Dis3 were determined by western blotting . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 01310 . 7554/eLife . 02112 . 014Figure 4—figure supplement 5 . Expression changes for different types of transcripts in 25F+Y cells compared to 26r cells . Values are shown in cumulative distribution function ( CDF ) curves . Transcript type is indicated in the graph . Expression change was based on log2 ratio of the Read Per Million total PASS reads ( RPM ) value . We used only genes that had at least 20 reads in two samples combined for this analysis . Expression of uaRNAs ( purple line ) is significantly upregulated compared to sense transcripts ( other lines ) . Genes with upregulated uaRNAs ( red line ) tend to be slightly downregulated as compared to genes with non-regulated uaRNAs ( blue line ) or no detectable uaRNAs ( gray line ) . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 01410 . 7554/eLife . 02112 . 015Figure 4—figure supplement 6 . Tyr1-P ChIP analysis . Data from Figure 4D were reanalyzed by normalizing the Tyr1-P signals to Rpb1 levels . N = 3 . Error bars display standard deviation . DOI: http://dx . doi . org/10 . 7554/eLife . 02112 . 015 We next investigated the mechanism underling the enhanced accumulation of uaRNAs in 25F+Y cells . uaRNAs are low-abundant , usually rapidly degraded by the nuclear exosome ( Preker et al . , 2008; Seila et al . , 2009; Wei et al . , 2011; Ntini et al . , 2013 ) . However , protein levels of four exosome subunits were comparable in 26r and 25F+Y cells ( Figure 4—figure supplement 4 ) , suggesting that the increase in uaRNAs was unlikely due to decreased exosome levels . Also , poly ( A ) sites of uaRNAs were unchanged ( Figure 4—figure supplement 3B ) , indicating that enhanced accumulation did not reflect altered poly ( A ) site utilization . Another possibility was that transcription of these transcripts was increased . However , ChIP assays indicated that Rpb1 levels were in fact reduced upstream of the ARGLU1 , METTL14 , SH3BP5 and WEE1 genes in 25F+Y cells ( Figure 4D; see also Figure 4—figure supplement 5 ) . Finally , ChIP analyses showed more Tyr1-P on these upstream genes than on the corresponding downstream sense genes ( Figure 4—figure supplement 6 ) . Our results point to a role for Tyr1-P in regulating accumulation of uaRNAs by contributing to their rapid turnover . In this study , we described two important functions for Tyr1 residues: Protecting the CTD from proteolysis and ensuring turnover of uaRNAs . Both these functions are likely conserved throughout vertebrates , as Descostes et al . ( Descostes et al . , 2014 ) report remarkably similar results in human cells . While CTD stabilization requires only a single Tyr1 residue , and Tyr1 phosphorylation , more Tyr1 residues are required for other essential functions . As shown previously ( Baskaran et al . , 1993 , 1999 ) , c-Abl can phosphorylate the CTD in vitro , which stabilizes it from proteasomal degradation . But the role of c-Abl in Tyr1 phosphorylation in vivo is unclear , as c-Abl inhibitors had at most modest effects on Try1-P levels in cells ( unpublished data ) . In any event , our results add additional and unexpected complexity to the multiple functions played by the CTD in controlling RNAP II activity .
DT40 cells and HEK293 cells were cultured at 37°C with 5% CO2 in RPMI1640 medium containing 10% FBS and 1% chicken serum , and in DMEM medium containing 10% FBS , respectively . Rpb1 CTD derivatives were cloned as previously described ( Hsin et al . , 2011 ) . Briefly , a fragment of beta-actin promoter and FLAG tag was inserted into pBlueScript containing Neomycin resistance gene . The human Rpb1 body without the CTD was inserted immediately after the FLAG tag , and various CTD fragments were inserted directly 3′ to the Rpb1 body . Procedures for complementation assays and for constructing stable cell clines were followed as previously described ( Hsin et al . , 2011 ) . Briefly , 107 cells were transfected with linearized DNA , and selected in the presence of appropriate antibiotics . Surviving cell clones were isolated , and the identity of these cells was further confirmed using western blotting . Cells lysates , prepared by dissolving washed cells directly in SDS sample buffer , were resolved by SDS-PAGE with indicated percentage of acrylamide . Western blotting was performed using standard protocols . For quantification , western blots were analyzed by ImageJ . Antibodies used in this paper as follows: Flag tag ( M2; Sigma , St . Louis , MO ) , actin ( Sigma ) , phospho CTD Tyr1 ( 3D12; Active Motif , Carlsbad , CA ) , U2AF65 ( Sigma ) , histone H3 protein ( abcam , Cambridge , MA ) , phospho CTD Ser2 ( 3E10; Millipore , Billerica , MA ) , phospho CTD Ser5 ( 3E8; Millipore ) , phospho CTD Ser 7 ( 4E12 , Millipore ) , Rpb1 CTD ( 8WG16; abcam ) , GST tag ( Invitrogen , Carlsbad , CA ) , Rpb1 ( N20; Santa Cruz , Santa Cruz , CA ) , Exosc10 ( Rrp6 ) ( Novus , Littleton , CO ) , Exosc9 ( Rrp45 ) ( Novus ) , Exosc3 ( Rrp40 ) ( Novus ) , and Dis3 ( Novus ) . Subcellular fractionation was performed using a modified protocol as described ( Mapendano et al . , 2010 ) . Briefly , cells ( 1 ∼ 2 × 107 ) were harvested , washed in PBS , and resuspended in 0 . 5 ml of RSB100 ( 50 mM Tris–HCl PH 7 . 4 , 100 mM NaCl ) containing 40 µg/ml digitonin . Cell extracts were incubated on ice for 5 min . The cytoplasmic fraction was separated from nuclear fraction by centrifugation ( 2000×g , 5 min ) . The pellets were resuspended in 0 . 5 ml of RSB100 containing 0 . 5% Triton X-100 , and the reactions were incubated on ice for 5 min . Separation of soluble nuclear proteins from insoluble chromatin bound proteins was carried by centrifugation ( 2000×g , 5 min ) . The pellets containing chromatin bound proteins were resuspended in 0 . 5 ml of RSB100 ( 0 . 5% Triton-X100 ) , and sonicated briefly . In vitro proteasome assays were performed as described ( Asher et al . , 2005 ) with the following modifications . Briefly , GST-CTD or GST-CTD phosphorylated by abl tyrosine kinase was incubated with 2–10 nM bovine 20S proteasome ( UBPBio , Aurora , CO ) in a buffer ( 50 mM Tris–HCl PH 7 . 4 , 100 mM NaCl , 0 . 5 mM EDTA ) at 37°C for 1 hr . Reactions were stopped by adding equal volume of 2X SDS PAGE sampling buffer . CDK7 complexes were expressed in insect cells , and purified using Ni-NTA agarose ( QIAGEN , Valencia , CA ) as described ( Larochelle et al . , 2006 ) . GST-CTD was expressed in E . coli and purified using glutathione Sepharose 4B ( GE Healthcare ) . Phosphorylation of GST-CTD by CDK7 complexes was carried out at 30°C for 1 hr in a kinase buffer ( 25 mM Hepes PH 7 . 5 , 10 mM MgCl2 , 150 mM NaCl , 1 mM ATP ) . GST-CTD phosphorylation by recombinant c-Abl kinase ( NEB , Ipswich , MA ) was performed as described in manual . Briefly , 500 nM GST-CTD was incubated with 25 U c-Abl at 30°C for 2 hr . Phosphorylated GST-CTD was purified using glutathione Sepharose 4B ( GE Healthcare , Pittsburgh , PA ) . RNA was extracted using Trizol ( Invitrogen ) , and further treated with DNase I . Reverse transcription and qPCR analysis were performed as previously described ( Hsin et al . , 2011 ) . Primer sequences are listed in Supplementary file 1 . About 2 × 107 cells were collected , and washed with PBS . Then , 1 ml cold RIPA ( 150 mM NaCl , 1 mM EDTA , 50 mM Tris–HCl pH 7 . 4 , 0 . 5% NP-40 , 0 . 25% sodium deoxycholate ) buffer containing 1X PhosSTOP ( Roche , Madison , WI ) , and 1X protease inhibitors ( 1 . 4 µg/ml Pepstatin A , 0 . 35 µg/ml Leupeptin , and 1 . 7 µg/ml Aprotinin ) . After brief sonication , debris was centrifuged at 12 , 000×g , 4°C , for 10 min , and the supernatant was removed to a new tube . 50 µl of the lysate were kept for input control , and the rest of the extract was incubated with 20 µl of pre-washed protein G Sepharose and 1–4 µg of antibody . Samples were rotated at 4°C for 1–2 hr , and beads were washed with cold RIPA buffer for 3 min three times , and then were resuspended in 100 ul of 1X SDS sample buffer for western blotting . Cells were grown to 70% confluence ( ∼2 × 106/ml ) , cross-linked with 1% formaldehyde for 10 min , and processed for ChIP as previously described ( Hsin et al . , 2011 ) . ChIP was performed using antibody against Flag tag ( M2; Sigma ) , phospho CTD Tyr1 ( 3D12; Active Motif ) . Primers sequences are listed in Supplementary file 1 . Total RNA was extracted from cells treated with tet for 24 hr . RNA was further processed , and analyzed by 3′READs , a deep sequencing method to analyze poly ( A ) + RNAs using 3′ end regions , as described ( Hoque et al . , 2013 ) with some modifications for RNA fragmentation . Briefly , poly ( A ) RNA was selected using oligo d ( T ) 25 magenetic beads ( NEB ) , followed by fragmentation of RNA on-bead using RNaseIII ( NEB ) . We generated two libraries for each cell type ( biological replicates ) , and ∼4 million reads for each library . cDNA insert size range corresponded to RNA fragments of ∼100–200 nt . The reads were mapped to the chicken genome ( version galGal4 ) , and those with at least two non-genomic As at the 3′ end were considered as poly ( A ) site-supporting ( PASS ) reads . PASS reads were assigned to protein coding genes defined by Refseq . The 3′ end of each gene was extended by 4 kb if there was no gene on the same strand within this region . The PASS reads mapped to genic regions are called sense strand reads . Those mapped to the 2 kb upstream region of transcription start site ( TSS ) on the antisense strand were called upstream antisense ( ua ) RNA reads . We also required that uaRNA reads could not be assigned to any other annotated genes as sense strand reads . To examine expression change of uaRNAs vs sense RNAs , we grouped all uaRNA reads and compared them to all sense reads of a gene between two samples , for example , 25F+Y vs 26r . Genes with p-value <0 . 05 ( Fisher′s exact test ) were selected . | When a gene is expressed , the DNA is first transcribed to produce an intermediate molecule called a messenger RNA ( mRNA ) , which is then translated to produce a protein . RNA Polymerase II is an enzyme that makes mRNA molecules in organisms as diverse as plants , animals and yeast . RNA Polymerase II is a complex made of a number of proteins . The largest protein in this complex includes a ‘carboxy-terminal domain’ that has multiple repeats of seven amino acids one after the other . The first amino acid in each repeat , a tyrosine , is referred to as tyrosine-1 . Adding various chemical tags to the amino acids in these repeats co-ordinates the steps involved in the transcription of genes . In yeast , for example , adding a phosphate groups to tyrosine-1 seems to help the polymerase to proceed to make long mRNA molecules . However , it is not known what these chemical tags do in humans or other animals . Now Hsin et al . ( and independently Descostes , Heidemann et al . ) have shown that the same phosphate groups on tyrosine-1 perform functions in vertebrates ( animals with backbones ) that are different to those performed in yeast . These functions include protecting the carboxy-terminal domain from being broken down inside cells , and transcribing the DNA that is upstream of genes . Hsin et al . replaced tyrosine-1 in RNA Polymerase II from chicken cells with a related amino acid that cannot have phosphate groups added to it . This mutant RNA Polymerase II was unstable and degraded by the molecular machinery in cells that breaks down damaged or unneeded proteins back into amino acids . Hsin et al . also compared the mRNA molecules that are made by the wild-type RNA Polymerase II with those produced by a related mutant . This comparison revealed an unexpected accumulation of RNA molecules that are transcribed in the opposite direction from mRNAs . These RNA molecules , known as ‘upstream antisense RNAs’ , have been described only recently . And while the function of these RNAs remains mysterious , the results of Hsin et al . suggest that tyrosine-1 helps to ensure that these RNA molecules are rapidly broken down . The results of Hsin et al . raise a number of important questions , and foremost among these questions is: how do these newly discovered properties of tyrosine-1 contribute to the control of gene expression in animals ? Further work is needed to answer this question . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"cell",
"biology",
"short",
"report",
"biochemistry",
"and",
"chemical",
"biology"
] | 2014 | RNAP II CTD tyrosine 1 performs diverse functions in vertebrate cells |
Fluorescent transcriptional reporters are widely used as signaling reporters and biomarkers to monitor pathway activities and determine cell type identities . However , a large amount of dynamic information is lost due to the long half-life of the fluorescent proteins . To better detect dynamics , fluorescent transcriptional reporters can be destabilized to shorten their half-lives . However , applications of this approach in vivo are limited due to significant reduction of signal intensities . To overcome this limitation , we enhanced translation of a destabilized fluorescent protein and demonstrate the advantages of this approach by characterizing spatio-temporal changes of transcriptional activities in Drosophila . In addition , by combining a fast-folding destabilized fluorescent protein and a slow-folding long-lived fluorescent protein , we generated a dual-color transcriptional timer that provides spatio-temporal information about signaling pathway activities . Finally , we demonstrate the use of this transcriptional timer to identify new genes with dynamic expression patterns .
Changes in gene expression are one of the key mechanisms that organisms use during both development and homeostasis . Gene expression is a highly dynamic process , which not only bears critical information about regulatory mechanisms but also controls the fate of many biological processes ( Purvis and Lahav , 2013; Yosef and Regev , 2011 ) . For example , oscillatory or constant expression of the Notch effector Hes1 dictates the choice of neuron stem cells between proliferation and differentiation ( Isomura and Kageyama , 2014 ) . In addition , defining the exact ‘on’ and ‘off’ timing of a relevant signal is vital to control different developmental events ( Doupé and Perrimon , 2014 ) . For example , during the development of fly compound eyes , simultaneous activation of EGF and Notch signals determines a cone cell fate ( Flores et al . , 2000 ) , while cells that experience sequential expression of EGF and the Notch-ligand Delta differentiate into photoreceptor cells ( Tsuda et al . , 2002 ) . Documenting precisely the spatio-temporal changes in gene expression that occur in response to intrinsic and extrinsic signals is a challenging problem in cell and developmental biology . Traditionally , transcriptional reporters that drive expression of fluorescent proteins ( FPs ) under the control of signaling response elements ( SREs ) have been widely used to visualize the activities of transcriptional events; however , the slow degradation ( half-life >20 hr ) of FPs makes it hard to achieve the temporal resolution needed to dissect the dynamic nature of gene expression . Recently , this problem has been addressed by the application of a fluorescent timer , a slow maturing fluorescent protein that changes its from blue to red in ~7 hr ( Bending et al . , 2018 ) . Despite the still relatively long conversion time , this fluorescent timer has two additional limitations: the signal is hard to fix for long-term storage , and because it can be photoconverted from blue to red , this timer only allows one single image and prohibits live-imaging application ( Subach et al . , 2009 ) . Another strategy is the development of a destabilized version of GFP with a half-life of ~2 hr , which is achieved by fusing GFP with a PEST peptide signal for protein degradation ( Li et al . , 1998; Rogers et al . , 1986 ) . However , despite many in vitro successes , this strategy has met a major limitation when applied in vivo due to substantial loss of fluorescent intensity . Therefore , regular stable FPs are still the primary choice for generating transcriptional reporters to study gene expression patterns in vivo . Here , we address this problem by using translational enhancers to boost production of the destabilized reporters and demonstrate the advantages of using short-lived FPs to study dynamic gene expression in vivo . In addition , we generate a transcriptional timer that can be readily applied to study spatio-temporal activation of signaling pathways . Finally , we document how this transcriptional timer can be used , either using the UAS/Gal4 system or in an enhancer trap screen , to identify genes with dynamic expression .
Matching reporter dynamics with the activity of target genes is essential to faithfully recapitulate signaling activities ( Doupé and Perrimon , 2014 ) . Two primary kinetic properties dictate reporter activities: the ‘switch-on’ and ‘switch-off’ speeds . The ‘on’ kinetic of FPs have been improved by engineering fast folding FPs , which shorten the maturation time of FPs from more than 1 hr to less than 10min ( Pédelacq et al . , 2006 ) ( Gordon et al . , 2007 ) . The ‘off’ kinetics of FPs has been improved from more than 20 hr to around 2 hr by fusing FPs with PEST peptides that promote degradation ( Li et al . , 1998 ) ( Figure 1a ) . Although the advantages of using a short-lived reporter have been previously reported ( Li et al . , 1998 ) , a systematic analysis of the differences between long-lived and short-lived reporters is still lacking . Using a protein synthesis and degradation model ( Figure 1—figure supplement 1a–c ) , we first simulated the dynamics of the reporters and demonstrated a significant improvement by decreasing the half-life ( Tp1/2 ) of the reporters from 20 hr to 2 hr ( Figure 1b–f ) . Specifically , we illustrate this problem using simulated responses of FP reporters to four basic types of promoter activities: switch on , switch off , pulse activation , and oscillation . Compared to FPs with a half-life of 20 hr , FPs with a half-life of 2 hr have a 90% shorter response time ( time to achieve 50% maximal intensity ) during the ‘on’ or ‘off’ events , and up to four times larger dynamic range in the case of oscillatory expression ( Figure 1e , f , Figure 1—figure supplement 1a–c ) . Although the destabilization strategy successfully leads to an FP with a shorter half-life , it is problematic as it causes significant loss of the signal ( Figure 1g–i ) . Thus , at constant expression , the intensity of the maximum signal is linearly proportional to the protein half-life , and decreasing the half-life from 20 hr to 2 hr causes a 90% signal loss ( Figure 1g , black curve ) . The reduction of maximum signal intensity is also affected by different types of transcriptional activation . In the case of short pulsatile activation , a more transient activation ( a shorter duration Td ) is less sensitive to a reduction in the protein half-life ( Figure 1g colored curves ) ; however , transient activation also triggers weaker reporter activity , which makes it more vulnerable to intensity reduction . We further analyzed the effects of reducing mRNA half-life ( Tm1/2 ) compared to protein destabilization ( Figure 1h , i ) . Measurements of Tm1/2 of FPs in the literature are highly variable , ranging from several minutes to hours ( Baker and Parker , 2006; Sacchetti et al . , 2001; Houser et al . , 2012 ) , which is probably due to differences in the 3’ UTR used or different mRNA expression levels relative to the mRNA degradation machinery . According to our measurements ( Figure 2—figure supplement 1e ) , the Tm1/2 of FP reporters is about 0 . 5 hr . Therefore , we used 0 . 5 hr to 3 hr in our modeling . According to the model , the mRNA half-life significantly influences reporter intensity ( Figure 1h ) , which is approximately proportional to the Tp1/2* Tm1/2 . In contrast , reducing mRNA half-life has much less effect on reporter response time , which is mainly controlled by Tp1/2 + Tm1/2 ( Figure 1i ) . Because in our system , Tp1/2 is much longer than Tm1/2 , shortening the mRNA lifetime will significantly reduce signal intensity without endowing the reporter with more range in dynamic detection . Therefore , we decided to primarily use destabilized FPs in our study . For a system with a large Tm1/2 relative to Tp1/2 , strategies to shorten mRNA lifetime by addition of RNA destabilizing sequence can be used ( Voon et al . , 2005 ) . The significant loss of signal limits implementation of the destabilization strategy , especially in systems like Drosophila where a transgene is usually present in 1 or 2 copies per genome . To overcome this obstacle , we searched for ways to increase the signal of destabilized FPs . One possible solution is to use FPs with high intrinsic brightness . To test this , we used a fly codon-optimized sfGFP for its fast folding and bright fluorescence ( Pédelacq et al . , 2006; Venken et al . , 2011 ) . The effectiveness of destabilization was first tested in cultured fly S2 cells . Adding the PEST sequence from mouse ornithine decarboxylase ( MODC ) effectively reduced the half-life of sfGFP to ~3 hr ( Figure 2—figure supplement 1 ) . Next , to test the approach in vivo , we generated transgenic flies with destabilized sfGFP ( dGFP ) for two widely used signaling reporters: STAT ( containing the STAT response element from Socs36E; Bach et al . , 2007 ) and Notch ( containing the Notch response element Su ( H ) Gbe; Furriols and Bray , 2001 ) . GFP signals were examined in tissues previously reported to show high STAT and Notch activities ( embryo for STAT and wing imaginal disc for Notch ) . Destabilization reduced the signal intensity of these reporters to near background ( Figure 2a–c ) . As further increasing the intrinsic brightness of FPs is challenging—even with the brightest FPs currently available the increase in signal intensity is still limited ( less than two fold; Cranfill et al . , 2016 ) —we decided to test other strategies to increase the FP signal . One potential solution is to increase expression of the FP by expressing multiple tandem FPs ( Shearin et al . , 2014; Genové et al . , 2005 ) . However , this strategy makes cloning cumbersome and may render the insertion unstable due to recombination . Thus , we instead decided to use translational enhancing elements that have been demonstrated to increase protein production from mRNA by up to 20 fold ( Pfeiffer et al . , 2012; Pfeiffer et al . , 2010 ) . These elements include a short 87 bp intervening sequence ( IVS ) from myosin heavy chain to facilitate mRNA export to the cytoplasm ( Pfeiffer et al . , 2010 ) , a synthetic AT-rich 21 bp sequence ( Syn21 ) to promote translational initiation ( Pfeiffer et al . , 2012; Suzuki et al . , 2006 ) , and a highly-efficient p10 polyadenylation ( polyA ) signal from baculovirus ( van Oers et al . , 1999 ) . To test if these elements can be used to increase the reporter signals , we inserted the translational enhancing elements into reporter constructs containing dGFP ( destabilized sfGFP ) ( Figure 2a–c ) . Transgenic flies were generated by phiC31-mediated site-directed integration into the same genomic locus ( attP40 ) to avoid potential position effect ( Groth et al . , 2004 ) . Strikingly , the addition of translational enhancers successfully increased the reporter signal with an expression pattern similar to that of previously reported stable reporters ( Figure 2a–c ) ( Furriols and Bray , 2001; Rodrigues et al . , 2012 ) . We further measured the in vivo half-life of the dGFP in live tissues by blocking transcription with Actinomycin ( 10 μM ) and monitored degradation of the GFP ( Figure 2d ) . The result shows a reporter half-life ( TGFP1/2 ~2 . 6 ± 0 . 3 hr ) similar to what was observed in cultured cells ( ~3 . 7 ± 0 . 7 hr ) ( Figure 2d-f; Figure 2—figure supplement 1e ) . We also tested the absolute signal intensity of regular GFP with the translational enhancer ( eGFP ) and destabilized GFP with the same enhancer ( edGFP ) . The total signal intensity from edGFP is about 8% of what is observed for eGFP , consistent with a 91% reduction in protein half-life . The effective increase of FP signal allows us to directly evaluate the activity of short-lived and long-lived traditional reporters . To achieve this , we generated a STAT::eRFP reporter with the same translational enhancing elements and inserted it into the same locus ( attP40 ) ( Figure 3—figure supplement 1a ) . The activities of STAT::edGFP and STAT::eRFP reporters ( as transheterozygotes ) in developing embryos were examined under live imaging conditions ( Figure 3a ) . Compared to the stable RFP , the destabilized reporter showed a transient increase in STAT activity in tracheal pits ( Tp ) , pharynx ( Pr ) , proventriculus ( Pv ) , posterior spiracles ( Ps ) , and hindgut ( Hg ) ( Figure 3a , Figure 3—figure supplement 1b , Videos 1 and 2 ) ( Johansen et al . , 2003 ) . To quantify the temporal changes of dGFP and RFP , the total fluorescent signals of both reporters were measured over time at the indicated region ( arrowhead in Figure 3a ) ; the dGFP signal shows a definite improvement in response time ( Figure 3b ) . Using the half-lives estimated from the in vivo measurement ( 2 . 1 hr for dGFP and 18 . 5 hr for RFP ) ( Figure 3—figure supplement 1c–e ) and the reporter synthesis and degradation model ( Figure 1—figure supplement 1 ) , we further estimate the actual transcriptional activity of the reporter , which happens at an earlier stage ( ~2 hr in advance of detectable dGFP reporter activity ) ( Figure 3b ) . This result is consistent with the previously observed temporal expression of the STAT ligand upd from stage 9 to 12 ( Johnson et al . , 2011 ) . We also notice that the degradation reaction deviated from first-order kinetics at a high concentration of FP ( Figure 3—figure supplement 1e ) . Meanwhile , the degradation speed of the FPs , which depends on the availability of relevant enzymes , may also vary under different conditions . Therefore , the direct interpretation of the difference between dGFP and RFP is the relative rather than absolute ratio . The absolute relationship between the reporters to the actual transcripts should be determined experimentally on a case by case basis . Short-lived reporters show clear advantages in revealing expression dynamics in live tissues . However , not all tissues are amenable to live imaging . Further , live imaging experiments are time-consuming and hard to adapt for large-scale studies . From our previous analyses , we noticed that dynamic information , including the initiation , maintenance , and reduction of transcriptional activities , can be directly estimated by comparing the ratio of fluorescence from stable vs destabilized reporters from a still image . Because the GFP matures faster than RFP ( maturation time ~0 . 1 hr for sfGFP [Khmelinskii et al . , 2012] vs . ~1 . 5 hr for TagRFP [Merzlyak et al . , 2007] used in this study ) , when the promoter switches on , a green signal is detected first . As the promoter remains active , both GFP and RFP signals reach a balance and such that their overlay produces a yellow color ( when the max intensities of GFP and RFP are normalized ) . Furthermore , when transcription switches off , because the GFP signal decreases more quickly ( the half-life of dGFP is ~2 hr , and the half-life of RFP is ~20 hr ) , only the RFP signal is left ( Figure 4a ) . Previously , ratiometric imaging of FPs based on their different properties , such as maturation time and sensitivity to specific ions , has been used successfully to measure protein life-time ( Khmelinskii et al . , 2012 ) , pH changes ( Gjetting et al . , 2012; Zhou et al . , 2012 ) , and Mg2+ concentration ( Koldenkova et al . , 2015 ) . We reason that dynamic transcriptional activities might also be detected similarly using the ratio between dGFP and RFP . To test this strategy in vivo in a multicellular tissue , we examined STAT activity in the larval optic lobe . During development , STAT activity has been shown to act as a negative signal that antagonizes progression of the cell differentiation wave , which triggers the transition of neuroepithelial cells ( NEs ) into neuroblasts ( NBs ) ( Yasugi et al . , 2008 ) . A stable STAT reporter shows the spreading of STAT activity within the neuroepithelial region , similar to what we observe with STAT::eRFP alone . In contrast , the signal from STAT::edGFP together with STAT::eRFP revealed a clear ‘green front’ and ‘red rear’ in the same region ( Figure 4b ) . This dynamic pattern , with a higher STAT activity at the boundary between NEs and NBs , is consistent with previous proposed wave-like STAT activity that propagates from the lateral to medial region ( Yasugi et al . , 2008 ) . In addition , analysis of STAT::edGFP and STAT::eRFP in fixed larval optic lobes at different developmental stages further support this wave-like propagation model ( Figure 4—figure supplement 1a ) . Another example of the dynamic information revealed by combining edGFP and eRFP is the expression of the Notch reporter in larval brain NBs . NBs undergo an asymmetric cell division to generate smaller progeny ( Figure 4c ) . Previous studies have shown that the Notch suppressor complex PON/Numb is preferentially localized to progeny cells ( Suzuki et al . , 2006; van Oers et al . , 1999 ) and that ectopic activation of Notch generates a brain tumor phenotype attributable to excess NBs , suggesting that Notch activity is required for NB self-renewal ( Wang et al . , 2006 ) . Strikingly , whereas the stable Notch reporter accumulates in both NBs and their progeny , the destabilized reporter is preferentially expressed in NBs ( Figure 4c , Figure 4—figure supplement 1b ) , consistent with functional studies ( Wang et al . , 2006 ) . Combining the information from the edGFP and eRFP reporters reveals a clear Notch activity gradient that decreases as NBs differentiate . Importantly , in addition to its superior spatial resolution , the destabilized reporter also shows improved temporal resolution in NBs under live imaging conditions ( Figure 4—figure supplement 2 ) . Our data suggest that combining edGFP and RFP creates a useful tool to study transcriptional dynamics , even in fixed samples . To facilitate this , we generated a multicistronic reporter containing both edGFP and RFP connected by the ‘self-cleaving’ 2A peptide ( Szymczak and Vignali , 2005 ) , edGFP-2A-RFP , which we refer to as a transcriptional timer or ‘TransTimer’ ( Figure 4—figure supplement 3a ) . Larval optic lobes expressing the transcriptional timer controlled by STAT response element show similar expression pattern as transheterozygous STAT::edGFP and STAT::eRFP , indicating that the multicistronic system is effective ( Figure 4—figure supplement 3b ) . Destabilization of dGFP depends on protein degradation . Potential changes in the degradation speed of dGFP could affect the intensity of dGFP , which might distort our estimation of the real transcriptional activities . To test this possibility , we generated a TransTimer under the control of the fly Ubiquitin promoter ( a constitutively active promoter in most fly tissues ) . Ubi::edGFP-2A-RFP shows no significant variation in the green and red ratios in fly embryos or the larval brain ( Figure 4—figure supplement 3c ) . In addition , under control of other constitutive promoters , TransTimer also shows a relatively stable ratio between the two colors in different tissues ( Supplementary file 2 ) , suggesting that changes in the FP ratios observed with TransTimer are primarily due to changes in transcriptional activity in the tested tissues , not cell type or tissue-specific differences in protein degradation . However , for specific organ or developmental stage , a control with a constitutive promoter for protein degradation changes is advisory . Creating a TransTimer reporter for new signaling or target genes requires cloning of different signal response elements . Meanwhile , several large transgenic collections ( up to several thousands ) of Gal4 lines under the control of enhancers of different genes have been generated ( Brand and Perrimon , 1993; Jenett et al . , 2012; Lee et al . , 2018 ) . A TransTimer controlled by UAS would provide a quick way to test expression dynamics of Gal4 lines using a simple genetic cross ( Figure 5a ) . Thus , we generated UAS::TransTimer transgenic flies and tested the UAS-controlled version in the adult Drosophila gut . TransTimer under control of esg-Gal4 , a fly intestinal stem cells ( ISCs ) and enteroblasts ( EBs ) driver ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2007 ) , revealed particular cells within the stem cell group that turned ‘red , ’ distinguishing them from other ‘yellow’ stem cells ( Figure 5b ) . Further analysis of these ‘red’ cells showed that they down-regulate esg expression and up-regulate the enteroendocrine cell marker Prospero ( Pros ) , suggesting that they are differentiating ( Figure 5—figure supplement 1a ) . In addition , the esg-controlled TransTimer shows significantly more dynamics in the developing intestine and regenerating intestine following Bleomycin treatment ( a DNA damaging agent ) ( Amcheslavsky et al . , 2009 ) , consistent with the different level of activity of stem cells under these conditions ( Figure 5b , Figure 5—figure supplement 1b ) . Next , we used TransTimer to examine the cell heterogeneity of different intestine tumors . In ISC tumors induced by knocking down O-fut1 , an enzyme required for Notch maturation ( Micchelli and Perrimon , 2006; Ohlstein and Spradling , 2007 ) , the regular GFP reporter showed only a moderate variation of the stem cell marker esg . By contrast , TransTimer revealed an evident decrease of dGFP compared to RFP in ~70–60% of the cells in the cluster , suggesting that a substantial heterogeneity in the tumor is caused by down-regulation of esg over time ( Meacham and Morrison , 2013 ) ( Figure 5b ) . Tumors generated by knocking down either Delta , the ligand of Notch receptor , or Domeless ( Dome ) , the transmembrane receptor of JAK/STAT signaling pathway , grow into a similar multilayered cell cluster ( Figure 5c ) ( Jiang et al . , 2009 ) . Interestingly , compared to Dl mutant tumors , where all multilayered cells maintain constant levels of esg ( dGFP/RFP ratio ) , the inner layer of Dome mutant tumors shows clear reduction of esg expression ( lower dGFP/RFP ratio ) relative to the basal layer . This result suggests that Dome mutant tumors , unlike Dl mutant tumors , require direct contact with the basal membrane to keep their stemness ( Figure 5d ) . As demonstrated above , UAS-TransTimer is an effective tool to discover expression changes when crossed with a Gal4 driver of interest . To further test the power of this approach to discover new genes with interesting expression dynamics , we screened ~450 Gal4 lines using UAS-TransTimer ( Marianes and Spradling , 2013 ) . 37 lines ( ~8% ) showed clear dynamic activities ( substantial variation in dGFP/RFP ratio ) in either larval brain , imaginal disc , or adult intestine ( Figure 6a , Supplementary file 1 ) , whereas the remaining Gal4s showed essentially uniform dGFP/RFP ratios , suggesting stable expression ( images of representative control Gal4 lines are shown in Supplementary file 2 ) . Among the genes with dynamic expression patterns , we discovered the mechanosensitive channel Piezo , which is expressed in the posterior midgut specifically in EE precursor cells ( He et al . , 2018 ) . TransTimer driven by Piezo-Gal4 displays a spatially dynamic expression pattern ( separation between the ‘green’ and ‘red’ signals ) ( Figure 6b ) . In addition , the ‘red’ cells , which down-regulate Piezo expression , are positive for the EE cell marker Pros , consistent with the results of our previous study showing that Piezo + cells differentiate into EE cells ( Figure 6c ) . In addition to Piezo , we also identified new uncharacterized genes with dynamic expression patterns in a subpopulation of esg+ cells ( Figure 6d , Figure 6e ) . Further studies of these genes will be required to determine whether they are markers of partially differentiated cells like Piezo or if their expression levels oscillate in the stem cells . Although UAS-TransTimer is a useful tool to explore the hidden transcriptional dynamics using the Gal4 collection , the addition of Gal4 as an intermediate also alters the signal property of TransTimer . According to our model , the presence of Gal4 has two major effects: first , as an amplification mechanism , it can enhance the output signal , which may be advantageous for some weak enhancer ( Figure 6—figure supplement 1a , b , f ) ; second , it delays the dynamic of the reporter , as the new half-life of TransTimer controlled by Gal4 is generally proportional to the sum of TG1/2 ( half-life of Gal4 ) and TFP1/2 ( half-life of FPs ) ( Figure 6—figure supplement 1c–e ) . Because the exact half-life of Gal4 in vivo is unknown , we estimated the effect of Gal4/UAS system by comparing the dGFP and RFP controlled directly by the Notch responding element Su ( H ) Gbe and UAS-TransTimer controlled by Su ( H ) Gbe-Gal4 . Consistent with the predictions of the model , the UAS-TransTimer that is driven by Su ( H ) Gbe-Gal4 shows a similar but considerable broader Notch activation pattern in the third instar larval wing discs than Su ( H ) Gbe-TransTimer , which is probably due to both longer signal retention and stronger signal amplification ( Figure 6—figure supplement 1g ) . In the larval brain , the UAS-TransTimer shows a similar activation gradient in neuroblast cells and their progenies but with significantly stronger retention in daughter cells due to the slow signal reduction ( Figure 6—figure supplement 1h ) . To overcome the delay effect of Gal4 , we decided to control TransTimer directly by endogenous promoters . To achieve this goal , we generated transgenic flies with TransTimer controlled by a minimal synthetic Drosophila promoter that is silent unless activated by nearby enhancers ( Pfeiffer et al . , 2008 ) , and randomly mobilized the transgene in the fly genome to identify endogenous enhancers with dynamic activities ( Figure 7a , Figure 7—figure supplement 1 ) . After screening ~400 independent enhancer trap lines , we identified 46 unique lines that showed fluorescent signals in the larval brain , imaginal disc , or adult intestine . 17 of these 46 lines show clear expression dynamics , suggesting that TransTimer can detect expression changes at endogenous levels ( Figure 7b , Figure 7—figure supplement 1b , Supplementary file 3 ) . To validate the screen , we tested the expression and function of new genes identified in this enhancer trap screen . Since we are particularly interested in new lines that show exclusive expression in stem cells , we chose TransTimer insertions near the promoters of Tsp42Ea , a Tetraspanin protein , and CG30159 , an evolutionarily conserved gene with unknown function - as the function of these genes had yet not been characterized in fly intestine . A Gal4 line ( NP1176-Gal4 ) , located closely ( within 250 bp ) to the TransTimer insertion site at the promoter of Tsp42Ea and CG30159 , also shows specific expression in both larval and adult intestine stem cells ( Figure 7d ) , which is very similar with the expression pattern revealed by TransTimer ( Figure 7b ) . Knocking down CG30159 significantly reduces stem cell numbers , suggesting that CG30159 is required for maintenance of intestinal stem cell ( Figure 7e , f ) . The human homolog of CG30159 is C3orf33 , which has been identified as a regulator of the extracellular signal-regulated kinase ( ERK ) and predicted to be a secreted peptide due to the presence of signal peptide at its N-terminus ( Hao et al . , 2011 ) . Its function in intestinal stem cells requires further investigation . As we have shown above , the enhancer trap screen with TransTimer can effectively detect expression dynamics in vivo . However , this screen can only detect gene expression and cannot be used to manipulate the target cell population . To extend the application of TransTimer , we replaced the RFP with lexA , a yeast transcriptional factor used as a binary expression system together with its binding sequence lexA operator ( lexAop ) ( Yagi et al . , 2010 ) . This dGFP-P2A-lexA construct can not only detect expression dynamics when crossed with lexAop-RFP but also manipulate gene expression in labeled cells in the presence of an additional lexAOP-controlled transgene ( Figure 7g ) . We refer to dGFP-P2A-lexA as ‘TransTimerLex’ . To test the feasibility of this strategy , we generated transgenic flies containing the TransTimerLex insertion and randomly mobilized the element in the fly genome . From our pilot screen ( ~20 independent lines ) , we identified one insertion under control of Larp , a transcriptional factor , which shows clear expression dynamics in the larval intestine ( Figure 7h ) . This result suggests that the TransTimeLex system will be a useful way to both identify new genes and manipulate gene expression in the corresponding cells .
In this study , we described a general and straightforward strategy to use destabilized transcriptional reporters in vivo and demonstrated its power in revealing the spatio-temporal dynamics of gene expression , which is missed by conventional transcriptional reporters . In addition , we generated a dual-color TransTimer that encodes the transcriptional dynamics into a green-to-red color ratio which can be analyzed in fixed tissues . This TransTimer provides a unique opportunity for large-scale screens for in vivo expression dynamics in all types of tissues . Further , our study indicates that TransTimer is effective for the discovery of new genes with interesting expression patterns , either using a candidate gene approach or random genome-wide screening . Our reporter system may also be combined with other techniques such as FACS-seq and signal cell sequencing techniques to provide a time-dependent change of the transcriptome in vivo . In fact , a similar strategy has recently been successfully applied to provide the temporal information for signal cell sequencing of the mouse intestinal stem cell system ( Gehart et al . , 2019 ) . Therefore , we expect that this new method will widely facilitate studies in Drosophila and other organisms .
The following fly lines were obtained from the Bloomington Drosophila Stock Center: Delta2-3 ( 99B ) ( BL3629 ) , Dorothy-Gal4 ( BL6903 ) , Dome-RNAi ( BL32860 ) , CG30159-RNAi ( BL61888 ) , Tsp42Ea-RNAi ( BL39044 ) , fz-Gal4 ( BL66817 ) , Per-Gal4 ( BL7127 ) , Trx-Gal4 ( BL40367 ) , ZnT41F-Gal4 ( BL66859 ) , Ogre-Gal4 ( BL49340 ) , Rho-Gal4 ( BL45254 ) , Gcm-gal4 ( BL35541 ) , igl-Gal4 ( BL76744 ) , dMyc-Gal4 ( BL47844 ) , Hh-Gal4 ( BL49437 ) , Antp-Gal4 ( BL26817 ) , Plc21C-Gal4 ( BL76142 ) , anchor-Gal4 ( BL66861 ) , tutl-Gal4 ( BL66824 ) , Act5C-Gal4 ( BL4414 ) , zip-Gal4 ( BL48187 ) , dMyc-Gal4 ( BL47844 ) , piezo-Gal4 ( BL58771 , BL59266 ) , ppk-Gal4 ( BL32078 , BL32079 ) , Ubi-Gal4 ( BL32551 ) , MESK2-Gal4 ( BL67434 ) , fz-Gal4 ( BL66817 ) , Mip-Gal4 ( BL51984 ) , CG7744-Gal4 ( BL76662 ) , CG4467-Gal4 ( BL66843 ) , CG2082-Gal4 ( BL76181 ) , CG7510-Gal4 ( BL66861 ) , CG10283-Gal4 ( BL76152 ) , CG33964-Gal4 ( BL76742 ) , CG14995-Gal4 ( BL76721 ) , CG5521-Gal4 ( BL76180 ) , CG8177-Gal4 ( BL77781 ) , CG34347 ( BL76674 ) , CG13175-Gal4 ( BL76742 ) , CG8270-Gal4 ( BL77741 ) , CG15270-Gal4 ( BL76649 ) , CG43980-Gal4 ( BL66863 ) , CG40006-Gal4 ( BL ) , CG43980-Gal4 ( BL66863 ) , and dMef2-Gal4 ( BL27390 ) . NP1176-Gal4 was from DGRC ( Kyoto Stock Center ) . Dl-RNAi ( v37287 ) was from Vienna Drosophila Resource Center . GMR-Gal4 , Da-Gal4 , tubGal4 , and esg-Gal4 were from lab stocks . Insc-Gal4 , ase-Gal80ts was a gift from Dr . Dong Yan ( Zhu et al . , 2011 ) , and Dl-Gal4 and Su ( H ) Gbe-Gal4 were from Dr . Steve Hou ( Amcheslavsky et al . , 2014 ) . All flies were maintained on cornmeal-yeast-agar media . Stocks were kept at room temperature with a 12/12 light/dark cycle . Drosophila S2R + cells were grown in Schneider's Drosophila Medium ( SDM ) ( Invitrogen ) containing 10% heat-inactivated fetal bovine serum ( FBS ) at 25°C . Sub-confluent S2R + cells were seeded in 6-well plates and subsequently transfected using Effectene Transfection Reagent ( QIAGEN ) . Cells were cultured for 48 hr before experiments . 10 μM ( final concentration ) Actinomycin D was used to block RNA synthesis , and 100 μg/ml ( final concentration ) cycloheximide was used to block protein synthesis . Cells were treated with the indicated drugs up to 4 . 5 hr before significant cell death was observed . Plasmids expressing pUbi-dGFP-Myc ( 0 . 03 ug ) , and pUbi-RFP-HA ( 0 . 01 ug ) , together with empty plasmids ( to reach a total of 0 . 3 ug of DNA ) were added in each 6-well plate during transfection . The dilution of the expression plasmid was important: we observed that too much protein expression saturates the degradation machinery and prolongs the observed half-life . S2R + cells were harvested by centrifugation and lysed in RIPA buffer . Proteins were separated on a 10% SDS-PAGE gel and analyzed by Western blotting . Quantitative Western blots were performed as previously described ( Eaton et al . , 2014 ) . Images were acquired using a LI-COR Odyssey Classic imager and analyzed using NIH ImageJ . Dynamic reporters were integrated into the fly genome using P-element mediated transformation by injection into w1118 embryos ( Rubin and Spradling , 1982 ) . Transgenic lines were balanced and mapped using w*; Sco/CyO; MKRS/TM6B . Then , six independent lines were generated by crossing with w*;Sp/CyO; Sb , P ( Delta2-3 ) 99B/TM6B , Tb+ ( BL3629 ) . Males from the F1 generation with red eyes and carrying the CyO balancer were further crossed individually with w1118 females . F2 males with red eyes that co-segregate with the CyO balancer were used in the initial screen . Detailed crosses for the enhancer screen are shown in Figure 7—figure supplement 1a . ~400 fly lines were recovered from the F2 generation . Third instar larval brains , imaginal discs , and adult intestines of these flies were dissected and examined for GFP and RFP signal using a Zeiss LSM 780 confocal microscope . P-element insertion sites were mapped by Splinkerette PCR ( Horn et al . , 2007 ) . PCR primers specific for 5 and 3 prime ends of P-elements were used as previously described ( Potter and Luo , 2010 ) . Genomic sequences flanking the P-element insertion sites were recovered and shown in Supplementary file 5 . These sequences were used in BLAST searches against the Drosophila Genome Database . Immunostainings of Drosophila intestines were performed as previously described ( Micchelli and Perrimon , 2006 ) . The following antibodies were used: mouse anti-Prospero ( 1:50 , Developmental Studies Hybridoma Bank ) , mouse anti-HA ( 1:500 , Abcam , ab18181 ) , rabbit anti-Myc tag ( 1/250 , Cell signaling ) , goat anti-mouse IgGs conjugated to Alexa 647 ( 1:500 , Molecular Probes ) , mouse IgGs conjugated to Alexa 488 ( 1:500 , Molecular Probes ) , IRDye 800CW Goat anti-Rabbit IgG ( 1:10 , 000 LI-COR P/N 926–32211 ) , and IRDye 680RD Goat anti-Mouse IgG ( 1:20 , 000 LI-COR P/N 926–68070 ) . Dissected fly tissues were mounted in Vectashield with DAPI ( Vector Laboratories ) . In all micrographs , the blue signal shows the nuclear marker DAPI . Fluorescence micrographs were acquired with a Zeiss LSM 780 confocal microscope . All images were adjusted and assembled in NIH ImageJ . The model used to calculate reporter synthesis , maturation , and degradation was modified from previously described equations ( Wang et al . , 2008 ) with the addition of an equation for mRNA degradation . Briefly , degradation of mRNA and the synthesis rate of premature ( nonfluorescent ) protein ( NP ) is proportional to the mRNA concentration ( R ) . Generation of the mature reporter ( MP ) , modeled as a first-order chemical reaction , only depends on the concentration of NP . Protein degradation is modeled independently of the maturation process . The degradation rates of mRNA and proteins are first modeled based on Michaelis-Menten ( MM ) function ( Figure 1-figure supplement b , equations 1-3 ) , which considers the potential saturation of the degradation machinery . When the substrate concentration is significantly smaller than the Michaelis constant Km , the equations can be simplified with the half-life of the mRNA and protein explicitly displayed ( Figure 1—figure supplement 1b , equations 1’−3’ ) . Dilution by cell division is not included in this model because the fluorescent signal is analyzed in a cell cluster rather than in individual cells , and cell division does not affect the total intensity from the entire cell group and no significant changes in degradation speed have been observed between different cells ( Figure 3—figure supplement 1 ) . With this first-order kinetic model , the transcriptional activity of the promoter F ( x ) can be calculated through equation from the observed fluorescent reporter signal [MP] ( 4 ) . For sfGFP , the maturation time is ~0 . 1 hr , which is much smaller than its protein half-life , such that equation 4 can be further simplified as 4’ . To calculate F ( x ) , the dGFP signal was fitted with a polynomial function ( order = 4 ) to generate the first and second derivatives . Live-cell imaging of developing embryos and dissected larval brains was performed as previously described ( Tomer et al . , 2012; Lerit et al . , 2014; Lemon et al . , 2015 ) . Images were captured on a Zeiss Lightsheet Z1 microscope using a 20X ( N . A . 1 . 0 ) lens . A z-stack of the dual-color image ( 488 nm excitation/500–550 nm detection for GFP , and 561 nm excitation/580–650 nm detection for RFP ) was recorded at 10 min intervals . This interval was chosen empirically to minimize photobleaching without losing temporal information . Photobleaching was measured by continuously imaging of the sample for 50 frames for 10 min and adjusted during image processing . Images of fixed tissue were captured on a Zeiss LSM 780 confocal microscope . Total fluorescent intensity in 3D volume was acquired using Imaris image analysis software ( Bitplane ) . The rest of the analysis was completed using NIH ImageJ with customized macros . Simulation of the model was completed in MATLAB . The Student’s unpaired , two-tailed t-test was used to determine statistical significance between samples . | Fruit flies and other animals have complex body plans containing many different types of cells . To make and maintain these body plans , individual genes must be switched on and off at specific times in particular cells to control how the animal grows . Some of these genes may be switched on for long periods of time , while others may be rapidly switched on and off on repeated occasions . Fluorescent reporter proteins have been extensively used to study gene activity in cells . Typically , this involves linking the gene encoding the fluorescent reporter to a gene of interest , so that when the gene is switched on a fluorescent protein will also be produced . The fluorescent protein emits light of a particular color and measuring this light provides a way to monitor a gene’s activity . Unfortunately , fluorescent proteins tend to break down slowly , and the level of fluorescence emitted cannot fluctuate quickly enough to reflect rapid changes in gene behavior . One way to overcome this limitation is to use destabilized fluorescent proteins that degrade more rapidly inside cells . However , current strategies for creating these proteins cause them to emit less light , making fluorescence more difficult to detect . To address this issue , He et al . developed a new green destabilized protein , adding elements that increase production of the protein so a greater amount of light can be emitted . The green destabilized protein was then combined with a red fluorescent reporter that degrades more slowly to develop a new tool called TransTimer . When the gene linked to the reporter switches on , the green destabilized protein turns on before the red reporter turns on . But , as the gene switches off , the destabilized protein will degrade until only the red signal remains . This allows the ratio of green to red color emitted from the TransTimer to indicate the timing of a gene’s activity . Using this tool , He et al . uncovered new details about the patterns of activity of two signals , known as Notch and STAT , that were largely missed by studies using traditional fluorescent reporters . Further experiments demonstrated that TransTimer can be used to carry out large-scale screens in living fruit flies , which have not been possible with more time consuming live-cell imaging techniques . The fluorescent reporter developed by He et al . will be a useful tool to understand when and where genes are switched on during the lives of fruit flies . In the future , TransTimer could be adapted for use in other model animals or plants . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"tools",
"and",
"resources"
] | 2019 | In vivo study of gene expression with an enhanced dual-color fluorescent transcriptional timer |
The diverse cell types and the precise synaptic connectivity between them are the cardinal features of the nervous system . Little is known about how cell fate diversification is linked to synaptic target choices . Here we investigate how presynaptic neurons select one type of muscles , vm2 , as a synaptic target and form synapses on its dendritic spine-like muscle arms . We found that the Notch-Delta pathway was required to distinguish target from non-target muscles . APX-1/Delta acts in surrounding cells including the non-target vm1 to activate LIN-12/Notch in the target vm2 . LIN-12 functions cell-autonomously to up-regulate the expression of UNC-40/DCC and MADD-2 in vm2 , which in turn function together to promote muscle arm formation and guidance . Ectopic expression of UNC-40/DCC in non-target vm1 muscle is sufficient to induce muscle arm extension from these cells . Therefore , the LIN-12/Notch signaling specifies target selection by selectively up-regulating guidance molecules and forming muscle arms in target cells .
Functional neural circuits are generated through coordinated events during the development of the nervous system including cell type specification , neuronal process formation and synaptogenesis . Many studies have demonstrated that stereotyped wiring exists between different cell types . Insights from studies in spinal cord and neocortex development strongly suggest that a combinatorial code of transcription factors mediates cell specification and defines cellular identities among different cells ( Jessell , 2000; Shirasaki and Pfaff , 2002 ) . An emerging literature indicates that precise synaptic connections are specified by diverse molecular mechanisms . Both positive and negative regulators of synapse formation can specify local synaptic connectivity ( Williams et al . , 2010; Maeder and Shen , 2011 ) . Homotypic and heterotypic adhesion molecules can determine the synaptic lamina formation and even synaptic partner choice ( Yamagata et al . , 2003; de Wit et al . , 2011 ) . For example , in Drosophila , many transmembrane molecules , including semaphorin , capricious and teneurins , have been suggested to guide synaptic target selection between olfactory receptor neurons ( ORN ) and projection neurons ( PN ) ( Komiyama et al . , 2007; Hong et al . , 2009 , 2012 ) . In vertebrate spinal cord , a semaphorin-plexin pathway regulates the connection specificity in the sensory motor circuit ( Pecho-Vrieseling et al . , 2009 ) . In vertebrate retina , retinal ganglion cells ( RGCs ) form synapses with retinal interneurons in the inner plexiform layer ( IPL ) , generating several synaptic laminae ( Sanes and Yamagata , 2009 ) . This laminar specificity is directed by both the homophilic interactions of several immunoglobin superfamily ( IgSF ) proteins , including Sidekicks and Dscam , as well as the inhibitory action of semaphorins ( Yamagata et al . , 2002; Yamagata and Sanes , 2008; Matsuoka et al . , 2011 ) . Similarly , in C . elegans , the heterophilic interaction of two IgSF proteins , SYG-1 and SYG-2 , guides the localization of the en passant synapse formation of HSN neurons ( Shen and Bargmann , 2003; Shen et al . , 2004 ) . Another IgSF protein UNC-40/DCC , the receptor of UNC-6/Netrin , has been shown to regulate axon guidance toward ventral UNC-6 as well as locally promote presynaptic assembly in the C . elegans interneuron AIY ( Hedgecock et al . , 1990; Colon-Ramos et al . , 2007 ) . In addition , UNC-40 plays critical roles in the formation of dendritic spine-like postsynaptic muscle arms of the body wall muscles in worms ( Dixon and Roy , 2005; Alexander et al . , 2009 ) . Intriguingly , this function appears to be independent of UNC-6 . While it is likely that transcription factors ultimately regulate the expression of cell surface molecules to determine the target specificity , few examples are well characterized . In one such example , the even-skipped transcription factor impacts long-range axon guidance choices through regulating a Netrin receptor , UNC-5 ( Labrador et al . , 2005 ) . However , it is largely unknown how cell fate decisions affect local synaptic development and target choices . One of the conserved developmental pathways to generate cellular diversity is through the lateral signaling system involving the Notch receptor and its ligands . Through contact-dependent , reciprocal feedback loops , Notch and its ligand Delta can generate different cell fates among identical neighboring cells ( Louvi and Artavanis-Tsakonas , 2006; Greenwald , 2012 ) . In C . elegans , lin-12 encodes one of the two homologs of Notch receptor ( Greenwald et al . , 1983; Greenwald , 1985; Wharton et al . , 1985 ) . Extensive literature showed that lin-12 is required for at least two cell fate decisions: the AC/VU decision and the vulval precursor cell ( VPC ) specification ( Greenwald , 2005 ) . In both cases , lin-12 and its ligands , including lag-2 , apx-1 and dsl-1 , are required to specify alternative cell fates ( Greenwald et al . , 1983; Seydoux and Greenwald , 1989; Wilkinson et al . , 1994; Shaye and Greenwald , 2002; Chen and Greenwald , 2004 ) . Notch signaling has also been implicated in the development of the nervous system . For example , the mammalian Notch 1 protein is asymmetrically inherited by one daughter cell of an active cortical progenitor during mitosis , which inhibits the neural differentiation but allows this daughter to retain the progenitor fate ( Ishibashi et al . , 1994; Chenn and McConnell , 1995 ) . The accumulation of nuclear Notch in many post-mitotic neurons is also striking ( Ahmad et al . , 1995; Sestan et al . , 1999; Redmond et al . , 2000 ) , prompting the speculation that Notch signaling might also be important for terminally differentiated neurons . Besides functioning in mature nervous system to regulate synaptic specificity in both invertebrates and vertebrates ( de Bivort et al . , 2009; Alberi et al . , 2011 ) , Notch has also been found to act in the post-mitotic neurons in multiple developmental contexts . Both the canonical Notch pathway and a cytosolic pathway have been shown to play important roles in axonal guidance ( Giniger , 1998; Endo et al . , 2007; Song and Giniger , 2011 ) . In the cytoplasmic pathway , Notch functions by regulating the activity of abl , enable and Rac GTPases ( Giniger , 1998; Song et al . , 2010; Song and Giniger , 2011 ) . Studies in Drosophila indicate that activation of Notch signaling lead to promotion or sometimes inhibition of axonal growth . In cultured primary Drosophila neurons , Notch is localized in developing axons and growth cones and interacts with an axonal abl tyrosine kinase to promote axon extension ( Giniger , 1998 ) . An opposite example is the dorsal cluster neurons ( DCN ) of the Drosophila brain ( Hassan et al . , 2000 ) . Reduction of Notch activity results in overbranching of DCN axons . Interestingly , the overbranching phenotype cannot be rescued by expressing wild-type Notch in DCN , indicating its non-autonomous requirement . Similarly , dendritic branching is also sensitive to the level of Notch signaling . Several observations indicate a Notch-mediated , contact-dependent inhibition of dendritic branching ( Berezovska et al . , 1999; Sestan et al . , 1999; Redmond et al . , 2000 ) . In mammalian brain , significant nuclear localization of Notch 1 was observed in differentiated neurons . Perturbation of Notch 1 activity leads to increased dendritic length and decreased number of branches ( Redmond et al . , 2000 ) . Other in vitro studies further indicated that the intercellular receptor–ligand interaction and the canonical Notch signaling pathway are required for the Notch-dependent dendritic branching inhibition ( Berezovska et al . , 1999; Sestan et al . , 1999 ) . It is generally assumed that the canonical pathway regulates post-mitotic phenotypes by controlling the expression of guidance molecules . However , the molecular targets of this pathway have not been identified . With the well-described cell lineage and the stereotyped connectivity between neurons , the nervous system of the nematode C . elegans provides an opportunity to study the molecular links between cell fate specification and relevant synaptic target selection . In the egg-laying circuit , the presynaptic neurons , HSNs and VC4/5 , form synapses exclusively with type 2 vulval muscles ( vm2 ) but not type 1 vulval muscles ( vm1 ) ( White et al . , 1986 ) ( Figure 1A ) . These synapses are not formed on the cell bodies of the muscles but instead onto the dendritic spine-like postsynaptic membrane protrusions called muscle arms ( White et al . , 1986; Leung et al . , 1999; Collins and Koelle , 2012 ) . The HSN presynaptic development occurs before vm2 muscle arm formation and is independent of vulval muscles . Instead , surrounding guidepost cells instruct local HSN presynapse formation with heterophilic interaction of SYG-1 and SYG-2 ( Shen and Bargmann , 2003; Shen et al . , 2004 ) . SYG-2 is expressed in the primary epithelial cells and recruits HSN-expressed SYG-1 to the future presynaptic domain . The development of postsynaptic specializations on the vulval muscles has not been studied . Both the vm1 and vm2 muscle cells are generated from a pair of sex myoblasts ( Foehr and Liu , 2008 ) . While the lineage , cell morphology and positions of both muscle types are very similar , only vm2 cells , not the vm1 cells , generate muscle arms and serve as the direct postsynaptic target of the HSN and VC neurons . vm1 is connected to vm2 through gap junctions ( White et al . , 1986 ) . 10 . 7554/eLife . 00378 . 003Figure 1 . The muscle arms of type 2 vulval muscles ( vm2 ) are missing in lin-12 ( wy750 ) , apx-1 ( wy755 ) and sel-12 ( wy760 ) mutants . ( A ) An illustration showing ventral ( top ) and lateral ( bottom ) views of C . elegans egg-laying circuit . vm1 ( orange ) and vm2 ( red ) are vulval muscles . HSN ( green ) and VC4/5 ( blue ) are presynaptic motoneurons . Each of the four vm2 cells extends a dendritic spine-like muscle arm laterally , on which it receives synapses with HSN and VC neurons ( boxed areas ) . Top row of ( A–C ) are ventral views; bottom row of ( A–C ) are left lateral views . Anterior ( A ) , posterior ( P ) , left ( L ) , right ( R ) , dorsal ( D ) , ventral ( V ) . ( B ) Ventral ( top ) and lateral ( bottom ) views of vulval muscles labeled by SER-1B::GFP transgene in a young adult . HSN neurons are labeled by a mCherry::RAB-3 transgene . The boxed areas indicate synaptic regions . Arrows indicate the HSN cell bodies . Note that the laterally extended muscle arms on vm2 cells are colocalized with the HSN presynaptic specializations . ( C ) Ventral ( top ) and lateral ( bottom ) views of vulval muscles labeled by a UNC-103E::GFP transgene and VC4/5 neurons labeled by a mCherry::RAB-3 transgene . The boxed areas indicate synaptic regions . Arrowheads indicate cell bodies of VC4 and VC5 . Note that the laterally extended muscle arms on vm2 cells are colocalized with the VC4/5 presynaptic specializations . Scale bar is 20 μm . ( D ) Ventral views of the vulval muscles and presynaptic regions of HSN neurons in wild-type , lin-12 ( wy750 ) , apx-1 ( wy755 ) and sel-12 ( wy760 ) animals . The boxed areas indicate synaptic regions . Note the absence of vm2 muscle arms in mutants . Scale bar is 20 μm . ( E ) Quantification of the vm2 muscle arm defects in wild-type , lin-12 ( wy750 ) , apx-1 ( wy755 ) and sel-12 ( wy760 ) animals . * p<0 . 0001 , n . s . no significant difference , chi-squared test , n = 90–114 animals . ( F ) High magnification lateral views of the synaptic region in wild type and lin-12 ( wy750 ) . vm2 muscle arms are visualized by a single-copy UNC-103E::GFP transgene . Vulval muscle morphology is labeled by mCD8::mCherry . The boxed areas indicate synaptic regions . Scale bar is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00310 . 7554/eLife . 00378 . 004Figure 1—figure supplement 1 . vm2 muscle arms are the postsynaptic specializations . ( A ) Lateral views of the vulval muscles . The vulval muscles are labeled by mCD8::mCherry ( top ) . vm2 muscles arms are labeled by a single-copy unc-103e::gfp transgene ( middle ) . Note that the UNC-103E::GFP is enriched on the muscle arms . Boxed areas indicate synaptic regions . Anterior ( A ) , posterior ( P ) , dorsal ( D ) , ventral ( V ) . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00410 . 7554/eLife . 00378 . 005Figure 1—figure supplement 2 . vm2 muscle arm development is a guidance event independent of presynaptic specialization . ( A ) Ventral views of the vulval muscle morphology in wild-type , unc-104 ( e1265 ) , egl-1 ( n986 ) and lin-3 ( e1417 ) mutant animals . Boxed areas indicate synaptic regions . Scale bar is 20 μm . ( B ) Ventral views of the vulval muscle morphology in wild-type ( left ) and egl-1 ( n986 ) ; lin-39 ( n709 ) ( right ) mutant animals . Boxed areas indicate synaptic regions . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00510 . 7554/eLife . 00378 . 006Figure 1—figure supplement 3 . Schematic model of LIN-12/Notch signaling pathway . ( A ) Schematic model showing the core LIN-12/Notch signaling pathway . The binding of the ligand APX-1/DSL to the receptor LIN-12/Notch triggers proteolytic cleavages . SEL-12/Presenilin releases the Notch intracellular domain ( NICD ) , which enters the cell nucleus and interacts with LAG-1/CSL and other transcription factors to activate target gene transcription . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00610 . 7554/eLife . 00378 . 007Figure 1—figure supplement 4 . lin-12 ( wy750 ) has specific vm2 muscle arm defects . ( A ) Vulval morphology in wild-type ( left two panels ) and lin-12 ( wy750 ) ( right two panels ) animals . Vulval morphogenesis is examined by visualizing the apical boundaries of vulval epithelial cells with AJM-1::GFP or by DIC microscopy . Scale bar is 5 μm . ( B ) VC4 and VC5 axons in wild-type ( left ) and lin-12 ( wy750 ) ( right ) animals . VC axons are labeled with CAT-1::GFP . Scale bar is 20 μm . ( C ) Visualization of vm1 and vm2 in wild-type ( left ) and lin-12 ( wy750 ) ( right ) animals . Vulval muscles ( both vm1 and vm2 ) are labeled with mCD8::mCherry . egl-15::GFP labels vm1 but not vm2 in both wild type and lin-12 ( wy750 ) animals . Scale bar is 20 μm . ( D ) HSN presynapstic regions in wild-type ( top ) and lin-12 ( wy750 ) ( bottom ) animals . HSN presynaptic regions are labeled by SNB-1::YFP . Boxed areas indicate synaptic regions . Arrows indicate the HSN cell bodies . Scale bar is 10 μm . ( E ) Visualizations of undifferentiated M lineage cells in wild-type ( top ) and lin-12 ( wy750 ) ( bottom ) mutants by hlh-8::GFP in mid-L4 animals ( lateral view ) . Note that the pattern of the M lineage cells do not change in the lin-12 ( wy750 ) mutant . Scale bar is 20 μm . ( F ) Ventral view of vulval muscles and uterine muscles in wild-type ( up ) and lin-12 ( wy750 ) ( down ) animals . Vulval muscles ( both vm1 and vm2 ) are labeled with mCD8::mCherry . rgs-2::GFP labels uterine muscles in both wild type and lin-12 ( wy750 ) animals . Note that the uterine muscle differentiation and morphology do not change in lin-12 ( wy750 ) mutant . Scale bar is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00710 . 7554/eLife . 00378 . 008Figure 1—figure supplement 5 . Schematic demonstrations of the quantitative measurements of vulval muscle morphology . ( A ) Schematic demonstration of vulval muscles . The intersection angles between vulval opening and vulval muscles are shown with dashed blue lines . vm1 and vm2 are delineated by dashed green lines . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 00810 . 7554/eLife . 00378 . 009Figure 1—figure supplement 6 . Quantifications of the vulval muscle morphology . ( A ) Quantification of the intersection angles between vulval opening and vulval muscles with different genotypes indicated on the X-axis . n . s . , no significant difference , t test , n = 48–80 pieces of vulval muscles . ( B ) Quantification of the relative size of vm2 to vm1 ( Svm2/Svm1 ) with different genotypes indicated on the X-axis . n . s . , no significant difference , t test , n = 48–80 pairs of vulval muscles . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 009 To understand how cell specification was coordinated with synaptic target selection , we explored the cellular and molecular mechanisms underlying vm2 postsynaptic muscle arm development . We found that postsynaptic muscle arm development in egg-laying circuit is mediated by LIN-12/Notch-dependent cell specification . The canonical LIN-12/Notch signaling pathway is specifically activated in the target vm2 cells by the ligand APX-1/DSL expressed in two adjacent cell types , the secondary vulval epithelial cells and the non-target vm1 cells . lin-12 functions cell-autonomously to up-regulate the activity of the guidance molecules UNC-40/DCC and MADD-2 , which in turn generates and guides the development of vm2 postsynaptic muscle arms . The ectopic expression of UNC-40 in vm1 cells that normally do not form muscle arms is sufficient to induce the extension of muscle arm-like membrane structures from vm1 , suggesting a deterministic role of LIN-12/Notch and UNC-40/DCC in postsynaptic target selection .
The C . elegans egg-laying circuit consists four type 1 vulval muscle cells ( vm1 ) and four type 2 vulval muscle cells ( vm2 ) , which form two X-shaped structures near the vulval opening ( Figure 1A ) . While vm1 and vm2 are from a common lineage , are physically adjacent and share many molecular components , the egg-laying motoneurons HSN and VC4/5 form serotoninergic and cholinergic chemical synapses exclusively on vm2 ( White et al . , 1986; Waggoner et al . , 1998; Duerr et al . , 2001; Kim et al . , 2001 ) . vm1 is mainly controlled by vm2 via gap junctions ( White et al . , 1986 ) . Besides forming major synaptic connection on HSN and VC4/5 , vm2 also receive some additional synapses with the ventral nerve cord ( Personal communication with Michael Koelle , November 2012 ) . To understand how vm2 is differentiated as the postsynaptic target and specifies the synaptic specializations , we first developed tools to visualize the egg-laying synapses . We fused presynaptic protein RAB-3 with mCherry ( mCherry::rab-3 ) to label presynaptic vesicles in HSN or VC neurons , receptively ( Baumeister et al . , 1996; Lickteig et al . , 2001; Shen and Bargmann , 2003 ) . On the postsynaptic side , we expressed ser-1b::gfp or unc-103e::gfp , a GFP tagged serotonin receptor or a voltage-gated potassium channel , in both vm1 and vm2 to visualize the postsynaptic specializations as well as the muscle cell morphology ( Figure 1B , C ) . Consistent with EM reconstruction , we found that the HSN presynaptic specializations were closely juxtaposed against the dendritic spine-like membrane protrusions ( muscle arms ) from the vm2 cell bodies when viewed from both the lateral and ventral sides ( Figure 1B ) . The muscle arms always extend from stereotyped subcellular locations , medially towards the center of the vulva . The muscle arms from the anterior and posterior vm2 cells often contact each other and form a ‘bridge-like’ structure that is best visualized from the ventral view ( Figure 1B , C , top row ) . Similar apposition was also observed between VC4/5 presynaptic specializations and the vm2 muscle arms ( Figure 1C ) . While high level of UNC-103E::GFP expression labels both the muscle arms and the vm cell bodies ( Figure 1C ) , low level of transgenic expression of UNC-103E::GFP with an integrated single-copy transgene showed that UNC-103E was dramatically enriched on vm2 muscle arms ( Figure 1—figure supplement 1 , middle ) , indicating that the muscle arms represent functional postsynaptic specializations ( Collins and Koelle , 2012 ) . Because most egg-laying synapses are formed onto the vm2 muscle arms , we investigated the cellular and molecular mechanisms mediating the vm2 muscle arm formation to gain insight into the postsynaptic development and synaptic target selection . Developmental studies showed that HSN axon guidance and presynaptic formation preceded both axon growth of VC neurons and vulval muscle differentiation ( data not shown ) . We therefore asked whether the HSN neurons are required for vm2 specification and muscle arm development by examining the vm2 morphology in the egl-1 ( n986 ) mutant , in which the HSN neurons undergo programmed cell death in embryonic stage ( Conradt and Horvitz , 1998 ) . Surprisingly , in all of the examined animals , the morphology and position of vm2 muscle arms were indistinguishable from the wild type counterparts ( Figure 1—figure supplement 2A ) . Several results further suggested that neither HSN nor VC neurons contributed to the vm2 muscle arm position . First , in a portion of the egl-1 ( n986 ) ;lin-39 ( n709 ) double mutant animals which lacked both HSN and VC neurons ( Clark et al . , 1993 ) , the vm2 muscle arms still existed ( Figure 1—figure supplement 2B ) . Second , in unc-104 ( e1265 ) mutant animals , in which the synaptic vesicles were absent from the presynaptic regions in both HSN and VC neurons ( Shen and Bargmann , 2003 ) , vm2 muscle arm development was not affected ( Figure 1—figure supplement 2A ) . These results indicate that presynaptic components are dispensable for vm2 postsynaptic development . If vm2 postsynaptic differentiation and muscle arm development are independent of presynaptic components , we asked whether the surrounding vulval epithelial cells that physically contact vm2 provide guidance for vm2 specification and muscle arm development . We examined the vm2 muscle arm morphology in lin-3 ( e1417 ) mutant animals , in which vulval epithelial cell fate is abnormal ( Hill and Sternberg , 1992 ) , and observed severe defects on both vulval muscle morphology and muscle arm development ( Figure 1—figure supplement 2A ) . This result suggests that the vulval epithelial cells are required for normal vulval muscle differentiation and morphogenesis , and therefore might be also important for vm2 postsynaptic muscle arm development . To understand the molecular mechanism of vm2 postsynaptic muscle arm formation , we performed a forward genetic screen on the transgenic strain labeling both HSN presynaptic specializations and vm2 postsynaptic muscle arms . We searched for mutants with normal vm2 differentiation and normal vulval morphology but with abnormal vm2 muscle arms . In total , six individual mutants representing three complementation groups were isolated in which the vm2 muscle arms were largely absent ( Figure 1D , E and Table 1 ) . Mapping and molecular identification of these mutations showed that they affected genes in LIN-12/Notch signaling pathway including lin-12/Notch , apx-1/DSL and sel-12/presinilin ( Table 1 and Figure 1—figure supplement 3 ) . 10 . 7554/eLife . 00378 . 010Table 1 . Mutants isolated from screen for genes required for vm2 postsynaptic target selectionDOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 010Complementation groupAlleleMutationHomologlin-12wy750G473R ( g1417a ) Notchapx-1wy754G208E ( g623a ) Dslwy755G159E ( g476a ) wy766C217Y ( g650a ) sel-12wy756W184stop ( g552a ) Presenilinwy760G373D ( g1118a ) In lin-12 ( wy750 ) mutant animals , while the muscle morphology of both vm1 and vm2 were largely normal , the SER-1B::GFP labeled vm2 muscle arm structures were completely absent ( Figure 1D , E ) . Furthermore , the single-copy insertion of the transgene UNC-103E::GFP potassium channel which specifically labels the vm2 muscle arms , as well as the artificial membrane marker mCD8::mCherry that delineates the entire outline of vulval muscles , was also completely absent in the muscle arm areas , suggestive of a complete loss of the muscle arm structure ( Figure 1F ) . Both HSN and VC presynaptic markers appeared normal in the lin-12 ( wy750 ) mutant , indicating that the mutation specifically affected postsynaptic differentiation and development ( Figure 1—figure supplement 4B , D ) . Surprisingly , we found no vulval morphogenesis defect in lin-12 ( wy750 ) mutants either by differential interference contrast ( DIC ) microscopy or the localization pattern of the ajm-1::gfp apical junction marker at several development stages ( Figure 1—figure supplement 4A and data not shown ) ( Koppen et al . , 2001 ) . Two other mutants isolated from the same screen , apx-1 ( wy750 ) and sel-12 ( wy760 ) , were very similar to the lin-12 ( wy750 ) mutant in their specific defective vm2 muscle arm development and normal vulval muscle development ( Figure 1D and data not shown ) . Vulval muscles , together with the uterine muscles , are produced through three rounds of cell divisions from a single sex myoblast cell , which is derived from the postembryonic mesodermal lineage ( M lineage ) . We considered the possibility that the absence of muscle arms in vm2 could be due to the change of vm2 cell fate to a different cell . We therefore examined several cell-type-specific markers for each differentiation stage: the hlh-8::gfp that is specifically expressed in undifferentiated cells of M lineage ( Harfe et al . , 1998b ) and an egl-15::gfp reporter which is exclusively expressed in vm1 cells in egg-laying system ( Harfe et al . , 1998a; Huang and Stern , 2004 ) . We also examined rgs-2::gfp that labeled uterine muscles ( Dong et al . , 2000 ) . In lin-12 ( wy750 ) mutants , all three markers exhibits wild type pattern , suggesting that many molecular signatures of the M lineage are intact in the lin-12 ( wy750 ) mutants , while the development of the postsynaptic specializations completely fails ( Figure 1—figure supplement 4C , E , F ) . Furthermore , since both vm2 and vm1 adopt complex cell morphology and stereotyped positions , we reasoned that the morphology and position of these muscles must reflect the execution of cell-specific developmental programs . We therefore quantitatively analyzed the morphology and relative position of vm1 and vm2 ( Figure 1—figure supplement 5 ) . In both wild type and lin-12 ( wy750 ) mutant , the position where vm1 is linked to the body wall is always on the dorsal side of the connection between vm2 and body wall . No significant difference was found on the intersection angle between the vulval muscle and the vulval slit , as well as the relative size of vm2 to vm1 ( Svm2/Svm1 ) ( Figure 1—figure supplement 6 ) . In summary , all these observations supports the notion that the muscle defects in lin-12 mutants appear to be restricted to the postsynaptic development in the vm2 cells . If vm2 muscle arms play crucial functional roles in the synaptic connections of the egg-laying circuit , mutants with defective vm2 muscle arm development should show altered egg-laying behavior . As a measure of this behavior , we examined the eggs freshly-laid by the wild-type animals and the muscle arm defective mutants . First , the muscle arm defective mutants laid notably fewer eggs compared to the wild-type animals ( Figure 2A ) . A more direct and sensitive way to analyze egg-laying is to quantify the developmental stages of the freshly-laid eggs . Indeed , we found that these mutants laid more late-stage eggs and fewer early-stage eggs compared with wild-type animals ( Figure 2B ) , indicative of defects in the egg-laying circuit . 10 . 7554/eLife . 00378 . 011Figure 2 . Muscle arm defective mutants have egg-laying behavioral and physiological defects . ( A ) Relative numbers of embryos freshly-laid by wild-type and mutant animals . wyIs333 is the marker that double-labels the HSN presynaptic specializations and vulval muscles . *p<0 . 0001 , Fisher's exact test , n = 62–303 embryos . ( B ) Percentage of different stages of embryos freshly-laid by wild-type and mutants animals . *p<0 . 0001 , Fisher's exact test , n = 62–303 embryos . ( C ) and ( D ) Ratiometric Ca2+ imaging in the vulval muscles in behaving wild-type ( C ) and lin-12 ( wy750 ) ( D ) animals . GCaMP3 ( top ) and mCherry ( middle ) were co-expressed in the vulval muscles and the GCaMP3/mCherry fluorescence ratio ( bottom ) was used to record Ca2+ transients ( arrowheads ) . Time points are shown from Video 1 ( Wild type ) and Video 2 ( lin-12 ( wy750 ) mutant ) . Vertical lines ( II ) indicate the vulval muscles at rest , and arrows indicate vulval muscle twitches ( small ) and egg-laying ( large ) contractions . Anterior ( A ) , posterior ( P ) , left ( L ) , right ( R ) , dorsal ( D ) , ventral ( V ) . Scale bars are 10 μm . ( E–G ) traces of vulval muscle GCaMP3/mCherry ratio change ( ΔR/R ) from the same wild-type ( E ) and lin-12 ( wy750 ) mutant ( F ) animals shown above ( horizontal bars ) and in the apx-1 ( wy755 ) mutant ( G ) . Also indicated are egg-laying events ( * ) , single transients limited to the anterior or posterior ( A or P ) vulval muscles ( | ) , double transients occurring simultaneously in both anterior and posterior ( A + P ) vulval muscles ( ^ ) , or delayed double transient where anterior and posterior transients within a body bend are separated by a visually discernible interval ( horizontal bracket ) . ( H–J ) Quantitations of vulval muscle Ca2+ signaling ( 6-min recording per animal , 11 or 12 animals per genotype; error bars indicate 95% confidence intervals ) . ( H ) lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants have more frequent Ca2+ transients than does the wild type . *p<0 . 0001 , one-way ANOVA , n = 215–488 . ( I ) Fewer synchronous double ( A + P ) Ca2+ transients in lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants . *p<0 . 0001 , chi-squared test , n = 200–305 . ( J ) Consecutive anterior and posterior Ca2+ transients are delayed in lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants . *p<0 . 0001; one-way ANOVA , n = 123–190 . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 011 We then investigated the functional consequences of the vm2 muscle arm defects seen in lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants , and the physiological basis for their reduced egg-laying . By co-expressing the Ca2+ reporter GCaMP3 and the calcium-insensitive fluorescent protein mCherry in the vulval muscles , we were able to perform ratiometric Ca2+ imaging in behaving wild-type , lin-12 ( wy750 ) , and apx-1 ( wy755 ) mutant animals . In wild-type animals , simultaneous Ca2+ transients in the anterior and posterior vulval muscles , which we call ‘double’ transients , drive synchronous muscle contraction and release of eggs ( Collins and Koelle , 2012 ) ( Figure 2C ) . Smaller Ca2+ transients ( magnitude < 150% ΔR/R ) drive small vulval muscle twitching contractions that do not result in release of eggs . Such twitch transients are sometimes ‘single’ transients , limited to either the anterior or posterior muscles , or smaller magnitude double transients . Double transients result in larger peaks in our ratiometric trace recordings because the anterior and posterior signals are summed ( Figure 2E and Video 1 ) . 10 . 7554/eLife . 00378 . 012Video 1 . Synchronous Ca2+ transients in the vulval muscles of wild-type animals drive egg-laying behavior . Intensity-modulated ratiometric imaging of wild-type C . elegans expressing the Ca2+ sensor GCaMP3 and soluble mCherry in the vulval muscles from the unc-103e promoter at 20 fps . Change in the GCaMP3 to mCherry fluorescence ratio is indicated by a rainbow scale from 0 ( dark blue ) to 1 . 2 ( red ) . Egg-laying events are observed at 23 and 56 s . Still images from this video are shown in Figure 2C , and the traces of ΔR/R and vulval muscle area are shown in Figure 2E . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 012 Despite their lack of postsynaptic muscle arms , we found that lin-12 ( wy750 ) and apx-1 ( wy755 ) vm2 muscle arm mutants were still able to execute single anterior and posterior vulval muscle Ca2+ transients , which was presumably due to the intact minor synaptic inputs from the ventral nerve cord alongside the vulval muscles ( Collins and Koelle , 2012 ) . However , double Ca2+ transients were often asynchronous , with significant delays between anterior and posterior vulval muscle transients that resulted in uncoordinated vulval muscle contractions that were not able to release eggs ( Figure 2D and Video 2 ) . In the ratiometric traces , these ‘delayed double’ Ca2+ transients could be resolved into two or more peaks instead of one large peak as in the wild type ( Figure 2E–G ) . As a result , lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants had a twofold increase in the frequency of Ca2+ transient peaks ( Figure 2H ) . We also found that lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants had significantly fewer synchronous double transients compared to the wild type ( Figure 2I ) . In the rare instances when strong single transients or asynchronously initiated anterior/posterior Ca2+ transients of lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants led to opening of the vulva , egg-laying would still occur , indicating that their egg-laying defect was not caused by structural problems in the vulva itself ( Video 3 ) . 10 . 7554/eLife . 00378 . 013Video 2 . lin-12 mutants have asynchronous vulval muscle Ca2+ transients leading to uncoordinated contractions . Intensity-modulated ratiometric imaging of lin-12 ( wy750 ) mutants expressing the Ca2+ sensor GCaMP3 and soluble mCherry in the vulval muscles from the unc-103e promoter at 20 fps . Change in the GCaMP3 to mCherry fluorescence ratio is indicated by a rainbow scale from 0 ( dark blue ) to 1 . 2 ( red ) . Still images from this video are shown in Figure 2D , and the traces of ΔR/R and vulval muscle area are shown in Figure 2F . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 01310 . 7554/eLife . 00378 . 014Video 3 . Asynchronous vulval muscle Ca2+ transients in lin-12 and apx-1 mutants can still lead to uncoordinated vulval opening and egg laying . Intensity-modulated ratiometric imaging of apx-1 ( wy755 ) mutants expressing the Ca2+ sensor GCaMP3 and soluble mCherry in the vulval muscles from the unc-103e promoter at 20 fps . Change in the GCaMP3 to mCherry fluorescence ratio is indicated by a rainbow scale from 0 ( dark blue ) to 1 . 2 ( red ) . At 6 and 18 s , there was sufficient vulval opening to permit an egg-laying event . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 014 To quantify the anterior/posterior coordination in double Ca2+ transients , we determined the delay between consecutive anterior and posterior Ca2+ transient peaks that occurred during a single locomotor body bend . In wild-type animals , the average anterior/posterior delay was only ∼100 ms ( Figure 2J ) , as nearly all double transients were not resolved into separate anterior/posterior peaks at the temporal and spatial resolution of our ΔR/R traces ( Figure 2E ) . In both lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants , the average anterior/posterior delay in double Ca2+ transients increased to ∼800 ms ( Figure 2J ) and anterior/posterior transients were easily resolved ( Figure 2F , G ) . Together , these results show that the vm2 muscle arms lacking in lin-12 ( wy750 ) and apx-1 ( wy755 ) mutants are not necessary for excitation of the vulval muscles , but suggest they are required for the synchronous excitation of anterior and posterior vulval muscles needed to successfully execute egg-laying behavior . LIN-12/Notch pathway was originally discovered through its effects on anchor cell and vulval precursor cell fates ( Greenwald et al . , 1983; Sternberg , 1988; Seydoux and Greenwald , 1989 ) . It was surprising that we isolated multiple alleles of genes in the LIN-12/Notch pathway , which did not show obvious phenotypes in vulval morphogenesis , except for one sel-12 early-stop allele wy756 ( Table 1 ) . To understand how vm2 muscle arm development was specifically affected , we identified the molecular lesions of these mutant alleles . Sequence analyses revealed that lin-12 ( wy750 ) contained a single missense mutation ( G473R ) in one of the conserved extracellular EGF domains of the LIN-12 protein ( Table 1 ) . Two additional results indicated that lin-12 ( wy750 ) was a loss-of-function allele . First , trans-heterozygotes between lin-12 ( wy750 ) and a canonical loss-of-function allele lin-12 ( n676n930 ) presented similar muscle arm defects ( data not shown ) . Second , a lin-12 genomic fragment including 5 kb of upstream regulatory sequence and the entire coding region of lin-12 strongly rescued the muscle arm phenotype of lin-12 ( wy750 ) mutants ( Figure 4A ) . Since strong loss-of-function alleles of lin-12 , apx-1 and sel-12 cause lethality , dramatic vulval morphogenesis or general muscle morphology phenotypes , which are all absent in lin-12 ( wy750 ) mutant animals , it is likely that the missense mutation we isolated represents a partial loss-of-function allele . Three apx-1 mutant alleles independently identified from the screen also contain missense mutations in the extracellular domain . One early-stop mutation in the sel-12 coding region leads to additional pleiotropic effects on both vulval muscle patterning and vulval morphogenesis , while the other missense mutation of sel-12 causes specific muscle arm defect ( Table 1 ) . Together , these data suggest that partial loss of LIN-12/Notch signaling activity spares vulval morphogenesis but specifically impairs vm2 specification and muscle arm development , and that muscle arm development might require a higher level of LIN-12/Notch signaling activity compared with earlier developmental events ( see ‘Discussion’ ) . To further distinguish whether LIN-12/Notch signaling is directly involved in the vm2 muscle arm development or indirectly affects vm2 development due to its role in vulval morphogenesis , we investigated the spatial and temporal requirements of lin-12 . Previous studies suggest a late requirement for lin-12 activity to allow proper egg-laying behavior ( Sundaram and Greenwald , 1993 ) . The ‘late egg-laying defect’ appears to be independent of AC/VU decision and vulval precursor cell ( VPC ) specification regulated by LIN-12/Notch signaling ( Sundaram and Greenwald , 1993 ) . We hypothesized that the late egg-laying defect could be caused by the defects of vm2 muscle arm development . In order to test this idea , we examined a temperature sensitive allele , lin-12 ( n676n930ts ) , in which LIN-12 activity is normal at the permissive temperature ( 15°C ) but disrupted at the restrictive temperature ( 25°C ) . Consistent with previous behavioral studies , lin-12 ( n676n930 ) mutant animals grown at the permissive temperature had no egg-laying defect and showed intact vulval muscle morphology and vm2 muscle arms , whereas animals grown at the restrictive temperature retained significantly more unlaid eggs inside their bodies and displayed vulval morphogenesis and vulval muscle arm defects . We then shifted lin-12 ( n676n930 ) mutant animals from the permissive temperature to the restrictive temperature at different larval stages and examined the vulval muscle morphogenesis and vm2 muscle arm development in the resulting adults . Most of the animals shifted from 15°C to 25°C prior to mid-L4 stage showed muscle arm defects similar to those of lin-12 ( wy750 ) , whereas the wild-type muscle arm phenotype was observed in most of lin-12 ( n676n930 ) animals that were shifted after mid-L4 stage ( Figure 3A ) . 10 . 7554/eLife . 00378 . 015Figure 3 . Temporal requirement of LIN-12 in vm2 muscle arm formation . ( A ) and ( B ) Temperature up-shift ( A ) and down-shift ( B ) experiments of lin-12 ( n676n930 ) . The percentages of animals with normal or defective muscle arm phenotype are indicated by dark or light gray bars , respectively . The time points when animals are shifted are shown on X-axis . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 015 In a temperature down-shift experiment , designed to circumvent LIN-12's role in VPC specification , we first kept lin-12 ( n676n930 ) mutant animals at the permissive temperature until the early-L3 stage , when the AC/VU decision and VPC specification processes were complete but vulval muscle cells had not yet been generated ( Sundaram and Greenwald , 1993; Foehr and Liu , 2008 ) . Early-L3 animals were then transferred to 25°C and shifted back to 15°C at different development stages . In all the animals we examined , no severe vulval morphogenesis or general vulval-muscle patterning defects were observed . More than 80% of the animals still showed defective muscle arms when shifted from 25°C to 15°C at mid-L4 ( Figure 3B ) . The upshift and downshift experiments together suggest a novel function of lin-12 at the L4 stage in regulating the postsynaptic muscle arm development and synaptic target choice . This lin-12 function is independent of other known roles of lin-12 , all of which occur prior to early-L4 stage . We next investigated where lin-12 functions to regulate the vm2 muscle arm development by analyzing the expression of lin-12 . A type II short trans-splicing leader sequence followed by mCherry ( SL2::mCherry ) was inserted into a fosmid containing the lin-12 genomic sequence at the 3′ end of the lin-12 coding region ( lin-12::SL2::mCherry ) . This recombineered fosmid fully rescued the lin-12 muscle arm phenotype ( Figure 4A and data not shown ) , suggesting that this construct should contain all upstream , downstream and intronic regulatory elements crucial for proper expression and function of lin-12 . Transgenic animals carrying this construct showed mCherry expression in a number of cell types including both the vm2 muscle cells ( outlined by dashed lines in Figure 4A , B ) and the vulval epithelial cells ( Figure 4A , B ) . Strikingly , no detectable lin-12 expression was observed in vm1 muscle cells . To further understand whether lin-12 is required in the vulval epithelial cells or vm2 , we injected this construct into the lin-12 ( wy750 ) mutant and analyzed mosaic animals in which expression of this transgene was lost in some of the tissues . Forty-five of 48 animals with lin-12 expressed in the vm2 cells ( vm2+ ) showed a rescued , wild-type muscle arms ( Figure 4A , C ) , whereas the muscle arm defect was never rescued in any of the examined animals ( >50 ) lacking lin-12 expression in the vm2 cell ( vm2- ) ( Figure 4B , C ) . On the contrary , lin-12 expression in vulval epithelial cells was not correlated with the rescue of the muscle arm defects ( data not shown ) . The mosaic rescue analyses argue that lin-12 functions cell-autonomously in vm2 to specify vm2 postsynaptic development through regulating muscle arm formation . 10 . 7554/eLife . 00378 . 016Figure 4 . Cell-autonomous requirement of LIN-12 in vm2 . ( A ) and ( B ) Representative images of lin-12 ( wy750 ) mutant animals with non-mosaic ( A ) and mosaic ( B ) expression of lin-12::SL2::mCherry . Left column shows the vulval muscles labeled by SER-1B::GFP . Middle column shows the lin-12 expression patterns . Right column shows the overlaid images . Note that the complete lin-12 expression ( vm2+ ) rescues muscle arm defects of lin-12 ( wy750 ) ( A ) . Lacking of lin-12 expression in vm2 ( vm2- ) abolishes its capacity to rescue muscle arm defects ( B ) . Boxed areas indicate synaptic regions . Dashed lines indicate the morphology of vm2 . Scale bar is 20 μm . ( C ) Quantification of the vm2 muscle arm phenotypes in the mosaic transgenic animals . *p<0 . 0001 , Fisher's exact test , n = 48–50 . ( D ) Vulval muscle morphology in the apx-1 ( wy755 ) ; lin-12 ( w750 ) mutant animals . Boxed area indicates synaptic region . Arrows indicate HSN cell bodies . Scale bar is 20 μm . ( E ) Epifluorescence and DIC images showing the apx-1 expression pattern in young adult . The four cells in the center of the image are 2° vulval epithelial cells . The four cells at the periphery are vm1 cells . Scale bar is 20 μm . ( F ) Cell-autonomous requirement of apx-1 , sel-12 and lag-1 . Quantification of muscle arm phenotypes in animals with different genotypes indicated on the X-axis . * , P<0 . 0001 , Fisher's exact test , n = 40–57 . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 01610 . 7554/eLife . 00378 . 017Figure 4—figure supplement 1 . APX-1 expression pattern and the relative position of vm2 muscle arms . ( A ) Overlaid epifluorescence and DIC images showing the identity of the apx-1 expressing cells . The egl-17::mCD8::mCherry transgene labels the 2° vulval epithelial cells in young adult . Scale bar is 20 μm . ( B ) Overlaid epifluorescence and DIC images showing the identity of the apx-1 expressing cells . The unc-103e::mCD8::mCherry transgene labels the vm1 cells in young adult . Scale bar is 20 μm . ( C ) Single confocal slices showing the position of muscle arms and the vulval epithelial cells . Vulval muscles are labeled by mCD8::mCherry . Secondary vulval epithelial cells are labeled by egl-17::gfp transgene ( left ) . Both primary and secondary vulval epithelial cells are labeled by F47B8 . 6::gfp transgene ( right ) . Note that the muscle arms are located between the 1o and the 2o vulval epithelial cells and directly contact these cells . Arrowheads indicate muscle arms . Scale bar is 10 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 017 LIN-12 is a membrane-associated protein that can be activated by binding with its ligand . Based on the cell-autonomous function of lin-12 in vm2 , we made two predictions . First , the ligand ( s ) of LIN-12 should be localized in the surrounding cells that directly contact vm2 to activate LIN-12/Notch pathway . Second , downstream components of LIN-12/Notch signaling pathway should also function in vm2 cells to regulate vm2 specification and muscle arm development . APX-1/Dsl is a good candidate as the relevant ligand because apx-1 ( wy755 ) , but not other known LIN-12 ligand mutants including dsl-1 and lag-2 , showed the muscle arm phenotype similar to that of lin-12 ( wy750 ) ( Figure 1D and data not shown ) . A lin-12 ( wy750 ) ; apx-1 ( wy755 ) double mutant did not show an enhanced muscle arm defect ( Figure 4D ) or more severe egg-laying phenotype ( Figure 2A ) , further suggesting that apx-1 and lin-12 function in the same pathway . To further understand where apx-1 functioned , we examined the expression pattern of apx-1 with an integrated transgenic expression construct ( Personal communication with Iva Greenwald , June 2012 ) . We observed the expression of apx-1 in both secondary vulval epithelial cells and vm1 muscle cells during the L4 stage , when LIN-12/Notch signaling was activated to promote vm2 muscle arm development ( Figure 4E and Figure 4—figure supplement 1A , B ) . Both secondary vulval epithelial cells and vm1 cells contact vm2 muscle cells from early L4 stage when vulval muscles are generated ( Figure 4—figure supplement 1C , data not shown , and personal communication with Kelly Liu , July 2012 ) . In fact , we observed that the vm2 muscle arms consistently tracked the cell junctions between the primary and secondary vulval epithelial cells ( Figure 4—figure supplement 1C and data not shown ) . To further understand the cellular requirement of apx-1 , we created two constructs in which the apx-1 cDNA was cloned after two different promoters . The egl-15 promoter supports expression specifically in the vm1 cells ( Harfe et al . , 1998a; Eimer et al . , 2002; Huang and Stern , 2004 ) , while the egl-17 promoter drives expression in the secondary vulval epithelial cells from early-L4 stage ( Shaye and Greenwald , 2002 ) . Cell-specific expression of apx-1 in either secondary vulval epithelial cells or vm1 muscle cells rescued the muscle arm defects of apx-1 ( wy755 ) mutants ( Figure 4F ) . These data strongly suggest that apx-1 functions in the vm1 and secondary vulval epithelial cells to activate LIN-12/Notch signaling in vm2 , the postsynaptic target cells . 10 . 7554/eLife . 00378 . 018Figure 5 . vm2 muscle arm development requires APX-1-induced LIN-12/Notch signaling activity . ( A ) hlh-29::GFP expression in wild-type ( top ) and lin-12 ( wy750 ) ( bottom ) animals . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows the hlh-29::GFP expression patterns . Right column shows the overlaid images . Arrowheads indicate VC4/5 cell bodies . Note that hlh-29::GFP is specifically expressed in vm2 but not vm1 in the wild type , and is down-regulated in the lin-12 ( wy750 ) mutant . Scale bar is 20 μm . ( B ) and ( C ) lin-12::SL2::mCherry expression in wild-type ( B ) and apx-1 ( wy755 ) ( C ) animals . Left column shows the vulval muscles labeled by SER-1::GFP . Middle column shows the lin-12 expression patterns . Right column shows the overlaid images . Boxed areas indicate synaptic regions . Dashed lines delineate the morphology of vm2 . Note that lin-12 transcription is specifically down-regulated in vm2 cells in apx-1 ( wy755 ) mutant animals . Scale bar is 20 μm . ( D ) apx-1 expression in wild-type ( left ) and lin-12 ( wy750 ) ( right ) animals . The four cells in the center of the image are 2o vulval epithelial cells . The four cells at the periphery are vm1 cells . Note that the apx-1 expression pattern does not change in the lin-12 ( wy750 ) mutant . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 01810 . 7554/eLife . 00378 . 019Figure 5—figure supplement 1 . vm2 muscle arm development requires high LIN-12/Notch signaling activity . ( A ) hlh-29::gfp expression in wild-type ( left ) , apx-1 ( wy755 ) ( middle ) and sel-12 ( wy760 ) ( right ) mutant animals . Vulval muscles are labeled by mCD8::mCherry . Note that hlh-29::gfp is specifically expressed in vm2 but not vm1 in the wild type , and is down-regulated in the apx-1 ( wy755 ) and sel-12 ( wy760 ) mutants . Scale bar is 20 μm . ( B ) Schematic illustration of the molecular lesion ( top ) and muscle arm phenotype ( bottom ) of lin-12 ( e2621 ) . lin-12 promoter and coding sequence are indicated by yellow and orange colors . Black box and bracket show the deletion in e2621 . LAG-1 consensus binding sites are indicated by green arrowheads . Note that there are two LAG-1 consensus binding sites located in the deletion region . Boxed area indicates synaptic region . Arrow indicates HSN cell body . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 019 Our genetic screen also yielded two alleles ( wy760 and wy756 ) of sel-12 , the C . elegans homolog of presenilin ( Table 1 and Figure 1—figure supplement 3 ) . sel-12 encodes the γ-secretase that releases the LIN-12/NOTCH intracellular domain ( NICD ) by ligand-induced cleavage ( Levitan and Greenwald , 1995; Ray et al . , 1999a , 1999b; Selkoe and Kopan , 2003 ) . NICD is then transported into cell nucleus where it serves as a transcription factor together with LAG-1/CSL to promote gene expression ( Christensen et al . , 1996; Andersson et al . , 2011 ) ( Figure 1—figure supplement 3 ) . To understand where sel-12 was required , we performed cell-specific rescue experiments . We found that expression of sel-12 in vulval muscle cells completely rescued the vm2 muscle arm phenotype of sel-12 ( wy760 ) , consistent with the cell-autonomous requirement of LIN-12/Notch signaling in vm2 ( Figure 4F ) . Additionally , specific expression of a dominant-negative lag-1 construct ( Kato et al . , 1997 ) in the vulval muscles phenocopied the muscle arm defects observed in lin-12 ( wy750 ) , apx-1 ( wy750 ) and sel-12 ( wy760 ) mutants ( Figure 4F ) , suggesting that the canonical LIN-12/Notch pathway is likely involved by controlling the transcription of downstream genes . Moreover , the expression of a hlh-29 transcription reporter ( hlh-29::gfp ) , a known target gene of the canonical LIN-12/Notch pathway in vm2 , was also dramatically down-regulated in lin-12 ( wy750 ) , apx-1 ( wy755 ) and sel-12 ( wy760 ) mutants ( Figure 5A and Figure 5—figure supplement 1A ) ( Fischer and Gessler , 2007; McMiller et al . , 2007 ) . Together , these results strongly support the model that APX-1 expressed in the secondary vulval epithelial cells and vm1 cells activates the canonical LIN-12/Notch pathway in vm2 to promote the formation of the postsynaptic muscle arms . The specific activation of the LIN-12/Notch pathway in vm2 but not in vm1 thus appears to dictate the choice of the synaptic target . We next investigated the mechanism by which apx-1 activated LIN-12/Notch signaling in vm2 . A reciprocal inhibition mechanism as well as a lateral signaling inducing the differentiation of a Notch cell and a Delta cell , mediates the AC/VU decision and VPC specification , respectively ( Greenwald et al . , 1983; Sternberg , 1988; Seydoux and Greenwald , 1989; Wilkinson et al . , 1994; Sundaram , 2004 ) . We wondered whether a similar mechanism is also used in specifying vm2 identity and regulating the muscle arm development . We first examined the lin-12 expression pattern in an apx-1 ( wy755 ) mutant . Compared with wild-type animals , apx-1 ( wy755 ) mutant animals showed significantly reduced expression of lin-12 in vm2 cells , whereas the lin-12 expression in vulval epithelial cells was not affected ( Figure 5B , C ) . On the other hand , we did not observe any change of apx-1 expression in the lin-12 ( wy750 ) mutant ( Figure 5D ) . These results argue that apx-1 unilaterally promotes lin-12 expression and activates the LIN-12/Notch pathway in vm2 . The positive feedback mechanism in the Notch cells is mediated by a CSL induced Notch transcription ( Wilkinson et al . , 1994 ) . Indeed , we found three LAG-1 consensus binding sites in the promoter region of lin-12 ( Christensen et al . , 1996; Maier and Gessler , 2000 ) . To understand if this positive feedback loop is important for vm2 muscle arm formation , we searched for mutations in the lin-12 promoter region . One lin-12 loss-of-function allele , e2621 , was previously known to have egg-laying defects without vulval morphogenesis phenotype ( Wu et al . , 1998 ) . We sequenced the lin-12 promoter in this allele and found a 541-bp deletion in the promoter region leading to the deletion of two of the three LAG-1 binding sites ( Figure 5—figure supplement 1B , top ) . Interestingly , the muscle arm development was similarly affected in lin-12 ( e2621 ) ( Figure 5—figure supplement 1B , bottom ) , suggesting that the muscle arm development might require the positive transcriptional feedback of the LIN-12/Notch pathway . Given the specific activation of the LIN-12/Notch pathway in vm2 and the cell-autonomous requirement of LIN-12 for muscle arm development , we searched for candidate downstream genes in this pathway . UNC-40 , which is a homolog of Deleted in Colorectal Carcinoma ( DCC ) and neogenin in vertebrates serving as a receptor for the UNC-6/Netrin ligand ( Serafini et al . , 1994; Chan et al . , 1996; Keino-Masu et al . , 1996 ) , was reported to play a role in muscle arm formation in the body wall muscles ( Dixon and Roy , 2005; Alexander et al . , 2009 ) . More recently , MADD-2 , a C1-TRIM protein encoding a homolog of human MID1 , was shown to directly bind and function together with UNC-40 to regulate body wall muscle arm extension as well as axon branching and patterning ( Alexander et al . , 2010; Hao et al . , 2010; Morikawa et al . , 2011 ) . We therefore examined the vm2 muscle arms in these two mutants . Interestingly , both unc-40 ( n324 ) and madd-2 ( ok2226 ) null mutants showed specific vulval muscle arm defects with largely normal vulval muscle morphology ( Figure 6A and Figure 1—figure supplement 6 ) . A significant proportion of the mutant animals completely lacked the vm2 muscle arms , while some other animals showed abnormal muscle arms that extended from incorrect positions or flimsy muscle arms that were thinner and dimmer than the wild-type muscle arms ( Figure 6B ) . These defects were strongly rescued by expressing unc-40 or madd-2 specifically in vulval muscles in respective mutants , suggesting that both unc-40 and madd-2 function cell-autonomously in vm2 to promote and guide muscle arm growth ( Figure 6B ) . Consistent with this notion , we found that the expression of both unc-40 and madd-2 could be robustly detected in the vm2 cells but not in vm1 cells ( Figure 6C , D ) . Moreover , UNC-40::GFP is enriched on the muscle arms , further suggesting that it might function locally to promote muscle arm growth ( Figure 6C ) . These results indicate that the levels of UNC-40 and MADD-2 critically determine the capacity of vulval muscles to grow muscle arms and be selected as the postsynaptic target by the egg-laying motoneurons . 10 . 7554/eLife . 00378 . 020Figure 6 . LIN-12/Notch signaling instructs vm2 muscle arm by regulating unc-40/DCC and madd-2 . ( A ) Representative images showing the vulval muscle morphology in wild type ( left ) , unc-40 ( e271 ) ( middle ) and madd-2 ( ok2226 ) ( right ) animals . Boxed areas indicate synaptic regions . Note that unc-40 ( e271 ) and madd-2 ( ok2226 ) show missing ( middle ) or abnormal ( right ) muscle arm phenotypes . Scale bar is 20 μm . ( B ) Quantification of muscle arm phenotypes in animals with different genotypes indicated on the X-axis . ‘Wild type’ indicates the animals with normal muscle arms . ‘Missing arms’ indicates the animals with missing muscle arms . ‘Abnormal arms’ indicates the animals with flimsy muscle arms or abnormal muscle arms extending from incorrect positions . ‘Other’ indicates animals with severe vulval muscle morphology phenotype that the muscle arms could not be scored . *p<0 . 0001 , chi-squared test , n = 73–194 . ( C ) Double-labeling of the expression of UNC-40 and vulval muscle morphology in wild type ( top ) and lin-12 ( wy750 ) animals ( bottom ) . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows the UNC-40::GFP expression patterns . Right column shows the overlaid images . Note that UNC-40::GFP is preferentially expressed in vm2 and enriched on the muscle arms in the wild type , and is down-regulated in the lin-12 ( wy750 ) mutant . Scale bar is 20 μm . ( D ) Double-labeling of the expression of MADD-2 and vulval muscle morphology in wild type ( top ) and apx-1 ( wy755 ) animals ( bottom ) . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows the MADD-2::GFP expression patterns . Right column shows the overlaid images . Note that MADD-2::GFP is preferentially expressed in vm2 in the wild type , and is down-regulated in the apx-1 ( wy755 ) mutant . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 02010 . 7554/eLife . 00378 . 021Figure 6—figure supplement 1 . LIN-12/Notch signaling instructs vm2 muscle arm by regulating unc-40/DCC and madd-2 . ( A ) Ventral views of the double-labeling of the transcription of unc-40 and vulval muscle morphology in wild type ( top ) and lin-12 ( wy750 ) animals ( bottom ) . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows the unc-40 transcription patterns . Right column shows the overlaid images . Note that unc-40 is preferentially transcribed in vm2 in wild type , and down-regulated in lin-12 ( wy750 ) mutant . Scale bar is 20 μm . ( B ) Ventral views of the double-labeling of the transcription of madd-2 and vulval muscle morphology in wild type ( top ) and lin-12 ( wy750 ) animals ( bottom ) . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows the madd-2 transcription patterns . Right column shows the overlaid images . Note that madd-2 is preferentially transcribed in vm2 in wild type , and down-regulated in lin-12 ( wy750 ) mutant . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 021 To investigate whether unc-40 and madd-2 functioned downstream of the LIN-12/Notch pathway , we compared the expression pattern of UNC-40 and MADD-2 in wild-type animals and lin-12 ( wy750 ) mutants . For these analyses , we used two types of expression constructs: translational reporters and transcriptional reporters . The translational reporters contain the promoter , coding and 3′UTR sequences in which the GFP is inserted in frame with the coding sequences . The fluorescence signal from this type of reporter reflects transcriptional activity , translational regulation and protein stability . The transcriptional reporters only contain the promoter regions and report the transcriptional activity . In wild-type animals , both UNC-40 and MADD-2 translational reporters were preferentially expressed in vm2 cells ( Figure 6C , D ) . UNC-40::GFP showed membrane-associated pattern and was enriched on vm2 muscle arms . MADD-2 was distributed diffusely in vm2 cells . The expression of both UNC-40 and MADD-2 in vm2 was significantly down-regulated in the lin-12 ( wy750 ) mutant ( Figure 6C , D ) , supporting the idea that lin-12 was required for the expression of unc-40 and madd-2 in vm2 . To further explore the mechanism by which the expression of unc-40 and madd-2 was regulated by lin-12 , we examined the transcriptional reporters for unc-40 and madd-2 to measure the transcriptional activity of these genes . Similar transcription pattern was observed for unc-40 and madd-2 . Both unc-40 and madd-2 were preferentially transcribed in vm2 cells , and the transcriptional activities of both promoters were dramatically suppressed in the lin-12 ( wy750 ) mutant ( Figure 6—figure supplement 1A , B ) . These data indicate that the canonical LIN-12/Notch signaling pathway acts to promote the transcriptions of both unc-40 and madd-2 directly . Consistent with this model , two LAG-1 consensus binding sites were found on the proximal promoter region of the madd-2 gene . Although no LAG-1 binding site was found in unc-40 promoter region , two were found in the third intron of unc-40 that is essential for its expression ( Chan et al . , 1996; Christensen et al . , 1996; Maier and Gessler , 2000; Personal communication with Daniel Colon-Ramos , January 2013 ) . To further examine whether unc-40 or madd-2 expression was an output of LIN-12/Notch signaling sufficient for muscle arm formation , we asked whether the overexpression of unc-40 or madd-2 with an exogenous promoter in vulval muscles could bypass the requirement of lin-12 . We found that the expression of unc-40 , but not madd-2 , in the lin-12 ( wy750 ) mutant was sufficient to induced muscle arm-like membrane extension in vm2 cells in ∼41% ( 36 of 88 ) of the animal ( Figure 7A and data not shown ) , suggestive of a deterministic role of unc-40 in promoting postsynaptic muscle arm development . 10 . 7554/eLife . 00378 . 022Figure 7 . unc-40 and madd-2 function partly in parallel in regulating vm2 muscle arm development . ( A ) Expression of UNC-40::GFP ( middle ) in vulval muscles in the lin-12 ( wy750 ) animals . Vulval muscles are labeled by mCD8::mCherry ( left ) . Boxed area indicates synaptic region . Arrowheads indicate rescued muscle arms . Note that the forced expression of UNC-40 in vm2 from the unc-103e promoter induces the formation of muscle arms in the lin-12 ( wy750 ) mutant . Scale bar is 20 μm . ( B ) Ectopic expression of unc-40::gfp ( top ) or mCD8::gfp ( bottom ) transgene in vm1 . Left column shows the vulval muscles labeled by mCD8::mCherry . Middle column shows forced transgene expression patterns . Right column shows the overlaid images . Boxed areas indicate synaptic regions . Arrowheads indicate ectopic muscle arm-like structures . Note that the ectopic expression of UNC-40 in vm1 induces the muscle arm-like membrane extensions . Scale bar is 20 μm . ( C ) Quantification of muscle arm phenotype in animals with different genotypes indicated on the X-axis . ‘Wild type’ indicates the animals with normal muscle arms . ‘Missing arms’ indicates the animals with missing muscle arms . ‘Abnormal arms’ indicates the animals with flimsy muscle arms or abnormal muscle arms extending from incorrect positions . ‘Other’ indicates animals with severe vulval muscle morphology phenotype that the muscle arms could not be scored . *p<0 . 0001 , chi-squared test , n = 105–311 . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 02210 . 7554/eLife . 00378 . 023Figure 7—figure supplement 1 . Ectopic expression of UNC-40 in vm1 induces muscle arm-like structures . ( A ) UNC-40 expression in the wild-type ( left ) and madd-2 ( ok2226 ) ( right ) animals . Vulval muscles are labeled by mCD8::mCherry . Note that in both wild-type and madd-2 ( ok2226 ) mutant animals , UNC-40 are preferentially expressed in vm2 cells . Scale bar is 20 μm . ( B ) MADD-2 expression in the wild-type ( left ) and unc-40 ( n324 ) ( right ) animals . Vulval muscles are labeled by mCD8::mCherry . Note that in both wild-type and unc-40 ( n324 ) mutant animals , MADD-2 are preferentially expressed in vm2 cells . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 02310 . 7554/eLife . 00378 . 024Figure 7—figure supplement 2 . unc-40 and madd-2 are not mutually regulated . ( A ) Three examples showing the vm1 morphology in animals with the expression of a single-copy transgene egl-15::unc-40::gfp . vm1 is visualized by egl-15::mCD8::gfp . Arrowheads indicate the induced muscle arm-like membrane protrusions . Scale bar is 20 μm . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 02410 . 7554/eLife . 00378 . 025Figure 7—figure supplement 3 . Schematic model of muscle arm development and vm2 postsynaptic specification . ( A ) Schematic model showing the physical organization of the vulva organ and the mechanism of vm2 postsynaptic target selection . Note that vm2 muscle arms ( red ) directly contact the primary ( 1° , dark blue ) and secondary ( 2° , dark gray ) vulval epithelial cells . APX-1/DSL is expressed in vm1 cells ( orange ) and the secondary vulval epithelial cells . LIN-12/NOTCH is expressed in vm2 cells ( red ) . UNC-40/DCC ( brown circles ) and MADD-2 ( black circles ) interact with each other and are localized on vm2 muscle arms . DOI: http://dx . doi . org/10 . 7554/eLife . 00378 . 025 vm1 and vm2 share similar lineage , physical location and morphology . However , only vm2 receives synaptic connections from the egg-laying neurons HSN and VC4/5 . Based on the data presented above , we speculated that the LIN-12/Notch signaling pathway elevates the level of UNC-40 and MADD-2 in vm2 , which in turn are directly involved in generating muscle arms . To test the sufficiency of UNC-40 and MADD-2 in conferring postsynaptic specificity , we created transgenic animals that expressed UNC-40 ectopically in the vm1 cells . We found that misexpression of UNC-40 was sufficient to induce the dendritic spine-like membrane extensions from vm1 cells ( Figure 7B ) . To circumvent the occasional silencing of the UNC-40 transgene , we achieved UNC-40 expression in vm1 by single-copy insertion and observed the similar ectopic muscle arm-like structure ( Figure 7—figure supplement 1 ) . Although the ectopic membrane extension from vm1 was usually thinner than the wild-type vm2 muscle arm , the vm1 muscle arms always formed from the medial surface of vm1 , the identical location as for the normal vm2 muscle arms , suggesting that UNC-40 responds to specific environmental cues to guide the formation of muscle arms . Similar ectopic expression of madd-2 in vm1 was not able to induce muscle arm formation ( data not shown ) , hinting that the activity of unc-40 in vm1 might be required . Since both unc-40 and madd-2 mutants showed muscle arm phenotypes , we investigated whether they functioned in the same or parallel pathways . We generated an unc-40 ( n324 ) ; madd-2 ( ok2226 ) double mutant and found slightly enhanced muscle arm defects in the double mutants ( Figure 6B ) . We also found that the overexpression of madd-2 partially rescued the muscle arm defects of unc-40 ( n324 ) null mutant ( Figure 7C ) . Conversely , overexpression of unc-40 largely rescued the muscle arm defects of madd-2 ( ok2226 ) mutant , suggesting that madd-2 likely functioned synergistically with unc-40 to promote muscle arm formation ( Figure 7C ) . Expression of madd-2 and unc-40 is not mutually regulated ( Figure 7—figure supplement 2A , B ) , consistent to the hypothesis that unc-40 and madd-2 at least partly function in parallel . We conclude that the postsynaptic development and synaptic target selection in the egg-laying circuit are mediated by the LIN-12/Notch signaling pathway . The secondary vulval epithelial cells and non-target vm1 cells express APX-1/Dsl , which activates its receptor LIN-12/Notch in the vm2 synaptic target cells . The LIN-12/Notch pathway promotes the expression of UNC-40/DCC and MADD-2 , which cooperates with each other to generate postsynaptic muscle arms .
The EM reconstruction work proposed that the vm2 muscle arms serve as the postsynaptic specializations , which are similar to the dendritic spines in the vertebrate neurons ( White et al . , 1986 ) . We found that both the postsynaptic potassium channel UNC-103E and a neurotransmitter receptor SER-1B were enriched on vm2 muscle arms , further supporting that the muscle arm structure is functionally relevant postsynaptic specializations ( Figure 1B , C and Figure 1—figure supplement 1 ) . In addition , the mutants lacking vm2 muscle arms show significant defects in the synchronized activation of vm2 and egg-laying behavior ( Figure 2 ) , indicating the necessity of the vm2 muscle arm for the integrated functional synapse . Therefore , the ability to form muscle arms during development dictates the choice of vm2 , but not its sister cell vm1 , to become the postsynaptic partner of the egg-laying motor neurons . It has been generally accepted that LIN-12/Notch signaling is involved in multiple cell specification processes . Particularly , LIN-12/Notch signaling is crucial for vulval morphogenesis in C . elegans ( Greenwald , 1998 , 2005 ) . Canonical lin-12 loss-of-function alleles show defective vulval morphogenesis . However , additional roles of LIN-12/Notch signal have also been suggested because the down-regulation of lin-12 activity at late developmental stage causes defects in the egg-laying behavior without any vulval morphology phenotype ( Sundaram and Greenwald , 1993 ) . However , little was known about the underlying mechanism . From our study , we definitively showed that the egg-laying defects in lin-12 , apx-1 , sel-12 weak loss-of-function alleles were due to the absence of vm2 muscle arm , which caused asynchronous activation of the anterior and posterior vulval muscles and inefficient egg-laying . The vm2 muscle arm phenotype in the lin-12 ( wy750 ) weak allele could be caused directly by the reduction of LIN-12/Notch activity , or indirectly by the disruption of other signaling ( s ) due to slight alternations of LIN-12/Notch-dependent vulval morphogenesis or other LIN-12/Notch-dependent events during early development . Using a temperature sensitive allele , we showed that disruption of lin-12 activity at a very late developmental stage caused vm2 muscle arm defects along with egg-laying defects . We never observed the loss or alternation of vulval muscles in these new weak alleles by examining several cell-type-specific markers ( Figure 1—figure supplement 4C , E , F ) , indicating that the function of LIN-12/Notch signaling in specifying the sex myoblasts , the sex muscle precursor cells , is intact ( Greenwald et al . , 1983; Foehr and Liu , 2008 ) . This developmental timing matches the relatively late window of vulval muscle specification and muscle arm formation . Furthermore , activation of LIN-12/Notch signaling in vm2 is necessary and sufficient to rescue the vm2 muscle arm development , indicating that LIN-12/Notch signaling is directly required in vm2 . These results suggest a novel role of LIN-12/Notch signaling in regulating vm2 postsynaptic muscle arm formation and target selection . Therefore , we conclude that LIN-12/Notch signal instructs both vulval organogenesis and synaptic development of the egg-laying circuit by regulating at least three sequential but independent developmental processes: the AC/VU decision , in which LIN-12/Notch signaling promotes the differentiation of AC and VU cells; VPC specification , in which it determines the vulval precursor cell fates; and vm2 postsynaptic specification , in which it specifies the postsynaptic target cell and further regulates the muscle arm development . While lin-12 , apx-1 and sel-12 null alleles show severe vulval morphogenesis and general vulval muscle morphology phenotypes , all the partial loss-of-function alleles isolated from our screen have specific muscle arm phenotype without other defects . Hence , the muscle arm phenotype shown in lin-12 ( wy750 ) mutant is not likely the consequence of a complete cell-fate conversion . Even in the muscle lineage , several cell-type specific reporters displayed wild type patterns in lin-12 ( wy750 ) mutant ( Figure 1—figure supplement 4 ) , supporting the notion that many aspects of the vm2 differentiation remain intact while only a subset of genes that determines specific postsynaptic development are affected . We can envision two hypotheses to explain the specificity of these alleles . First , vm2 muscle arm development requires a high level of LIN-12/Notch signaling activity , while vulval morphogenesis and overall muscle development require low activity of LIN-12 . The new alleles represent partial loss-of-function mutants that specifically disrupt muscle arm development . Alternatively , LIN-12 could execute two different downstream pathways , one required for vulval morphogenesis and the other required for muscle arm development . Since the lesions in almost all the new alleles are point mutations , they might specifically disable the muscle arm development pathway . We favor the first dosage hypothesis for several reasons . First , the new mutations in lin-12 and apx-1 were found in the extracellular domains , but not in the signal transducing cytosolic domain of LIN-12 . In addition , a nonsense mutation in sel-12 ( wy756 , W184stop ) caused strong vulval phenotypes , while a missense mutation in sel-12 ( wy760 , G373D ) showed specific vm2 muscle arm phenotype , further arguing for the difference in the dosage requirement for these developmental events . Along the same line , a lag-1 dominant negative construct expressed in the vulval muscles led to muscle arm defects , suggesting that the same canonical LIN-12/Notch signaling pathways is required for the muscle arm development . Second , the lin-12 ( e2621 ) allele , showed muscle arm defects ( Figure 5—figure supplement 1B ) with intact vulval morphogenesis ( Wu et al . , 1998 ) . Sequencing of the lin-12 locus of this mutant revealed a 541-bp deletion in the promoter region with no mutation in the coding sequence . The deleted sequence in the promoter region contains two LAG-1 consensus binding sites and therefore is likely to disrupt the positive feedback loop , which could in turn reduce the level of LIN-12/Notch signaling . In fact , the expression of the hlh-29 transcription reporter was down-regulated in lin-12 ( e2621 ) ( data not shown ) . Together , these data strongly argue that vulval morphogenesis requires a low level of LIN-12/Notch signaling activation , while vm2 postsynaptic specification requires a relatively high level of LIN-12/Notch activity . Consistent with this model , mir-61::gfp , another transcription reporter expressed in VPCs during its specification process , does not change in the lin-12 ( wy750 ) mutant ( data not shown ) , suggesting that the alternation of LIN-12/Notch signaling activity in lin-12 ( wy750 ) has little or no effect in VPC specification . The function of LIN-12 in the AC/VU decision is a case of classic lateral inhibition . A small initial difference in LIN-12/Notch activity between two adjacent cells is amplified by reciprocal feedback mechanisms . The enhanced difference of LIN-12/Notch activity eventually leads to alternative cell fates ( Greenwald et al . , 1983; Seydoux and Greenwald , 1989; Wilkinson et al . , 1994 ) . However , in vm2 muscle arm development , the lin-12 expression is down-regulated in the apx-1 ( wy755 ) mutant , while the apx-1 expression does not change in the lin-12 ( wy750 ) mutant . These results indicate a different unilateral induction of LIN-12/Notch signaling activated by APX-1 . The stereotyped vm2 muscle arm structure indicates a specific guidance mechanism regulating the muscle arm extension . Given the fact that lin-12 is expressed and cell-autonomously required in the postsynaptic target cell vm2 , and that the activity of canonical LIN-12/Notch signaling pathway is involved , we speculate that the putative guidance receptors could be the downstream target genes of the canonical LIN-12/Notch pathway . While the canonical LIN-12/Notch signaling has been demonstrated to function in axon guidance and dendrite morphogenesis in fly and mammals ( Berezovska et al . , 1999; Endo et al . , 2007 ) , its transcriptional targets remain elusive ( Giniger , 2012 ) . UNC-40/DCC is a good candidate of the putative guidance receptor . unc-40/DCC encodes a C . elegans homolog of UNC-6/Netrin receptor ( Serafini et al . , 1994; Chan et al . , 1996; Keino-Masu et al . , 1996 ) . It has been suggested to serve as a receptor to guide the growth of axon as well as the extension of dendrite towards the ligand UNC-6/Netrin ( Hedgecock et al . , 1990; Teichmann and Shen , 2011 ) . More recently , UNC-40 , together with its cytoplasmic activator MADD-2 , was shown to regulate body wall muscle arm extension and axon guidance ( Alexander et al . , 2010; Hao et al . , 2010; Morikawa et al . , 2011 ) . Indeed , we found that unc-40 and madd-2 loss-of-function mutants showed vm2 muscle arm defects , similar to the phenotype found in the lin-12 ( wy750 ) mutants . Both UNC-40 and MADD-2 are expressed in the postsynaptic vm2 cells but show little or no expression in the non-target vm1 cells . Cell-autonomous rescue of UNC-40 and MADD-2 in the vm2 cells further supports the notion that these two proteins function in vm2 to direct muscle arm development . Interestingly , the expression of the translational reporters for both genes in vm2 is dramatically down-regulated in lin-12 ( wy750 ) mutant , suggesting that both genes function downstream of LIN-12/Notch signaling pathway . Consistent to the previous studies in the C . elegans body wall muscle arms ( Dixon and Roy , 2005; Alexander et al . , 2009 ) , we did not observe muscle arm defects in unc-6 mutants ( data not shown ) , suggesting the existence of a novel ligand of UNC-40 that functions in guiding the muscle arm extension . The putative ligand provided by vulval epithelial cells could be either a short-range cue that is associated on membrane or locally secreted , or a globally diffusive cue . We favor the former prediction based on the observation that the relative distance between the vulval muscles and the vulval opening is crucial . SM migration is affected in egl-15 ( n484 ) mutant , which causes posteriorly displaced vulval muscles descended from an SM that stops prematurely during the migration ( Goodman et al . , 2003 ) . Different phenotypes were observed , which appeared to correlate with the location of the vulval muscles . The muscle arms that were located far away from the vulva displayed severe morphogenesis defects , whereas when the vulval muscles did not reach but were still closed to the correct places , we observed muscle arms growing with longer shafts to reach the correct locations defined by HSN presynaptic regions ( data not shown ) . These observations suggest that vulval epithelial cells might provide a membrane-associated or locally secreted cue to guide the vm2 muscle arm extension . Our additional experiments showed that both unc-40 and madd-2 are regulated by LIN-12/Notch signaling at the level of gene transcription . The transcription of unc-40 and madd-2 are both specifically activated in vm2 and dependent on lin-12 , indicating that they could be the direct transcriptional targets of LIN-12/Notch signaling . Consistent with this hypothesis , sequence analysis of madd-2 promoter and unc-40 regulatory elements uncovered several LAG-1 consensus binding sites . Although multiple lines of evidence indicated that UNC-40 and MADD-2 physically interact with each other and function in the same pathway ( Alexander et al . , 2010 ) , the mechanism by which unc-40 and madd-2 function is still elusive . In both the Netrin-mediated axon attracting process and Netrin-independent body wall muscle arm extension process , madd-2 potentiates the activity of unc-40 ( Hedgecock et al . , 1990; Dixon and Roy , 2005; Alexander et al . , 2009 ) . Consistent to the previous observations , a madd-2 null allele showed quantitatively less dramatic defects in vm2 muscle arm extension than did unc-40 or lin-12 mutants . In addition , we observed only marginally significant enhancement of these muscle arm defects in an unc-40; madd-2 double mutant . Furthermore , the expression of unc-40 significantly rescued the vm2 extension phenotype in madd-2 mutant , while the expression of madd-2 only showed slight rescue of the unc-40 mutant . Forced expression of unc-40 in non-target vm1 cells was sufficient to induce muscle arms , whereas the overexpression of madd-2 could not achieve or enhance this phenotype . Taken together , these results demonstrate that unc-40 , potentiated by madd-2 , plays a deterministic role in guiding the muscle arm extension and the postsynaptic target selection . Our studies of the C . elegans egg-laying circuit have yielded surprises in the mechanisms of synapse formation in vivo . Although many in vitro studies suggest that direct interactions between pre- and postsynaptic cells trigger the mutual differentiation processes leading to synaptogenesis , our experiments revealed a significantly different picture . First , the formation of pre- and postsynaptic specializations can occur at different times . The HSN presynaptic specializations form at the early L4 stage , significantly before the maturation of the postsynaptic targets . The extension of the postsynaptic muscle arms takes place 5–6 hr later . Second , the development of the pre- and postsynaptic specializations appears to be independent of each other . Eliminating the postsynaptic cells does not affect the timing and location of synaptic vesicle accumulation in HSN neurons ( Shen and Bargmann , 2003 ) . Similarly , ablation of the presynaptic neurons has no obvious effect on the timing and direction of the muscle arm growth ( Figure 1—figure supplement 2A , B ) . Third , pre- and postsynaptic differentiations are guided by two different but related guidepost cells . Through genetic analysis , we previously found that the primary vulval epithelial cells serve as guidepost cells for the HSN presynaptic development ( Shen and Bargmann , 2003; Shen et al . , 2004 ) . These primary cells express a transmembrane IgSF protein SYG-2 , which clusters and activates a related molecule , SYG-1 on the HSN axon . The intercellular interaction between SYG-1 and SYG-2 defines the location of HSN synapses and initiates the presynaptic formation ( Shen and Bargmann , 2003; Shen et al . , 2004 ) . Interestingly , postsynaptic differentiation is likely guided by the secondary vulval epithelial cells ( Figure 7—figure supplement 3 ) . APX-1/DSL expressed in the secondary vulval epithelial cells and the non-target vm1 cells activates the LIN-12/Notch signaling pathway in the vm2 cells . Subsequently , UNC-40/DCC and MADD-2 are up-regulated leading to the selective muscle arm formation in vm2 . The muscle arms appear to grow along the interface between the primary and the secondary vulval epithelial cells through possibly an UNC-40-dependent mechanism , which ensures that the pre- and postsynaptic compartments will contact each other . The epithelial cells not only function as the channel for egg-laying but also play organizational roles for the development of the entire vulval organ , coordinating multiple events including vulval muscle migration , axon guidance of HSN , VC axon branching and synapse formation . Our discoveries reveal that many of the principles of synaptic organization in vivo might involve cells other than the pre- and postsynaptic neurons .
Worms were maintained at 20°C on OP50 Escherichia coli-seeded nematode growth medium ( NGM ) plates . N2 Bristol strain worms were used as the wild-type reference . The following mutants were used in this study: unc-104 ( e1265 ) II , egl-1 ( n986 ) V , lin-3 ( e1417 ) IV , lin-39 ( n709 ) III , egl-15 ( n484 ) X , unc-40 ( e271 ) I , unc-40 ( n324 ) I . The mutants strains unc-32 ( e189 ) ; lin-12 ( n676n930 ) III , lin-12 ( n941 ) /eT1 ( III ) ; him-5 ( e1467 ) /eT1[him-5 ( e1467 ) ] ( V ) , apx-1 ( or3 ) /uT1[unc- ? ( n754 ) ;let- ? ] ( IV , V ) , sel-12 ( ty11 ) X , dsl-1 ( ok810 ) IV , glp-1 ( e2141 ) III and madd-2 ( ok2226 ) V were obtained from Caenorhabditis Genetics Center . lin-12 ( e2621 ) III was kindly provided by Iva Greenwald . Fluorescence images were captured in live C . elegans using a Plan-Apochromat 63 3 1 . 4 objective on a Zeiss LSM710 confocal microscope . Worms were immobilized using a mixture of 200 mM 2 , 3-butanedione monoxime ( Sigma-Aldrich , St Louis , MO ) and 2 . 5 mM levamisole ( Sigma-Aldrich ) in 10 mM Na HEPES . For quantifying the vulval muscle sizes and the intersection angles between vulval slit and vulval muscles , 12–16 confocal images were used for each genotype . Each vulval muscle was lineated by polygonal lines in ImageJ to obtain the size . The middle points of the connection lines where vm1 was linked to the body wall , and the tips of vm2 cells were used as the landmarks to measure the intersection angles . Twenty L4-stage worms of each genotype were isolated and allowed to develop 30 hr at 20°C into gravid adults . The worms were transferred to a fresh plate and allowed to lay eggs for 2 hr at 20°C . The adult worms were removed before the total number of each plate and the stage of each egg were scored . The eggs were classified into one of three developmental stages: 1–8 cell , eight-cell to comma , or post-comma . Three independent assays were performed with at least 50 eggs quantified in total for each genotype . Fisher's exact test was performed to compare statistical significance between two groups . To perform Ca2+ imaging in the vulval muscles , we used as a control strain LX1890 vsIs153; lite-1 ( ce314 ) , which expresses GCaMP3 and mCherry from the unc-103e promoter and is blue light insensitive , as previously described ( Collins and Koelle , 2012 ) . KG1180 animals carrying the lite-1 ( ce314 ) mutation ( Edwards et al . , 2008 ) were crossed to lin-12 ( wy750 ) or apx-1 ( wy755 ) animals to generate LX1892 lin-12 ( wy750 ) ; lite-1 ( ce314 ) and LX1893 apx-1 ( wy755 ) ; lite-1 ( ce314 ) , respectively . LX1890 vsIs153; lite-1 ( ce314 ) animals were crossed with LX1892 lin-12 ( wy750 ) ; lite-1 ( ce314 ) and LX1893 apx-1 ( wy755 ) ; lite-1 ( ce314 ) to generate LX1889 lin-12 ( wy750 ) ; vsIs153; lite-1 ( ce314 ) and LX1890 vsIs153; apx-1 ( wy755 ) ; lite-1 ( ce314 ) . Ratiometric Ca2+ imaging of behaving animals was performed as previously described ( Collins and Koelle , 2012 ) . Single worms aged 24 hr after the L4 stage were picked with a small amount of OP50 bacteria to an unseeded NGM plate . From this plate , a ∼20 × 20-mm chunk was placed worms-side down onto a 24 × 60 mm #1 coverslip and overlaid with a 22 × 22 mm #1 coverslip . The worms were allowed to recover for 1–2 hr in a humidified chamber at 20°C before imaging . Two-channel confocal slices ( 18-µm-thick , 256 × 256 pixel , 16-bit ) were collected for 6 min at 20 Hz through a 20× Plan-Apochromat objective ( 0 . 8NA ) using the LIVE detector of an inverted Zeiss 710 Duo confocal microscope . The stage and focus were adjusted manually to keep the egg-laying system in view and focused during recording periods . Ratiometric analysis was performed in Volocity ( version 5 . 4; Perkin Elmer ) . Briefly , separate GCaMP3/mCherry ratio and intensity-modulated ratio channels were calculated . Voxels with mCherry fluorescence intensities 2 standard deviations above background were selected as objects for measurement ( typically ∼500 voxels per time point ) . The average ratio of these voxels were smoothened using a 150 ms ( three time points ) rolling average , and the lowest 10% of the GCaMP3/mCherry ratio values were averaged to establish a ΔR/R ( change in GCaMP3/mCherry fluorescence ratio/absolute ratio ) baseline . Ca2+ transients were identified by visual inspection of intensity-modulated ratio movies and ratio traces , and ΔR/R peaks above 15% were analyzed . Transients were classified as single when the transient was evidently confined to the anterior or posterior vulval muscle and double when both anterior and posterior muscles had obvious transients . Double transients that resolved to a single peak in the ratiometric trace were defined as having no delay ( 0 s , Figure 2K ) . When two or more transients within a single body bend were separable both spatially and temporally in the intensity-modulated ratio movies and had independent peaks separated by Δ250 ms in the ratiometric trace , the delay between the anterior and posterior peaks was determined . | The development of the nervous system involves the formation of complex networks of connections between diverse cell types , such as motor neurons , interneurons and pyramidal cells . However , the mechanisms by which individual cells are programmed to acquire particular identities , and how they are instructed to form connections with other specific cells , remain unclear . In many species , the Notch signaling pathway has a role in setting up these networks . Notch is a transmembrane protein , which means that it has one component inside the cell and another outside . When a ligand binds to the extracellular part of Notch , this causes the receptor to break in two . The intracellular domain then travels to the nucleus where it can influence gene expression . The nematode worm ( C . elegans ) , which has two Notch receptors , is often used to study the formation of neuronal networks because each worm has only around 300 neurons , and they are connected in roughly the same way in each worm . C . elegans relies on two types of cell that are very similar to each other—type-1 and type-2 vulval muscle cells—to lay eggs , and the neurons that trigger egg-laying form synaptic connections on specialized structures called muscle arms . However , these structures are found only in type-2 vulval muscle . To investigate the mechanisms underlying the formation of the egg-laying circuit , Li et al . screened large numbers of mutant worms to find animals that lacked muscle arms . They identified a number of such mutants , which laid fewer eggs compared to wild-type worms , and found that they all had mutations in genes that encode for proteins or ligands that are involved in the LIN-12/Notch pathway . This pathway mediates cell–cell interactions that help to specify cell fates . Li et al . showed that type-2 vulval muscle cells develop muscle arms when their neighbors—type-1 vulval muscle cells and vulval epithelial cells—produce enough ligand to activate the LIN-12 Notch receptor on the type-2 vulval muscle cells . They also identified two of the downstream targets of LIN-12 , and found that artificially expressing one of these in type-1 vulval muscle cells is sufficient to trigger the formation of muscle arms . The work of Li et al . provides further evidence that the Notch signalling pathway , which is well known for its role in early development , also acts at later developmental stages to determine cell fate and patterns of connectivity . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"developmental",
"biology",
"neuroscience"
] | 2013 | LIN-12/Notch signaling instructs postsynaptic muscle arm development by regulating UNC-40/DCC and MADD-2 in Caenorhabditis elegans |
Learning requires neural adaptations thought to be mediated by activity-dependent synaptic plasticity . A relatively non-standard form of synaptic plasticity driven by dendritic calcium spikes , or plateau potentials , has been reported to underlie place field formation in rodent hippocampal CA1 neurons . Here , we found that this behavioral timescale synaptic plasticity ( BTSP ) can also reshape existing place fields via bidirectional synaptic weight changes that depend on the temporal proximity of plateau potentials to pre-existing place fields . When evoked near an existing place field , plateau potentials induced less synaptic potentiation and more depression , suggesting BTSP might depend inversely on postsynaptic activation . However , manipulations of place cell membrane potential and computational modeling indicated that this anti-correlation actually results from a dependence on current synaptic weight such that weak inputs potentiate and strong inputs depress . A network model implementing this bidirectional synaptic learning rule suggested that BTSP enables population activity , rather than pairwise neuronal correlations , to drive neural adaptations to experience .
Activity-dependent changes in synaptic strength can flexibly alter the selectivity of neuronal firing , providing a cellular substrate for learning and memory . In the hippocampus , synaptic plasticity plays an important role in various forms of spatial and episodic learning and memory ( Nakazawa et al . , 2004 ) . The spatial firing rates of hippocampal place cells have been shown to be modified by experience and by changes in environmental context or the locations of salient features ( O’Keefe and Conway , 1978; Mehta et al . , 1997; Lever et al . , 2002; Dupret et al . , 2010; Zaremba et al . , 2017; Turi et al . , 2019; Ziv et al . , 2013; Muller and Kubie , 1987; Bostock et al . , 1991; Fyhn et al . , 2007; Leutgeb et al . , 2005 ) . These modifications can occur rapidly , even within a single trial ( Hill , 1978; Mehta , 2015; Monaco et al . , 2014; Bittner et al . , 2015; Bittner et al . , 2017; Diamantaki et al . , 2018; Jezek et al . , 2011; Geiller et al . , 2017; Bourboulou et al . , 2019; Zhao et al . , 2020 ) . Here , we investigate the synaptic plasticity mechanisms underlying such rapid changes in the spatial selectivity of hippocampal place cells . Various forms of Hebbian synaptic plasticity have been considered for decades to be the main , or even only , synaptic plasticity mechanisms present within most brain regions of a number of species ( Magee and Grienberger , 2020 ) . The core feature of such plasticity mechanisms is that they are autonomously driven by repeated synchronous activity between synaptically connected neurons , which results in either increases or decreases in synaptic strength depending on the exact temporal coincidence ( Gerstner et al . , 2018; Keck et al . , 2017; Shouval et al . , 2010; Song et al . , 2000 ) . This includes the so-called ‘three-factor’ plasticity rules that , in addition to pre- and postsynaptic activity , depend on a third factor that extends the time course over which plasticity can function ( Magee and Grienberger , 2020; Gerstner et al . , 2018; He et al . , 2015; Yagishita et al . , 2014 ) . To implement these three-factor plasticity rules , it has been proposed that correlated pre- and postsynaptic activity drives the formation of a synaptic flag or eligibility trace ( ET ) that is then converted into changes in synaptic weights by the delayed third factor , usually a neuromodulatory signal ( Gerstner et al . , 2018; Sajikumar and Frey , 2004; Frey and Morris , 1997 ) . Recently , we reported a potent , rapid form of synaptic plasticity in hippocampal CA1 pyramidal neurons that enables a de novo place field to be generated in a single trial following a dendritic calcium spike ( also called a plateau potential ) ( Bittner et al . , 2015; Bittner et al . , 2017; Diamantaki et al . , 2018 ) . This form of synaptic plasticity , termed behavioral timescale synaptic plasticity ( BTSP ) , rapidly modifies synaptic inputs active within a seconds-long time window around the plateau potential . This relatively long time course suggests that BTSP may be similar to the above-mentioned three-factor forms of plasticity , with synaptic activity generating local signals marking synapses as eligible for plasticity ( ETs ) , and plateau potentials acting as the delayed factor that converts synaptic ETs into changes in synaptic strength . However , BTSP was shown to strengthen many synaptic inputs whose activation did not coincide with any postsynaptic spiking or even subthreshold depolarization detected at the soma ( Bittner et al . , 2017 ) , suggesting that changes in synaptic weight might be independent of correlated pre- and postsynaptic activity , and that BTSP may be fundamentally different than all variants of Hebbian synaptic plasticity ( Gerstner et al . , 2018; Keck et al . , 2017; Shouval et al . , 2010; Mehta , 2004; Golding et al . , 2002 ) . Such a non-standard plasticity rule could enable learning to be guided by delayed behavioral outcomes , rather than by short timescale associations of pre- and postsynaptic activity . In this study , we tested the effect of dendritic plateau potentials on the spatial selectivity of CA1 neurons that already express pre-existing place fields , and therefore exhibit substantial postsynaptic depolarization and spiking prior to plasticity induction . We found that dendritic plateau potentials rapidly translocate the place field position of hippocampal place cells , both by strengthening inputs active near the plateau position and weakening inputs active within the original place field . In order to determine if the increased postsynaptic activity in place cells is causally related to the synaptic depression observed within the initial place field , we performed a series of voltage perturbation experiments , which indicated that the direction of plasticity induced by plateau potentials is independent of postsynaptic depolarization and spiking . Next , we inferred from the data a computational model of the synaptic learning rule underlying this bidirectional form of plasticity , which suggested that it is instead the current weight of each synaptic input that controls the direction of plasticity such that weak inputs potentiate and strong inputs depress . Finally , we implemented this weight-dependent learning rule in a network model to explore the capabilities of bidirectional BTSP to adapt network-level population representations to changes in the environment .
We first examined how plasticity induced by dendritic plateau potentials changes the intracellular membrane potential ( Vm ) dynamics in neurons already exhibiting location-specific firing ( i . e . place cells ) . Intracellular voltage recordings from CA1 pyramidal neurons were established in head-fixed mice trained to run for a water reward on a circular treadmill decorated with visual and tactile cues to distinguish spatial positions ( ~185 cm in length ) . Brief step currents ( 700 pA , 300 ms ) were injected through the intracellular electrode for a small number ( Nakazawa et al . , 2004; O’Keefe and Conway , 1978; Mehta et al . , 1997; Lever et al . , 2002; Dupret et al . , 2010; Zaremba et al . , 2017; Turi et al . , 2019; Ziv et al . , 2013 ) of consecutive laps to evoke plateau potentials at a second location that was between 0 and 150 cm from the initial place field ( labeled ‘Induction 2’ in Figure 1A and B; n = 26 plasticity inductions in 24 neurons ) . In 8/24 neurons a ‘natural’ pre-existing place field was expressed from the start of recording , while in 16/24 the initial place field was first experimentally induced by the same procedure ( labeled ‘Induction 1’ in Figure 1A and B ) . In 2/24 neurons the induction procedure was repeated a third time with plateaus evoked at a different location , resulting in a total of 26 plasticity inductions in cells with pre-existing place fields ( see Figure 1—figure supplement 1E and Materials and methods ) . In most cases the evoked dendritic plateaus shifted the location of the neuron’s pre-existing place field toward the position of the second induction site ( Figure 1A and B ) . Place field firing is known to be driven by a slow , ramping depolarization of Vm from sub- to supra-threshold levels ( Bittner et al . , 2015; Harvey et al . , 2009 ) . Isolation of these low-pass filtered Vm ramps ( Figure 1—figure supplement 2A-H; Materials and methods ) revealed that plateau potentials likewise shifted the neuron’s Vm ramp toward the position of the plateau , such that the new Vm ramp peaked near the plateau position in most neurons ( average distance = 19 . 5 ± 4 . 7 cm; n = 26; Figure 1C–E and Figure 1—figure supplement 2; example cells shown in Figure 1C are indicated with matching colored arrows in Figure 1D ) . We also observed similar shifts in place field position to be induced by spontaneous , naturally occurring plateau potentials in a separate set of recordings ( n = 5; Figure 1—figure supplement 2I-M ) . The spatial profile of plateau-induced Vm changes ( ΔVm ) ( Figure 2A ) was obtained by subtracting the average Vm ramp for trials occurring before plateau initiation ( Figure 1C; before ) from the average Vm ramp for trials occurring after ( Figure 1C; after ) . These data indicate that plateaus induced both positive and negative changes to Vm ramp amplitude ( Figure 2A and B ) . In general , the increases in Vm depolarization peaked near the position of the plateau , while the negative changes peaked near the initial place field ( Figure 2A and B , and Figure 2—figure supplement 1A and B ) . Although these changes varied considerably in magnitude across cells , the peak change in the positive direction was greater than the peak change in the negative direction ( mean positive change± SEM vs . mean negative change± SEM: 6 . 73 ± 0 . 73 mV vs . 3 . 89 ± 0 . 32 mV , n = 26 inductions; p = 0 . 0001 , paired two-way Student’s t-test; Figure 2—figure supplement 1A ) . Aligning each ΔVm trace to the position of the plateau ( Figure 2A and B ) demonstrates that the increases in Vm depolarization observed near the plateau position decay with distance , eventually becoming hyperpolarizing decreases in Vm . At even greater distances from a plateau , ΔVm decays back to zero ( Figure 2B ) . To summarize the data presented thus far , dendritic plateau potentials change the location of place field firing by depolarizing Vm around the plateau position and hyperpolarizing Vm at positions within a pre-existing place field . Previously we showed that location-specific increases in Vm depolarization induced by plateau potentials are the result of synapse-specific increases in the strength of spatially tuned excitatory inputs ( Bittner et al . , 2017 ) . The above results suggest that , in addition to this synaptic potentiation , BTSP is also capable of inducing synaptic depression to cause location-specific decreases in Vm depolarization . In analyzing the spatial extent of the Vm changes induced by plateaus , we observed a strong linear relationship between the width of the resulting ΔVm and the running speed of the animal during plateau induction laps ( Figure 2C and D ) , which had a slope on the order of seconds . This suggested that the run trajectory of the animal ( Figure 2C ) affected the spatial extent of the plasticity ( Figure 2A and B ) by determining which positions were traversed within a fixed seconds-long temporal window for plasticity , as we previously reported ( Bittner et al . , 2017 ) . Therefore , we next analyzed the temporal relationship between plateau potentials and location-specific potentiation and depression . To do this , we used the running trajectory of the mice during plateau induction trials ( Figure 2C ) as a time base for ΔVm ( Figure 2E; see also Figure 1—figure supplement 1 , Figure 1—figure supplement 2A-H and Materials and methods ) . This analysis showed that the positive and negative changes to Vm induced in place cells occurred over a timescale of multiple seconds ( Figure 2F ) , with the positive changes appearing to be asymmetric with respect to the onset time of the plateaus ( ratio of potentiation duration before/after plateau onset: 2 . 2; black circles and crossmarks in Figure 2F mark the time points when ΔVm crosses zero ) . This asymmetry was similar to that observed for the positive Vm changes induced by BTSP in silent cells ( Bittner et al . , 2017 ) . The negative changes ( i . e . the hyperpolarizations indicative of synaptic depression ) occurred within a time window between ±2 and ±6 s from the plateau in many neurons that expressed pre-existing place fields ( Figure 2E and F ) . Notably , this hyperpolarization was greatly reduced , or even absent , in a set of place cells where the time delay between plateau onset and the initial place field Vm ramp was greater than 4–5 s ( red traces in Figures 1C , 2A and E; see also Figure 2—figure supplement 1C-F ) , further indicating the time delimited aspect of the depression component . These data reinforce the idea that BTSP is a bidirectional form of synaptic plasticity with a seconds-long timescale that enables dendritic plateau potentials to shift the locations of hippocampal place fields by inducing both synaptic potentiation and depression . We next sought to understand why dendritic plateaus induce both Vm depolarization and Vm hyperpolarization in cells expressing pre-existing place fields ( Figures 1 and 2 ) , but induce only Vm depolarization in spatially untuned silent cells ( Figure 3—figure supplement 1; Bittner et al . , 2017 ) . Figure 3A shows that the initial temporal profile of Vm in place cells with pre-existing place fields was highly variable across neurons , as plateaus were experimentally induced at different temporal intervals from the existing place field in different neurons . In contrast , the change in Vm ( ΔVm ) induced by plateaus showed a more consistent shape in time that appeared to depend on the initial level of Vm depolarization at each time point prior to plasticity ( Figure 3B ) . Large positive changes occurred at time points with relatively hyperpolarized initial Vm , while time points with more depolarized initial Vm were associated with less positive and more negative ΔVm . These changes resulted in final Vm profiles that were highly similar across neurons , regardless of the initial Vm ( Figure 3C ) . These results indicate that BTSP induces variable changes in synaptic strength that reshape the selectivity of neurons toward a common target shape – a place field centered near the location of evoked plateau potentials that decays toward baseline over many seconds in each direction . In Figure 3D-F , we examined this further by comparing data from initially hyperpolarized silent cells ( black; n = 29 inductions , see Figure 3—figure supplement 1 and Materials and methods ) to data from place cells ( dark red; n = 26 inductions ) . Place cells were on average more depolarized before plasticity than silent cells ( Figure 3D ) , and more depression occurred in place cells compared to silent cells ( Figure 3E ) . However , each place cell had both spatial positions where it was depolarized within its place field , and positions where it was hyperpolarized out-of-field . To determine if spatial positions that were initially depolarized were associated with larger depression , we grouped Vm ramp data from all place cells , considering only spatial bins where each cell was more depolarized than a threshold of –56 mV ( light red traces labeled ‘PCs ( within-field ) ’ in Figure 3D-F ) . Indeed , more depression and less potentiation was induced in place cells at those spatial positions that were initially most depolarized ( Figure 3E ) . However , the final Vm ramps after plasticity were less sensitive to the initial state of depolarization across spatial bins of place cells ( Figure 3F ) . This analysis further supported the findings that , while changes in Vm induced by plateaus were highly dependent on initial Vm , these changes drove the resulting final Vm ramp toward a common target shape ( Figure 3F ) . Indeed , when all spatial bins from all place cells were analyzed , ΔVm showed a strong inverse correlation with initial Vm ( m = –0 . 91; Figure 3G ) . In contrast , final Vm showed a very weak positive correlation with initial Vm ( m = 0 . 04; Figure 3H ) , which reflects that some spatial bins show no change in Vm during plasticity , either because they were traversed outside the temporal window for plasticity or because the Vm at those positions had already reached a final Vm target value . That BTSP induces variable changes in Vm that reshape the Vm ramp toward a particular target shape is further evident from a heatmap depicting the relationships of ΔVm to both initial Vm ramp depolarization and time from plateau onset ( Figure 3I , positive ΔVm in red , and negative ΔVm in blue; see Materials and methods ) . The white regions of this plot trace out a temporal profile of Vm that corresponds to the final target place field shapes shown in Figure 3C and F . All initial deviations from this equilibrium Vm profile resulted in either positive or negative changes to approach this target place field shape ( see dashed arrows ) . It should also be noted that the depression of Vm in place cells appeared to be weaker than the potentiation , leaving some residual depolarization at positions distant from the peak ( Figure 3F ) . The functional significance of this is unclear , but may suggest that BTSP induces synaptic depression at a slower rate than potentiation ( Cone and Shouval , 2021 ) . To summarize , BTSP induces precise changes in synaptic strength that modify pre-existing place fields with any initial shape such that they approach a target shape that peaks near the location where dendritic plateaus were evoked . Altogether these data revealed that , in general , the magnitude and direction of ΔVm depended on the time from the plateau potential , and correlated inversely with the initial Vm ramp amplitude prior to plasticity induction . Does this anti-correlation reflect a causal relationship between postsynaptic depolarization and changes in synaptic weight induced by BTSP ? This possibility would require that small depolarizations induce synaptic potentiation and large depolarizations induce synaptic depression , which is actually opposite to what has been observed in CA1 pyramidal cells with a variety of other plasticity protocols ( Shouval et al . , 2010; Yang et al . , 1999; Graupner and Brunel , 2012; Clopath et al . , 2010; Clopath and Gerstner , 2010; Jedlicka et al . , 2015 ) . Furthermore , the increased Vm depolarization within a cell’s place field also reflects the activation of strongly weighted synaptic inputs , which have been potentiated by prior plasticity ( Bittner et al . , 2015; Bittner et al . , 2017; Figure 3J ) . Thus , a causal dependency on either Vm or synaptic weight could explain the data so far . To discriminate between these two possibilities , we next devised a set of voltage perturbation experiments . We reasoned that , if increased depolarization and spiking within a cell’s place field causes synaptic depression , then artificially increasing Vm and inducing spiking in otherwise silent cells would cause plateau potentials to induce negative ΔVm . Likewise , artificially decreasing Vm and preventing spiking in place cells would prevent plateau potentials from inducing negative ΔVm ( Figure 3K ) . On the contrary , if the direction of plasticity depended instead on the initial strengths of synapses prior to plasticity , these voltage manipulations would have no effect on the balance between positive and negative ΔVm ( Figure 3K ) . It is important to note that these somatic voltage manipulations are not expected to strictly control or even completely overwhelm Vm at the synaptic sites relevant to plasticity induction ( Magee and Johnston , 1997; Koester and Sakmann , 1998; Froemke et al . , 2005 ) due to attenuation of current and voltage along the dendritic cable ( Magee , 1998; Golding et al . , 2005 ) , and compartmentalization of synaptic voltage in dendritic spines ( Harnett et al . , 2012 ) . However , by either increasing or decreasing the generation of somatic action potentials , this manipulation will unequivocally alter the number of action potentials that back-propagate into dendrites , which will in turn influence the activation of voltage-gated channels in dendrites and spines ( e . g . Na+ channels , Ca2+ channels , and NMDA-Rs ) ( Magee and Johnston , 1997; Takahashi and Magee , 2009 ) . Expected changes to the mean Vm in active dendritic spines were supported by simulations of a biophysically and morphologically detailed CA1 place cell model expressing voltage-gated ion channels and receiving rhythmic excitation and inhibition to mimic the in vivo recording conditions ( Figure 4—figure supplement 1; Grienberger et al . , 2017 ) . Moreover , manipulation of somatic Vm and spike timing is widely used to successfully influence plasticity induction in vitro and in vivo ( Malinow and Miller , 1986; Jacob et al . , 2007; Schulz et al . , 2010 ) . According to the above scheme , we first recorded from spatially untuned silent cells , and injected current ( ~100 pA ) through the intracellular pipette to depolarize the neurons’ Vm by ~10 mV and to increase spiking during plasticity induction trials ( Figure 4A; baseline trials mean AP rate: 0 . 26 ± 0 . 25 Hz; first induction trial mean AP rate: 4 . 8 ± 1 . 4 Hz , n = 8; blue trace in Figure 4B ) . In all neurons tested , we observed plateau potentials to induce large positive ΔVm at spatial positions surrounding the plateau location , and no negative ΔVm at any spatial positions ( Figure 4A; blue trace in Figure 4C ) . This result is inconsistent with a causal dependence on initial Vm ( Figure 3K ) , which predicted a ΔVm profile similar to that of control place cells at their most depolarized positions within their pre-existing place fields ( red traces , ‘control PCs ( within-field ) ’ in Figure 4B and C repeated from Figure 3D and E for comparison ) . Next , we performed the inverse manipulation by recording from place cells and injecting current ( ~–150 pA ) to hyperpolarize the neurons’ Vm by ~–15 mV and prevent spiking at spatial locations surrounding their pre-existing place fields while plasticity was induced at a second location ( Figure 4D; baseline trials in-field mean AP rate 10 . 66 ± 0 . 93 Hz; first induction trial in-field mean AP rate 0 . 06 ± 0 . 06 Hz , n = 5; green trace in Figure 4E ) . This manipulation did not prevent negative ΔVm at positions within the original place field ( Figure 4D; green trace in Figure 4F ) , again incompatible with synaptic depression requiring elevated postsynaptic depolarization and spiking ( Figure 3K ) . In fact , full amplitude synaptic depression was observed at locations within the original place field despite the somatic Vm being more hyperpolarized than either the silent cell ( black traces in Figure 4E and F ) or control place cell groups ( red traces , ‘control PCs’ in Figure 4E and F ) . These data clearly show that the direction of plasticity induced by dendritic plateau potentials is not determined by the activation state of the postsynaptic neuron . Instead , the results of these voltage perturbation experiments support the alternative hypothesis that it is the initial strength of each synapse that controls whether an input will be potentiated or depressed by BTSP ( Figure 3K ) . However , the magnitude of potentiation and depression was slightly affected by the voltage perturbations ( e . g . potentiation was slightly but significantly increased in silent cells during artificial depolarization compared to control , Figure 4C ) . This is consistent with the previously reported finding that BTSP induction requires activation of voltage-dependent ion channels , including NMDA-type glutamate receptors ( NMDA-Rs ) and voltage-gated calcium channels ( Bittner et al . , 2017 ) , which would have predicted BTSP to depend on postsynaptic depolarization . To examine this further , we performed an additional set of experiments in which silent cells were strongly hyperpolarized by somatic current injection ( ~–50 mV for ~3 s just before plateau initiation ) during plasticity induction ( Figure 4—figure supplement 2 ) . This manipulation decreased synaptic potentiation ( Figure 4—figure supplement 2 ) , consistent with a requirement for activation of voltage-dependent NMDA-Rs . That such a large , non-physiological level of global Vm hyperpolarization was required to alter BTSP reinforces the finding that , operationally , the dependence is not on voltage signals associated with neuronal activation state ( sustained somatodendritic Vm and action potentials ) , but rather on those associated with synaptic input ( transient local spine depolarization ) ( Beaulieu-Laroche and Harnett , 2018 ) . Finally , these experiments do not support a role for synaptic depolarization in determining the direction of changes in synaptic strengths . The above voltage perturbation experiments suggested that the form of synaptic plasticity underlying BTSP does not depend on the activation state of the postsynaptic neuron ( Figure 4 ) . This contrasts with Hebbian plasticity rules that typically depend on either the firing rate or depolarization of the postsynaptic cell to determine the amplitude and direction of changes in synaptic weight . Another difference is that BTSP appears to be inherently stable , converting synaptic potentiation into depression when input strengths exceed a particular range , whereas most models of Hebbian learning require additional homeostatic mechanisms to counteract synaptic potentiation in highly active neurons ( Oja , 1982; Bienenstock et al . , 1982; Abbott and Nelson , 2000; Zenke et al . , 2013; Turrigiano and Nelson , 2004 ) . To better understand the synaptic learning rule underlying BTSP and its functional consequences , we next sought a mathematical description of BTSP to account for the following features of the in vivo recording data: As mentioned previously , ‘three-factor’ plasticity models propose a mechanism for the strengths of activated synapses to be modified after a time delay – a biochemical intermediate signal downstream of synaptic activation marks each recently activated synapse as ‘eligible’ to undergo a plastic change in synaptic weight . This ‘ET’ decays over a longer timescale than synaptic activation , and while it does not induce plasticity by itself , it enables plasticity to be induced upon the arrival of an additional modulatory biochemical signal . While ‘three-factor’ models consider synaptic ETs to be generated by a coincidence of presynaptic spikes ( factor 1 ) and postsynaptic spikes or sustained depolarization ( factor 2 ) , the results of the above voltage perturbation experiments suggest that if BTSP involves the generation of synaptic ETs , these signals depend only on a single factor – local synaptic activation . In the context of BTSP , the modulatory or ‘instructive signal’ ( IS ) could be instantiated by a dendritic plateau potential . To model this , we assumed that the large magnitude dendritic depolarization associated with a plateau potential ( ~60 mV ) effectively propagates to all synapses ( Xu et al . , 2012 ) , activating an IS at each synapse and allowing a spatially and temporally local interaction between ET and IS to drive plasticity independently at each individual synapse ( Figure 5A ) . To account for plasticity that occurs at inputs activated up to multiple seconds after a plateau , this IS would have to decay slowly enough to overlap in time with ETs generated after the end of the plateau ( Figure 5A ) . Accordingly , we modeled changes in synaptic weights as a function of the time-varying amplitudes of these two biochemical intermediate signals , ET and IS . For simplicity , we first considered how BTSP would change the weight W of a single synapse activated by a single presynaptic spike with precise timing relative to the onset of a plateau potential ( Figure 5A ) . We modeled the synaptic ET as a signal that increases upon synaptic activation at time ts and decays exponentially with time course τET ( see Figure 5A and Materials and methods ) . The IS was modeled as a signal that increases during a plateau potential with onset at time tp and duration d and decays exponentially with time course τIS ( see Figure 5A and Materials and methods ) . Next , we modeled bidirectional changes in synaptic weight dWdt as a function of the temporal overlap or product of these two signals , ET*IS . To account for the observation that BTSP favors synaptic potentiation at weak synapses and synaptic depression at strong synapses , we expressed dWdt in terms of two separate plasticity processes q+ and q− with opposite dependencies on the current synaptic weight W: ( 1 ) dWdt= ( Wmax−W ) ∗k+∗q+ ( ET∗IS ) −W∗k−∗q− ( ET∗IS ) where W is saturable up to a maximum weight of Wmax , and k+ and k− are learning rate constants that control the magnitudes of synaptic potentiation and depression per plateau potential . This formula can be obtained from a two-state model of finite synaptic resources ( see Materials and methods ) . When the current synaptic weight W is near Wmax , the potentiation rate becomes zero , and when W is near zero , the depression rate becomes zero . To calculate the net change in synaptic weight ∆W after plasticity induction , dWdt was integrated in time for the duration of plasticity induction laps . Experimental evidence suggests that synaptic potentiation and depression processes involve biochemical interactions between enzymes ( e . g . phosphokinases-like CaMKII and phosphatases-like calcineurin ) and synaptic protein substrates ( e . g . AMPA-type glutamate receptors ) ( Herring and Nicoll , 2016; Mansuy , 2003 ) . Such concentration-limited reactions are typically saturable and nonlinear ( Graupner and Brunel , 2012 ) . Accordingly , we defined the plasticity processes q+ and q− as saturable ( sigmoidal ) functions of the signal overlap ET*IS ( see Materials and methods ) . If the depression process q− has a lower threshold for activation than the potentiation process q+ ( Graupner and Brunel , 2007; Inglebert et al . , 2020 ) , the resulting change in synaptic weight dWdt is positive and increases monotonically when initial weights are low , but is negative and non-monotonic when initial weights are high ( Figure 5B ) . At intermediate weights , dWdt transitions from negative ( depression ) to positive ( potentiation ) for values of signal overlap ET*IS that are beyond a threshold ( Figure 5B ) . Thus , the largest negative changes in synaptic weight occur when inputs are initially large in weight and signal overlap ET*IS is intermediate in amplitude . This is consistent with the in vivo data , which showed that negative changes in place field ramp Vm were largest at intermediate delays from a plateau ( Figure 3B and I ) . We tested this weight-dependent model of bidirectional BTSP by varying both the timing of a single presynaptic spike relative to a plateau ( Figure 5A ) and the initial weight of the activated synapse ( Figure 5B ) . Model parameters were calibrated ( see Materials and methods ) such that synapses with an initial weight less than a baseline weight of 1 undergo only potentiation , while synapses with higher weight undergo either potentiation or depression , depending on the timing of their activation relative to the plateau ( Figure 5C ) . This produced a profile of changes in synaptic weight similar to the profile of changes in intracellular Vm measured in vivo ( Figure 3I ) . This model also recapitulated the finding that the positive and negative changes in weight induced by BTSP appear to drive synaptic inputs toward a stable target weight , after which additional plateaus do not induce any further changes in strength ( indicated in white , compare Figures 3I and 5C ) . We next exploited the mathematical formulation of the model to analyze these equilibrium conditions in more detail . We defined Weq as the stable equilibrium value of W where potentiation and depression processes are exactly balanced , and the change in weight ∆W is zero over the course of a trial from times t0 to t1 : ( 2 ) ΔW=0= ( Wmax−Weq ) ∗k+∗∫t0t1q+ ( ET∗IS ) dt−Weq∗k−∗∫t0t1q− ( ET∗IS ) dt If we abbreviate the integrated potentiation and depression terms as: ( 3 ) ΔQ+=∫t0t1q+ ( ET∗IS ) dt ( 4 ) ΔQ−=∫t0t1q− ( ET∗IS ) dt then Weq can be expressed as: ( 5 ) Weq=Wmax∗K+∗ΔQ+K+∗ΔQ++k−∗ΔQ− Note that the quantities ΔQ+ and ΔQ− , and therefore the value of Weq , will vary with the activation time of the input ( ts ) , and the onset time ( tp ) and duration ( d ) of a plateau . For a plateau with fixed onset time and duration , this produces a distribution of target equilibrium weights that varies only with the timing of synaptic activation relative to plateau onset ( dashed line in Figure 5C ) , and matches the asymmetric shape of place fields induced by BTSP . In contrast , an alternative version of the model in which the potentiation and depression processes were defined to be linear instead of sigmoidal , predicted a single value for Weq regardless of the timing of synaptic activation ( Figure 5D ) , thus failing to account for the data . Finally , we verified that the model requires long timescales for ET and IS by testing the model with shorter values ( 100 ms ) for the decay time constants τET and τIS ( Figure 5E ) . This was unable to explain changes in synaptic weight at inputs activated at seconds-long time delays to a plateau . Having demonstrated that this weight-dependent model of plasticity at single synapses captures the essential features of BTSP , we next tested if the model can account quantitatively for the in vivo place field translocation data ( Figures 1—3 ) . For this purpose , we assumed that the Vm ramp depolarization measured in a CA1 pyramidal cell during locomotion on the circular treadmill reflects a weighted sum of presynaptic inputs that are themselves place cells with firing rates that vary with spatial position ( see Figure 3J and Materials and methods ) . As a population , the place fields of these inputs uniformly tiled the track , and the firing rate of an individual input depended on the recorded run trajectory of the animal ( Figure 6A , first and second rows ) . In this case , presynaptic activity patterns were modeled as continuous firing rates rather than discrete spike times . For each cell in the experimental dataset ( n = 26 inductions from 24 neurons , Figures 1—3 ) , the initial weight Wi of each presynaptic input i was inferred from the recorded initial Vm , and the changes in weight ∆Wi during plasticity induction laps containing evoked plateau potentials were computed as above ( Equation 1; see Materials and methods ) . The relevant signals modeled for an example lap from a representative cell from the dataset are shown in Figure 6A . Note that , at inputs activated before the onset time of the plateau , changes in synaptic weight ( bottom row ) do not begin until after plateau onset when the instructive signal IS and the signal overlap ET*IS are nonzero . The parameters of the model were optimized to predict the final synaptic weights ( Figure 6B ) and reproduce the final Vm ramp ( Figure 6C ) after multiple plasticity induction laps ( Figure 6—figure supplement 1 , Materials and methods ) . Across all cells , these predictions quantitatively matched the corresponding experimental data ( Figure 6D ) . Finally , the sensitivity of changes in Vm to initial Vm and time to plateau predicted by the model recapitulated that measured from the in vivo intracellular recordings ( Figure 6E and Figure 6—figure supplement 2 ) . The above modeling results help to clarify the differences between BTSP and previously characterized forms of associative synaptic plasticity based on input-output correlations over short timescales ( Gerstner et al . , 2018; He et al . , 2015; Brzosko et al . , 2015; Brzosko et al . , 2017 ) . First , the model supports the hypothesis that a dependence on initial synaptic weight is the actual source of the observed inverse relationship between initial Vm and plasticity-induced changes in Vm ( Figure 3 ) . Second , the scaling of both potentiation and depression by synaptic weight produces a balanced form of plasticity that rapidly stabilizes during repeated inductions ( Figure 1 , and Figure 6—figure supplement 1A and B; Shouval et al . , 2010; Jedlicka et al . , 2015; Bienenstock et al . , 1982; Abraham , 2008; Cooper and Bear , 2012 ) . Third , the time course of BTSP is determined by temporal overlap between slow eligibility signals associated with synaptic activity and slow IS associated with plateau potentials . This selects a subpopulation of synaptic inputs activated with appropriate timing to undergo a change in synaptic strength ( Figure 6—figure supplement 3 ) . Finally , IS are internal signals activated by dendritic plateau potentials , rather than by spiking output , arguing that BTSP is not simply a variant of Hebbian plasticity that depends on input-output correlations over a longer timescale . The above observations imply that BTSP could enable spatial representations to be shaped non-autonomously by delayed behavioral outcomes , if dendritic inputs carrying information about those outcomes are able to evoke plateau potentials ( Muller et al . , 2019 ) . To evaluate the feasibility and implications of this theory , we next considered the conditions that are required for dendritic plateau potentials to be generated in the context of the hippocampal neural circuit . Previous work has shown that ( 1 ) plateau potentials are positively regulated by excitatory inputs from entorhinal cortex ( Bittner et al . , 2015; Takahashi and Magee , 2009; Milstein et al . , 2015 ) , ( 2 ) they are negatively regulated by dendrite-targeting inhibition ( Grienberger et al . , 2017; Milstein et al . , 2015; Lovett-Barron et al . , 2012; Royer et al . , 2012; Palmer et al . , 2012 ) , ( 3 ) they occur more frequently in novel environments ( Cohen et al . , 2017 ) and precede the emergence of new place fields ( Sheffield et al . , 2017 ) , and ( 4 ) introduction of a fixed reward site induces large shifts in the place field locations of many place cells in a population , as assayed by calcium imaging ( Turi et al . , 2019 ) . In order to explore the consequences of these regulatory mechanisms on memory storage by BTSP at the network level , we next constructed a network model of the CA1 microcircuit that incorporates these critical elements to regulate plateau initiation ( Figure 7A ) and implements the above-described weight-dependent model of BTSP ( Figures 5 and 6 ) at each input to the network . In a population of 500 firing rate model CA1 pyramidal neurons , plateaus were positively regulated by a long-range feedback input from entorhinal cortex and negatively regulated by local feedback inhibition ( Figure 7A and B; Stefanelli et al . , 2016 ) . Generation of plateau potentials within the population of CA1 neurons in the model was stochastic , which would result from fluctuations in inputs from entorhinal cortex that occasionally cross a threshold for the generation of a plateau potential in different cells at different times . The presence of reward delivered at a fixed goal location was implemented as an increase in input from entorhinal cortex ( Boccara et al . , 2019; Butler et al . , 2019 ) , although an equivalent increase in plateau generation could result instead from neuromodulatory input that directly increased dendritic excitability or reduced dendritic inhibition ( Sjöström et al . , 2008; Pi et al . , 2013; Tyan et al . , 2014; Guerguiev et al . , 2017 ) . During goal-directed navigation , hippocampal neurons have been shown to preferentially acquire new place fields near behaviorally relevant locations , and to translocate existing place fields toward those locations ( Dupret et al . , 2010; Zaremba et al . , 2017; Turi et al . , 2019; Hollup et al . , 2001; Gauthier and Tank , 2018; Lee et al . , 2020 ) . We modeled this situation by simulating a virtual animal running on a circular treadmill for three separate phases of exploration ( Figure 7C ) . At each time step ( 10 ms ) , instantaneous plateau probabilities were computed for each cell ( Figure 7B ) , determining which neurons would initiate a dendritic plateau and undergo plasticity . During the first few laps of simulated exploration , CA1 pyramidal neurons rapidly acquired place fields that , as a population , uniformly tiled the track ( Figure 7C and D ) . As neurons increased their activity over time , feedback inhibition increased proportionally and prevented further plasticity ( Figure 7A–C ) . During the next phase a goal was presented at a fixed location , resulting in both acquisition of new place fields nearby the goal location in a population of initially silent neurons , and translocation of place fields toward the goal location in a separate population of cells with pre-existing fields ( Figure 7E , left; Figure 7—figure supplement 1 ) . Overall , this resulted in an increased proportion of place cells with fields near the goal position ( Figure 7E , right ) , recapitulating experimentally observed modifications in CA1 network activity during goal-directed behavior ( Zaremba et al . , 2017 ) . The asymmetric time course of BTSP caused the population representation of the goal in the model to peak before the goal location itself , producing a predictive memory representation of the path leading to the goal ( Mehta et al . , 1997; Stachenfeld et al . , 2017 ) . Simulated place cell activity remained stable in a final phase of exploration without reward ( Figure 7C and Figure 7—figure supplement 1 ) . These network modeling results demonstrate that plasticity regulated by local network activity and long-range feedback , rather than by pairwise correlations , can enable populations of place cells to rapidly adapt their spatial representations to changes in the environment without any compromise in selectivity .
In summary , we observed translocation of hippocampal place fields by dendritic plateau potentials and characterized the underlying synaptic learning rule . We found that BTSP is bidirectional , inducing both synaptic potentiation and synaptic depression in neurons expressing pre-existing place fields . The direction of plasticity is determined by the synaptic weight of each excitatory input prior to a plateau potential , and the time interval between synaptic activity and a plateau . The large magnitude of synaptic weight changes enables BTSP to rapidly reshape place field activity in a small number of trials . These results corroborate recent work showing that changes in place field firing in CA1 could be induced by juxtacellular current injection , which was correlated with the occurrence of long duration complex spikes ( Diamantaki et al . , 2018 ) . Here , we used intracellular stimulation and recording to reliably evoke dendritic calcium spikes with precise timing and duration , and to monitor subthreshold changes in Vm dynamics , which enabled inference of the underlying synaptic learning rule . The time and synaptic weight dependence of BTSP suggests that it is driven by an input-specific process rather than nonselective heterosynaptic ( Lynch et al . , 1977 ) or homeostatic plasticity ( Mendez et al . , 2018; Hengen et al . , 2016 ) , or modulation of cellular excitability ( Chandra and Barkai , 2018; Titley et al . , 2017 ) . A significant role for changes in inhibitory synaptic weights is unlikely given that ( 1 ) inhibitory neurons in CA1 exhibit low levels of spatial selectivity ( Grienberger et al . , 2017 ) , ( 2 ) homosynaptic potentiation of excitatory inputs by dendritic plateau potentials can be induced with GABAergic inhibition blocked ( Bittner et al . , 2017 ) , and ( 3 ) inhibitory input to CA1 neurons does not change following induction of synaptic potentiation by BTSP ( Grienberger et al . , 2017 ) . The voltage perturbation experiments we performed ( Figure 4 ) showed that BTSP does not depend on the activation state of the postsynaptic neuron . These results point to a fundamental difference between BTSP and existing Hebbian models of plasticity . In most previous models , including the aforementioned ‘three-factor’ plasticity models , the firing rate ( He et al . , 2015; Brzosko et al . , 2015; Markram et al . , 1997; Bi and Poo , 1998 ) , or sustained level of global depolarization ( Clopath et al . , 2010; Artola et al . , 1990; Brandalise and Gerber , 2014 ) at the time of presynaptic spiking primarily determines whether a synaptic weight increases or decreases ( Gerstner et al . , 2018; Abbott and Nelson , 2000; Caporale and Dan , 2008 ) . Our voltage perturbation experiments ( Figure 4 and Figure 4—figure supplement 2 ) show that the direction of plasticity is not determined by either global depolarization or spiking . This lack of dependence on the postsynaptic activity or output could enable plasticity to be robust to fluctuations in postsynaptic state due to noise or network oscillations ( e . g . theta or gamma ) ( Buzsáki and Moser , 2013 ) , and may allow the postsynaptic state to subserve other functions , such as temporal coding , without interfering with ongoing synaptic weight modifications . Furthermore , while in traditional Hebbian models of plasticity , short timescale synchrony between pre- and postsynaptic activity modifies weights to reinforce pre-existing correlations , BTSP instead provides a mechanism to either create new pairwise activity correlations ‘from scratch’ , or remove pre-existing ones based on delayed outcomes . Our network model ( Figure 7 ) highlights how this fundamental element of BTSP could shape spatial memory storage at the network level , allowing neuronal circuits to rapidly acquire population-level representations of previously unencountered environmental features , and to modify outdated representations . This model also demonstrated that , if plateau potentials are generated by a mismatch between local circuit output and target information relayed by long-range feedback , BTSP can implement objective-based learning ( Richards et al . , 2019b; Sacramento , 2018; Payeur et al . , 2020 ) . Together our experimental and modeling results establish BTSP as a potent mechanism for rapid and reversible learning . In addition to providing insight into the fundamental mechanisms of spatial memory formation in the hippocampus , these findings suggest new directions for general theories of biological learning and the development of artificial learning systems ( Guerguiev et al . , 2017; Payeur et al . , 2020; Bono and Clopath , 2017; Richards and Lillicrap , 2019a; Lillicrap et al . , 2020 ) .
All experimental methods were approved by the Janelia or Baylor College of Medicine Institutional Animal Care and Use Committees ( Protocols 12–84 and 15–126 ) . All experimental procedures in this study , including animal surgeries , behavioral training , treadmill and rig configuration , and intracellular recordings , were performed identically to a previous detailed report ( Bittner et al . , 2017 ) in an overlapping set of experiments , and are briefly summarized here . In vivo experiments were performed in 6- to 12-week-old mice of either sex . Craniotomies above the dorsal hippocampus for simultaneous whole-cell patch clamp and local field potential ( LFP ) recordings , as well as affixation of head bar implants were performed under deep anesthesia . Following a week of recovery , animals were prepared for behavioral training with water restriction , handling by the experimenter , and addition of running wheels to their home cages . Mice were trained to run on the cue-enriched linear treadmill for a dilute sucrose reward delivered through a licking port once per lap ( ~187 cm ) . A MATLAB GUI interfaced with a custom microprocessor-controlled system for position-dependent reward delivery and intracellular current injection . Animal-run velocity was measured by an encoder attached to one of the wheel axles . Plasticity was induced in vivo by injecting current ( 700 pA , 300 ms ) intracellularly into recorded CA1 neurons to evoke dendritic plateau potentials at the same position on the circular treadmill for multiple consecutive laps . In most cases , plateaus were evoked on five consecutive laps ( Figure 1—figure supplement 1E , left ) . However , during some experiments , large changes in the spatial Vm ramp depolarization could be observed to develop after as few as one plateau ( consistent with the observation that plasticity could be induced by a single spontaneously-occurring plateau ) , and so fewer induction laps were used . In other experiments , plateaus were induced on more than five consecutive laps if place field expression remained weak after the first five trials ( Figure 1—figure supplement 1E , left ) . The source of this variability across cells/animals is not yet clear , and requires future investigation . Overall , this procedure induced changes in spatial Vm ramp depolarization in 100% of cells in which it was attempted by three investigators . In some cells , the initial place field was first induced by this procedure , and then the procedure was repeated a second or third time in the same cell with plateaus induced at different locations . In those cases , there was no systematic difference in the number of plateaus required to induce the first place field compared to subsequent fields ( Figure 1—figure supplement 1E , right ) . Since the time window for plasticity induction by BTSP extends for seconds around each plateau , and plateaus were typically evoked on multiple consecutive laps , the changes in synaptic weights induced by BTSP depended on the run behavior of the animals across all induction laps . We showed in Figure 3D that the spatial width of place fields induced by BTSP varied with the average velocity of animals across all plasticity induction laps . Another factor that contributed to the spatial width of induced fields is the proximity of the evoked plateaus to the reward site , as animals tended to stop running briefly to lick near the fixed reward site . Variability across laps in either the run velocity or the duration of pauses could pose a challenge in trying to relate spatial changes in Vm ramp depolarization to the time delay to the plateau ( see below ) . Figure 1—figure supplement 1 shows the full run trajectories of animals during all plasticity induction laps for the five representative example cells shown in Figure 1 . While some variability across induction laps was observed , each animal tended to run consistently at similar velocities across laps . To establish whole-cell recordings from CA1 pyramidal neurons , an extracellular LFP electrode was lowered into the dorsal hippocampus using a micromanipulator until prominent theta-modulated spiking and increased ripple amplitude was detected . Then a glass intracellular recording pipette was lowered to the same depth while applying positive pressure . The intracellular solution contained ( in mM ) : 134 K-gluconate , 6 KCl , 10 HEPES , 4 NaCl , 0 . 3 MgGTP , 4 MgATP , 14 Tris-phosphocreatine , and in some recordings , 0 . 2% biocytin . Current-clamp recordings of intracellular membrane potential ( Vm ) were amplified and digitized at 20 kHz , without correction for liquid junction potential . The silent-cell population of neurons ( n = 29 ) contained recordings from 17 neurons that have been previously reported ( Bittner et al . , 2017 ) . In a subset of experiments ( Figure 4 and Figure 4—figure supplement 2 ) , in addition to position-dependent step current to evoke plateau potentials , additional current was injected either to depolarize neurons beyond spike threshold or to hyperpolarize neurons below spike threshold , during plasticity induction laps . While these perturbations to Vm at the soma are expected to attenuate along the path to distal dendrites ( Golding et al . , 2005 ) , the pairing of back-propagating action potentials with synaptic inputs has been shown to significantly amplify dendritic depolarization ( Jarsky et al . , 2005; Stuart and Häusser , 2001; Migliore et al . , 1999; Schiller and Schiller , 2001 ) . Simulations of a biophysically detailed CA1 place cell model with realistic morphology and distributions of dendritic ion channels ( Grienberger et al . , 2017 ) suggest that somatic depolarization of a silent CA1 cell increases distal dendritic depolarization , and that somatic hyperpolarization of a place cell substantially reduces distal dendritic depolarization at the peak of its place field ( Figure 4—figure supplement 1 ) . To analyze subthreshold Vm ramps , action potentials were first removed from raw Vm traces and linearly interpolated , then the resulting traces were low-pass filtered ( <3 Hz ) . For each of 100 equally sized spatial bins ( ~1 . 85 cm ) , Vm ramp amplitudes were computed by averaging across 10 laps of running on the treadmill both before and after plasticity induction . The spatially binned ramp traces were then smoothed with a Savitzky-Golay filter with wrap-around . Ramp amplitude was quantified as the difference between the peak and the baseline ( average of the 10% most hyperpolarized bins ) . For cells with a second place field induced , the same baseline Vm value determined from the period before the second induction was also used to quantify ramp amplitude after the second induction . Plateau duration was estimated as the duration of intracellular step current injections , or as the full width at half maximum Vm in the case of spontaneous naturally occurring plateaus . Vm ramp half-width ( Figure 2D and Figure 6—figure supplement 1C ) was calculated from the ΔVm traces as the time ( s ) or distance ( cm ) between the plateau and the final return of ΔVm to zero ( or at least to 25% of min; see Figure 2—figure supplement 1G ) . In most cases this only occurred on one side of the plateau , during either the running period before or after the plateau . In 5/26 inductions , the mouse ran so quickly that the ΔVm did not have time to reach 25% of min on either side of the plateau ( Figure 2—figure supplement 1G ) , resulting in an underestimation of the ramp half-width . The average velocity was calculated as the mean velocity of the mouse from the plateau to the end of the plasticity ( Figure 2—figure supplement 1 ) . In order to relate spatial changes in Vm ramp depolarization to the time delay to a plateau ( e . g . Figures 2E , F—4B , C , E , F–6E ) , we assigned to each spatial position the shortest time delay to plateau that occurred across multiple induction laps ( Figure 1—figure supplement 1 ) . This is a conservative estimate , as the shortest delay between presynaptic activity and postsynaptic plateau will generate the largest overlap between ET and IS , and will result in the largest changes in synaptic weight . While this method is imperfect and did discard variability in running behavior across laps , it enabled direct comparison of the time course of BTSP across neurons . We also note that , to generate the modeling results shown in Figure 6 , the full run trajectory of each animal during all induction laps , including pauses , was provided as input to the model ( see details below ) . This resulted in good quantitative agreement between experimentally recorded and modeled spatial Vm ramps ( Figure 6D ) . Since not all possible pairs of initial ramp amplitude and time delay relative to plateau onset were sampled in the experimental dataset , expected changes in ramp amplitude ( e . g . Figure 3I ) were predicted from the sampled experimental or model data points by a two-dimensional Gaussian process regression and interpolation procedure using a rational quadratic covariance function , implemented in the open-source Python package sklearn ( Abraham et al . , 2014; Rasmussen and Williams , 2006 ) . To statistically compare ΔVm vs . time plots among groups each individual induction trace was binned in time ( average of values in 80 , 100 ms , bins from –4 to +4 s ) . The number of points in each bin for each group is as follows: silent cells ( −4 , + 4 s ) : n = 19 , 19 , 19 , 0 , 20 , 20 , 21 , 21 , 25 , 26 , 26 , 27 , 27 , 27 , 27 , 27 , 28 , 28 , 28 , 28 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 29 , 27 , 25 , 25 , 25 , 24 , 21 , 20 , 19 , 17 , 16 , 14 , 14 , 10 , 9 , 9 , 8 , 8 , 7 , 7 , 7 , 7 , 7 , 6 , 6 , 6 , 6 , 6 . Silent + depolarization ( −4 , + 4 s ) : n = 2 , 2 , 4 , 5 , 5 , 6 , 6 , 6 , 6 , 6 , 6 , 7 , 7 , 7 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 7 , 7 , 6 , 6 , 6 , 6 , 5 , 5 , 5 , 5 , 3 , 3 . Depolarized PCs ( –4 to +4 s ) : n = 6 , 6 , 6 , 6 , 6 , 5 , 5 , 5 , 4 , 4 , 5 , 6 , 6 , 6 , 7 , 7 , 6 , 6 , 6 , 7 , 7 , 8 , 7 , 7 , 7 , 7 , 6 , 6 , 5 , 5 , 5 , 4 , 4 , 4 , 4 , 4 , 4 , 4 , 4 , 4 , 4 , 5 , 5 , 5 , 5 , 5 , 5 , 5 , 6 , 7 , 8 , 8 , 8 , 9 , 10 , 14 , 14 , 15 , 15 , 15 , 15 , 15 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 14 , 13 , 12 , 12 , 10 , 9 , 9 . All PCs ( –1 to +4 s ) : n = 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 26 , 25 , 24 , 24 . PCs + hyperpolarization ( –1 to +4 s ) : n = 5 , 6 , 6 , 7 , 7 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 , 8 . Silent+ large hyperpolarization ( –4 to +4 ) : n = 4 , 4 , 4 , 4 , 5 , 5 , 5 , 6 , 6 , 6 , 6 , 6 , 6 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 7 , 6 , 6 , 6 , 6 , 5 , 5 , 5 , 5 , 5 , 4 , 4 , 4 , 4 , 4 , 4 , 3 , 3 , 3 , 3 , 2 , 2 , 2 , 2 , 2 , 2 , 2 , 2 , 2 , 2 . Statistical details of experiments can be found in the figure legends . Unless otherwise specified , measured values and ranges reflect mean ± SEM . Significance was defined as p < 0 . 05 . Sample sizes were not determined by statistical methods , but efforts were made to collect as many samples as was technically feasible . No data or subjects were excluded from any analysis . | A new housing development in a familiar neighborhood , a wrong turn that ends up lengthening a Sunday stroll: our internal representation of the world requires constant updating , and we need to be able to associate events separated by long intervals of time to finetune future outcome . This often requires neural connections to be altered . A brain region known as the hippocampus is involved in building and maintaining a map of our environment . However , signals from other brain areas can activate silent neurons in the hippocampus when the body is in a specific location by triggering cellular events called dendritic calcium spikes . Milstein et al . explored whether dendritic calcium spikes in the hippocampus could also help the brain to update its map of the world by enabling neurons to stop being active at one location and to start responding at a new position . Experiments in mice showed that calcium spikes could change which features of the environment individual neurons respond to by strengthening or weaking connections between specific cells . Crucially , this mechanism allowed neurons to associate event sequences that unfold over a longer timescale that was more relevant to the ones encountered in day-to-day life . A computational model was then put together , and it demonstrated that dendritic calcium spikes in the hippocampus could enable the brain to make better spatial decisions in future . Indeed , these spikes are driven by inputs from brain regions involved in complex cognitive processes , potentially enabling the delayed outcomes of navigational choices to guide changes in the activity and wiring of neurons . Overall , the work by Milstein et al . advances the understanding of learning and memory in the brain and may inform the design of better systems for artificial learning . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2021 | Bidirectional synaptic plasticity rapidly modifies hippocampal representations |
In response to environments that cause cellular stress , animals engage in sleep behavior that facilitates recovery from the stress . In Caenorhabditis elegans , stress-induced sleep ( SIS ) is regulated by cytokine activation of the ALA neuron , which releases FLP-13 neuropeptides characterized by an amidated arginine-phenylalanine ( RFamide ) C-terminus motif . By performing an unbiased genetic screen for mutants that impair the somnogenic effects of FLP-13 neuropeptides , we identified the gene dmsr-1 , which encodes a G-protein coupled receptor similar to an insect RFamide receptor . DMSR-1 is activated by FLP-13 peptides in cell culture , is required for SIS in vivo , is expressed non-synaptically in several wake-promoting neurons , and likely couples to a Gi/o heterotrimeric G-protein . Our data expand our understanding of how a single neuroendocrine cell coordinates an organism-wide behavioral response , and suggest that similar signaling principles may function in other organisms to regulate sleep during sickness .
When acutely ill , animals engage in a behavioral sequence that includes cessation of feeding and body movements as well as reduced responsiveness to the environment . During acute infectious illness in mammals , electrophysiological correlates of sleep behavior are observed ( Toth and Krueger , 1988 , 1989 ) , indicating that the behavioral sequence is sleep . In the arthropod Drosophila melanogaster ( Williams et al . , 2007; Lenz et al . , 2015 ) and the nematode Caenorhabditis elegans ( Hill et al . , 2014 ) , acute illness results in cellular stress , which then induces a sleep behavior: the animals stop moving and feeding , do not respond to weak stimuli but move normally in response to strong stimuli ( Hill et al . , 2014 ) . This stress-induced sleep ( SIS , also known as sickness sleep ) is beneficial to the animal and helps it recover from the acute injury ( Spiegel et al . , 2002; Hill et al . , 2014; Kuo and Williams , 2014; Fry et al . , 2016 ) . In fruit flies and round worms , environments that induce SIS include bacterial pathogens , bacterial toxins , heat shock , cold shock , osmotic shock , and ultraviolet light exposure ( Hill et al . , 2014; Lenz et al . , 2015 ) . A comparison between mammalian sickness sleep and invertebrate SIS was recently reviewed ( Davis and Raizen , 2016 ) . Nematode SIS is a distinct sleep state from a larval sleep state known as developmentally timed sleep ( DTS ) ( Trojanowski et al . , 2015 ) , which is regulated by a homolog of the core circadian protein PERIOD ( Monsalve et al . , 2011 ) . In the absence of stress , nematodes experience sleep only when they transition between larval stages but do not sleep in the adult stage . Since C . elegans does not have an identifiable circadian rhythm of sleep , adult nematodes are an ideal system to study SIS in the absence of the circadian and homeostatic effects of animals that require daily sleep . The mechanism of SIS is poorly-understood , yet a few common themes have emerged from studies across phylogeny . The acute illness can occur outside of the brain yet affect behavior , suggesting that communication occurs between non-neural and neural tissues . Cytokine signaling is involved . For example , in mammals , the cytokines interleukin-1 beta and tumor necrosis factor alpha , whose levels increase during an infectious challenge , are each sufficient to induce sleep when injected into the brain ( reviewed in ( Krueger , 2008 ) . In nematodes ( Van Buskirk and Sternberg , 2007 ) , arthropods ( Foltenyi et al . , 2007 ) , and mammals ( Kushikata et al . , 1998; Kramer et al . , 2001 ) , signaling by epidermal growth factor ( EGF ) is sufficient to induce sleep behavior and , at least in nematodes , EGF signaling is necessary for SIS ( Hill et al . , 2014 ) . These cytokines act on central nervous system ( CNS ) neurons , which then induce sleep . In mammals , CNS neurons that regulate sleep reside in the hypothalamus ( Saper et al . , 2005b ) . In C . elegans , the target for EGF action is a single interneuron called ALA ( Van Buskirk and Sternberg , 2007 ) , whose developmental program has similarities to the developmental program of mammalian neuroendocrine cells ( Van Buskirk and Sternberg , 2010 ) . With EGF activation , ALA depolarizes to release neuropeptides encoded by the gene flp-13 ( FMRFamide-Like Peptide-13 ) to promote sleep ( Nelson et al . , 2014 ) . FLP-13 peptides are characterized by an amidated Arginine-Phenylalanine ( RFamide ) motif at their C-termini . RFamide neuropeptides are involved in many physiological functions in both invertebrates ( López-Vera et al . , 2008; Peymen et al . , 2014 ) , and vertebrates ( Rőszer and Bánfalvi , 2012; Kim , 2016 ) . In fruit flies , several RFamide neuropeptides regulate sleep ( He et al . , 2013; Shang et al . , 2013 ) , including the RFamide neuropeptide FMRFamide , which regulates SIS ( Lenz et al . , 2015 ) . In this study , we focused on understanding the downstream mechanism of the sleep-promoting activity of FLP-13 RFamide peptides . Both locomotion quiescence and feeding quiescence induced by flp-13 can be reversed by activation of motor neurons ( Trojanowski et al . , 2015 ) , suggesting that flp-13 mediates quiescence at the level of the nervous system . Furthermore , quiescence induced by flp-13 requires the G protein alpha subunit GOA-1 ( Trojanowski et al . , 2015 ) , suggesting that these peptides signal through a G protein-coupled receptor ( GPCR ) . There are more than 150 genes in the C . elegans genome predicted to encode neuropeptide receptor GPCRs ( Frooninckx et al . , 2012; Hobert , 2013 ) . In prior detailed analysis of one of these GPCRs ( Nelson et al . , 2015 ) , we showed that while FRPR-4 can be activated by FLP-13 peptides in cell-based assay , its genetic removal does not abrogate flp-13 induced quiescence in vivo , suggesting that it is not the receptor mediating the quiescence-inducing effects of FLP-13 peptides in response to cellular stress . Since we had no strong a priori reason to implicate other specific GPCRs , we took a hypothesis-independent forward genetic screen approach to identify the FLP-13 receptor ( Yuan et al . , 2015 ) . We here identify the GPCR DMSR-1 ( DroMyoSuppressin Receptor related-1 ) as required for flp-13 somnogenic effects . DMSR-1 is expressed in about one tenth of all C . elegans neurons and localizes diffusely to membranes . FLP-13 peptides can activate the receptor directly , indicating that DMSR-1 is likely an in vivo receptor for FLP-13 neuropeptides . Inhibition of neurons where dmsr-1 is expressed enhances the effect of flp-13 , suggesting that DMSR-1 transduces the FLP-13 signal by reducing activity of wake-promoting neurons .
Our goal was to characterize the downstream mechanism for the sleep-inducing FLP-13 neuropeptides . We induced sleep by over expressing the flp-13 gene in somatic cells under the control of the heat-inducible promoter hsp-16 . 2 ( heat-shock protein-16 . 2 ) . Phsp-16 . 2:flp-13 transgenic animals are quiescent for body and pharyngeal movements two hours after exposure to a heat pulse to induce flp-13 gene expression ( Nelson et al . , 2014 ) . As previously described ( Yuan et al . , 2015 ) and illustrated in Figure 1A , we performed a forward mutagenesis screen to identify mutants with defective quiescence in response to flp-13 overexpression . By performing genetic complementation tests , we determined that five of the identified mutants ( qn45 , qn49 , qn51 , qn52 and qn53 ) were alleles of the same gene ( Yuan et al . , 2015 ) . We found two additional alleles ( qn40 and qn44 ) in this gene by manually screening for mutants with defective feeding quiescence induced by overexpression of flp-13 . 10 . 7554/eLife . 19837 . 003Figure 1 . Mutations in the seven-transmembrane domain protein DMSR-1 suppress flp-13 induced quiescence . ( A ) Mutagenesis approach to identify the downstream signaling mechanism for FLP-13 peptides . The grand-daughters of Phsp-16 . 2:flp-13 worms that were mutagenized with the chemical ethyl methanesulfonate ( EMS ) were screened for rare animals that continued to feed and move after induction flp-13 overexpression . Moving animals were selected using a microfluidics automated assay ( Yuan et al . , 2015 ) and feeding animals were identified by direct observation of pharyngeal pumping movements . ( B ) The DMSR-1 protein is predicted to have seven transmembrane domains with a C-terminus tail of either 145 amino acids ( isoform A ) or 87 amino acids ( isoform B ) . Five mutations in DMSR-1 result in premature stop codons , one mutation results in an alanine to valine change in the second extracellular loop , and one mutation results in the removal of the C-terminal half of the protein . ( C ) Phylogenetic tree relationship between 30 C . elegans proteins previously demonstrated or predicted to be RFamide receptors , three Drosophila melanogaster RFamide receptors ( NM_139501; NP_647713; NP_647711 ) , and one receptor from each Anopheles gambiae ( XP_314133 ) and Rhodnius prolixus ( Lee et al . , 2015 ) . Unless indicated otherwise , all proteins are from Caenorhabditis elegans . We drew a box to highlight DMSR-1 ( isoform A ) . The evolutionary history was inferred by using the Maximum Likelihood method based on the JTT matrix-based model ( Jones et al . , 1992 ) . The bootstrap consensus tree inferred from 1000 replicates is taken to represent the evolutionary history of the taxa analyzed . The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test ( 1000 replicates ) is shown next to the branches . Initial tree ( s ) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model , and then selecting the topology with superior log likelihood value . The analysis involved 35 amino acid sequences . All positions containing gaps and missing data were eliminated . There were a total of 181 positions in the final dataset . Evolutionary analyses were conducted in MEGA7 ( Kumar et al . , 2016 ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 00310 . 7554/eLife . 19837 . 004Figure 1—figure supplement 1 . DNA rearrangement of dmsr-1 exon 4 in the qn45 allele . ( A ) The qn45 mutant allele involved a deletion of 332 nucleotides and a duplication of 54 nucleotides flanking the deletion , with a net loss of 278 nucleotides . This insertion/deletion ( indel ) is predicted to cause a frameshift resulting in an early stop codon . This indel was identified using Sanger sequencing . ( B ) Predicted amino acid sequence of wild type ( N2 ) DMSR-1 and qn45 DMSR-1 proteins . * indicates STOP codon . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 004 To identify the gene mutated to cause the defect in flp-13 induced quiescence , we sequenced the genomes of six of the allelic mutants . Each of the six mutants had a mutation in dmsr-1 , a gene predicted to encode a GPCR ( Figure 1B ) . Four of the mutants ( qn44 , qn49 , qn51 and qn53 ) contained premature stop codons . One mutant ( qn45 ) contained a complex rearrangement in exon 4 consisting of deletion of 332 nucleotides coupled with insertion of 54 nucleotides ( Figure 1—figure supplement 1 ) . This rearrangement results in a frame shift that eliminates the three transmembrane domains at the C-terminus of the protein . One mutant ( qn52 ) contained an alanine to valine mutation in exon 3 . We Sanger sequenced the dmsr-1 gene in qn40 , and found a premature stop mutation in exon 4 ( Figure 1B ) . The six mutations which caused premature stops or a deletion are predicted to eliminate gene function . We do not know whether the missense mutation in qn52 eliminates genes function; however , based on the phenotype and the relatively conservative alanine to valine mutation , the second extracellular loop where this mutation occurred likely plays an important role in the function of DMSR-1 . We performed subsequent analyses using the qn45 allele because it was a convincing null mutation and because the genotype could be easily determined using polymerase chain reaction ( PCR ) . DMSR-1 is one of 41 C . elegans proteins in the RFamide receptor family ( Frooninckx et al . , 2012 ) and is orthologous to Drosophila receptors for the neuropeptides dromyosuppressin and FMRFamide . Like FLP-13 peptides , Dromyosuppressin and FMRFamide are neuropeptides with an amidated Arginine-Phenylalanine ( RFamide ) motif at their C-terminus ( Nachman et al . , 1993 ) . We grouped the C . elegans DMSR and other RFamide receptor proteins based on their sequence homology to the Drosophila myosuppressin receptors MsR1 and MsR2 , which are known to respond to RFamide peptides ( Figure 1C ) . Other RFamide receptors in C . elegans include FRPR ( FMRFamide Peptide Receptor family ) proteins , and the neuropeptide receptor family ( NPR ) , although these FRPR and NPR receptors are more closely related to the Drosophila FMRFaR than they are to the myosuppressin related receptors . We included in the phylogenetic tree other identified RFamide receptors from the arthropods Anopheles gambiae and Rhodnius prolixus , which also share homology to the Drosophila myosuppressin receptors . Based on its mutant phenotype and its homology to known RFamide receptors , we hypothesized that DMSR-1 is a receptor for FLP-13 neuropeptides . To measure the degree to which these dmsr-1 mutations suppressed the behavioral quiescence induced by flp-13 overexpression , we subjected Phsp-16 . 2:flp-13 transgenic animals to a heat pulse . We quantified feeding and locomotion quiescence two hours after induction of flp-13 overexpression – well after the acute effects of heat on quiescence have dissipated ( Nelson et al . , 2014 ) . All wild-type animals pumped their pharynxes rapidly two hours after the heat pulse whereas the Phsp-16 . 2:flp-13 transgenic animals showed little to no feeding movements ( Figure 2A–B ) . By contrast , a significant fraction of each of the dmsr-1 mutants had pharyngeal pumping movements despite overexpression of flp-13 ( Figure 2A ) . In the absence of a heat pulse , dmsr-1 ( qn45 ) mutants pumped rapidly at a rate no different from wild-type animals ( Figure 2B ) . Following induction of flp-13 overexpression , the rate of pumping in dmsr-1 ( qn45 ) mutants was not changed . 10 . 7554/eLife . 19837 . 005Figure 2 . dmsr-1 mutations suppress flp-13 induced quiescence . ( A ) Fraction of animals quiescent for pharyngeal pumping two hours after induction of flp-13 overexpression by exposure to a 30 min 33°C heat pulse . Statistical significance was assessed using a Fisher’s exact test . N = 15–20 for each genotype . Asterisks indicate significant difference ( p<0 . 0001 ) compared to wild type animals two hours after heat pulse . ( B ) Rate of pharyngeal pumping in wild type or dmsr-1 ( qn45 ) mutants with or without the Phsp16 . 2:flp-13 transgene comparing the behavior pre-heat pulse to two hours post-heat pulse . ( C ) Body movement quiescence in a two hour period starting two hours after induction of flp-13 overexpression . ( D ) Total body movement activity in the same two-hour period as ( C ) . Activity is the sum of pixels changed between sequential images . Statistical significance in panels B-D was assessed using a 2-Way ANOVA with post-hoc pairwise comparisons made using Bonferroni correction method . Error bars denote Mean ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 005 To quantify body movements , we recorded worm activity and quiescence using the WorMotel ( Churgin and Fang-Yen , 2015 ) . The WorMotel is a polydimethylsulfoxane ( PDMS ) device consisting of 48 individual wells filled with agar nematode growth medium ( NGM ) and used to track individual worms over several hours . Two hours after induction of flp-13 overexpression , we recorded body movement activity for two hours by taking images every 10 s . We used image subtraction analysis ( Raizen et al . , 2008 ) to determine both the total activity of the animals and the amount of time that each worm spent quiescent following induction of flp-13 overexpression . Following induction of flp-13 overexpression , dmsr-1 ( qn45 ) mutants had increased body movement activity and reduced body movement quiescence in comparison to animals that were wild type for dmsr-1 ( Figure 2C–D ) . We can make a number of predictions based on the hypothesis that DMSR-1 is the receptor for FLP-13 neuropeptides . The first prediction is that , like flp-13 ( Nelson et al . , 2014 ) , dmsr-1 should be required for quiescence during SIS . The second prediction is that FLP-13 neuropeptides will activate DMSR-1 in a cell-culture system . The third prediction is that dmsr-1 should be expressed in the nervous system . Finally , DMSR-1 should be active in neurons that affect sleep/wake behavior . In order to test whether dmsr-1 is required for SIS , we first artificially activated the SIS pathway . In response to stress , the C . elegans EGF ( called LIN-3 ) activates the EGF-receptor ( called LET-23 ) on the ALA neuron ( Van Buskirk and Sternberg , 2007; Hill et al . , 2014 ) . Overexpression of LIN-3/EGF produces a robust sleep state that mimics SIS ( Van Buskirk and Sternberg , 2007 ) . While dmsr-1 mutants showed wild-type levels of feeding quiescence in response to EGF overexpression ( Figure 3A ) , they had a small but highly significant defect relative to wild-type worms in body movement quiescence ( Figure 3B ) . This result provides evidence that dmsr-1 is required downstream of or in parallel to EGF . The observation of only partial suppression of the EGF-induced quiescence suggests that there are other ALA-regulated signaling pathways working in parallel to dmsr-1 . 10 . 7554/eLife . 19837 . 006Figure 3 . dmsr-1 mutants are defective in quiescence associated with stress-induced sleep . ( A ) Rate of pharyngeal pumping before and up to two hours following heat pulse induction of EGF/LIN-3C overexpression . dmsr-1 does not suppress EGF induced feeding quiescence . ( B ) dmsr-1 ( qn45 ) partially suppresses body movement quiescence induced by EGF overexpression ( Phsp-16 . 2:lin-3 ) . Body movements were measured for two hours starting one hour after induction of EGF overexpression . ( C ) Rate of pharyngeal pumping in dmsr-1 ( qn45 ) mutants and dmsr-1 genomic rescue during the first hour following 35°C heat shock to induce SIS . Rescue construct is the operon-based reporter shown in Figure 5 . Asterisks denote significant difference compared to dmsr-1 ( qn45 ) . ( D ) Body movement quiescence during 90 min after a 35°C heat shock . The dmsr-1 ( qn45 ) mutation suppresses body movement quiescence in response to heat shock . This defect in quiescence is rescued by a genomic fragment containing dmsr-1 . Statistical significance was assessed using a 2-Way ANOVA with post-hoc pairwise comparisons made using Bonferroni correction method . Error bars denote Mean ± SEM . *p<0 . 05 , **p<0 . 01 , ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 00610 . 7554/eLife . 19837 . 007Figure 3—figure supplement 1 . dmsr-1 mutants are defective in SIS triggered by different stressors but are not defective in developmentally timed sleep during lethargus . ( A ) Total body movement quiescence during the first hour after exposure to the indicated temperature for 30 min . Error bars denote SEM . ***p<0 . 001 . ( B ) Total body movement quiescence each hour after exposure to ultraviolet C ( UVC ) light . Error bars denote SEM . ***p<0 . 001 . ( C ) Total body movement quiescence in the three hours after the start of L4 lethargus . NS denotes p>0 . 05 . In panels A and B , statistical significance was assessed 2-Way ANOVA with post-hoc pairwise comparisons made using Bonferroni correction method , and in C with a two-tailed Student’s t test . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 007 To test the hypothesis that dmsr-1 mediates SIS , we induced sleep by exposing the animals to a heat stress of 35°C for 30 min . dmsr-1 mutations suppressed SIS with respect to both body movement quiescence and feeding quiescence ( Figure 3C–D ) . To test whether dmsr-1 was required for SIS after exposure to other temperatures , we exposed cohorts of animals to temperatures ranging from 29 to 37 degrees Celsius . The dmsr-1 ( qn45 ) mutation suppressed SIS at temperatures ranging from 34 to 37 degrees ( Figure 3—figure supplement 1A ) . To test whether dmsr-1 was required for sleep in response to stressors other than heat , we exposed animals to ultraviolet C ( UVC ) radiation . UVC induced a strong body movement quiescent response , which was attenuated in the dmsr-1 ( qn45 ) mutant ( Figure 3—figure supplement 1B ) . We partially restored SIS in dmsr-1 mutants by expressing a genomic fragment of dmsr-1 including the coding region and 4 kb of upstream regulatory DNA ( Figure 3C , D ) . It is possible that the incomplete rescue of SIS is due to an absence of dmsr-1 in the full complement of cells where it acts . This incomplete expression may be due to the nature of our extrachromosomal DNA array , which is not inherited by every somatic cell . Alternatively , it is possible that the genomic DNA fragment does not contain all required dmsr-1 regulatory elements . While dmsr-1 mutants were defective in quiescence associated with SIS , they had normal quiescence associated with DTS ( Figure 3—figure supplement 1C–D ) , as previously shown for flp-13 mutants ( Nelson et al . , 2014 ) , supporting the notion that the DTS and SIS are regulated at least partially differently ( Trojanowski et al . , 2015 ) . These experiments therefore demonstrate that , in response to cellular stress , dmsr-1 is required for quiescence of feeding movements and body movements . To test the prediction that FLP-13 peptides activate DMSR-1 in cell culture , we cloned DMSR-1 isoform A into a mammalian expression vector and transiently expressed the protein in Chinese hamster ovarian ( CHO ) culture cells . These cells stably express a promiscuous Gα16 subunit as well as an aequorin reporter activated by intracellular calcium fluxes . Gα16 causes a Ca2+ flux in response to receptor activation regardless of the type of G-protein that couples to the receptor in vivo ( Figure 4A ) ( Beets et al . , 2011 ) . 10 . 7554/eLife . 19837 . 008Figure 4 . FLP-13 peptides activate DMSR-1 in a cell-based system . ( A ) Signal transduction components in a cell system used to test for receptor activation by peptide ligands ( L ) . DMSR-1A was expressed in CHO cells along with the calcium sensitive bioluminescent protein aequorin . The receptor was paired with the promiscuous human Gα16 protein , which causes calcium release from intracellular storage sites upon receptor activation . The calcium response elicits a luminescent signal from aequorin , a calcium-activated protein . ( B ) FLP-13 peptide sequences . flp-13 encodes seven distinct peptides ( * indicates that this peptide is encoded by the gene sequence twice ) . EC50 indicates the concentration of neuropeptide required to elicit 50% of the maximum luminescent response from the aequorin protein ( as shown in C ) . ( C ) Dose response curves for the seven FLP-13 peptides . Error bars represent SEM from 8–12 trials of neuropeptide treatment . Line represents non-linear regression fit of a variable slope line using four parameters . The x-axis is shown on a logarithmic scale . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 008 We measured the response of cells expressing DMSR-1 isoform A to a range of FLP-13 peptide concentrations . Since the FLP-13 protein is processed into seven distinct RFamide peptides ( Figure 4B ) , we performed a dose-response study with each of these seven peptides . Each of these peptides was capable of activating DMSR-1 . The most potent activator was FLP-13–5 , which elicited a 50% maximal response ( EC50 ) at a concentration of 2 . 3 nM ( Figure 4B–C ) . The observed potent activation of DMSR-1 by FLP-13 peptides in this cell based assay supports the hypothesis that DMSR-1 is an in vivo receptor for FLP-13 peptides . The third prediction made by our model is that dmsr-1 is expressed in the nervous system . This prediction arises from prior studies indicating that the mechanism of feeding and movement quiescence observed with flp-13 overexpression occurs through the inhibition of cholinergic motor neurons ( Trojanowski et al . , 2015 ) . To test this prediction , we attached the rescuing genomic DNA of dmsr-1 to a trans-splice acceptor site followed by the coding sequence of the red fluorescent protein dsRED ( Figure 5A ) . This Pdmsr-1:dmsr-1:SL2:dsRED transgene rescued the dmsr-1 mutant phenotype ( Figure 3C–D ) , indicating that the construct contains regulatory elements required for appropriate dmsr-1 expression . We observed red fluorescence in several neurons both in the head and tail of the animals ( Figure 5B ) . Based on the location of the cell body , the neuronal process morphology , the co-expression of the transgene with other , well-characterized , green fluorescent protein ( GFP ) reporters , and the co-localization of the red fluorescence with green DiO fluorescence , we concluded that dmsr-1 is expressed in the RID neuron ( Figure 5C ) , the paired AIY neurons ( Figure 5D ) , 12 other head neurons , the paired PHA and PHB neurons in the tail ( Figure 5E ) , and in five other tail neurons . 10 . 7554/eLife . 19837 . 009Figure 5 . DMSR-1 is expressed non-synaptically in the nervous system . ( A ) Schematic of operon ( top ) and promoter fusion ( middle ) transcriptional reporters used to determine expression pattern of dmsr-1 . Schematic overview of dmsr-1 expression pattern , with identified neurons labelled . ( B ) Example images of dmsr-1 expression in head and tail . The images were processed by 3D deconvolution ( Leica Application Suite X , Leica ) , and presented as a maximum projection of a Z-stack . In this and subsequent images , anterior is to the left . ( C ) Colocalization of dmsr-1 promoter reporter with kal-1 ( Wenick and Hobert , 2004 ) in the RID neuron . The dmsr-1 promoter:mCherry reporter gave the same expression pattern as the operon-based transcriptional reporter shown in panel A . ( D ) Colocalization of dmsr-1 mCherry reporter with ttx-3 ( Hobert et al . , 1997 ) in an AIY neuron . Images were captured as a z-stack and processed by 3D deconvolution . One individual slice is shown . ( E ) DiO staining of tail phasmid neurons PHA and PHB , which colocalizes with the dmsr-1 red reporter . ( F ) Lack of colocalization of dmsr-1 promoter reporter with nmr-1 ( Brockie et al . , 2001 ) in command interneurons . Images were captured as a z-stack and processed by 3D deconvolution . One deconvolved slice , which shows all three neuron types , is shown . There is a red neuron that partially overlaps the green AVE neuron but our close evaluation of multiple worms shows that these neurons are distinct . ( G ) Membrane localization of GFP attached to the C-terminus of DMSR-1 using a fosmid construct ( TransgeneOme project ) . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 00910 . 7554/eLife . 19837 . 010Figure 5—figure supplement 1 . Rescue of the dmsr-1 mutant quiescence-defective phenotype with a fosmid that encodes a GFP tag at the c-terminus of DMSR-1 . ( A ) Body movement quiescence following UV exposure calculated in one hour intervals . Two-way repeated measures ANOVA with Bonferroni corrected comparisons for post-test . Comparisons are displayed as differences between the indicated line and dmsr-1 ( qn45 ) without the rescue construct . *p<0 . 05; **p<0 . 01; ***p<0 . 001 . ( B ) Body movement quiescence during the 8-hr period following UV exposure . One-way ANOVA with Tukey HSD post-test . *p<0 . 05; **p<0 . 01; ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 01010 . 7554/eLife . 19837 . 011Figure 5—figure supplement 2 . Effects of selective expression of dmsr-1 in the paired AIY head neurons or in the paired PHA tail neurons . ( A ) Fraction of animals quiescent for pharyngeal pumping two hours after induction of flp-13 overexpression using a 30 min 33°C heat pulse of hsp16 . 2:flp-13 or control animals . Statistical significance was assessed using a Fisher’s exact test . N = 15–20 for each genotype . ( B ) Pharyngeal pumping rates two hours after induction of flp-13 overexpression using a 30 min 33°C heat pulse of hsp16 . 2:flp-13 or control animals . Statistical significance was assessed using a 1-way ANOVA with Dunnett’s post-hoc pairwise comparisons with hsp16 . 2:flp-13; dmsr-1 ( qn45 ) animals . ( C ) and ( D ) Body movement quiescence in a two-hour period starting at two hours after induction of flp-13 overexpression . Statistical comparisons were made using 1-way ANOVA with Tukey HSD post-test . ( E ) Body movement quiescence for 60 min following a 30 min 35°C exposure to induce SIS . Statistical comparisons were made using 1-way ANOVA with Dunnett’s post-hoc pairwise comparisons with dmsr-1 ( qn45 ) animals . ( F ) Pharyngeal pumping rate following a 30 min 35°C exposure to induce SIS . Two-way repeated measures ANOVA with Bonferroni corrected comparisons for post-test . Comparisons are displayed as differences between the indicated line and dmsr-1 ( qn45 ) . In all panels , bars denote the mean and SEM , ns denotes not significant , * denotes p<0 . 05 , **p<0 . 01 , and ***p<0 . 001 . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 011 Based on additional analysis , we concluded that dmsr-1 is not expressed in several other neurons , which are listed in Supplementary file 1 . Importantly , we did not see expression of dmsr-1 in AVE or AVA neurons ( Figure 5F ) . These command interneurons are the only post-synaptic partners to the ALA neuron based on the presence of ultrastructurally-defined chemical synapses ( White et al . , 1986 ) . The expression in neurons that do not make direct synaptic connections with ALA suggested that ALA signals in a neuroendocrine fashion to promote SIS . In support of this suggestion , we found that a DMSR-1 protein with a GFP fluorescent tag at its C-terminus was localized diffusely to neuronal cell membranes and was not localized to synapses ( Figure 5G ) . This GFP-tagged DMSR-1 protein rescued the mutant phenotype of dmsr-1 ( Figure 5—figure supplement 1 ) , indicating that it was functional . Taken together , these results support a model whereby ALA releases FLP-13 peptides in a neuroendocrine fashion to directly activate DMSR-1 expressed in neurons and thus promote sleep in response to cellular stress . The notion that ALA can signal via non-synaptic means was previously proposed based on the observation that a mutant with disrupted ALA axonal outgrowth retained normal ALA function ( Van Buskirk and Sternberg , 2007 ) . In an attempt to determine which of the neurons observed to express dmsr-1 is sufficient to relay the sleep inducing effect of flp-13 overexpression , we transgenically restored dmsr-1 gene expression in specific neurons in animals otherwise lacking dmsr-1 function . We tested two neuron types—AIY and PHA—where we had the tools to easily express DMSR-1 selectively in a single neuron type . Expression of dmsr-1 in the paired PHA tail sensory neurons was not sufficient to restore feeding or body movement quiescence induced by flp-13 overexpression ( Figure 5—figure supplement 2A–C ) . Expression of dmsr-1 in the paired head AIY interneurons conferred a small but statistically-significant ( p<0 . 01 ) rescue of the body movement quiescence defect ( Figure 5—figure supplement 2D ) but not the feeding quiescence defect observed in dmsr-1 mutants during flp-13 overexpression ( Figure 5—figure supplement 2A–B ) . AIY specific dmsr-1 expression did not however rescue the feeding or body movement quiescence defects observed in dmsr-1 mutants during SIS ( Figure 5—figure supplement 2E–F ) . The fourth prediction from our model is that DMSR-1 signaling affects neurons that control sleep/wake behavior . DMSR-1 may regulate sleep either by activating sleep-promoting neurons or by inhibiting wake-promoting neurons . To distinguish between these possibilities , we inhibited the dmsr-1-expressing cells to test their effects on sleep . If DMSR-1 activation promotes sleep by activating sleep-promoting neurons , then inhibiting these neurons should impair sleep . Conversely , if DMSR-1 activation promotes sleep by inhibiting wake-promoting neurons , then inhibiting these cells would promote sleep . We inhibited DMSR-1 expressing neurons by using histamine-gated chloride ( HisCl ) channels . We can use this approach because histamine does not appear to act as a neurotransmitter in C . elegans ( Hobert , 2013 ) . In the presence of histamine , HisCl channels allow chloride ions to pass into the neurons to render them less excitable ( Liu and Wilson , 2013; Nelson et al . , 2014; Pokala et al . , 2014 ) . By applying histamine to worms expressing HisCl channels under the control of the dmsr-1 promoter ( Pdmsr-1:HisCl ) , we inhibited dmsr-1 expressing neurons in a spatially- and temporally- controlled fashion . Because non-transgenic animals lack endogenous receptors for histamine ( Pokala et al . , 2014 ) , they behaved similarly in the presence and absence of histamine ( Figure 6 ) . When transgenic worms expressing Pdmsr-1:HisCl worms were placed on histamine , their movement quiescence and total activity were not affected ( Figure 6 ) , indicating that in the absence of cellular stress , silencing these neurons has minimal effect on behavior . 10 . 7554/eLife . 19837 . 012Figure 6 . DMSR-1 has an inhibitory effect on neurons . ( A ) Pdmsr-1:HisCl worms were placed on histamine-containing agar , and sleep was induced by ultraviolet C ( UVC ) light irradiation . Worms were compared to the wild type strain ( N2 ) on histamine , as well as to worms not exposed to histamine . Gray lines indicate control worms that were not exposed to UVC light , whose quiescence was not affected by activation of HisCl channels . Statistical comparisons were performed using a 2-way repeated measures ANOVA using time and experimental group as factors . Asterisks indicate lowest level of significance in comparisons between Pdmsr-1:HisCl on histamine with each of the other groups in pairwise comparison using Bonferroni correction . *p<0 . 05 , **p<0 . 005 , ***p<0 . 0005 . ( B ) Total amount of sleep over eight hours from Figure 6A for each individual worm . There was a significant interaction between histamine and HisCl channels such that HisCl channels only increased sleep in the presence of histamine . ***p<0 . 0005 . There were no significant effects found for worms that were not irradiated with UV . NS denotes not significant . ( C ) Total activity analysis across an eight-hour period . HisCl channels reduced activity only in animals exposed to histamine , and there were no significant effects for worms that were not irradiated with UV . ***p<0 . 0005 . NS denotes not significant . In all three panels , triangles represent animals that expressed Pdmsr-1:HisCl , circles represent animals that do not express Pdmsr-1:HisCl , filled symbols represent animals that have been exposed to histamine , empty symbols represent animals that have not been exposed to histamine , black symbols represent animals that have been exposed to UV irradiation , and gray symbols represent animals that were not exposed to UV irradiation . In panels B and C , statistical significance was assessed with a 2×2 Factorial ANOVA with Bonferroni post-hoc correction . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 012 We then tested whether activating the HisCl channels affected activity and quiescence after triggering cellular stress . For these experiments , we triggered cellular stress by exposing the animals to ultraviolet ( UV ) light ( Figure 3—figure supplement 1 ) . UV light exposure triggers a robust sleep state that requires ALA and FLP-13 signaling ( H . Debardeleben and D . Raizen , unpublished data ) . The long-lasting effects of UV induced sleep provide a sensitive tool for measuring the effects of neuronal silencing in this experiment . When HisCl worms were placed on histamine following exposure to UV stress , they had reduced total activity and enhanced body movement quiescence compared to control worms that lacked the HisCl channels ( Figure 6A–C ) . Importantly , HisCl channels did not affect activity or quiescence in the absence of histamine . These observations of elevated body movement quiescence with HisCl activation suggest that DMSR-1 promotes sleep by inhibiting wake-promoting cells . One neuropeptide ( or a group of neuropeptides encoded by the same gene ) may activate multiple receptors . Similarly , a given receptor may be activated by multiple neuropeptides . Does the somnogenic effect of FLP-13 neuropeptides result exclusively from the activation of DMSR-1 ? Do other neuropeptides activate DMSR-1 to induce sleep ? We used a double mutant analysis to answer these questions . We compared the SIS phenotypes of flp-13; dmsr-1 double mutant animals to those of dmsr-1 and flp-13 single mutant animals . We reasoned that if flp-13 and dmsr-1 have sleep inducing effects that are independent of one another , then the double mutant will have a greater defect in SIS compared to either of the single mutants . The double mutant animals indeed showed a greater defect in quiescence than either the dmsr-1 or the flp-13 single mutant animals ( Figure 7 ) . This result suggests that ( 1 ) FLP-13 peptides activate receptors in addition to DMSR-1 , and ( 2 ) DMSR-1 is either activated by ligands in addition to those encoded by flp-13 , or DMSR-1 has ligand-independent activity . 10 . 7554/eLife . 19837 . 013Figure 7 . DMSR-1 and FLP-13 neuropeptides have parallel activity . ( A ) Double mutant analysis between flp-13 and dmsr-1 . Total amount of body movement quiescence was measured following UV irradiation . Statistical comparisons were made using 2-way repeated measures ANOVA with post-hoc pairwise Bonferroni correction method and restricted to comparisons made with the double mutant . ( B ) Total quiescence during first four hour period following UV irradiation shown in Figure 7A . Statistical comparisons were made using 1-way ANOVA with Dunnett’s post-hoc pairwise comparisons with the double mutant . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 013
Our findings expand our understanding of how stress induced sleep is regulated in C . elegans ( Figure 8 ) . Cellular stress leads to the release of the EGF/LIN-3 ( Hill et al . , 2014 ) either from the stressed cells directly or from other cells . EGF activates the ALA neuron through the EGF-receptor LET-23 ( Van Buskirk and Sternberg , 2007 ) . This activation requires depolarization of ALA ( Nelson et al . , 2014 ) . ALA releases neuropeptides encoded by the flp-13 gene ( Nelson et al . , 2014 ) as well as other neuropeptides ( Nath et al . , 2016 ) . FLP-13 neuropeptides then act in a neuroendocrine fashion to activate the G-protein coupled receptor DMSR-1 and thus to inhibit wake-promoting neurons , which include the AIY interneurons . The G-protein that is coupled to DMSR-1 may be the Gi/o alpha subunit GOA-1 , since goa-1 mutants are defective in SIS and in flp-13 induced sleep ( Trojanowski et al . , 2015 ) and since goa-1 is expressed widely in the nervous system ( Ségalat et al . , 1995 ) . GOA-1 signaling may ultimately down regulate the activity of the Gq ( EGL-30 ) signaling pathway , since a gain of function mutation in EGL-30 suppresses flp-13 induced sleep ( Trojanowski et al . , 2015 ) . 10 . 7554/eLife . 19837 . 014Figure 8 . Model for the regulation of stress-induced sleep . Cellular stress leads to the release of EGF ( LIN-3 ) either directly from the stressed cells or indirectly via other cells receiving a signal from the stressed cells . EGF activates the ALA neuron by binding to the EGF receptor ( LET-23 ) . ALA depolarizes and releases FLP-13 neuropeptides , among other sleep inducing signals . FLP-13 peptides signal in a non-synaptic fashion via the seven-transmembrane domain receptor DMSR-1 and the G protein alpha subunit Gi/o to inhibit several wake-promoting neurons . These neurons include AIY , PHA , PHB , RID , and other neurons . DOI: http://dx . doi . org/10 . 7554/eLife . 19837 . 014 The observation that the suppression of quiescence by most of our genetic manipulations is only partial suggests that there are additional complexities in the sleep promoting system downstream of ALA activation . Our double mutant analyses support the notion that FLP-13 peptides act on other receptors in addition to DMSR-1 , and that DMSR-1 is activated by peptides other than FLP-13 peptides . The additional receptors responding to FLP-13 peptides may include one or more of the 15 other DMSR proteins encoded in the C . elegans genome ( Figure 1C ) , and the additional peptides acting on DMSR-1 may include one or more of the other neuropeptides released from ALA ( Nath et al . , 2016 ) . We showed that selective dmsr-1 expression in the AIY neurons conferred a small but significant rescue of the flp-13 overexpression-induced movement quiescence but did not confer a rescue of SIS . Three possibilities could reconcile these observations . ( 1 ) The endogenous FLP-13 peptides that act on the AIY neurons are released from a neuron other than ALA . ( 2 ) FLP-13 peptides act on AIY only when expressed at high levels and do not normally do so under physiological conditions . ( 3 ) Our SIS assay conditions were not sufficiently sensitive to identify a small rescue in SIS . Regardless of the explanation of these results , since the AIY interneurons are known to promote worm locomotion ( Gray et al . , 2005; Shtonda and Avery , 2006 ) , our observations that dsmr-1 is expressed in AIY and may function in AIY are consistent with the model that FLP-13/DMSR-1 signaling functions to inhibit wake-promoting neurons . Although selective dmsr-1 expression in neither AIY nor PHA neurons fully rescued the mutant quiescence phenotype , our findings do not exclude the possibility that dmsr-1 functions selectively in a neuron type that we have not yet tested . Several of the other neurons expressing dmsr-1 remain unidentified and may also promote waking activity . One possibility is the neuroendocrine RID neuron , which promotes forward locomotion ( Lim et al . , 2016 ) . The alternative to the model that dmsr-1 functions in a small subset of the neurons where it is expressed is that dmsr-1 is simultaneously required in several neurons to promote sleep in response to FLP-13 signaling . In this model , which we favor , restoring dmsr-1 gene expression in only a subset of the neurons in which dmsr-1 is required would not be sufficient to fully rescue the mutant phenotype . In mammals , wake-promoting centers are distributed broadly across brainstem , basal forebrain , midbrain , and diencephalic structures , but sleep promoting neurons are anatomically restricted ( Saper et al . , 2005b ) . The ventrolateral preoptic area of the hypothalamus , which is proposed to be sleep-promoting , contains inhibitory projections to several wake-promoting neurons ( Saper et al . , 2005a ) . It is possible that C . elegans similarly has a diffuse wake-promoting circuit that can be inhibited by a central sleep-promoting system , the single ALA neuron . It remains to be seen whether sleep in response to acute illness in other animals functions in a similar fashion , but since RFamide and EGF signaling regulates sleep in arthropods too , it would not be unreasonable to propose that the molecular details and signaling concepts elucidated here are conserved across phylogeny .
Worms were cultivated on the agar surface of NGM medium ( 1 . 7% agar ) , fed the OP50 derivative bacterial strain DA837 ( Davis et al . , 1995 ) , and maintained at 20 degrees Celsius unless noted otherwise . We found that the type of agar used affected the behavior of animals after heat shock . For experiments reported here we used only granulated agar ( Apex , catalog number 20–275 ) . N2 NQ570: qnIs303[Phsp-16 . 2:flp-13; Phsp-16 . 2:GFP; Prab-3:Cherry] NQ793: dmsr-1 ( qn40 ) V; qnIs303 NQ810: dmsr-1 ( qn44 ) V; qnIs303 NQ792: dmsr-1 ( qn45 ) V; qnIs303 NQ814: dmsr-1 ( qn49 ) V; qnIs303 NQ51: dmsr-1 ( qn51 ) V; qnIs303 NQ52: dmsr-1 ( qn52 ) V; qnIs303 NQ53: dmsr-1 ( qn53 ) V; qnIs303 NQ915: dmsr-1 ( qn45 ) V NQ602: flp-13 ( tm2427 ) NQ943: flp-13 ( tm2427 ) IV; dmsr-1 ( qn45 ) V PS5009: pha-1 ( e2132ts ) III ( ? ) ; syEx723[Phsp16 . 2:LIN-3C; Pmyo-2:GFP; pha-1 ( + ) ] NQ978: pha-1 ( e2132ts ) III ( ? ) ; dmsr-1 ( qn45 ) V; syEx723 NQ990: qnEx514[Pdmsr-1:HisCl; Pttx-3:GFP] NQ1006: qnEx526[Pdmsr-1:mCherry; Punc-122:GFP; pha-1 ( + ) ; ladder] NQ1083: unc-119 ( ed3 ) III; qnEx585[dmsr-1 ( fosmid ) ::GFP; pCFJ151] NQ1055: dmsr-1 ( qn45 ) V; qnEx569[dmsr-1 ( fosmid ) ::GFP; Pttx-3:GFP; 1 kb DNA ladder] NQ1111: dmsr-1 ( qn45 ) V; qnEx602[dmsr-1:SL2:dsRed; Pttx-3:GFP; 1 kb DNA ladder] NQ1142: dmsr-1 ( qn45 ) ; qnEx610[Pttx-3:dmsr-1:SL2:dsRed; Pmyo-2:GFP; 1 kb DNA ladder] NQ1145: dmsr-1 ( qn45 ) V;qnIs303; qnEx610[Pttx-3:dmsr-1:SL2:dsRed; Pmyo-2:GFP; 1 kb DNA ladder]; NQ1147: dmsr-1 ( qn45 ) V; qnIs303; qnEx614[Pgcy-17:dmsr-1:SL2:dsRed;Pmyo-2:GFP; 1 kb DNA ladder] We used the overlap extension PCR method to generate constructs for transgenic analysis ( Nelson and Fitch , 2011 ) . Oligonucleotide sequences used for constructs and for sequencing are listed in Supplementary file 2 . Microinjection of extrachromosomal arrays was performed as described in Mello et al . ( 1991 ) . Whole genome sequencing was performed at the Wistar Institute Genomics Core Facility and bioinformatics analysis of the sequences was performed at the Wistar Institute Bioinformatics Core Facility . Crosses with dmsr-1 were performed using either the balancer nT1 , which expresses a dominant GFP , or by following the qn45 deletion by PCR analysis using the primers oNQ1480 and oNQ1515 . Except for data shown in Figure 3—figure supplement 1 , which was collected from fourth larval stage animals , all behavioral assays were performed in young adult worms . Worms were synchronized by picking larval stage four animals and performing the experiments the following day . Multi-factor analyses were performed using 2-way ANOVA with pairwise post-hoc comparisons using the Bonferroni correction . For post-hoc tests , all comparisons were made against one group . For time course analyses , we used a 2-way ANOVA with the time factor as a repeated measure . One factor analyses were performed using a 1-way ANOVA with pairwise comparisons made using Dunnett’s test . All statistics were performed and graphs were generated using GraphPad Prism 5 . 0 . 3 . 477 and 6 . 0 ( GraphPad Software , Inc . , La Jolla California ) . The assessment of FLP-13 peptide activation of DMSR-1 was done as described ( Nelson et al . , 2015 ) . Briefly , DMSR-1A cDNA was cloned into the pcDNA3 . 1 ( + ) TOPO expression vector ( Thermo Fisher Scientific , Waltham Massachusetts ) . Receptor activation was studied in Chinese hamster ovary cells ( CHO ) stably expressing apo-aequorin ( mtAEQ ) targeted to the mitochondria as well as the human Galpha16 subunit . We used the CHO-K1 cell line ( PerkinElmer , ES-000-A2 ) for receptor activation assays . Quality control and authentication was performed by determining the EC50 for reference agonists ( e . g . ATP ) in an AequoScreen calcium mobilization assay . A mycoplasma test was performed using the MycoAlert Mycoplasma ( Lonza ) detection kit . This cell line tested negative for mycoplasma . The CHO/mtAEQ/Galpha16 cells were cultured in Ham’s F12 medium ( Sigma ) , containing 10% fetal bovine serum ( FBS ) , 100 UI/ml of penicillin/streptomycin , 250 μg/ml Zeocin and 2 . 5 μg/ ml Fungizone ( Amphoterin B ) . Cell lines were grown at 37°C in a humidified atmosphere of 5% CO2 and were diluted fifteen-fold every third day . CHO/mtAEQ/Galpha16 cells were transiently transfected with the DMSR-1 cDNA construct or the empty pcDNA3 . 1 ( + ) vector using the Lipofectamine transfection reagent ( Thermofisher Scientific ) , according to the manufacturer’s instructions . Cells expressing the receptor were shifted to 28°C 1 day later , and collected 2 days post-transfection in BSA medium ( DMEM/HAM’s F12 with 15 mM HEPES , without phenol red , supplemented with 0 . 1% BSA ) and loaded with 5 μM coelenterazine h ( Thermo Fisher ) for 4 hr to reconstitute the holo-enzyme aequorin . Cells ( 25 , 000 cells/well ) were exposed to synthetic peptides in BSA medium , and aequorin bioluminescence was recorded for 30 s on a MicroBeta LumiJet luminometer ( PerkinElmer , Waltham Massachusetts ) in quadruplicate . For dose-response evaluations , after 30 s of ligand-stimulated calcium measurements , Triton X-100 ( 0 . 1% ) was added to the well to obtain a measure of the maximum cell Ca2+ response . BSA medium without the peptides was used as a negative control and 1 μM ATP was used to check the functional response of the cells . Cells transfected with the pcDNA3 . 1 empty vector were used as a negative control for the effect of the receptor . EC50 values were calculated from dose-response curves , constructed using a computerized nonlinear regression analysis , with a sigmoidal dose-response equation ( Prism 6 . 0 ) . | People often feel fatigued and sleepy when they are sick . Other animals also show signs of sleepiness when ill – they stop eating , move less , and are less responsive to changes in their environment . Sickness-induced sleep helps both people and other animals to recover , and many scientists believe that this type of sleep is different than nightly sleep . Studies of sickness-induced sleep have made use of a simple worm with a simple nervous system . In this worm , a single nerve cell releases chemicals that cause the worm to fall asleep in response to illness . Animals exposed to one of these chemicals , called FLP-13 , fall asleep even when they are not sick . As such , scientists would like to know which cells in the nervous system FLP-13 interacts with , what receptor the cells use to recognize this chemical , and whether it turns on cells that induce sleep or turns off the cells that cause wakefulness . Now , Iannacone et al . show that FLP-13 likely causes sleep by turning down activity in the cells in the nervous system that promote wakefulness . The experiments sifted through genetic mutations to determine which ones cause the worms not to fall asleep when FLP-13 is released . This revealed that worms with a mutation that causes them to lack a receptor protein called DMSR-1 do not become sleepy in response to FLP-13 . This suggests that DMSR-1 must be essential for FLP-13 to trigger sleep . About 10% of cells in the worm’s nervous system have the DMSR-1 receptor . Some of these neurons tell the worm to move forward or to forage around for food . The experiments also showed that FLP-13 is probably not the only chemical that interacts with the DMSR-1 receptor , but the identities of these other chemicals remain unknown . Additional experiments are now needed to determine if sickness-induced sleepiness in humans and other mammals is triggered by a similar mechanism . If it is , then drugs might be developed to treat people experiencing fatigue associated with sickness as well as other unexplained cases of fatigue . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2017 | The RFamide receptor DMSR-1 regulates stress-induced sleep in C. elegans |
Foot-and-mouth disease virus ( FMDV ) is an economically devastating viral disease leading to a substantial loss to the swine industry worldwide . A novel alternative strategy is to develop pigs that are genetically resistant to infection . Here , we produce transgenic ( TG ) pigs that constitutively expressed FMDV-specific short interfering RNA ( siRNA ) derived from small hairpin RNA ( shRNA ) . In vitro challenge of TG fibroblasts showed the shRNA suppressed viral growth . TG and non-TG pigs were challenged by intramuscular injection with 100 LD50 of FMDV . High fever , severe clinical signs of foot-and-mouth disease and typical histopathological changes were observed in all of the non-TG pigs but in none of the high-siRNA pigs . Our results show that TG shRNA can provide a viable tool for producing animals with enhanced resistance to FMDV .
Foot-and-mouth disease ( FMD ) is an acute and highly contagious disease of cloven-hoofed animals , including cattle , pigs , sheep and goats and more than 70 wildlife species , and is devastating especially in young animals ( Grubman and Baxt , 2004 ) . The etiological agent of FMD is foot-and-mouth disease virus ( FMDV ) , which belongs to the genus aphthovirus of the family Picornaviridae ( Bachrach , 1968 ) . Some control strategies including eradication , vaccination , selective test and slaughter have been widely used for preventing FMDV infection ( Leforban , 1999; Barnett and Carabin , 2002 ) , but diseases caused by FMDV remain prevalent in pigs and cattle all over the world owing to the absence of reciprocal protection among several FMDV serotypes ( Haydon et al . , 2001 ) . RNA interference ( RNAi ) is a post-transcriptional process initiated by double-stranded RNA ( dsRNA ) homologous to a target gene sequence ( Meister and Tuschl , 2004 ) . Specific gene silencing can also be triggered in mammalian cells by using synthetic short interfering RNA ( siRNA ) , and plasmid or virus-mediated short hairpin RNA ( shRNA ) ( Elbashir et al . , 2001; Hammond et al . , 2001; Paul et al . , 2002; Michel et al . , 2005; Kim and Rossi , 2007 ) . The shRNA was proposed as a therapy for suppressing the infection of FMDV in vitro and in vivo ( Chen et al . , 2006; Kim et al . , 2008 ) . Recently , we extended that finding by producing genetically engineered mice integrating shRNA targeting FMDV ( Pengyan et al . , 2008 , 2010 ) . The majority of transgenic ( TG ) mice infected with FMDV were resistant to infection and showed only slightly abnormal pathology compared with controls . Now , we report that TG pigs expressing siRNA against FMDV are resistant to viral challenge .
We constructed a total of 10 shRNA expression vectors ( Figure 1A ) targeting viral structural protein VP1 of FMDV type O and determined the efficacy of shRNAs for inhibiting FMDV replication in BHK cells by real-time RT-PCR . The V3 shRNA reduced the expression of viral RNA by 96 . 8% compared to scrambled control ( Figure 1B ) . The V3 shRNA expression vector was used to generate TG pigs by somatic cell nuclear transfer . A total of 42 pigs were born alive , eight of which survived at least 6 months . TG pigs contained the stably integrated transgene as evidenced by PCR ( Figure 2A ) and green fluorescent protein ( GFP ) expression ( Figure 2B and Figure 2—figure supplement 1 ) . The copy numbers of transgene were measured by real-time PCR . The copy numbers of the inserted vector were calculated to be 3–11 ( Figure 2C ) . Expression of siRNA in fibroblast cells isolated from TG pigs was examined by custom TaqMan small RNA assays ( Figure 2D ) . The siRNA expression in TG 11 , 19 , 69 and 101 was 10–30-fold that from TG 24 , 45 , 49 and 78 . After necropsy of TG 69 and 101 , expression of siRNA was detected in various tissues , including heart , lung , spleen , liver , kidney and muscle , although the siRNA levels were diverse among different tissues ( Figure 2E ) . 10 . 7554/eLife . 06951 . 003Figure 1 . Design and screening of shRNA expression vector . ( A ) Schematic diagram of shRNA expression vector ( pXL-EGFP-NEO ) used . This vector includes a mouse H1 RNase promoter driving ubiquitous expression of shRNA and a cytomegalovirus-immediate early ( CMV ) promoter driving GFP and neomycin fusion expression . The arrows denote the PCR primers spanning H1 promoter , shRNA and GFP elements used to identify transgene integration in cloned pigs . ( B ) Relative expression of viral RNA in shRNA-transfected BHK cells . Data are means of three replicates ±SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 00310 . 7554/eLife . 06951 . 004Figure 2 . Analysis of shRNA transgene in cloned pigs . ( A ) PCR for detecting shRNA expression cassette . PCR products spanned H1 promoter , shRNA and GFP cassette . P: plasmid as positive control . Non-TG: non-TG pig as negative control . 11 , 19 , 24 , 45 , 49 , 69 , 78 and 101: cloned pigs . ( B ) EGFP fluorescence of transgenic pigs . ( C ) The copy numbers of transgene were determined by real-time PCR . ( D ) Analysis of siRNA expression in fibroblast cells of all transgenic pigs . ( E ) Analysis of siRNA expression in various tissues of TG 69 and 101 . Data are presented as means of three replicates ±SD . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 00410 . 7554/eLife . 06951 . 005Figure 2—figure supplement 1 . Pictures of transgenic pigs and EGFP expression in the fibroblast cells . ( A ) EGFP of transgenic pigs ( red arrows ) under natural light . ( B ) Expression of EGFP in ear fibroblasts isolated from transgenic pigs . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 005 Next , we tested shRNA transgene resistance to FMDV infection in fibroblast cells isolated from high-siRNA TG ( 11 , 19 , 69 and 101 ) , low-siRNA TG ( 24 , 45 , 49 and 78 ) and non-TG pigs . Compared to non-TG cells , viral RNA expression was reduced by 30-fold in high-siRNA TG and 12-fold in low-siRNA TG cells at 36 hr after virus challenge ( Figure 3A ) . Inhibition of FMDV infection was a positive correlation with siRNA expression in fibroblast cells ( Figure 2B and Figure 3A ) . Moreover , TG fibroblasts visibly reduced FMDV-induced cytopathogenic effects as compared with non-TG fibroblasts ( Figure 3B ) . 10 . 7554/eLife . 06951 . 006Figure 3 . shRNA transgene resistance to FMDV infection in fibroblast cells of transgenic pigs . ( A ) Relative expression of viral RNA in fibroblast cells after FMDV infection . Data are presented as means ±SD . ( B ) Fibroblast cells were observed for development of cytopathogenic effect by bright-field microscopy at 12 , 24 and 36 hr post-infection . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 006 The resistance of TG pigs to FMDV infection was further tested by intramuscular injection of O serotypes of FMDV . The challenged animals included high-siRNA TG ( 11 and 19 ) , low-siRNA TG ( 24 , 49 and 78 ) and non-TG pigs ( n=5 ) . Prior to the day of infection , no animal tested was positive for FMDV . After FMDV challenge , all non-TG pigs developed high fever within 72 hr of challenge and severe clinical signs of FMD , the appearance of vesicles on the feet and nose ( Figure 4A and Figure 4—figure supplement 1 ) . All non-TG pigs became deteriorated and the lesion score reached 24 at 5 d after challenge ( Figure 4A and Figure 4—figure supplement 1 ) . Some smaller vesicles in low-siRNA TG ( 24 , 49 and 78 ) pigs were also observed until 7 d after challenge , as shown in Figure 4A . However , the body temperature of high-siRNA TG ( 11 and 19 ) pigs remained normal throughout the experiment ( Figure 4A and Figure 4—figure supplement 1 ) . TG pigs 11 and 19 developed one small vesicle at day 9 of challenge , but TG pig 11 recovered soon on the next day ( Figure 4A ) . We subsequently quantified the viral genome RNA in the serum of the infected animals . Consistent with the clinical signs data , the viral load in the serum of the high-siRNA TG and low-siRNA TG pigs was lower than that in the non-TG pigs ( Figure 4B ) . The viral RNA expression in serum was 42-fold lower in the high-siRNA TG group than that in the non-TG pigs at day 10 of challenge . 10 . 7554/eLife . 06951 . 007Figure 4 . Transgenic shRNA pigs resisted FMDV infection . ( A ) Clinical sign of TG and non-TG pigs challenged with O serotypes of FMDV . Body temperature was detected every day after infection . Body temperature 38–39 . 5°C ( no fever ) ; body temperature up to 39 . 5–40°C ( mild fever ) ; body temperature over 40°C ( high fever ) . Lesion score based on the appearance of vesicles on the feet and nose ( see ‘Materials and methods’ ) . None of vesicles on the feet and nose ( − ) . ( B ) Relative expression of viral RNA in serum of the infected animals . Data are presented as means ±SD . ( C ) Relative expression of viral RNA in various tissues of the infected animals . Data are presented as means ±SD . ( D ) HE staining of tissue sections from non-TG and TG pigs . HE staining revealed that there was hepatic multifocal necrosis in non-TG pigs and interstitial pneumonia and severe congestion in the lung of non-TG pigs . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 00710 . 7554/eLife . 06951 . 008Figure 4—figure supplement 1 . Body temperature curve of all infected pigs . The body temperatures of non-TG ( 120 , 141 , 159 , 191 and 211 ) and TG pigs ( 11 , 19 , 24 , 49 and 78 ) were measured at 24 hr intervals before FMDV challenge until the end of the experiment . DOI: http://dx . doi . org/10 . 7554/eLife . 06951 . 008 All animals were killed on the 10th day post-infection , and major tissues including lesions were collected for levels of virus RNA and histopathology analysis . Viral RNA was not detected in the heart , lung , spleen , liver , kidney and muscle in the challenged TG pigs , but viral RNA still maintained high levels of expression in lymph and lesions in the non-TG pigs , except heart , lung , spleen and liver ( Figure 4C ) . No viral RNA in the non-TG heart , lung , spleen and liver showed clearance of viral RNA from the tissues , consistent with prior findings ( Zhang and Alexandersen , 2004; Chen et al . , 2006 ) . Furthermore , lesions , as a potential source of virus transmission by aerosol , were well known to be the predominant tissue site of FMDV infection and amplification ( Zhang and Bashiruddin , 2009; Dillon , 2011 ) . Levels of viral RNA in foot lesions of TG pigs were much lower than those in the non-TG pigs ( Figure 4C ) , suggesting an encouraging result for blocking transmission . Haematoxilin/eosin ( HE ) staining of major tissues revealed that non-TG pigs had severe abnormal pathology compared to TG pigs . In particular , non-TG pigs showed multifocal necrosis in the liver , and interstitial pneumonia and severe congestion in the lung ( Figure 4D ) . None of the TG pigs showed typical histopathological changes , except one case of interstitial pneumonia . Recently , Lyall et al . ( 2011 ) in Science reported that onward transmission of avian influenza in TG shRNA chickens was prevented , although the TG chickens succumbed to the initial direct challenge , leading to a strategy for controlling avian influenza outbreaks . Our results showed that the TG pigs exhibited a marked resistence to FMDV infection after direct challenge . As encouraging as these results are , an onward transmission experiment will be performed in the future when producing enough high-siRNA TG pigs . The most important threat caused by FMDV is the high speed of viral replication , short incubation time , and high contagiousness . Although protective immune responses from vaccination against FMDV can be efficacious , the rapidity of virus replication and spread can outpace the development of immune defenses and overrun the immune system ( Summerfield et al . , 2009 ) . Current FMDV vaccines do not induce a protective response until 7 d post-vaccination ( Barnett et al . , 2002; Doel , 2003 ) . FMD signs in high-siRNA TG pigs in our study were delayed for at least 8 d after FMDV infection ( Figure 4A ) . siRNA expressed in TG pigs can also play a role as co-agent to induce rapid resistance before routine vaccination can evoke protective immunity . TG siRNA pigs immunized with current vaccines may achieve complete protection for an FMDV outbreak , which provides a novel strategy for preventing FMD in a disease-free country . The shRNA-based transgene strategy has substantial benefits over vaccination by offering potential sub-serotype protection when using multiple-shRNA expression systems targeting different viruses ( Cong et al . , 2010 ) . Our findings demonstrate that RNAi technology combining animal cloning offers the possibility to produce TG animal with improved resistance to viral infection .
Conserved sequences such as the siRNA target site had been reported as an alternative strategy preventing the escape mutants of virus ( Dave and Pomerantz , 2004 ) . Conserved target sequences were selected from the viral structural protein VP1 gene by sequence alignment of O , A and Asia 1 serotypes of FMDV . The shRNA was designed by using the Ambion website tool ( http://www . ambion . com/techlib/misc/siRNA_finder . html ) . These shRNA sequences are summarized in Supplementary file 1 . Oligonucleotides were annealed and cloned into the pXL-EGFP-NEO to generate a series of shRNA expression plasmids ( Figure 1A ) . BHK cells were seeded in 24-well plates ( CoStar , Cambridge , MA ) the day before transfection to achieve 90% confluency . The cells were transfected with 2 . 5 μg shRNA expression plasmids using Lipofectamine 2000 ( Invitrogen , Carlsbad , CA ) . After 12 hr of transfection , the transfection complex was removed and the cells were washed twice with DMEM . The transfected cells in per well plates were then infected with 100 TCID50 of FMDV ( OS/99 strain ) . After 1 hr of adsorption , the inoculum was removed and the cells were washed twice with DMEM . The infection then proceeded in DMEM supplemented with 2% fetal bovine serum . Virus samples were collected at designated time points and frozen at −80°C until assessment of viral RNA . Viral RNA was isolated using Trizol ( Invitrogen ) according to the manufacturer’s instructions . From purified RNA , complementary DNA was synthesized using random hexamer primers and was quantified by spectrophotometer at 260 nm . Real-time PCR was carried out using SYBR Green ( TaKaRa Biotech , Dalian , China ) following the manufacturer’s protocol . The following primers were used for FMDV VP1 gene amplification ( VP1-F: 5′-TCA AGC CAA AGG AAC AAGT-3′; VP1-R: 5′-TAG ACG GTC GCT AAG ACAC-3′ ) . GAPDH served as an internal control . The ΔΔCt method was used for relative quantification ( Livak and Schmittgen , 2001 ) . Pig primary fibroblasts were isolated as previously described ( Fan et al . , 2013 ) . The fibroblast cells were transfected with linearized shRNA expression vectors , and then were split into 12-well plates at an appropriate dilution ( 2000 cells/well ) for G418 selection ( 400 μg/ml; Promega , Madison , WI ) ( Cong et al . , 2010 ) . G418-resistant and GFP-positive colonies derived from individual cells were obtained after 14 d of culture . The positive cells were used for somatic cell nuclear transfer as described previously ( Li et al . , 2009; Ni et al . , 2014 ) . Approximately 200–300 embryos were transferred into each surrogate pig . Cloned pigs were delivered by natural birth at full term . Genomic DNA was isolated from ear biopsies of cloned pigs using the TIANamp genomic DNA kit ( Tiangen Biotech , China ) . Transgene integration was identified by PCR assays . PCR was performed on 400 ng of genomic DNA using specific primers ( H1-F: TGT CGC TAT GTG TTC TGGG; GFP-R: TGT CTT GTA GTT CCC GTC ATC ) for amplifying shRNA and GFP expression cassette . PCR reaction consisted of 95°C for 4 min; 30 cycles at 95°C for 30 s , 57°C for 30 s and 72°C for 50 s; an extension at 72°C for 5 min . PCR products were analyzed by 1% gel electrophoresis . The copy numbers of transgenes were determined by real-time PCR as previously described ( Kong et al . , 2009 ) . Briefly , a standard curve was produced with series of standard samples containing 1 , 2 , 4 , 8 , 10 , 12 , 16 copies of the shRNA gene , respectively , by mixing the wild-type genome of pig with shRNA expression vector . The absolute quantitative standard curve was drawn by plotting ÄCt ( ÄCt=CtshRNA−CtTFRC ) against the log of shRNA gene copies of corresponding standard samples . siRNA expression in TG pigs was determined by using TaqMan small RNA assays ( Applied Biosystems , Foster City , CA ) ( Chen et al . , 2005 ) . Small RNAs were isolated by using the mirVana miRNA isolation Kit ( Ambion , Austin , TX ) . Real-time RT-PCR was performed according to the manufacturer’s instructions . Endogenous U6 was used as a RNA quality and loading control . The shRNA expression was normalized to the expression of U6 using the 2−ΔΔCt method ( Ct of shRNA–Ct of U6 ) ( Livak and Schmittgen , 2001 ) . TG fibroblasts were isolated from ear biopsies of cloned pigs as previously described ( Li et al . , 2014 ) . The cells cultured in 96-well plates were inoculated with 100 TCID50 of O serotypes of FMDV ( OS/99 strain ) . After 1 hr absorption , the inoculum was removed and the infection then proceeded in DMEM supplemented with 2% fetal bovine serum . The infected cells were observed for cytopathic effects at 12 , 24 and 36 hr post-challenge . Virus samples were collected at 36 hr post-challenge . Relative expression of viral RNA was evaluated by real-time RT-PCR as described above . All experiments involving animals were conducted under the protocol ( SU-ACUC-12031 ) approved by the Animal Care and Use Committee of Shihezi University . Viral challenge was performed with O serotypes of FMDV ( OS/99 strain ) . The challenged pigs ( 10–13 months of age ) included high-siRNA TG ( 11 and 19 ) , low-siRNA TG ( 24 , 49 and 78 ) and non-TG controls ( n=5 ) . Before virus challenge , all animals were confirmed as negative for FMDV infection . All animals were housed in one room and challenged by intramuscular injection with 100 LD50 in 1 ml of phosphate-buffered saline ( PBS ) in the neck area . After challenge , animals were examined daily for clinical signs of FMD , including body temperature and the appearance of vesicles on the nose , mouth and feet . Body temperature remaining at 38–39 . 5°C was defined as no fever , body temperature up to 39 . 5–40°C was defined as mild fever , and body temperature over 40°C was defined as high fever . The lesion score was calculated by determining the number and size of vesicles on the nose , mouth and feet of each animal; 1 cm of each vesicle was recorded as 1 , 2 cm were recorded as 2 , and other larger vesicles were recorded as 3 ( if on the nose or mouth ) or 6 ( if on the feet ) . The observations were terminated on day 10 post-challenge when the animals were killed . Blood samples were collected at days 0 , 1 , 3 , 5 , 7 , 9 and 10 after challenge . Total RNA was extracted from blood and subjected to real-time RT-PCR as described above . All animals were killed on the 10th day post-infection , and major tissues were fixed in formalin for 10 hr followed by routine paraffin sectioning and HE staining . Histopathological changes were observed under microscope . | Foot-and-mouth disease regularly causes serious outbreaks in livestock . The virus that causes the disease can infect cattle , pigs , sheep , goats , and many species of wild animals; the disease is also highly contagious and spreads very quickly and easily . To control the spread of foot-and-mouth disease , farmers must often kill entire herds of animals that have been exposed . Wild animals that can spread the virus may also be killed in an effort to stop the spread of the disease . Vaccines that protect against foot-and-mouth disease are available and are often used to help prevent the spread of the disease . However , once an outbreak of foot-and-mouth disease begins it may be too late for vaccines to stop its spread . This is because the vaccines can take about a week to provide protection , and by that time an exposed animal may already be very ill . Previous work conducted in 2010 reported that mice could be genetically engineered to produce short stretches of RNA molecules that can switch off genes from the foot-and-mouth disease virus . Compared with normal mice infected with the foot-and-mouth disease virus , the genetically engineered mice showed little sign of the disease in their bodies . Now , Hu , Qiao , Fu et al . —including some of the researchers involved in the 2010 work—have genetically engineered pigs in the same way . The experiments show that when cells from these pigs are exposed to the foot-and-mouth disease virus in the laboratory , the virus grows much less than normal . Next , Hu , Qiao , Fu et al . injected genetically engineered pigs and non-genetically engineered pigs with the virus . All of the normal pigs developed severe symptoms very quickly , including the disease's characteristic mouth and foot sores . Additionally , examinations of these pigs' cells showed signs of the disease . But the genetically engineered pigs did not become seriously ill and their cells showed little sign of the disease . Some of the genetically engineered pigs developed a few sores but these sores appeared much later than normal . So far , the results suggest that it may be possible to develop pigs that are resistant to foot-and-mouth disease . Hu , Qiao , Fu et al . will next determine whether or not the genetically engineered pigs can pass the foot-and-mouth virus on to other pigs and livestock . | [
"Abstract",
"Introduction",
"Results",
"and",
"discussion",
"Materials",
"and",
"methods"
] | [
"short",
"report",
"microbiology",
"and",
"infectious",
"disease"
] | 2015 | Transgenic shRNA pigs reduce susceptibility to foot and mouth disease virus infection |
Birds land on a wide range of complex surfaces , yet it is unclear how they grasp a perch reliably . Here , we show how Pacific parrotlets exhibit stereotyped leg and wing dynamics regardless of perch diameter and texture , but foot , toe , and claw kinematics become surface-specific upon touchdown . A new dynamic grasping model , which integrates our detailed measurements , reveals how birds stabilize their grasp . They combine predictable toe pad friction with probabilistic friction from their claws , which they drag to find surface asperities—dragging further when they can squeeze less . Remarkably , parrotlet claws can undergo superfast movements , within 1–2 ms , on moderately slippery surfaces to find more secure asperities when necessary . With this strategy , they first ramp up safety margins by squeezing before relaxing their grasp . The model further shows it is advantageous to be small for stable perching when high friction relative to normal force is required because claws can find more usable surface , but this trend reverses when required friction shrinks . This explains how many animals and robots may grasp complex surfaces reliably .
While airplanes require runways and most aerial robots need flat landing areas , arboreal birds consistently land on perches with a wide range of geometries and surface properties . Recent advances in perching capabilities of aerial robots have improved their landing ability on engineered surfaces . However , they still lack the versatility and robustness that birds exhibit when landing on an extraordinarily wide range of complex engineered and natural surfaces ( Kovac , 2016; Roderick et al . , 2017a; Doyle et al . , 2013 ) . Previous studies have quantified leg and wing dynamics ( Provini et al . , 2012; Provini et al . , 2014; Chin and Lentink , 2017; Provini and Abourachid , 2018 ) and suggested visual control strategies ( Chin and Lentink , 2017; Lee et al . , 1993 ) that arboreal birds use during landings . However , it is unclear whether the same findings would apply with perches that have different geometric properties and challenging surface textures . Even less is understood at the actual interface between the bird’s foot and the perch surface . While previous studies have examined the morphology of bird feet in cadavers ( Hopson , 2001; Sustaita et al . , 2013 ) and claw shape in relation to body size and lifestyle ( Pike and Maitland , 1999; Birn-Jeffery et al . , 2012; Feduccia , 1993 ) , no in vivo studies quantify toe forces during grasping or how claws actually interact with perch surface features . Consequently , we do not know how birds actually grasp perches so reliably in a seemingly effortless manner . To elucidate the mechanisms and strategies responsible for the versatile perching capabilities of arboreal birds , we studied voluntary landings made by Pacific parrotlets ( Forpus coelestis ) on perches with different geometries and textures ( Video 1 ) . We tested nine landing perch variations in total , which included three natural variations , three diameter variations , and three man-made surface variations ( Figure 1a ) . The natural perches were constructed from ~0 . 75" ( 19 mm ) diameter branches from three different trees . The first tree was a Coast Live Oak ( Quercus agrifolia ) , which varies in roughness at different parts of its branches . The second was a Floss-Silk tree ( Ceiba speciosa ) , which has relatively smooth , soft branches and is native to the same forests of Ecuador and Peru ( Nasr et al . , 2018 ) as the Pacific parrotlets . The third was a Sweet Olive tree ( Osmanthus fragrans ) , which has rough , striated branches . Birch dowels , which are commonly used for commercial bird housing perches ( our parrotlet housing uses 0 . 625"-dia . birch dowels ) , were used to construct the three diameter variations: 1 . 5" ( 38 mm ) , too large for the parrotlets to wrap their feet more than a quarter of the way around; 0 . 75" ( 19 mm ) , which their feet can wrap about halfway around; and 0 . 25" ( 6 mm ) , which their front and rear toes wrap around completely and overlap . To test surface texture extremes , three 0 . 75" diameter perches were constructed by wrapping birch wooden dowels in foam ( squishy ) , Teflon ( slippery ) , and sandpaper ( rough ) . To characterize the surface properties of these perches , we made 3D structured-light scans to quantify roughness ( Supplementary file 1A ) and reconstruct surface profiles ( see Materials and methods ) . The profiles show the variation in surface roughness , from smoothest , Teflon , to roughest , sandpaper ( Figure 1A ) . To evaluate the effect of these properties on foot-surface interactions , we conducted toe and claw drag tests to measure friction forces on each surface . We also performed claw indentation tests to measure surface deformation ( see Materials and methods; Figure 1B ) . Finally , we made high-speed recordings of parrotlets landing on the nine different perches . The perches were split in half and instrumented on a pair of force/torque sensors for measuring net and squeeze forces exerted by their feet ( see Materials and methods; Figure 1C ) .
The parrotlets exhibit a stereotyped set of landing behaviors across all perch variations ( Figure 2A , Video 2 ) . Landing is initiated by braking with the wings ( ‘aerial’ ) , and then the legs absorb the remaining momentum upon impact with the perch ( ‘absorption’ ) . Both feet would make contact with the perch within a few milliseconds of each other , but initial contact tended to be led by one preferred foot ( bird 1 = 100% right foot , bird 2 = 83% left , bird 3 = 85% left ) . The parrotlets then wrap their toes and claws more securely around the perch to establish a static grasp ( ‘anchoring’ ) . After this stage , they often take additional steps ( 2 . 0 ± 0 . 7 , n = 78 landings ) to adjust their footing ( ‘adjustments’ ) . The birds would occasionally overshoot or undershoot the perch , leading to more variance in the foot angle at which they establish a static grasp ( Figure 2B ) . However , while the net leg force magnitudes ( Figure 2B ) and directions ( Figure 2C ) also exhibit some variance , the average landing force trends remain remarkably similar across the different perches . The variation seen during landing may be explained by the parrotlets’ landing strategy . According to Tau theory ( Lee et al . , 1993 ) , birds control their landings by visually estimating their time to contact , τ ( t ) . To land , they adjust their approach speed to maintain a constant τ˙ ( t ) , the time rate of change of tau . More specifically , tau is defined as the distance to the perch s divided by the speed of approach v , and is therefore a first-order approximation for the time to contact . If a bird brakes with constant deceleration a , then τ=sv=0 . 5at2at=0 . 5t , and τ˙ ( t ) =0 . 5 . If τ˙ ( t ) <0 . 5 , braking is decreasing until the bird stops at the landing perch , and if 0 . 5<τ˙<1 , braking is increasing until the bird makes a controlled collision with the landing perch ( Lee et al . , 1993 ) . Based on their aerial kinematics , the parrotlets maintain relatively constant τ˙ values ( Figure 2E ) consistent with those of controlled collisions ( Figure 2F ) . The smallest landing target , the 0 . 25”-dia . perch , had the smallest average τ˙ value of 0 . 80 . The foam perch had the highest at τ˙= 0 . 94 , which indicates that parrotlets increase their braking more when they are closer to the softer perch . Across all perches , the high τ˙ values ( >0 . 5 ) indicate that the birds have not yet completed the braking process when they touch down , which likely leads to greater variance in their landing dynamics . The parrotlets’ stereotyped wing and leg behavior suggest that they compensate for perch differences with their feet and claws . However , the foot kinematics reveal a similarly stereotyped set of landing stages ( Figure 3A , B ) until the toes engage with the surface . The feet remain closed during flight ( ‘resting’ ) , and then about 100 ms before touchdown the foot spread angle begins to increase ( ‘spreading , ’ 40 ms ± 8 ms duration ) until the toes are fully outstretched ( ‘open , ’ 21 ms ± 7 ms duration ) . This open stance may accommodate for some uncertainty in when and where the foot actually makes contact with the perch ( as in Figure 2B ) . Immediately before contact , the toes start closing in ( ‘pre-shaping , ’ 31 ms ± 10 ms duration ) , which suggests that the birds begin releasing the muscles holding their feet open when they sense that impact is more imminent . Pre-shaping the foot may also help the birds attain a secure grasp more quickly . After contact is made , the feet exhibit decreasing foot and claw angles as the toes finish closing around the perch ( ‘foot wrapping , ’ 19 ms ± 7 ms duration ) . After conforming to the surface , the claw angle generally decreases during the ‘claw curling’ phase ( 185 ms ± 11 ms duration ) . The decrease in claw angle can result from the claw tip curling inwards towards a relatively stationary toe pad or the toe pad slipping outwards towards a relatively stationary claw tip . In some cases , both the toe pad and claw tip may move towards each other . The different perch surfaces only affect the kinematics of the subsequent landing stages after perch contact ( Figure 3C ) . Before contact , foot and claw kinematics remain consistent across all flights , even as the feet begin to close during pre-shaping . If the birds adjusted for perch differences while in the air , differences would be most likely to appear in this final stage before impact . Instead , the parrotlets consistently begin pre-shaping ~30 ms before touchdown ( Figure 3D ) , suggesting that they have predictive control resolution within this time frame . Maintaining sufficiently large claw angles prior to this stage ensures that the claws are not damaged during the controlled collision with the perch . Only after contact do foot and claw kinematics begin to diverge . The extent of foot wrapping is governed by the diameter of the perch ( Figure 3E ) . The parrotlets wrap their feet completely around the 0 . 25" diameter perch , which enables them to use normal forces from their toes to stabilize themselves . Although some claws still engage with the surface , there is no clear claw curling phase . In contrast , the parrotlets exhibit foot wrapping and claw curling on all of the other surfaces . After contacting the perch , both the feet and claws can slip . As a result , the claw curl angle can change during the claw curling stage because the foot slips , the claw slips , or both slip , all of which can take place at varying speeds . Whereas the claw curling stage can last over 100 ms ( based on visual observation , see Materials and methods ) , individual claw curling movements initiated when a claw slips off an asperity can take place as quickly as 1–2 ms ( 2–4% of a full wingbeat; Figure 3F , Figure 3—figure supplements 1 , 2 ) . These superfast claw movements occur most frequently on intermediate surfaces that are neither extremely rough nor extremely smooth . This is likely because rough surfaces , such as Sweet Olive and sandpaper , have asperities that give sufficient friction so that there is little slipping ( Figure 3F , G ) . On the other hand , smooth surfaces , such as Teflon , have very few usable asperities to slip off of or latch onto ( Figure 1A , Figure 5E ) . Deformable surfaces , such as foam , do not have distinct asperities in the same way as the other surfaces , which explains why we did not observe any superfast claw movements on foam . The superfast claw movements are likely not governed by muscle contraction alone; muscle contractions in human hands are significantly slower ( ~50 ms ) ( Buchthal and Schmalbruch , 1970 ) . Even the fastest vertebrate locomotory muscles recorded to date are still 5–10 times slower; the Etruscan shrew’s extensor digitorum longus contracts in 11 ms ( Jürgens , 2002 ) and the superfast pectoralis muscle of a hummingbird contracts in 8 ms ( Ingersoll and Lentink , 2018 ) . Therefore , the speed of the parrotlet claw movements probably rely on the release of energy stored in an elastic tendon ( McNeill Alexander , 2002 ) and the low inertia of the claw . The superfast movements help ensure that the claws slip over timescales that are much shorter than the body dynamics timescales , which helps the bird remain more easily anchored on the perch . In general , we find that the extent of claw curling depends on both perch diameter and surface properties ( Figure 3G ) . Claw curling is the most pronounced on the 1 . 5" diameter perch ( Figure 3G ) . Foot wrapping is much more limited on the large perch ( Figure 3C , E ) , so birds must rely on shear forces from their toe pads and claws to generate sufficient stabilizing forces . The parrotlets tend to curl their claws less on perches that they can squeeze with more force , such as the rougher sandpaper and Sweet Olive perches . They curl their claws more on perches that they can squeeze less , such as the smooth Floss-Silk , the slippery Teflon , and the 1 . 5" diameter perch ( Figure 3H ) . In short , the birds curl their claws more on perches that are harder to grasp . To understand mechanistically how varying squeeze force or claw curling enables the birds to adapt to different perch surfaces , we characterized foot-surface interactions using our surface testing setup ( Figure 1B ) . Drag tests on the different surfaces reveal how toe pads ( Figure 4A ) and claws ( Figure 4C ) produce qualitatively similar shear force traces as they slip and stick . The maximum static shear forces correspond to the maximum force a bird could expect from a toe pad or claw at a given normal force . For toe pads , a Coulomb friction model is a suitable predictor of shear force as a function of normal force for all surfaces ( Figure 4B ) . For claws , the friction ratios ( tangential to normal force ) can reach over eight times those generated by toe pads ( Figure 4C vs . Figure 4A ) . These forces may also be affected by how the claw is oriented relative to the surface ( Figure 4D ) ; while the maximum static friction force stays relatively constant on the smoothest natural surface ( Floss Silk Tree ) , it increases with claw angle on the roughest natural surface ( Sweet Olive Tree ) . Even at a single angle , the variation of the maximum force ratios is much larger for claws than for toe pads ( Figure 4E ) . To compensate for the stochastic nature of claw engagement , birds can take advantage of load sharing and claw movement over the surface . Load sharing across multiple claws effectively narrows the spread of expected total claw forces ( Jiang et al . , 2018 ) ( Figure 4E ) . Birds can also drag their claws over longer distances to improve the likelihood of hitting better asperities . This increases the expected value of the friction force ( Figure 4F ) , which may explain why parrotlets curl their claws more on surfaces that are harder to squeeze ( Figure 3F ) . The greatest benefits of dragging longer distances can be gained on surfaces that have fewer but larger asperities . However , the parrotlets tend to curl more on surfaces that give less expected benefit with claw travel ( Figure 4F ) . This behavior suggests that the birds are not seeking out the best possible asperity when grasping . Rather , it suggests that they are instead curling their claws until they generate sufficient force to maintain a stable grasp or until they reach an upper limit . If the needed force is still not achieved by then , the bird may adjust its footing or take off . The greater variability in claw friction compared to toe pad friction may be explained by the local surface contact geometry . Toe pads , which are relatively soft , can conform to a surface . This allows them to distribute the frictional load across many contact points , leading to more consistent maximum friction forces . In contrast , claws , with a rigid structure and smaller contact area , rely on fitting into local surface geometries or hooking onto asperities . To quantify these geometric effects , we first characterized claw shape in the sagittal plane by width as a function of height from the claw tip ( Figure 5A ) . The wide shape variation among parrotlet claws may result from the growth itself or from claw wear . High load drag tests ( up to 3 times bodyweight , bw ) caused approximately 900 μm of wear on a relatively sharp claw . Although quite high , the resulting claw geometry was still within the range of claw geometries that we measured in some of the duller , untested claws ( Figure 5A , Figure 5—figure supplement 1 ) . Drag tests at up to 25% bw caused no noticeable wear , showing how birds benefit markedly from limiting the normal force load on their claw . In addition to using existing surface asperities , claws can also generate friction forces from surface deformation . To model this interaction , we combine measurements of claw penetration depth into different surfaces ( Figure 5B ) with the claw geometry . We find that the claw tip geometry can be modeled by a 50 μm radius sphere for loads up to 25% bw ( Figure 5C ) . For example , the surface deformation from a normal force of 25% bw in Coast Live Oak can be approximated by a partial sphere with a 10 μm depth and 50 μm diameter . Therefore , although a 10 μm penetration depth is only about 0 . 2% of the outer arc length of a parrotlet claw ( ~5 mm ) , surface deformation can play an important role in maintaining a stable grasp , particularly on soft surfaces . Next , to quantify the effect of claw size on claw engagement , we simulate claws with different tip radii dragging along measured surface profiles ( Figure 5D , model adapted from Asbeck et al . , 2006 ) . The trajectory of the claw generates a traced surface ( Asbeck et al . , 2006; Okamura , 2000 ) , a smoothed version of the original profile that only comprises asperities that the claw can reach . We define the usable surface na as the proportion of the traced surface with a slope greater than the minimum necessary for static friction ( θmin ) . Specifically , θmin is the angle formed between the vertical axis and the net force vector on the claw ( FN +FT ) when the tangential to normal force ratio is at its maximum . Starting with the case of no surface penetration , we find that rougher surfaces ( Figure 1A ) tend to have more usable surface ( Figure 5E ) for a fixed θmin . Still , because of the surface-specific distribution of asperity shapes and sizes , surfaces accommodate different claw sizes slightly differently . For example , despite exhibiting similar overall trends , birch offers more na at smaller tip radius ra while Coast Live Oak has slightly more na at larger ra . An exponential fit with an offset describes the relationship between na and ra well ( Figure 5E ) . We might expect that as ra approaches infinity , na should go to zero ( no offset ) because the traced surface flattens out . However , the offset can be understood by modeling the surface as a fractal; surface features with similar length scales to those shown here still exist at large ra ( Costa et al . , 2000; Greenwood , 1992a; Greenwood , 1992b ) . When surface penetration effects are added ( Figure 5F ) , we find that the usable surface increases , with larger effects at lower θmin . If claw tip radius and mass both scale isometrically , and exerted forces scale with mass , then the depth ratio ( DR , penetration depth divided by tip radius ) will also scale isometrically . Although active muscle forces scale with length to the second power ( McMahon , 1984 ) , we assume that the forces involved in perching scale to the third power . This is because we expect these forces to be governed by bodyweight since perched animals must apply forces just large enough to maintain static equilibrium . Similar trends hold for constant and isometrically scaled depth ratios , DR , but the latter case yields trends that are flatter and eventually reverse for smaller surface slope requirements , θmin . Thus , smaller animals tend to have more usable surface than larger animals when steeper surface features are required , but the trend begins to reverse when shallow asperities provide sufficient friction . We now zoom back out to see how the claws are integrated together with the feet and body in the perching behavior . We introduce a 2D rigid-body model of the parrotlet , in which aerodynamic and inertial forces and torques on the body must be matched by friction forces at the perch surface ( Figure 6A ) . By leveraging our foot-surface contact mechanics data ( Figures 3–5 ) , we can now apply a constrained optimization to determine the bird’s 3D ‘wrench space’ ( Ferrari and Canny , 1992 ) , the set of foot force and torque combinations that permit static grasping ( see Materials and methods ) ( Figure 6B ) . Force plane cross sections show how torque twists and warps the stable boundary region . Thus , both the magnitude and direction of a bird’s velocity when landing are important; if the bird’s velocity vector is not directed at or near the perch center , the bird may slip . The stability region can be actively expanded by the bird through an increase in squeeze force , and is also enlarged when the surface texture itself presents more available surface friction ( Figure 6C ) . Changing the angle of the foot relative to the perch rotates the wrench space ( Figure 6C ) . The effect of perch size on the stability region may be less well-defined in nature . Smaller branches tend to enable greater squeeze forces ( Figure 3G ) , and when the foot can fully wrap around a perch , the ability to sustain forces depends only on muscular and structural limits . Thus , smaller branches should enable birds to withstand larger force disturbances . However , smaller branches can bend , which may make perching more difficult , and larger branches offer a larger area on which birds could potentially perch . Finally , we apply this model to estimate how much more force and torque the birds apply during landing than would be necessary to maintain a static grasp . We define three safety margins , SMT , SMy , and SMx , which correspond to how much more torque or force could be sustained without slip in the following directions , respectively: the axial torque direction ( bz ) , along a vector pointing from the center of the perch to the ankle ( by ) , and the direction perpendicular to both bz and by ( bx ) ( Figure 6B ) . For all surfaces , most of the body forces and torques are absorbed while the bird is still slipping ( Figure 6D ) . Squeeze forces decay more slowly ( Figure 6E ) and appear to exert a greater influence over the safety margin trends ( Figure 6F ) . Except for the large diameter perch flights , SMx values are much larger than SMy values and do not decay as much . This trend arises because pulling the foot in either the forward ( +bx ) or aft ( -bx ) direction increases normal and shear forces on one half of the foot , so SMx will remain relatively high even as a bird relaxes its grip . In contrast , pulling in the by direction away from the perch decreases normal forces on the toes , so less force can be sustained in this direction without slipping . Safety margin peak values tend to be associated with both squeeze force and the quality of surface asperities for generating friction ( Figure 4B , E ) . Perches with longer average slipping times , like the large diameter and Teflon perches , tend to have lower peak safety margins . These perches also do not consistently follow the same fast rise and slow decay trends in the safety margins of the other perches . This suggests that the ability to generate friction is the limiting factor for maintaining a stable grasp . However , we see positive safety margins even when birds are slipping , which suggests that some slipping may be intentional . Indeed , slipping does not always imply a lack of control; rather , it may help absorb body energy during landing , increase the chance of finding high-friction asperities , and help the bird slide into a more stable position at the top of the perch .
When landing , birds must balance safety margin and reducing muscle activation energy expenditure from squeezing . If birds prioritized safety during landing , we would expect consistently high grasp forces until all of the body energy had been absorbed . If they prioritized minimizing energy expenditure from squeezing , we would expect low , constant safety margins . In all landings , we find that the squeeze force and safety margins tend to peak later and decay more slowly than the body forces ( Figure 6D–F ) . The reduction in squeeze force is likely aided by the high shear to normal force ratios that claws can achieve as soon as they lock onto an asperity ( Figure 4E ) , enabling a stable grasp with relatively low normal forces . Reducing normal forces also has the added benefit of limiting claw wear that could result from high loading ( Figure 5A ) . The later peak of the safety margins is likely a result of the delay between impact and wrapping the toes around the perch . These fast rise and slow decay safety margin trends on most of the surfaces indicate that the birds initially overcompensate – likely to help account for uncertainty due to both surface variability and landing variability ( Figure 2B ) from their controlled collisions ( Figure 2E ) – and then they slowly relax to equilibrium as they establish their grip . This ‘overcompensation and then relax’ strategy makes sense since landing safely is a critical requirement for preventing injury , while reducing muscular energy expenditure from squeezing improves energetics . Still , the safety margins are highly variable , likely due to variation between and within different surfaces , as well as variation in the bird’s landing dynamics ( orientation , velocity , etc . ) . It is still unclear how the birds use tactile feedback to adjust their grip during perching , which may drive how they decide to squeeze different surface textures . To understand how the implications of these results may drive the grasping performance of other perching animals and robots , we must consider the effects of scale . In particular , the traced surface model ( Figure 5D–F ) indicates that small animals like insects ( beetle tip radius = 8–10 μm; Dai et al . , 2002 ) or even robots , which generally rely on sharp fish hooks for ‘claws’ ( tip radius = 10–15 μm; Lussier Desbiens and Cutkosky , 2010; Roderick et al . , 2017b ) , have more usable surface on vertical surfaces ( high FT/FN ) ( see Materials and methods for details on the claw penetration depth testing ) . On the other hand , when more normal force is available ( as when squeezing with large forces or on horizontal surfaces ) , larger animals , such as dinosaurs ( Manning et al . , 2006 ) , have just as many or more usable surface options ( Figure 5F ) . Previous work ( Asbeck et al . , 2006; Labonte et al . , 2014 ) reported a directly inverse relationship between asperities per unit length and tip radius at all length scales on rough building materials . The use of this metric was motivated by modeling the probability of hitting an asperity for a given travel distance along the surface . In this work , we chose to use usable surface rather than asperities per unit length because usable surface is independent of the spatial frequency of the surface . In other words , this method , unlike previous methods , assumes that many distinct usable asperities on a given section of surface are just as valuable to a grasping animal as a smaller number of larger asperities , provided that the usable surface in each case is equal and usable asperities will be encountered during the course of claw curling . However , isometric scaling also suggests that structure strength ( ~length2 since stress scales with area ) diminishes with mass ( ~length3 ) , so larger animals may be unable to achieve high safety margins during grasping . On the other hand , they may not need to , because they generally deal with longer time scales and can tune control efforts more precisely ( Kovac , 2016; Roderick et al . , 2017a ) . Indeed , the safety margins employed by parrotlets ( Figure 6F ) are less than those used by beetles ( Voigt et al . , 2017 ) ( ~30 ) , roughly on the same order as those used by snakes ( Byrnes and Jayne , 2014; Marvi and Hu , 2012 ) ( ~5 ) and robots ( Estrada et al . , 2014 ) ( ~4 ) , and generally larger than those used by humans ( Westling and Johansson , 1984; Cadoret and Smith , 1996 ) ( ~1 . 5 ) . While these safety margins are calculated in widely different contexts and do not reflect the different risks involved in grasp failure , they do suggest an association with scale , likely due to the structural and control differences . In summary , parrotlets are able to use stereotyped aerial landing maneuvers across perch variations by adapting to perch differences with their feet and claws . During grasping , they use their toe pads to generate predictable friction forces , and their claws to generate higher but less predictable forces . Claws can be curled along the surface to increase the expected value of the friction force , and multiple claws used together reduce the variance of the average force generated per claw . Further , parrotlets employ an ‘overcompensation then relax’ grasping strategy which balances safety and reducing energy expenditure from squeezing . The separate , specialized functions of the legs and wings versus the feet and claws not only improve the reliability and efficacy of perching , but also simplify the aerodynamic and muscular control required . Many of these landing and grasping insights can be applied more generally to both animals and robotic end-effectors . Our findings lend support to ( 1 ) the utility of modularity for simplifying control and improving individual subsystem performance , ( 2 ) load sharing , ( 3 ) taking advantage of regression to the mean when using toes and claws to handle the statistical variation of surfaces , and ( 4 ) grasp strategies that prioritize first safety then efficiency . These benefits and considerations are equally relevant for animals of all sizes and likely played key roles in shaping landing behavior across arboreal animals . Further , these grasping strategies likely also extend to takeoff maneuvers; stereotyped takeoff behaviors have been observed in previous work ( Provini and Abourachid , 2018 ) as well , although more study is needed to similarly quantify takeoff dynamics over a range of complex surfaces . The parrotlets’ adaptive grasping strategies can also help inform the design of aerial robots , which are frequently equipped with small robotic arms to serve as landing ‘legs' ( Roderick et al . , 2017a ) . Our results suggest that the mechanical design and control of aerial and perching components of multimodal robots can function independently . Regardless of perch surface , aerodynamic surfaces can be controlled the same way using feedforward control with optic flow to make controlled collisions . Meanwhile , the perching mechanism can provide robustness for different perches and surface irregularities by using avian-like grasping features . These features include foot wrapping with multiple joints , claw curling , and tactile feedback for adjustments . The traced surface and wrench space models presented here can give insights into the design tradeoffs between using more contact points , using claws and/or frictional pads , and determining a suitable balance of strength and sharpness in claws . Incorporating the perching capabilities of arboreal birds into aerial robots will greatly improve their versatility and utility in different environments .
We fabricated nine different perches for this experiment: three ‘natural’ , three birch dowel , and three ‘artificial’ . The three natural perches were cut directly from similar diameter branches of three trees on Stanford’s campus: a Coast Live Oak ( Quercus agrifolia ) ( 15 mm dia . ) , a Floss-Silk tree ( Ceiba speciosa ) ( 16 mm dia . ) , and a Sweet Olive tree ( Osmanthus fragrans ) ( 17 mm dia . ) . To minimize moisture loss , the branch segments were kept in a sealed bag until testing , which was within three days of the branches being cut . The birch perches were cut from commercially available 0 . 25" ( 6 mm ) , 0 . 75" ( 19 mm ) , and 1 . 5" ( 38 mm ) diameter birch dowels , all of which were colored black with a marker to match the color of the artificial perches ( to control for color-based visiomotor bias ) . The three artificial perches were made from wrapping 0 . 125"-thick black ( sponge neoprene ) foam around a 0 . 5" ( 13 mm ) diameter birch dowel , transparent PTFE ( Teflon ) tape around a 0 . 75"-dia . birch dowel ( also colored black ) , and 80-grit black sandpaper around a 0 . 75"-dia birch dowel . A limitation of the split perch design was that the parrotlets could lodge a claw in the split when securing their grasp , but this happened infrequently ( less than 25% of recorded flights ) . These events did not involve the claw tracked for kinematics , and we did not observe any noticeable effect on grasping dynamics , so we did not change our treatment of these cases . We measured the external morphology of 32 claws from four parrotlets using a 5X microscope lens ( Meiji Techno MX8000 Metallurgical Microscope ) . All claws were imaged in the sagittal plane at a resolution of 2560 × 1920 pixels . Photos of the claws were taken with a QImaging MicroPublisher 5 . 0 RTV ( 01-MP5 . 0-RTV-R-CLR-10 Color , RTV 10 Bit ) camera and then post-processed in Matlab to determine the width of the claw as a function of the height from the claw tip . The specific steps of the algorithm are described in detail in Figure 5—figure supplement 3 . | Most of the flying vehicles designed by humans need to land on smooth , standardized surfaces such as runways . A bird , on the other hand , can use structures that vary widely in diameter and texture , from phone lines to branches to statues . Yet , few studies have focused on how these animals transition from the air to a perch , and especially on how they adapt to different surfaces . To fill this gap , Roderick , Chin et al . recorded how Pacific parrotlets landed on nine natural and man-made perches that varied in diameter and texture , ranging from smooth Teflon to rough sandpaper . High-speed cameras tracked each of the landings while sensors measured how hard the birds landed on and squeezed the perches . The experiments revealed that the first landing phase was the same regardless of the nature of the perch . The birds used their wings to slow down , unfurled their feet and claws in preparation for touchdown and then allowed their legs to absorb the landing impact . Once the feet had made contact with the surface , however , the birds used their toes and claws to adapt to different perches . First , they steadied their grasp by tightly squeezing the perches . Then , the parrotlets dragged their claws on the surface of the perches to find minuscule bumps and dips that allowed better stabilization . These movements could be remarkably fast – in the range of one to two milliseconds . The birds also curled their claws more on perches that were harder to grasp . Once secured on the branch , they relaxed their grip . The results by Roderick , Chin et al . will help biologists understand how birds , insects and even large tree-dwelling creatures can grab perches in various environments . This knowledge will also be relevant for engineers who are trying to create robots that can climb or land on diverse surfaces . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"evolutionary",
"biology"
] | 2019 | Birds land reliably on complex surfaces by adapting their foot-surface interactions upon contact |
Speech is a complex sensorimotor skill , and vocal learning involves both the basal ganglia and the cerebellum . These subcortical structures interact indirectly through their respective loops with thalamo-cortical and brainstem networks , and directly via subcortical pathways , but the role of their interaction during sensorimotor learning remains undetermined . While songbirds and their song-dedicated basal ganglia-thalamo-cortical circuitry offer a unique opportunity to study subcortical circuits involved in vocal learning , the cerebellar contribution to avian song learning remains unknown . We demonstrate that the cerebellum provides a strong input to the song-related basal ganglia nucleus in zebra finches . Cerebellar signals are transmitted to the basal ganglia via a disynaptic connection through the thalamus and then conveyed to their cortical target and to the premotor nucleus controlling song production . Finally , cerebellar lesions impair juvenile song learning , opening new opportunities to investigate how subcortical interactions between the cerebellum and basal ganglia contribute to sensorimotor learning .
Speech is a highly complex motor skill which requires precise and fast coordination between vocal , facial and respiratory muscles . Human infants learn to reproduce adult vocalizations and to progressively master speech motor coordination within their first few years of life through an imitation process that builds up on motor sequence learning and strongly relies on auditory feedback ( Kuhl and Meltzoff , 1996 ) . This process , called vocal learning , is widely believed to rely on similar mechanisms as sensorimotor learning in general ( Doupe and Kuhl , 1999; Kuhl and Meltzoff , 1996 ) . The neural mechanisms underlying this process remain , however , poorly understood . Brain circuits known to be essential for sensorimotor adaptation and learning , namely the basal ganglia-thalamo-cortical loop ( Krakauer and Mazzoni , 2011; Pekny et al . , 2015 ) and the cerebello-thalamo-cortical loop ( Brooks et al . , 2015; Izawa et al . , 2012 ) , are both crucial for vocal learning in humans ( Vargha-Khadem et al . , 2005; Ziegler and Ackermann , 2017 ) . The anatomical structure of these circuits and their function in sensorimotor learning are well conserved over vertebrate evolution ( Grillner and Robertson , 2016; Redgrave et al . , 1999; Sultan and Glickstein , 2007 ) . In particular , avian song learning has been used as a paradigm to study the neural mechanisms of vocal learning , as it shares striking similarities with human speech learning ( reviewed in Doupe and Kuhl , 1999 ) . The basal ganglia-thalamo-cortical network is involved in sensorimotor learning in several species , from lamprey to primates ( Hikosaka et al . , 2002; Stephenson-Jones et al . , 2013; Wickens et al . , 2007 ) . The basal ganglia are thought to rely on reward prediction error signals conveyed by dopaminergic neurons ( Gadagkar et al . , 2016; Schultz et al . , 1997; Wickens et al . , 2003 ) to drive reinforcement learning strategies ( Doya , 2000; Sutton and Barto , 1981 ) . In songbirds , a specialized circuit homologous to the motor loop of the mammalian basal ganglia ( McCasland , 1987; Doupe et al . , 2005 ) is critical for song learning in juveniles and plasticity in adults ( Brainard and Doupe , 2002 ) . This circuit is thought to correct vocal errors through reinforcement learning driven by an internal song evaluation signal conveyed by dopaminergic neurons ( Fee and Goldberg , 2011; Gadagkar et al . , 2016; Hoffmann et al . , 2016 ) . The cerebello-thalamo-cortical circuit also participates in sensorimotor learning in vertebrates , from fishes to primates ( Brooks et al . , 2015; Gómez et al . , 2010; Lewis and Maler , 2004 ) . It is believed to implement error-based supervised learning ( Albus , 1971; Ito , 1984; Knudsen , 1994; Marr , 1969; Raymond et al . , 1996 ) based on an error prediction denoting a mismatch between sensory prediction and actual sensory feedback ( Doya , 2000; Dreher and Grafman , 2002 ) . The cerebellum also drives on-line correction during movements building on the same sensory error prediction ( Tseng et al . , 2007 ) and controls the duration of movements and its prediction during sensorimotor learning ( Day et al . , 1998; Flament and Hore , 1988; Izawa et al . , 2012 ) . The existence of a pathway from the cerebellum to the song-related basal ganglia has been suggested by previous anatomical studies in songbirds ( Person et al . , 2008; Vates et al . , 1997; Nicholson et al . , 2018 ) , but whether cerebellar circuits are involved in avian song learning and production remains unknown . Beyond the indirect interaction via their respective loop with thalamo-cortical and brainstem networks , the basal ganglia and the cerebellum interact via a subcortical disynaptic pathway through the dentate nucleus , the motor part of the thalamus - more precisely the ventral anterior and ventral lateral nuclei of the thalamus in monkeys , and the centro-median nucleus of the thalamus in rodents - and the striatum ( Bostan et al . , 2010; Chen et al . , 2014; Hoshi et al . , 2005 ) . The cerebellum and the basal ganglia therefore may not simply act in parallel to shape cortical and brainstem activity during learning . Instead , we hypothesize that cerebellar signals may reach the basal ganglia to drive error correction and reinforcement learning through the same output pathway . We test this hypothesis in zebra finches . We show that ( i ) cerebellar inputs are conveyed to the basal ganglia in songbirds via the thalamus , ( ii ) they drive activity in the cortical target of the basal ganglia , and ( iii ) the cerebellar signals contribute to juvenile song learning , and to the timing of song elements .
We performed anatomical tracing experiments to confirm the previously reported ( Person et al . , 2008 ) indirect connection from the DCN to the song-related basal ganglia nucleus Area X , via the dorsal thalamic zone ( DTZ ) . In a first set of experiments ( n = 2 birds ) , we used two bidirectional tracers ( fluorescently tagged dextran ) injected both in Area X and the lateral DCN ( Figure 1A–C ) . We then injected in Area X a retrograde tracer captured by synapses , Cholera-toxin B , while a bidirectional tracer ( fluorescently tagged dextran ) was injected in the lateral DCN ( n = 1 bird , Figure 1D–G ) . In the cerebellum , the concomitant labeling of DCN and Purkinje cells indicated the proper location of the injection sites in the DCN ( Figure 1D and Figure 1—figure supplement 1 ) . As illustrated in both examples ( Figure 1B , C , E and F ) , we found fibers labeled with the DCN-injected tracer in DTZ , posterior to the thalamic nucleus involved in song learning and production , the dorsolateral nucleus of the anterior thalamus ( DLM ) . This provides evidence of axonal projections from the lateral DCN neurons to this region . Within the same DTZ area , cell somata of thalamic neurons were labeled with either the bidirectional or retrograde tracer injected in Area X ( Figure 1B , C , E and F ) . We observed a close association between the two types of tracers with anterogradely-labeled fibers making putative contacts on retrogradely-labeled cell bodies ( Figure 1G ) . This observation suggests that neurons in the lateral DCN project to DTZ thalamic neurons , which in turn project to Area X . We also injected bidirectional tracers ( fluorescently tagged dextran , n = 2 birds ) in DTZ ( Figure 1H ) . In the cerebellum , retrograde transport of the tracer was confined to large cell bodies within the DCN ( Figure 1I ) . These large cells likely correspond to the large glutamatergic DCN output neurons that project to premotor areas . Labeled cell bodies were located for the most part in the lateral DCN . We did not find dorso-ventral distinction in the labelling of the lateral DCN , suggesting that the projection from the lateral DCN to DTZ is not topographically organized ( Figure 1I ) . Some neurons in the interpositus nucleus were also labeled ( results not shown ) . This suggests that , even if the projection from the cerebellum to DTZ largely comes from the lateral DCN , the interpositus may also be partially involved in this cerebello-thalamic projection . Regarding the anterograde transport of tracers injected in DTZ ( Figure 1H ) , we found many labeled axonal fibers in Area X , confirming the direct projection from DTZ to Area X ( Figure 1J ) . Thus , as already suggested in a previous study ( Person et al . , 2008 ) , we found anatomical evidence for a disynaptic connection between the cerebellum and the song-related basal ganglia Area X: the lateral DCN sends projections to DTZ which in turn projects to Area X . Importantly , these anatomical results have been replicated very recently , confirming the existence of the DCN-DTZ-Area X pathway ( Nicholson et al . , 2018 ) . We then determined whether this DCN-DTZ-Area X pathway drives activity within the basal ganglia . To this end , we investigated the responses evoked by DCN electrical stimulation in Area X neurons in anaesthetized zebra finches . To this end , we investigated the responses evoked by DCN electrical stimulation in Area X neurons . Most neurons are silent or display very little spontaneous activity in Area X under isoflurane anesthesia , whereas a minority of them displays high spontaneous activity ( >25 spikes/sec , see Materials and methods ) . These spontaneously active neurons are most likely pallidal-like neurons ( Leblois et al . , 2009; Person and Perkel , 2007 ) . Hereafter , this population of neurons , at least some of which are Area X projection neurons ( Goldberg et al . , 2012; Leblois et al . , 2009 ) , will be referred as pallidal neurons . DCN stimulation provoked a strong increase in the firing rate of pallidal neurons , as illustrated in Figure 2B . When a response was evoked by single-pulse stimulation in at least one pallidal neuron in Area X , all subsequently recorded neurons were also responsive to the stimulation . However , their response profile at a given intensity differed from one another . This diversity of response profiles could be classified as follows: single excitatory responses ( observed in 71% of case , Figure 2C , two last bottom panels 0 . 2 and 0 . 5 mA ) , biphasic responses with excitation followed by inhibition ( observed in 19% of case , Figure 2C , middle panel 1 mA ) , or triphasic responses with a rapid inhibition followed by an excitation and a late inhibition ( observed in 10% of case , Figure 2C , top panel 2 mA ) . Interestingly , different response profiles were found in the same neuron depending on the stimulation intensity used: higher stimulation intensity induced biphasic or triphasic responses , while lower stimulation intensity only caused excitation . Previous studies have shown that excitatory inputs to Area X can drive such biphasic or triphasic responses in pallidal neurons due to feedforward inhibition mediated by local Area X inhibitory neurons ( Leblois et al . , 2009 ) . The response latencies between the onset of the stimulation pulse and the onset of the excitatory response ( see Materials and methods ) were broadly distributed from 10 to 50 ms ( 20 . 80 ± 4 . 56 ms , median: 21 ms , Figure 2D ) . While short latency responses ( 10–20 ms ) can be naturally explained by a disynaptic excitatory transmission from the DCN to Area X through DTZ , biphasic and triphasic responses involve longer latencies to which feedforward inhibition within Area X likely participates . Indeed , fast feedforward inhibition within Area X can delay the response of pallidal neurons to their excitatory inputs ( Leblois et al . , 2009 ) , as it is the case in the mammalian striatum ( Mallet et al . , 2005 ) . Altogether , these results show that stimulation of DCN neurons can drive the activity of pallidal neurons in Area X , confirming that the latter receive a functional input from the cerebellum . In the original version of the paper , the stimulation intensities applied in the deep cerebellar nuclei ( DCN ) to evoke neural responses in Area X were high relative to the intensities typically used in ortho- or anti-dromic functional mapping . The duration of stimulation pulses was also long ( 1 ms ) , leading to a high level of total stimulation current . The selectivity of such stimulation may therefore be questioned . To resolve this issue , we have pursued additional experiment to assess the responsiveness of Area X pallidal neurons following low intensity ( 50-200 µA ) and short duration ( 100 µs ) stimulation pulses in the lateral DCN . We show in a new figure ( Figure 3 of the present version of the article ) that Area X pallidal neurons show strong responses to these low-intensity and short-duration stimulation . Moreover , we provide evidence that even high-intensity ( 1 mA ) and long-duration ( 1 ms ) stimulation , when applied a few hundreds of microns away from the lateral DCN , do not evoke responses in Area X pallidal neurons . Altogether , our additional results confirm that pallidal neurons in Area X selectively respond to the stimulation of the lateral DCN . Our anatomical results suggest that DTZ relays Area X neuronal responses to cerebellar stimulation . To demonstrate this , we blocked glutamatergic transmission in DTZ while monitoring the responses in Area X to DCN stimulation . In mice and rats , the cerebellar projections to the thalamus are mediated by glutamate ( Kuramoto et al . , 2009; Kuramoto et al . , 2011 ) . We therefore pressure-injected a cocktail of AMPA/kainate and NMDA receptor antagonists , respectively 2 , 3-dihydroxy-6-nitro-7-sulfamoyl-benzo quinoxaline-2 , 3-dione ( NBQX ) and 2-amino-5-phosphonovaleric acid ( APV ) , to block glutamatergic transmission within DTZ ( see Materials and methods , Figure 4A ) . Figure 4B shows an example of the change in the response of a pallidal neuron to DCN stimulation following the injection of glutamatergic blockers in DTZ . As our hypothesis predicts , the excitation that DCN stimulation induced in this pallidal neuron was suppressed following drug injection . We then quantified the change in response induced by glutamatergic blockers in DTZ over the population of pallidal neurons recorded under this pharmacological protocol ( n = 16 pallidal neurons in 8 birds ) . The response strength and peak of the excitatory response ( see Methods ) were strongly reduced or totally suppressed when we blocked DTZ glutamatergic relay . Mean response strength decreased from 0 . 55 ± 0 . 13 spikes at baseline to 0 . 16 ± 0 . 04 spikes following drug injection ( paired Wilcoxon test , p=3e-004 , Figure 4C ) , and mean excitation peak decreased from 99 ± 23 Hz at baseline to 44 ± 10 Hz following drug injection ( paired Wilcoxon test , p=5e-004 ) . These results confirm that the responses to DCN stimulation in Area X pallidal neurons are relayed by glutamatergic transmission in DTZ . Thalamo-striatal projections are glutamatergic in most vertebrates ( Smith et al . , 2004 ) . It is thus natural to suppose that DTZ neuronal projections excite Area X neurons through glutamatergic transmission in zebra finches . We tested this hypothesis by blocking glutamatergic transmission around the recorded pallidal neuron upon injection of the same glutamatergic blockers as above ( Figure 4D ) . We indeed confirmed that responses of pallidal neurons to DCN stimulation were abolished by the drug injection ( Figure 4E and F; response strength decreased from 0 . 8 ± 0 . 3 spikes at baseline to 0 . 16 ± 0 . 05 spikes following drug injection , paired Wilcoxon test , p=0 . 008 , and mean excitation peak decreased from 125 ± 44 Hz at baseline to 30 ± 10 Hz following drug injection , paired Wilcoxon test , p=0 . 008 ) . We cannot completely exclude that drugs injected in DTZ could diffuse to DLM , which would block a response mediated by the well-known DLM-LMAN-Area X pathway . To rule this alternative hypothesis out , we applied in LMAN the cocktail of AMPA and NMDA receptor antagonists while monitoring pallidal responses to DCN stimulation ( Figure 5A ) . We found no significant difference in the excitatory response of pallidal neurons to DCN stimulation between baseline and drug application conditions ( Figure 5B and C , n = 12 neurons in 6 birds; response strength was 1 . 49 ± 0 . 5 spikes at baseline and 1 . 34 ± 0 . 38 spikes following drug injection , paired Wilcoxon test , p=0 . 5; mean excitation peak was 211 ± 50 Hz at baseline and 200 ± 47 Hz following drug injection , paired Wilcoxon test , p=0 . 5 ) . While LMAN does not appear to mediate the main response to DCN stimulation in Area X pallidal neurons , it may participate to a reverberation of the responses through the Area X – DLM – LMAN loop . In this respect , is interesting to note that all but one pallidal neurons underwent a slight decrease in their response upon glutamatergic blockade in LMAN , possibly reflecting a reduced reverberation in the loop . As the measured response strength only reflects the first peak of excitatory response in Area X , the slow response mediated by the propagation through the loop is unlikely to provide an important contribution to this measure ( see Methods ) . Following each drug injection in LMAN , we verified the efficacy of the pressure injection by moving the drug pipette in the vicinity of the recorded pallidal neuron ( Figure 5C , inset , n = 5 pallidal neurons in 5 birds ) . During those controls , DCN stimulation response strength decreased from 1 . 32 ± 0 . 59 spikes at baseline to 0 . 35 ± 0 . 16 spikes following drug injection ( paired Wilcoxon test , p=0 . 008 ) and mean excitation peak was reduced from 182 ± 82 Hz at baseline to 57 ± 26 Hz following drug injection ( paired Wilcoxon test , p=0 . 02 ) . This confirms that our pharmacological blockades were efficient , and we can therefore rule out a transmission from DCN to Area X through the DLM-LMAN pathway . In songbirds , the DTZ relays projections from the DCN to Area X and is composed of several thalamic regions as described previously by anatomical studies ( Person et al . , 2008; Vates et al . , 1997 ) . One of these regions , called the dorsal medial posterior thalamic zone ( DMP ) projects to the medial part of the magnocellular nucleus ( MMAN ) ( Foster et al . , 1997; Nicholson et al . , 2018 ) . MMAN is in turn implicated in a pathway ending in the song-related motor nuclei HVC ( used as a proper name ) and RA ( Williams et al . , 2012 ) . As HVC projects to Area X ( Nottebohm et al . , 1976; Nottebohm et al . , 1982 ) , we wondered whether the response we observed in Area X could be conveyed through this MMAN-HVC-X pathway . To rule out this possibility , we blocked glutamatergic transmission in MMAN while monitoring pallidal responses to DCN stimulation ( Figure 5D ) . We found no significant effect of the drug injection in MMAN on the responses of pallidal neurons to DCN stimulation ( Figure 5E and F , n = 5 pallidal neurons in 2 birds , response strength was 1 . 43 ± 0 . 24 spikes at baseline and 1 . 63 ± 0 . 43 spikes following drug injection , Wilcoxon test , p=0 . 8; mean excitation peak was 246 ± 110 Hz at baseline and 259 ± 116 Hz following drug injection , paired Wilcoxon test , p=0 . 4 ) . As previously , we checked the efficacy of the pressure injection through the glass pipette in Area X at the end of each experiment ( Figure 5F , inset , n = 5 pallidal neurons , decrease from 1 . 33 ± 0 . 66 spikes at baseline to 0 . 12 ± 0 . 06 spikes following drug injection , Wilcoxon test , p=0 . 03; mean excitation peak decreased from 239 ± 119 Hz at baseline to 35 ± 18 Hz following drug injection , paired Wilcoxon test , p=0 . 03 ) . This experiment ruled out the possible transmission from DCN to Area X pallidal neurons through the MMAN-HVC-Area X pathway . In conclusion , the results of our electrophysiological experiments provide strong evidence that the cerebellum is linked to the song-related basal ganglia nucleus Area X through a functional excitatory connection involving a glutamatergic projection from the DCN to DTZ , and a glutamatergic projection from DTZ to Area X . In songbirds , Area X is known to be part of the basal ganglia-thalamo-cortical circuit homologous to the motor loop of the basal ganglia-thalamo-cortical networks in mammals ( Brainard and Doupe , 2002 ) . This basal ganglia-thalamo-cortical loop affects song production and drives song learning and plasticity via its projection to the premotor nucleus RA ( Andalman and Fee , 2009; Bottjer et al . , 1984 ) . In the following experiments , we tested whether the responses observed in the pallidal neurons after DCN stimulation were conveyed to the output nucleus of the basal ganglia-thalamo-cortical loop , LMAN , and its efferent premotor nucleus RA ( Figure 6 ) . DCN stimulation elicited strong responses in LMAN neurons ( Figure 6B ) . Those responses could be composed of two excitatory components: a first strong and rapid one followed by a delayed and slow one . Such bimodal responses with two peaks were found in 10% ( n = 3/30 ) of the LMAN neurons recorded . The remaining LMAN neurons ( 90% , n = 27/30 ) displayed one of the two excitatory responses provoked by DCN stimulation . The latency of excitatory responses in LMAN neurons was therefore spread in a bimodal distribution ( Figure 6C ) with two distinct peaks: a first one between 10 and 50 ms ( 26 ± 7 . 8 ms , median: 19 ms , 28% of all recorded LMAN neurons , n = 8/30 ) , and a second one around 100 ms ( 125 ± 32 ms , median: 110 ms , 72% of all recorded LMAN neurons , n = 22/30 ) . Interestingly , these two peaks in the latency distribution of LMAN neurons mirrored the inhibitory responses observed in Area X pallidal neurons . Indeed , pallidal neurons displayed inhibitory responses either preceding or following the excitatory component of their response . Inhibition in Area X pallidal neurons , many of which project to the thalamic nucleus DLM ( Fee and Goldberg , 2011; Leblois et al . , 2009 ) , induces a fast excitatory response in DLM neurons ( Goldberg et al . , 2012; Leblois et al . , 2009; Person and Perkel , 2007 ) and thereby activates LMAN through DLM excitatory projections ( Leblois et al . , 2009 ) . The first excitation in LMAN neurons , around 20 ms latency , could therefore be mediated by the fast inhibition observed in pallidal neurons ( Figure 2C , top panel ) . Similarly , the slow inhibitory component of pallidal responses to DCN stimulation , with a mean latency around 30 ms ( 28 . 2 ± 9 . 5 ms ) , likely activates the DLM-LMAN pathway with longer latencies ( >50 ms ) and may therefore drive the second excitation in LMAN . To confirm that Area X relays the response of LMAN neurons to DCN stimulation , we first blocked glutamatergic transmission in Area X ( Figure 6D ) . The response strength was calculated as the total area of the response , containing one or two peaks of excitation when they are present . After application of the glutamatergic blockers to Area X , responses were suppressed in LMAN ( Figure 6E and F , n = 14 multiunit recordings , response strength decreased from 2 . 04 ± 0 . 54 spikes at baseline to 0 . 89 ± 0 . 23 spikes following drug injection , paired Wilcoxon test , p=0 . 001; mean peak excitation decreased from 27 . 3 ± 7 . 3 Hz at baseline to 10 . 9 ± 2 . 9 Hz following drug injection , paired Wilcoxon test , p=4e-004 ) . Finally , to confirm that the inhibitory components in the pallidal response to DCN stimulation mediate responses in LMAN , we blocked fast GABAergic transmission in Area X with the GABA-A receptor inhibitor gabazine while monitoring the response of LMAN neurons to DCN stimulation ( Figure 6G ) . We observed the suppression of LMAN neurons excitatory responses after GABAergic blockade in Area X ( Figure 6H and I , n = 7 multiunit recordings , response strength decreased from 0 . 94 ± 0 . 48 spikes at baseline to 0 . 06 ± 0 . 05 spikes following gabazine injection , paired Wilcoxon test , p=0 . 02; mean peak excitation decreased from 42 . 51 ± 9 . 66 Hz at baseline to 9 . 17 ± 6 . 68 Hz following gabazine injection , paired Wilcoxon test , p=0 . 03 ) . Altogether , our results strongly support the view that DCN inputs are transmitted through the basal ganglia-thalamo-cortical loop via the disinhibition of DLM thalamic neurons by Area X pallidal neurons , evoking an excitatory response in LMAN . We then tested whether DCN stimulation also drives responses in RA neurons via this loop ( Figure 6J ) . DCN stimulation induced strong excitatory responses in RA neurons ( Figure 6K ) with latencies in the 10 to 100 ms range ( Figure 6L , 30 . 2 ± 7 . 8 ms , median: 16 ms ) , consistent with a transmission through LMAN . Blocking glutamatergic transmission in LMAN significantly reduced the excitatory response to DCN stimulation in RA neurons ( Figure 6K and M , n = 6 neurons in 5 birds , response strength decreased from 0 . 8 ± 0 . 32 spikes at baseline to 0 . 29 ± 0 . 12 spikes following drug injection , Wilcoxon test , p=0 . 009; mean excitation peak decreased from 187 ± 76 Hz at baseline to 71 ± 29 Hz following drug injection , paired Wilcoxon test , p=0 . 02 ) . Our experiments provide evidence of a functional disynaptic cerebellum-thalamus-basal ganglia pathway in songbirds . This pathway drives the output nucleus of the basal ganglia-thalamo-cortical loop , LMAN , and drives activity in RA neurons . As song learning relies on the basal ganglia-thalamo-cortical loop ( Bottjer et al . , 1984; Brainard and Doupe , 2002; Nottebohm et al . , 1976; Scharff and Nottebohm , 1991 ) , we tested the hypothesis that the cerebellum contributes to song learning during development . Juvenile zebra finches were subjected to partial lesions in their lateral DCN , either electrolytic ( n = 7 ) or chemical using ibotenic acid ( n = 3 ) . Lesions were performed between 55 and 60 days post hatch ( 56 . 8 ± 7 . 5 dph for the lesion group , 57 . 0 ± 4 . 5 dph for the sham group ) , a period which corresponds to the end of the sensory phase of song learning , and to the beginning of the sensorimotor phase ( Figure 7B ) . Figure 7B displays the spectrograms of the song motifs produced by a tutor and its two fledglings ( pupils ) after crystallization phase ( 90 to 100 dph ) , one of them with a DCN lesion . The pupil that underwent the DCN lesion copied fewer syllables than his control brother . To test for a systematic effect of DCN lesions on song imitation , we compared the quality of tutor imitation of the pupils undergoing partial DCN lesion or sham surgery . To this end , we computed the average imitation score over multiple song bouts ( Mandelblat-Cerf and Fee , 2014 ) . The song bouts ( 50–100 in each condition ) were carefully sorted among 2 days of recordings before and after the surgery , and after crystallization ( 90 dph ) . This was done for birds of both the lesion and sham groups . We found a significant correlation between the proportion of the lateral DCN that was left intact and the relative increase in imitation score between the period preceding the surgery ( pre ) and the crystallization period ( Pearson’s correlation coefficient r = 0 . 7 , p=0 . 03; Figure 7D ) . Moreover , birds with large lesions ( <75% lateral DCN left intact , n = 7/10 ) displayed a lower imitation score than the sham group at crystallization ( large lesion group: imitation score of 0 . 39 ± 0 . 09 , n = 7 , sham group: imitation score of 0 . 51 ± 0 . 06 , n = 6 , t-test , df = 11 , p=0 . 04 , Figure 7E , F ) . We confirmed this effect of DCN lesions using a custom-written similarity score analysis based on the peak cross-correlation between the spectra of the tutor’s motifs and of the pupil’s songs ( see Figure 7—figure supplement 3 ) . In conclusion , partial lesions in the lateral DCN induced a subtle but significant effect on the song acquisition process in juvenile zebra finches , providing evidence that the cerebellum contributes to song learning . Imitation scores are affected by both acoustic and temporal features of the song . To understand in more details how the cerebellum may contribute to song learning or production , we compared temporal and acoustic features of the song before and after DCN lesion in juvenile and adult zebra finches . As exemplified on Figure 8B , DCN lesions in juvenile birds induced a consistent drift in syllable duration ( Figure 8B , see Figure 8A and Material and methods for details on how syllable duration is calculated ) . To determine if and how syllable duration was affected by DCN lesion , we report the relative change in syllable duration induced by the lesion between the baseline ( 2 days preceding the lesion ) and the following period ( days 5–6 after lesion , a period chosen to avoid contamination by transient short-term effects of surgery , Figure 8C , left panel ) . Relative changes in syllable duration are higher following DCN lesion than in the sham juvenile group ( Wilcoxon test , n = 21 syllables in the lesion group , n = 28 syllables in the sham group , p=0 . 003 ) , demonstrating that DCN lesions impact syllable duration in juvenile birds . In contrast , the variability of syllable duration was not affected by cerebellar lesions ( Figure 8C , Wilcoxon test , p=0 . 03 , non-significant when correcting for multiple tests , see Materials and methods ) . In adult birds , the effect of DCN lesions on syllable duration did not reach significance , although a similar trend to increase the relative change in syllable duration compared to sham was observed ( Figure 8—figure supplement 2A–B , Wilcoxon test , non-significant , see Supplementary file 1 for detailed statistical value ) . These results show that lateral DCN lesions performed at 60 dph do not completely prevent birds from copying a tutor or modifying song syllable duration over development . However , comparing the course of syllable duration of sham and lesion birds between the early sensorimotor phase and the crystallization period suggests that those lesions affect the developmental trajectory of song timing properties ( Figure 8B ) . To reveal this , we compared the relative change in syllable duration between the period post 5–6 ( after stabilization of acute lesion effects ) and 90 dph for the sham lesion groups . Figure 8D shows that sham birds display a change of 12 ± 2% during this period , revealing the normal syllable duration learning process at this stage . The group with DCN lesion , on the contrary , displayed a smaller change in syllable duration over the same time interval ( 4 ± 3% , Figure 9D , Wilcoxon test , n = 21 syllables in lesion group , n = 24 syllables in sham group , p=0 . 02 ) . A closer look at the change in syllable duration after lesion and at crystallization ( Figure 8—figure supplement 1A ) reveals that DCN lesions induce a small acute drift in duration but prevent further changes possibly related to the normal learning process . Thus , lateral DCN lesions performed during the sensorimotor stage impair the learning-related changes in syllable duration . Our analysis of syllable duration was based on threshold detection ( see Materials and methods and Figure 8A ) , and strongly depends on the sound amplitude during singing: a lower sound amplitude , for example , could induce an artifactual decrease in syllable duration in our analysis . We thus checked if DCN lesions affected the amplitude of syllables in adult and juvenile birds ( Figure 8—figure supplement 2A and D ) . DCN lesions induced no change in syllable amplitude or in its variability in adults ( Figure 8—figure supplement 2D , Wilcoxon test , non-significant for all periods ) or in juveniles ( Figure 8—figure supplement 1D , Wilcoxon test , non-significant for all periods , see legend for details ) , and we can thus rule out any artifactual change in duration due to an effect of the lesion on syllable amplitude . LMAN , the output nucleus of the song-related basal ganglia-thalamo-cortical loop , is known to drive learning-induced changes in the fundamental frequency of syllables ( Andalman and Fee , 2009; Warren et al . , 2011 ) and to affect its variability ( Kao et al . , 2005 ) . Because we showed that LMAN is under the influence of cerebellar input , lateral DCN lesions could also affect the fundamental frequency of the harmonic stacks in the song motif . Comparison of relative changes in the fundamental frequency of harmonic stacks between the two groups did not reveal any significant change during the early period after lesion ( Figure 8E and F , n = 19 harmonic stacks for the lesion group and n = 18 stacks for the sham group , Wilcoxon test , p=0 . 4 ) . We also found no effect of DCN lesion on the learning trajectories of fundamental frequency , measured as the change in frequency between the last period after lesion and the crystallization ( Figure 8G , sham group , n = 10 fundamental frequency syllable , mean: 2 . 7 ± 1 . 9% , lesion group , n = 19 fundamental frequency syllable , mean: 1 . 6 ± 1 . 5% , Wilcoxon test , non-significant , p=0 . 5 ) . Adult birds did not display any significant change in fundamental frequency following DCN lesions either ( Figure 8—figure supplement 2E–F ) . Finally , the variability of fundamental frequency was not affected by DCN lesion in adult or juveniles ( Figure 8F , Wilcoxon test , non-significant , see Supplementary file 1 for detailed statistical value ) . Altogether , our results suggest that the cerebellar output from the DCN is not required for the acquisition and adjustment of harmonic stacks fundamental frequency .
The DCN receive strong convergent Purkinje cell inputs from many functional territories in the cerebellar cortex ( Apps and Garwicz , 2005 ) . To avoid impairing global function or vital sensorimotor abilities ( potentially leading to a high post-operative mortality ) , we limited the extent of our lesions and monitored the animal state and gross motor functions in the days following the lesion . Our quantification of the effect of DCN lesions on song were performed once the transient motor impairments observed following surgery had disappeared and the birds had resumed perching and singing . Gross motor dysfunction was thus unlikely to significantly contribute to the specific changes observed in song . However , only specific lesions of the cerebello-thalamic projections achieved by pathway-specific ablation techniques will rule out this experimental limitation in the future . Our results indicate that the cerebellar input to the basal ganglia modulates the activity of putative pallidal neurons . We did not directly investigate the response of other neuronal types in this structure . The song-related basal ganglia nucleus , Area X , contains all the neuron types found in the striatum and pallidum in mammals ( Farries and Perkel , 2000; Farries and Perkel , 2002 ) : pallidal neurons , medium spiny neurons and several striatal interneuron types . Only pallidal neurons , however , project outside of the basal ganglia; these share physiological , biochemical and anatomical properties with mammalian pallidal neurons ( Carrillo and Doupe , 2004 ) . Songbirds pallidal neurons display strong spontaneous activity both in vitro ( Budzillo et al . , 2017; Farries and Perkel , 2000; Farries and Perkel , 2002 ) and in vivo ( Person and Perkel , 2007; Goldberg and Fee , 2010 ) and can therefore be distinguished from the other neuronal populations in the song-related basal ganglia nucleus , the spontaneous activity of which is much lower ( Person and Perkel , 2007; Leblois et al . , 2009; Goldberg and Fee , 2010 ) . Given the strongly bimodal distribution of spontaneous activity observed in our recording ( see Materials and methods ) and the relative scarcity of neurons displaying a low spontaneous activity in the song-related basal ganglia nucleus ( Goldberg and Fee , 2010 ) , our dataset likely contains mostly if not only pallidal neurons . A contribution from a small fraction of spontaneous striatal interneurons cannot , however , be ruled out in the absence of post-hoc histological verification of the recorded cell type . In mammals , a pathway connecting the cerebellum to the striatum through the thalamus was demonstrated in rodents ( Chen et al . , 2014 ) and monkeys ( Hoshi et al . , 2005 ) . However , it remains unknown whether and how these cerebellar inputs are conveyed to basal ganglia output neurons and to their thalamo-cortical targets ultimately affecting behavior ( Alexander et al . , 1990 ) . Here , we show in songbirds that the cerebellar signals travel through the basal ganglia-thalamo-cortical circuit and can drive firing in song-related premotor neurons . In monkeys , the dentate nucleus can be divided into two parts: the dorsal part , which has reciprocal projections with motor and premotor cortical areas via the motor thalamus , and the ventral part , which has reciprocal projections with associative and other non-motor cortical areas via non-motor thalamic regions ( Dum and Strick , 2003; Kelly and Strick , 2003; Orioli and Strick , 1989 ) . Additionally , anatomical tracing shows that some projections to the thalamus also come from the interpositus and the fastigial nuclei ( 25% ) ( Bostan et al . , 2010; Hoshi et al . , 2005 ) . In songbirds , our tracing experiments show that one part of the thalamus projects to the song-related basal ganglia nucleus and receives extensive axonal projections from the most lateral part of the DCN , that could be analogous to the dentate nucleus in mammals ( Arends and Zeigler , 1991; Sultan and Glickstein , 2007; Voogd and Glickstein , 1998 ) . However , we found no dorso-ventral contrast in the lateral DCN and we thus make no distinction between potential motor and non-motor parts of this nucleus . Bidirectional tracer injected in the dorsal thalamus revealed a weak , but consistent , projection from the intermediate nucleus , analogous to nucleus interpositus in mammals ( Arends and Zeigler , 1991; Sultan and Glickstein , 2007; Voogd and Glickstein , 1998 ) . Although the labeling was less intense in the intermediate nucleus as compared to the lateral one ( suggesting weaker projections to the thalamus ) , both cerebellar nuclei seem to project to the dorsal thalamus , as reported in Nicholson et al . ( 2018 ) . Both of them may , thereby , be involved in the cerebello-thalamo-basal ganglia pathway studied here . During our electrophysiological experiments , the stimulation electrode targeted the most lateral part of the DCN , as confirmed histologically . We could observe the activation of the cerebello-thalamo-basal ganglia pathway only with very specific and restrictive placement of the stimulation electrode ( see Materials and methods ) . It is thus unlikely that the responses we report were due to current spread to the neighboring intermediate nucleus . However , the size of the stimulated area can hardly be controlled ( Ranck , 1975; Tehovnik et al . , 2006 ) , and we cannot exclude a contribution of the intermediate nucleus to the neural responses we describe here . Further investigations will be necessary to assess this question and determine the role of the putative connections between the intermediate nucleus and the thalamus . Because striatal and pallidal neurons are intermingled in the song-related basal ganglia nucleus ( Farries and Perkel , 2000; Farries and Perkel , 2002 ) , we could not determine the direct targets of thalamic fibers: - the striatal neurons ( as in mammals , Smith et al . , 2004 ) - the pallidal neurons - or both . While we focused on the song-related basal ganglia nucleus , the thalamic projections may also reach other parts of the avian basal ganglia . Further investigation using multiple tracing techniques will be necessary to clarify this question and determine which thalamic area projects to which neurons in the basal ganglia . The cerebellum is a major contributor to timing processes in the brain , both by controlling the duration and variability of movement and by computing the timing prediction necessary to produce an accurate and adapted response during sensorimotor learning . More particularly , clinical observations have highlighted that sensorimotor timing is strongly impaired in patients with unilateral cerebellar lesions . These patients are not able to realize a task in a precise time ( Day et al . , 1998; Flament and Hore , 1988; Izawa et al . , 2012 ) or to conserve a temporal motor pattern in repetitive and synchronized tasks ( Ivry and Keele , 1989; Ivry et al . , 2002 ) . These observations were confirmed with transcranial magnetic stimulation ( Bijsterbosch et al . , 2011; Théoret et al . , 2001 ) . In repetitive tapping tasks , it has been also shown that motor variability is increased when the lateral cerebellum is inhibited ( Théoret et al . , 2001 ) and that compensatory mechanisms appear if patients are asked to do bimanual tasks ( Bijsterbosch et al . , 2011; Franz et al . , 1996; Théoret et al . , 2001 ) . In mammals , the cerebellum is also responsible for the correct perception of time and time intervals ( Moberget et al . , 2008; Rao et al . , 1997 ) . Conditioning of the eyeblink reflex , which relies on timing ( delay ) learning , is impaired following lesions of the cerebellum ( Woodruff-Pak and Thompson , 1985 ) . In our results , we reveal an involvement of the cerebellum in the duration of syllables but no effects on variability of syllable duration . The relatively small changes in syllable duration induced by DCN lesions may be highly significant behaviorally as zebra finches have been shown to discriminate syllable duration with millisecond precision ( Narula and Hahnloser , 2013 ) . Knowing which specific features of timing functions ( i . e . perception of time or movement timing ) is impaired in our songbird model remains an open question . We revealed a functional connection from the lateral nucleus of the cerebellum to the song-related basal ganglia thalamo-cortical loop , known to generate variability or systematic bias in the fundamental frequency of syllables ( Kao et al . , 2005; Olveczky et al . , 2005; Scharff and Nottebohm , 1991 ) . Thus , a putative role for the cerebellum in the control of fundamental frequency could be expected . No change in fundamental frequency could be detected here either in adults or in juveniles following DCN lesions . Given the relatively small extent of the lesions performed and that other circuits in the song system may compensate for the effect of DCN lesions , we cannot exclude a cerebellar contribution to fundamental frequency . We have revealed a subcortical connection between the cerebellum and the cortico-basal ganglia circuit involved in song learning and plasticity , indirectly affecting activity in the premotor song-related nucleus RA . A more direct connection may also exist from the cerebellum to the motor pathway from HVC to RA that could exert a direct influence on song production . The dorsal thalamus , which mediates cerebellar input to the basal ganglia that we have evidenced here , is also known to project to the pallial nucleus MMAN , which in turn projects to HVC ( Foster et al . , 1997; Nicholson and Sober , 2015; Williams et al . , 2012 ) . This new pathway remains to be characterized by anatomical and electrophysiological experiments to assess the impact of cerebellar input on the cortical pathway during song learning and production . In mammals , the cerebellum is known to project to the motor part of the thalamus , which in turn projects to the motor cortex ( Kelly and Strick , 2003 ) . This disynaptic connection between the cerebellum and the motor cortex is important in motor control and motor coordination ( Brooks , 1984 ) and we therefore hypothesize a contribution of the DCN-DTZ-MMAN-HVC pathway in the production of song in songbirds . We have shown that a cerebello-thalamo-basal ganglia pathway exists in songbirds , is functional and shares many similarities with the mammalian cerebello-thalamo-basal ganglia pathway . Knowing the role of the cerebellum and the basal ganglia , respectively in supervised and reinforcement learning ( Doya , 2000 ) , we hypothesize that the cerebellum can participate in basal ganglia functions by sending an error-correction signal related to a detected mismatch between actual and predicted sensory feedbacks . This error correction signal is integrated into the basal ganglia to drive the motor command output during the learning process . As recently reported , the song-related basal ganglia nucleus receives a reward prediction error from the ventral tegmental area that is necessary and sufficient to drive song learning ( Gadagkar et al . , 2016; Hoffmann et al . , 2016; Xiao et al . , 2018 ) . The reward prediction error signal from the VTA and the cerebellar error correction signal could cooperate within the basal ganglia to achieve faster and more efficient sensorimotor learning . In this context , the cerebellar input could modulate plasticity of the avian equivalent of the cortico-striatal connections , as described in mice ( Chen et al . , 2014 ) , and thereby regulate the learning rate in the basal ganglia circuits . In songbirds , the basal ganglia-thalamo-cortical loop is necessary for song learning and plasticity ( Brainard and Doupe , 2002; Olveczky et al . , 2005 ) . Our data suggest that these functions - presently attributed to the basal ganglia-thalamo-cortical loop only - may also be influenced by the cerebellum through its subcortical connection to the song-related basal ganglia nucleus . Finally , the subcortical pathway from the cerebellum to the basal ganglia is involved in dystonia ( Calderon et al . , 2011; Fremont et al . , 2017; Neychev et al . , 2008; Tewari et al . , 2017 ) . The existence of the cerebello-thalamo-basal ganglia pathway makes the songbird model , classically used as a model to study vocal learning , a good model for further investigations of the cooperation between cerebellum and basal ganglia in sensorimotor learning and its dysfunction in movement disorders .
All the experiments were performed in adult male zebra finches ( Taeniopygia guttata ) , >90 days post-hatch unless otherwise specified . Birds were either reared in our breeding facility or provided by a local supplier ( Oisellerie du Temple , L’Isle d’Abeau , France ) . All animals had constant access to seeds , crushed oyster shells and water . Seeds supplemented with fresh food and water were provided daily . Birds were housed on a natural photoperiod ( both in the aviary and in sound isolation boxes during the behavioral experiment ) . Animal care and experiments were carried out in accordance with the European directives ( 2010–63-UE ) and the French guidelines ( project 02260 . 01 , Ministère de l’Agriculture et de la Forêt ) . Experiments were approved by Paris Descartes University ethics committee ( Permit Number: 13–092 ) . Before surgery , birds were first food-deprived for 20–30 min , and an analgesic was administered just before starting the surgery ( meloxicam , 5 mg/kg ) . The anesthesia was then induced with a mixture of oxygen and 3–5% isoflurane for 5 min . Birds were then moved to the stereotaxic apparatus and maintained under anesthesia with 1% isoflurane . Xylocaine ( 31 . 33 mg/mL ) was applied under the skin before opening the scalp . Small craniotomies were made above the midline reference point , the bifurcation of the midsagittal sinus , and above the structures of interest . Stereotaxic zero in anteroposterior and mediolateral axis was determined by the sinus junction . To ease the access to the cerebellum , we used a head angle of 50° . The stereotaxic coordinates used for each brain structure are summed up in Table 1 . We performed iontophoretic injections of fluorescent dye using dextran conjugates with Alexa 594 and Alexa 488 ( Thermofischer , 5% in PBS 0 . 1M 0 . 9% saline ) in targeted cerebral structures ( lateral DCN and Area X nucleus ) using a glass pipette with a small ( 10 µm ) tip and ±5 μA DC pulses of 10 s duration , 50% duty cycle , applied for 3 min . In the cerebellum , to be sure that the injection was constrained to the lateral deep cerebellar nucleus , we verified that the retrograde labeling of Purkinje cells was limited the most lateral sagittal zone ( Figure 1D , and Figure 1—figure supplement 1 ) . In additional tracing experiments , 250 nL of cholera toxin tracers coupled with Alexa 488 ( Thermofischer , diluted in PBS 0 . 1M 0 . 9% saline ) were pressure-injected with a Hamilton syringe ( 1 µL , Phymep , Paris , France ) , at 100 nL per minute , at each injection site ( two injection sites per brain hemisphere ) . Birds were then housed individually for three days after injection to allow for dye transport . Recordings in Area X , LMAN , and RA were made with a tungsten electrode with epoxy isolation ( FHC , impedance varying from 3 . 0 to 8 . 0 MΩ depending on the type of neuron recorded ) . Acquisition of the signal was done with the AlphaOmega software , using low-pass ( frequencies below 8036 Hz ) and high-pass ( frequencies above 268 Hz ) filters to only detect the spike signal . The sampling frequency was 22 , 320 Hz . In Area X , the recorded neurons displayed a bimodal distribution of spontaneous firing rate , above 25 Hz or under 10 Hz . We considered neurons with frequency above 25 Hz as pallidal neurons in Area X ( Leblois et al . , 2009; Person and Perkel , 2007 ) . Other neurons in Area X with spontaneous firing rates under 10 Hz were not taken into account in the present study . Distribution of the pallidal-like neurons firing rate is represented in Figure 9C . Note that the level of spontaneous activity is different under anesthesia compared to what was seen in awake birds ( Goldberg and Fee , 2010 ) and can vary depending on the specific drug used ( Brooks , 1984 ) . This may explain the slight difference in spontaneous activity among neurons recorded here as pallidal , compared to previous studies performed under urethane anesthesia ( Leblois et al . , 2009; Person and Perkel , 2007 ) , known to preserve awake-like cortical activity ( Albrecht et al . , 1990 ) . A single-pulse electrical stimulation in the lateral deep cerebellar nucleus ( DCN ) was applied through a bipolar electrode during recording of different structures in the contralateral basal ganglia nucleus ( Area X ) , the lateral part of the magnocellular nucleus ( LMAN ) , the medial part of the magnocellular nucleus ( MMAN ) , and robust archopallium nucleus ( RA ) . The duration of the stimulation was 1 ms , with an inter-stimulation time of 1 . 6 s , and the intensity ranged from 0 . 1 to 4 mA . Despite long stimulation duration , observed responses in recorded neurons were stable over time . We aimed to place the stimulation electrode within the lateral cerebellar nucleus , and the positioning of the electrode was confirmed histologically ( see next paragraph ) . However , we cannot completely rule out that the stimulation current did spread to the nearby interpositus nucleus . Other possible confounds due to non-specific effects of stimulation could be that brainstem structures that communicate with the forebrain song system , and fibers of passage that descend from RA to the brainstem could be activated . However , such non-specific effects are highly unlikely due to the distance between the DCN and the song-related brainstem structure ( >1 mm ) , and their separation by the fourth ventricule . Most importantly , a small offset in the placement of the stimulating electrode most often led to the total disappearance of the responses evoked in the basal ganglia circuit , and it is thus unlikely that neurons or fibers away from the stimulation site are mediating the observed responses . During electrophysiological experiments , drugs were applied locally by pressure with small tip glass pipette ( 10 µm ) and nitrogen picospritzer ( Phymep , Paris , France ) during 5 ms . The volumes injected are around 100–200 nL , with a maximal total injected volume during one experiment of 500 nL . We used a mix of NBQX 5 mM ( Sigma Aldrich , diluted in PBS 0 . 1M 0 . 9% saline ) and APV 1 mM ( Sigma Aldrich , diluted in PBS 0 . 1M 0 . 9% saline ) to block glutamate receptors . Except for Area X blocking , that requires several coordinates injection in order to block a large part of this structure , all blockade are made in one location with two puff injections . To determine the drug dispersion , we injected NBQX/APV at several distance from the recorded neuron in Area X . We then compared neurons responses strength ( see Data analysis for the quantification protocol ) with and without drug injection to assess the percentage of resting response ( Figure 9C ) . Drug dispersion experiments indicate that excitatory responses were not impacted if the distance between the recorded neuron and the drug injection was more than 200 µm ( n = 3 neurons for 200 µm , mean resting response: 94 , 3 ± 9 , 6% ) . For distances between 150 and 50 μm , we observe a progressive decrease in excitatory responses , with a halving of excitatory responses for distances around 75 µm ( n = 3 neurons for 150 µm , mean resting response: 83 , 3 ± 15 , 5%; n = 2 neurons for 100 µm , mean resting response: 76 , 1 ± 11 , 5%; n = 2 neurons for 75 µm , mean resting response: 37 , 1% ± 33 , 2; n = 3 neurons for 50 µm , mean resting response: 51 ± 29 , 6% ) . Then , excitatory responses in pallidal neurons were totally prevented if the distance between the recorded neuron and the drug injection was less than 50 µm ( n = 0 . 5 neurons , mean resting responses: 0% ) . Moreover , glutamatergic blockade effect on the recorded neuron firing rate was quantified ( Figure 10 , see Data analysis for the quantification protocol ) during baseline , drug injection and washout conditions . No significant effect of the drug injection on the firing rate of recorded neurons was observed ( Figure 10A-E , paired Wilcoxon test , see Legend for p values ) , except for recordings in RA during LMAN glutamatergic blockade ( Figure 10F , paired Wilcoxon test , p=0 . 0313 ) . During LMAN recordings , we also blocked inhibition transmission in Area X . To do so , we used gabazine 1 mM ( Sigma , diluted in PBS 0 . 1M 0 . 9% saline ) . Analyses of recorded neurons after DCN stimulation were done using Spike2 and Matlab . Spike sorting was performed with the software Spike2 ( CED , UK ) , using principal components analysis of spike waveforms . For Area X neurons , and RA neurons , we managed to record single units , and we focus on these single unit neurons in the analysis . In the LMAN and MMAN nuclei , we chose to record mostly multiunit activity . Indeed , most neurons in these nuclei exhibit very low spontaneous activity ( ~1 sp/s ) , leading to wide fluctuation in the PSTH estimate of baseline activity preceding stimulation with high temporal resolution ( time bin: 10 ms ) and making it difficult to estimate response latency , strength and duration . Instead multi-unit activity with higher baseline levels allows better baseline statistics and narrower confidence intervals for the detection of the response to stimulation . Spike train analysis was then performed using Matlab ( MathWorks , Natick , MA , USA ) . We calculated peri-stimulus time histograms ( PSTH ) of recorded neurons after DCN stimulation . PSTHs were calculated with a 2 ms bin for neurons in Area X and RA . For structures with low firing rate ( LMAN and MMAN ) the time bin was 10 ms to limit bin-to-bin fluctuations in spike count . We calculated the mean and the standard deviation ( SD ) of the firing rate over the period preceding the stimulation ( 50 ms for Area X and RA , 100 ms for LMAN and MMAN ) , and we considered that a neuron exhibited a significant response to the stimulation when at least two consecutive bins of the PSTH were above ( for excitation ) or below ( for inhibition ) the spontaneous mean firing rate ±2 . 5*SD . The return of two consecutive bins at the spontaneous mean firing rate ±2 . 5*SD indicated the end of the response . We defined the latency of response as the time between the stimulation onset and the beginning of the first excitatory or inhibitory response . Response strength was calculated as the sum of the difference between the PSTH values and the mean baseline firing rate over the entire response period and represents the average number of excess ( default ) spikes induced by a single stimulation . For neurons in Area X and RA , the response strength was calculated over the first peak of excitation only ( as most responses did not elicit two peaks of excitation , see Results ) . For LMAN and MMAN neurons recording , neurons tended to display bimodal responses ( see Results ) and both the first and second excitation peaks were considered to calculate the response strength . We also report the peak firing rate in the response period as the maximal value of the PSTH . The PSTHs are displayed either as histograms or as solid curves with gray shaded area surrounding the curve representing the SD of the baseline firing rate . Lesions were performed in the DCN of juvenile zebra finches . We targeted the most lateral DCN , analogous to the dentate nucleus in mammals . In a first group of birds ( n = 7 ) , a partial electrolytic lesion was performed in the lateral deep cerebellar nucleus by passing 0 . 05mA during 30 s through a tungsten electrode . Lesions were made at three points ( see the stereotaxic coordinates in Table 1 , DCN coordinates , second row ) . In a second experimental group ( n = 3 ) , chemical partial lesion was performed using ibotenic acid in 1 µL Hamilton syringe , with a rate of 100 nL/min . We also performed injections at three locations ( see Table 1 , DCN coordinates ) injecting 150 nL per point . Sham lesions were performed in another group of age-matched juvenile birds . Sham birds underwent the same surgery as the lesion group , with a stimulating electrode placed at the lesion location but no current was applied . Both lesion and sham protocols were done around 57 days post hatch ( 56 , 8 ± 7 , 5 days post hatch for lesion group , 57 . 0 ± 4 , 5 days post hatch for sham group ) . Following surgery , the behavior of birds was closely monitored for a few days to ensure proper recovery . Many birds underwent temporary motor deficits ( postural and balance troubles ) for a couple of days but recovered very quickly and were all perching and feeding normally 48 hr after surgery . Singing usually resumed after 48 hr , or at most after 72 hr . Each juvenile ( sham and lesion ) was put in a recording box one week before the lesion experiment , and recorded using Sound Analysis Pro software ( SAP , Tchernichovski et al . , 2001 ) . To prevent any deficit due to the lack of tutor , we presented the tutor to the juvenile two hours per day until the bird underwent the surgery . All birds had same access to their respective tutors . After the surgery , each juvenile was recorded until the crystallization phase ( 30 days after the surgery experiment ) . For the anatomical tracing protocol: Birds were sacrificed with a lethal intraperitoneal injection of pentobarbital ( Nembutal , 54 . 7 mg/mL ) , perfused intracardially with PBS 0 . 01M followed by 4% paraformaldehyde as fixative . The brain was removed , post-fixed in 4% for 24 hr , and cryoprotected in 30% sucrose . We then cut 40 µm thick sections in the parasagittal plane with a freezing microtome . Slices were mounted with Mowiol ( Sigma Aldrich ) and observed under an epifluorescence ( Leica Microsystems , Leica DM 1000 , Nanterre , France ) or a confocal microscope ( Zeiss , LSM 710 , France ) . Images were analyzed using ImageJ software ( Rasband WS , NIH , Bethesda , Maryland , USA ) . After electrophysiological recordings , the bird was perfused as described above . Then , brain was removed , post-fixed one day in PFA 4% , store in sucrose 30% , and we did 60 µm slices with Nissl staining to control the stimulation electrode and recording electrode tracts . For the lesion protocol: All juvenile birds were sacrificed at 100 dph using the protocol previously described for tracing protocol . We then cut 60 µm-thick cerebellar sections in the horizontal plane with a freezing microtome . We did Nissl staining to check lesions locations . Slices were mounted with Mowiol ( Sigma Aldrich ) and observed under a transmitted-light microscope ( Leica Microsystems , Leica DM1000 , Nanterre , France ) . With ImageJ software ( Rasband WS , NIH , Bethesda , Maryland , USA ) , we calculated the area of lesion for each nucleus compared to the control nucleus in the other hemisphere . Songs were continuously recorded using Sound Analysis Pro software ( SAP , Tchernichovski et al . , 2001 ) . Songs were then sorted and analyzed using custom Matlab programs ( https://github . com/aleblois/Pidoux_et_al_2018 . git , Pidoux and Leblois , 2018; MathWorks , Natick , MA , USA; copy archived at https://github . com/elifesciences-publications/Pidoux_et_al_2018 ) . Briefly , the program detected putative motifs based on peaks in the cross-correlation between the sound envelope of the recorded sound file and a clean preselected motif . Putative motifs were then sorted based on their spectral similarity with the pre-selected clean motif , using thresholds set by the experimenter . Song bouts including one or more song motifs separated by less than 500 ms of silence were then cut based on the same sound amplitude threshold . This analysis allowed us to successfully sort >98% of the songs produced by a bird on a given day ( assessed by comparing hand sorting with the automated sorting by the program ) . We calculated the spectrogram of extracted song through fast Fourier transforms using 256-point Hanning windows moved in 128-point steps . Among all songs produced by a juvenile in each considered condition: before and after lesion , as well as at crystallization , 50 to 100 song clean song bouts with no noise contamination ( cage noise , wing flaps , … ) were randomly-selected songs to be compared to the tutor’s selected motifs using the procedure described in Mandelblat-Cerf and Fee ( 2014 ) . The corresponding Matlab program provides 3 outputs: an acoustic similarity index and a sequencing similarity index , which are compiled together into a single imitation score . We only reported the final imitation score in the present study as the relatively mild effect of DCN lesion did not allow to distinguish acoustic and sequencing effects . A custom-based analysis relying on the cross-correlation between spectrograms was also applied ( see Figure 8—figure supplement 2 and its legend for method ) to confirm the default in imitation revealed by the imitation score . For each bird undergoing DCN lesion , or sham-lesion experiments , spectrograms of 500 randomly-selected , manually-checked renditions of the stereotyped motif were stored . To determine the acute effect of the lesion , we analyzed several song features in the first week after the lesion , grouped values for two consecutive days and named these periods pre , post days 1–2 , post days 3–4 , post days 5–6 . Moreover , the same analysis was performed at days 90–91 ( after crystallization ) , to determine the learning trajectory of each song feature . For each considered day , roughly 500 motifs were used to calculate the duration , fundamental frequency and amplitude of each syllable using the following procedure . The sound envelop was generated , and a threshold was determined , corresponding to the lowest envelop signal value ( i . e . the smallest amplitude in the motif ) . The beginning and the end of each syllable was determined as the time at which the song envelop crossed this threshold . This process was performed for each syllable type in the motif ( generally 4 to 6 syllables per motif ) , on our spectrograms of 500 randomly-selected motifs . To pool the data from all syllable types , we normalized syllable duration by doing the absolute ratio between the syllable duration in the post periods ( post days 1–2 , post days 3–4 , post days 5–6 , crystallization ) over the duration syllable calculated in the period before lesion . This calculation reveals the relative duration changes compared to the pre-lesion period , i . e . how the duration evolved over the time . The variability of syllable duration between syllable types from a given bird and between birds were not significantly different ( Kruskal and Wallis test , n = 7 birds/21 syllables for lesion group , p=0 . 69 between birds and p=0 . 56 between syllable , and n = 10 birds/27 syllables , p=0 . 75 between birds and p=0 . 71 between syllables ) , allowing to compare all syllable types from a given group ( sham vs lesion ) in each condition . Relative syllable amplitude was determined as the peak sound envelop during the syllable divided by the peak sound envelop over the whole motif . The syllable fundamental frequency was determined for each syllable type displaying a clear harmonic structure based on peaks in the autocorrelation function , as in the study by Kao and Brainard ( 2006 ) . For some syllables , several sub-syllabic elements had a clear and distinct fundamental frequency , leading to several fundamental frequency measurements in the same syllable . Normalizations , identical to the one described for syllable duration , were applied for the amplitude and fundamental frequency of all syllable types . Finally , the learning trajectory was calculated for each group and each feature using the relative change at crystallization minus the relative change values for post days 5–6 . Numerical values are given as mean ± SD , unless stated otherwise . | Human infants learn to speak by imitating the speech of adults around them . Over time , they learn to coordinate movements of their vocal cords and breathing muscles to produce specific sounds . Juvenile songbirds go through a similar process while learning to sing . Fledglings mimic adult birds and each other as they learn to produce their own songs . Songbirds are therefore often used as a model for how the brain drives vocal learning – whether of speech or song . Circuits made up of similar brain regions support vocal learning in infants and in songbirds . These regions include areas of cortex , the outermost layer of the mammalian brain , as well as structures deep below the cortex . The latter include the basal ganglia , a set of structures that help mammals learn and perform fine motor skills . But there is one brain region that has been implicated in vocal learning in infants but not in songbirds . Known as the cerebellum or ‘little brain’ , this structure also helps with planning and performing movements . Anatomical studies in songbirds suggest a connection between the cerebellum and song-related circuits . But a direct role in birdsong has never been shown . Pidoux et al . now demonstrate that stimulating the cerebellum in anaesthetized zebra finches activates basal ganglia neurons involved in song learning . This activation spreads through a song-related circuit to neurons controlling the vocal cords . Disrupting the cerebellum , by contrast , makes it harder for juvenile birds to imitate adult song . This is the first direct evidence for a role of the cerebellum in the acquisition of birdsong . Beyond vocal learning , the results shed light on the circuits that support motor learning more generally . They also suggest that we can use songbirds to study the cerebellum and its interactions with the basal ganglia . Abnormal interactions between these regions occur in movement disorders such as Parkinson's disease . Studying these interactions in the healthy mammalian brain should provide clues to the pathology behind these conditions . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"neuroscience"
] | 2018 | A subcortical circuit linking the cerebellum to the basal ganglia engaged in vocal learning |
Subgenomic flaviviral RNA ( sfRNA ) accumulates during infection due to incomplete degradation of viral genomes and interacts with cellular proteins to promote infection . Here we identify host proteins that bind the Zika virus ( ZIKV ) sfRNA . We identified fragile X mental retardation protein ( FMRP ) as a ZIKV sfRNA-binding protein and confirmed this interaction in cultured cells and mouse testes . Depletion of FMRP elevated viral translation and enhanced ZIKV infection , indicating that FMRP is a ZIKV restriction factor . We further observed that an attenuated ZIKV strain compromised for sfRNA production was disproportionately stimulated by FMRP knockdown , suggesting that ZIKV sfRNA antagonizes FMRP activity . Importantly , ZIKV infection and expression of ZIKV sfRNA upregulated endogenous FMRP target genes in cell culture and ZIKV-infected mice . Together , our observations identify FMRP as a ZIKV restriction factor whose activity is antagonized by the sfRNA . Interaction between ZIKV and FMRP has significant implications for the pathogenesis of ZIKV infections .
Zika virus ( ZIKV ) is a mosquito-borne flavivirus that is closely related to dengue ( DENV ) and yellow fever viruses ( YFV ) . There are three ZIKV genotypes: one from Asia , which has caused the recent pandemic in the Americas , and two from Africa ( East and West African ) ( Lanciotti et al . , 2016 ) . ZIKV infection is characterized by fever , arthralgia and conjunctivitis ( Goeijenbier et al . , 2016 ) and was considered an exceedingly rare human infection since only 14 cases were reported prior to 2007 ( Faye et al . , 2014 ) . In 2007 , the first ZIKV outbreak caused by an Asian lineage was reported in Micronesia , followed by epidemics in French Polynesia in 2013 and the recent pandemic ( 2015–2016 ) in the Americas ( Weaver et al . , 2016; Aliota et al . , 2017 ) . During this period , it has been estimated that 1 . 5 million cases occurred in Brazil and more than 25 , 000 cases in Colombia ( Focosi et al . , 2016; Samarasekera and Triunfol , 2016 ) . As of January 2018 , the virus was widely distributed in 50 countries in the Americas ( PAHO/WHO , 2018 ) . The public health concern about Zika has been primarily driven by its maternal-fetal ( Brasil et al . , 2016a; Yockey et al . , 2016; Driggers et al . , 2016 ) and sexual modes of transmission ( Deckard et al . , 2016; Davidson et al . , 2016; Hills et al . , 2016 ) as well as its association with congenital abnormalities , especially microcephaly , and Guillain-Barré syndrome in adults ( Krauer et al . , 2017; Martines et al . , 2016; Miner et al . , 2016; Brasil et al . , 2016b ) . These unique aspects set ZIKV apart from other flavivirus infections and have spurred efforts to understand ZIKV host-pathogen interactions . The ZIKV life-cycle starts with virus attachment and receptor-mediated endocytosis ( Hamel et al . , 2015; Meertens et al . , 2017 ) . Upon endosome acidification , the viral envelope fuses with the endosomal membrane , releasing the nucleocapsid into the cytoplasm for uncoating and initial viral translation at the cytosolic surface of the endoplasmic reticulum ( ER ) . Translation produces a single polyprotein that is cleaved co- and post-translationally by cellular and viral proteases into 10 mature proteins: three structural proteins forming the virion ( capsid , C; pre-membrane , prM; and envelope , E ) and seven non-structural proteins ( NS1 , NS2A , NS2B , NS3 , NS4A , NS4B and NS5 ) required for viral replication and inhibition of host defense mechanisms ( Campos et al . , 2017; Barrows et al . , 2018 ) . Replication of ZIKV RNA and assembly of viral particles occur in close association with rearranged ER membranes ( Rossignol et al . , 2017 ) . Assembled virions are transported through the secretory pathway , where the furin protease cleaves prM into and M , resulting in mature virions that are secreted into the extracellular space ( Hasan et al . , 2018 ) . The flaviviral genome contains 5′ and 3′ untranslated regions ( UTR ) that are essential for genome cyclization and initiation of RNA synthesis ( Filomatori et al . , 2006; Alvarez et al . , 2005 ) . The ZIKV 3′ UTR is highly structured and consists of four domains: xrRNA1 , xrRNA2 , the dumbbell ( DB ) and the 3′ SL ( Zhu et al . , 2016 ) . The ZIKV 3′ UTR contains confirmed and predicted pseudoknot interactions located in xrRNA1 and xrRNA2 , respectively . These pseudoknots are important for stalling of the cellular 5′ to 3′ exonuclease , XRN1 , and accumulation of at least two sfRNA species ( sfRNA1 and sfRNA2 ) ( Akiyama et al . , 2016 ) . The sfRNA of a few different flaviviruses has been described to exert pro-viral functions , possibly through acting as a ‘sponge’ for antiviral host proteins ( Göertz et al . , 2018 ) . The DENV sfRNA has been shown to interact with stress granule associated proteins G3BP1 , G3BP2 and CAPRIN1 and the ubiquitin ligase TRIM25 to dampen innate immune responses ( Bidet et al . , 2014; Manokaran et al . , 2015 ) . Similarly , Moon et al reported that DENV-2 and KUNV sfRNAs inhibited XRN1 activity , leading to accumulation of uncapped transcripts in infected cells and disruption of mRNA homeostasis that could potentially deregulate antiviral responses ( Moon et al . , 2012 ) . Regarding ZIKV , Musashi-1 ( MSI1 ) was reported to be required for efficient infection and interact with viral genomes; additionally , ZIKV infection disrupted the activity of MSI1 ( 33 ) . However , whether or not ZIKV sfRNA interacts with MSI1 and modulates its activity is unclear . In this work , we identified the Fragile X Mental Retardation Protein ( FMRP ) as a host factor that binds to the ZIKV sfRNA and to viral genomes . Functional assays indicated that FMRP represses ZIKV infection by inhibiting viral translation . Analysis of an attenuated ZIKV vaccine candidate ( Δ10 ZIKV ) that is defective for sfRNA production suggests that sfRNA enhances ZIKV infection partly through antagonizing FMRP activity . Additionally , we observed that ZIKV infection blocks the canonical activity of FMRP and increases the expression of FMRP target genes . Finally , we present evidence implicating deregulation of FMRP activity in a mouse model of ZIKV infection , suggesting that ZIKV pathogenesis may involve virus-mediated FMRP inhibition .
Before this work , there was no knowledge of host proteins that bind and are modulated by ZIKV sfRNA . We previously used RNA affinity chromatography coupled with quantitative mass spectrometry ( RAC-MS ) to identify proteins that interact with DENV-2 sfRNA ( Bidet et al . , 2014; Manokaran et al . , 2015 ) . Here , we used a similar approach with lysates from JEG3 ( choriocarcinoma ) cells , which are permissive for ZIKV infection , to identify host proteins that interact with full-length ZIKV 3′ UTRs of the pandemic Asian lineage strain PRVABC59 and the African strain MR766 . This analysis revealed 27 proteins that preferentially interacted with the ZIKV RNAs compared to an RNA derived from the coding sequence of DENV ( Table 1-source data 1 ) . We also analyzed a second negative control RNA derived from vector backbone sequence and obtained similar results ( data not shown ) . We did not detect major differences in the protein-binding profiles of PRVABC59 and MR766 ZIKV sequences . Eight proteins that were identified by ≥2 unique peptides and were ≥2 times as abundant in both the ZIKV sequences relative to the control RNA were considered as high-confidence ZIKV 3′ UTR binding proteins ( Table 1 ) . The proteins showing highest preferential binding to ZIKV sfRNAs were FMRP and its two paralogs , Fragile X Related proteins 1 and 2 ( FXR1 and FXR2 ) for both ZIKV 3′ UTRs . In order to map the region ( s ) of the ZIKV 3′ UTR that mediate binding of FXR proteins , we performed an RNA affinity chromatography experiment using HeLa lysate and fragments of the ZIKV- PRVABC59 3′ UTR , and probed for proteins by western blotting ( WB ) . As expected , the complete 3′ UTR interacted with FMRP , FXR1 and FXR2 ( Figure 1 , lane 1 ) . Deletion of xrRNA1 dramatically reduced interaction with all three proteins ( lane 2 ) and deletion of xrRNA1 and 2 eliminated detectable binding . Additionally , xrRNA1 alone was sufficient to bind FMRP , FXR1 and FXR2 ( lane 4 ) . These data indicate that xrRNA1 is necessary and sufficient for efficient binding to these proteins . We also probed for DDX6 , G3BP1 and PTB . DDX6 interacted specifically with the ZIKV dumbbell region in contrast to FMRP . Interestingly , PTB and G3BP1 , which were previously reported to interact with the DENV 3′ UTR ( Bidet et al . , 2014; De Nova-Ocampo et al . , 2002 ) , did not strongly interact with ZIKV 3′ UTR ( Figure 1 ) . We chose to focus on FMRP due to its strong association with ZIKV RNA in binding experiments ( Figure 1B ) , well-established association with human neurodevelopmental disease , and high level of expression in tissues relevant to ZIKV , such as brain and reproductive tract ( Devys et al . , 1993 ) . To validate the interaction of FMRP with ZIKV 3′ UTR in the context of ZIKV infection , we performed RNA immunoprecipitation ( RIP ) with anti-FMRP antibody or isotype control and detected ZIKV RNAs using northern blot ( NB ) ( Figure 2A ) . HeLa cells were infected with Dakar 41525 ( ZIKV-Dakar ) , Cambodia FSS13025 ( ZIKV-Cambodia ) or PRVABC59 ( ZIKV-Puerto Rico ) strains , cell lysates were collected and FMRP was immunoprecipitated . Total RNA interacting with FMRP was isolated and the presence of ZIKV RNAs was detected by NB using a DNA probe that hybridizes to the ZIKV 3′ UTR . Both ZIKV sfRNA and viral genome co-precipitated with anti-FMRP but not control IgG ( Figure 2A ) for all three virus strains . Interestingly , quantification of sfRNA and genomic RNA signals indicated that FMRP preferentially interacts with the ZIKV sfRNA ( Figure 2B ) . We additionally performed RIP with anti-PTB and isotype control antibodies and showed that viral genomes and sfRNA minimally co-precipitated with PTB and enrichment was much weaker than that for FMRP RIP ( Figure 2C ) . We also analyzed the integrity of unbound viral RNAs present in supernatants after RIP . This analysis revealed that viral RNAs did not degrade during the RIP procedure ( Figure 2C ) . We determined whether or not the interaction of FMRP with ZIKV RNA occurs in testes of infected mice . We selected testes for analysis because this tissue exhibits relatively high viral load in the A129 mouse model ( IFNAR1 knockout ) of ZIKV infection ( Dowall et al . , 2016; Rossi et al . , 2016 ) . FMRP was subjected to IP from testes lysates of mice infected with ZIKV-Cambodia for 6 days and RT-qPCR was used to measure co-precipitating viral RNA using qPCR assays targeting the viral genome exclusively or both the genome and sfRNA ( Figure 2D ) . We observed enrichment of viral genomes by ~2 fold for FMRP IP compared to the negative control . Importantly , the enrichment was nearly twice as robust when using a qPCR assay that detects both genomes and the sfRNA . These results validate the interaction of FMRP with ZIKV RNA in vivo and , as observed in infected cultured cells , indicate preferential binding of FMRP to the ZIKV sfRNA . Given the binding of FMRP to ZIKV sfRNA , we hypothesized that this protein would play an antiviral role for ZIKV . To analyze the functional relevance of FMRP in ZIKV infection , we infected HeLa cells in which FMRP was knocked down ( KD ) with individual siRNAs or a pool of four different siRNAs ( Figure 3A ) . Forty-eight hours post-infection with ZIKV-Dakar at low MOI ( 0 . 01 ) , we measured viral titers using plaque assay and rate of infection by flow cytometry . Both viral titers and rate of infection were increased by 2- to 5-fold due to FMRP KD ( Figure 3B and C , Figure 3—figure supplement 1A ) . The pooled siRNAs effectively reduced FMRP levels and this correlated with large increases in ZIKV infection . The antiviral effect of FMRP was corroborated by immunofluorescence ( IF ) of infected cells to detect viral antigen and analysis by high-content imaging ( Figure 3D ) . Moreover , we observed that FMRP KD enhanced infection with Asian strains ( ZIKV-Cambodia and ZIKV-Puerto Rico ) to different extents , but did not impact infection by DENV-2 ( Figure 3—figure supplement 1B and C ) . We additionally tested the effect of FXR1 and FXR2 knockdown on ZIKV infection . We observed a slight increase in virus infection rate for FXR1 depletion while FXR2 knockdown significantly reduced infection ( Figure 3—figure supplement 2 ) . Together , these observations suggest that FMRP acts as a specific restriction factor for multiple ZIKV strains . We also tested the effect of FMRP KD under experimental conditions that interrogate virus infection at higher MOI and shorter infection time , and observed that FMRP KD increased infection rate from ~50% to~80% ( Figure 3—figure supplement 3A and B ) . Analysis of fluorescence intensity ( FI ) histograms ( Figure 3—figure supplement 3C ) revealed the presence of two distinct populations of infected cells: one of low FI and a second of high FI for expression of viral envelope ( E ) protein . Cells transfected with non-targeting siRNA ( blue histogram ) were predominantly in the low-infected cell population . In contrast , cells transfected with siRNAs targeting FMRP ( red histogram ) were mostly in the highly-infected population . We calculated FI mode values , indicating the maximum FI peak for each condition and found that FMRP knockdown led to a ~ 10 fold increase , indicating elevated levels of E protein expression ( Figure 3—figure supplement 3D ) . WB analysis revealed increased NS4B and NS2B protein levels in FMRP KD cells infected with ZIKV at MOIs of 1 and 5 ( Figure 3—figure supplement 3E and F ) . Together , these data indicate that during a single ZIKV replication cycle , the absence of FMRP significantly increases both the rate of infection and the level of viral protein accumulation per cell . Because FMRP is known to be a repressor of cellular mRNA translation ( Zalfa et al . , 2003; Laggerbauer et al . , 2001 ) , we hypothesized that translation of ZIKV is inhibited by FMRP early after infection . To test this hypothesis , we used an infectious ZIKV that expresses NanoLuc luciferase and analyzed early accumulation of NanoLuc in control and FMRP KD cells . Treatment with the translation elongation inhibitor cycloheximide ( CHX ) was used to control for background signal present in the virus stock . The Nanoluc signal observed in control cells was well above that detected in the CHX-treated control at 3 . 5 hr post infection ( hpi ) indicating measurable viral protein synthesis at this early time point . Importantly , we observed a 4-fold increase in NanoLuc signal in FMRP depleted cells ( siFMR1 ) compared to control siRNA transfected cells , suggesting that FMRP depletion results in increased of ZIKV translation ( Figure 4A and B ) . Similar effects were observed in the presence of NITD008 , a potent inhibitor of the flaviviral NS5 RNA-dependent RNA polymerase ( Deng et al . , 2016; Yin et al . , 2009 ) , ruling out a possible contribution of RNA replication to the increased luciferase signal in FMRP KD cells ( Figure 4A and B ) . Consistent with this , we did not observe significant differences in viral RNA accumulation between control and FMRP KD cells in the absence or presence of NITD008 ( Figure 4C ) . Finally , we calculated a 3 . 5-fold increase of ZIKV translation efficiency ( NanoLuc levels normalized to ZIKV RNA ) in FMRP KD cells compared to the negative control in both the absence and presence of NITD008 ( Figure 4D ) . Together , these results strongly suggest that FMRP inhibits ZIKV infection by reducing viral translation . Because FMRP binds preferentially to ZIKV sfRNA , we tested the hypothesis that sfRNA , which accumulates to high levels during the course of infection ( Figure 2A ) , is capable of attenuating FMRP-mediated anti-ZIKV activity . For this purpose , we analyzed a ZIKV mutant ( Δ10 ZIKV ) that contains a 10-nt deletion within the 3′ UTR DB structure and is highly attenuated in mice and non-human primates ( Shan et al . , 2017a; Shan et al . , 2017b ) . We first asked whether Δ10 ZIKV is deficient in sfRNA production . Analysis of viral RNA from infected HeLa cells demonstrated that Δ10 ZIKV has a defect in sfRNA accumulation . An sfRNA/gRNA ratio of 2 . 7 was observed in WT ZIKV infected cells but the ratio was only 0 . 6 in Δ10 ZIKV infected cells . These results suggest that Δ10 ZIKV is attenuated , as least in part , through loss of sfRNA accumulation caused by the 10-nt deletion ( Figure 5A and B ) . Next , we asked whether the WT and mutant viruses exhibit differential sensitivity to FMRP depletion . We observed that Δ10 ZIKV infection rate in HeLa cells is reduced compared to WT ZIKV under negative control conditions ( ~11% vs~25% infected cells ) and , importantly , FMRP KD disproportionately enhanced Δ10 ZIKV infection rate ( Figure 5C ) : whereas WT virus was enhanced by approximately 3-fold , the infection rate for Δ10 ZIKV rose by nearly 5-fold ( Figure 5D ) . Representative scatter plots depicting the infected cell populations are shown in Figure 5—figure supplement 1 . Additionally , similar effects were observed on levels of cell-associated ZIKV NS2B and NS4B proteins ( Figure 5E to G ) . Together , these results suggest that sfRNA antagonizes FMRP and thus enhances ZIKV infection . Given the key role of FMRP in neurodevelopment , its mechanism of action has been extensively studied and several cellular mRNAs are known targets for translational repression by FMRP ( Zalfa et al . , 2003; Ascano et al . , 2012; Korb et al . , 2017 ) . We examined expression levels of ten reported targets in control and FMRP-depleted HeLa cells: ARC RHOA , SOD1 , RAC1 , PNPLA6 , KDM5C , TSC2 , FXR2 , TLN1 and BRD4 . Protein levels of FXR2 , TLN1 and BRD4 increased after transfection of siRNA targeting FMRP ( Figure 6—figure supplement 1 ) suggesting that the mRNAs encoding these proteins are genuine FMRP targets in HeLa cells . The remaining proteins were either undetectable ( ARC ) or were unaffected ( RAC1 , Rho A , PNPLA6 , KDM5C , TSC2 , SOD1 ) which suggests that FMRP may have cell-type specific effects ( Ascano et al . , 2012; Darnell and Klann , 2013 ) . To probe for antagonistic effects of ZIKV sfRNA on FMRP activity , the expression of FMRP targets was evaluated in the context of infection . First , we analyzed expression levels of FXR2 by WB in HeLa cells infected with WT ZIKV and Δ10 ZIKV . In order to achieve similar infection rates for the two viruses , cells were infected at MOI of 3 for WT ZIKV and MOI of 4 . 5 for the mutant virus . Interestingly , these experiments revealed a 2 . 7-fold-increase of FXR2 protein levels in cells infected with WT ZIKV ( Figure 6A and B ) . Although Δ10 ZIKV infection also increased the mean level of FXR2 , this effect was not statistically significant ( p=0 . 277 ) compared to uninfected cells . In parallel , we measured the expression of FXR2 by flow cytometry in ZIKV-infected cells using a double staining protocol . Figure 6C shows the distributions of the cells based on infection with either WT ZIKV or Δ10 ZIKV , indicating similar levels of infection for both viruses ( ~86% for WT and ~90% for Δ10 ZIKV ) . FXR2 mean fluorescence intensity ( MFI ) was enhanced by WT virus and , to a lesser extent , by the Δ10 ZIKV mutant . Quantitatively , there was a two-fold increase of FXR2 MFI in WT ZIKV infected cells and a 1 . 5-fold increase for Δ10 ZIKV compared to uninfected cells ( Figure 6D ) . We also conducted IF microscopy analysis and observed FXR2 accumulation specifically in ZIKV-infected cells ( Figure 6E ) . Interestingly , infection of cells by Δ10 ZIKV caused FXR2 to redistribute into discrete cytoplasmic puncta which was also evident at reduced frequency in WT ZIKV-infected cells . FXR2 has been previously reported to concentrate in stress granules ( Gonçalves et al . , 2011 ) , suggesting that infection with the attenuated virus may induce formation of stress granules more efficiently than WT ZIKV . In addition to FXR2 , we also observed elevated expression of TLN1 by WB in cells infected with WT ZIKV compared with uninfected and Δ10 ZIKV infected cells ( Figure 6—figure supplement 2A and B ) . We further assayed the expression of BRD4 in ZIKV-infected cells . Although the anti-BRD4 antibody used recognizes the largest known BRD4 isoform , we observed a second and faster migrating band of ~175 kDa that appeared only in ZIKV-infected cells ( Figure 6—figure supplement 2C ) . Since there are no annotated isoforms of this predicted mass , the data suggest that BRD4 is cleaved during ZIKV infection . We analyzed the sum of signals for both 204 kDa and 175 kDa BRD4 species and observed a 1 . 6-fold increase of BRD4 expression only in WT ZIKV infected cells ( Figure 6—figure supplement 2D ) . To further test the role of sfRNA we asked whether or not this noncoding RNA could modulate expression of FMRP targets in the absence of infection . Electroporation was used to transfect full-length , in vitro synthesized sfRNA or a deletion mutant lacking xrRNA1 and xrRNA2 ( DBSLIII ) . We electroporated RNAs with either 5′ triphosphate or monophosphate into HeLa cells and evaluated RNA integrity and abundance side-by-side with sfRNA from infected cells . We detected three distinct sfRNA species in infected cells that have been previously reported for ZIKV: sfRNA1 , sfRNA2 and sfRNA3 ( Figure 7A and B ) ( Filomatori et al . , 2017 ) . Electroporated RNAs were partially degraded although some intact sfRNA remained even at 48 hr post-electroporation . In parallel , we analyzed the expression of the FMRP targets FXR2 and BRD4 by flow cytometry ( Figure 7B to F ) . Compared to the DBSLIII RNA , electroporation of intact sfRNA resulted in modest , but highly significant , increases in both FXR2 and BRD4 levels . These results suggest expression of ZIKV sfRNA is sufficient to interfere with FMRP . We next asked whether ZIKV infection alters the expression of FMRP targets in mice ( Figure 8 ) . Male A129 mice were infected with WT ( n = 4 ) or Δ10 ZIKV ( n = 4 ) using 1 × 105 FFU . At 6 days post-infection , protein expression levels of FXR2 and PNPLA6 [a validated FMRP target ( Ascano et al . , 2012 ) ] were measured in testes . In WT ZIKV-infected mice , we observed statistically significant increases in expression of both FXR2 ( 1 . 6-fold ) and PNPLA6 ( 2 . 5-fold ) compared to uninfected mice ( Figure 8A , C and D ) . For Δ10 ZIKV-infected mice , expression levels of FXR2 and PNPLA6 were elevated to lesser extents , but these effects were not statistically significant . Measurements of viral genomes indicated a higher burden of infection for the WT virus than Δ10 ZIKV ( Figure 8B ) which may contribute to the differential effects on FXR2 and PNPLA6 levels . Nevertheless , these data taken together with observations made in cultured cells ( Figures 6 and 7 and Figure 6—figure supplement 2 ) , implicate the ZIKV sfRNA as an antagonist of FMRP function .
Here , we report functional interactions between ZIKV and FMRP , an important regulatory factor in neurodevelopment ( Harlow et al . , 2010; Hoeft et al . , 2010 ) . FMRP interacted with ZIKV RNA , with particular affinity for the sfRNA , in both infected cultured cells and mouse testes . FMRP interaction with the viral genomic RNA presumably limits infection through inhibiting early synthesis of viral proteins . Importantly , accumulation of ZIKV sfRNA suppressed the anti-viral activity of FMRP and , consequently , resulted in de-repression of endogenous FMRP target mRNAs . Together , our observations have important implications for ZIKV infection and pathogenesis . Multiple studies have been performed to characterize the RNA-binding specificity of FMRP ( Ascano et al . , 2012; Darnell et al . , 2001; Darnell et al . , 2011; Maurin et al . , 2018; Ray et al . , 2013 ) . Bioinformatic analysis of FMRP CLIP-seq datasets identified WGGA ( W = T/A ) as the top sequence motif ( Anderson et al . , 2016 ) which is found at six sites in the ZIKV 3′ UTR , each of them located in the dumbbell regions . Our mapping experiments , however , showed that the dumbbells are dispensable for FMRP binding to ZIKV 3′UTR , suggesting that these WGGA motifs do not mediate interaction with FMRP . Interestingly , using in vitro RNA selection Darnell et al . identified a highly structured target for the FMRP KH2 domain containing a pseudoknot ( Darnell et al . , 2005 ) . To date , no FMRP target mRNA has been identified that contains this motif but it is notable that the ZIKV xrRNA1 , which is required for FMRP binding in vitro , folds into a complex structure containing several tertiary interactions ( Akiyama et al . , 2016 ) . We speculate that the artificial RNA identified by Darnell and colleagues shares similar three-dimensional structural features with the sfRNA which allow it to bind FMRP with high affinity . FMRP inhibited infection of multiple ZIKV strains but did not restrict the related flavivirus , DENV , even though we previously identified FMRP as a DENV RNA-binding protein by RNA affinity chromatography ( Ward et al . , 2011 ) , suggesting that FMRP acts as an intrinsic restriction factor for ZIKV . It is not clear why DENV infection is immune to FMRP . It is possible that DENV eludes physical interaction with FMRP in infected cells or that ribosomes translating DENV RNA do not effectively recruit FMRP , which is thought to be a prerequisite for translational repression ( Chen et al . , 2014; Blackwell et al . , 2010 ) . Notably , FMRP was recently described as a proviral host factor for influenza A virus that promotes assembly of viral RNPs , and FXR proteins were shown to redundantly promote replication of specific alphaviruses ( Zhou et al . , 2014; Kim et al . , 2016 ) . Hirano et al . showed interaction of FMRP with the genomic RNA of Tick-Borne Encephalitis Virus ( TBEV ) , accumulation of FMRP at sites of local TBEV replication , and that FMRP depletion reduces TBEV infection ( Hirano et al . , 2017; Muto et al . , 2018 ) . Thus , FMRP can play a positive , negative or no role in infection , depending on viral species . Mechanistically , we determined that FMRP inhibits ZIKV translation . FMRP is widely expressed and has been characterized to repress translation of specific neuronal mRNAs but the precise mechanism is unknown . Chen et al . demonstrated that drosophila FMRP ( dFMRP ) can bind directly to the ribosome in the absence of mRNA ( Chen et al . , 2014 ) . The authors proposed that FMRP docks on the 80S ribosome using KH1/2 domains and simultaneously binds to mRNA via its RGG motif . The binding location of dFMRP on the ribosome suggests that it occludes recruitment of elongation factors and tRNA , leading the ribosome reversibly stall as observed by Darnell et al . ( Darnell et al . , 2011 ) . Based on this proposed mechanism , we speculate that FMRP binds the ZIKV genome within the viral open reading frame and causes elongating ribosomes to stall , leading to a deficiency in viral protein synthesis . Alternatively , it is possible that FMRP interaction with the 3′ UTR of the viral genome could lead to impaired translation . Finally , changes in the expression of FMRP targets may indirectly contribute to enhanced ZIKV infection upon FMRP knockdown . For example , the increase in FXR2 subsequent to FMRP depletion could positively impact ZIKV infection ( Figure 3—figure supplement 2 ) . We found that the recently developed ZIKV vaccine strain , Δ10 ZIKV , is compromised for sfRNA production and is replication-deficient in HeLa cells ( Shan et al . , 2017a; Shan et al . , 2017b ) . Interestingly , infection with Δ10 ZIKV , but not WT ZIKV , caused relocalization of FXR2 to cytoplasmic granular structures reminiscent of stress granules ( SG ) , suggesting that sfRNA may prevent SG formation . This is consistent with previous reports that show very low SG formation in flavivirus infected cells ( Bidet et al . , 2014; Emara and Brinton , 2007; Ruggieri et al . , 2012 ) . Importantly , Δ10 ZIKV infection was disproportionately increased by FMRP depletion compared to WT ZIKV , suggesting that one function of ZIKV sfRNA is to antagonize FMRP . In support of this , our RNA-IP experiments showed that FMRP preferentially co-IPs sfRNA compared to viral genomes . This may be partly explained by stoichiometry of sfRNA to genomes which reaches nearly 4 to 1 at the height of infection , depending on ZIKV strain . It is also possible that sfRNA structure differs from the corresponding region present in the viral genome and this dictates preferential FMRP interaction . We speculate that ZIKV sfRNA works as a ‘sink’ to saturate FMRP and prevent its repressive interactions with both viral RNA and cellular target mRNAs . This is reminiscent of recently described roles for the DENV sfRNA as a molecular sink for proteins involved in interferon responses ( G3BP1 , G3BP2 , CAPRIN1 , TRIM25 ) ( Bidet et al . , 2014; Manokaran et al . , 2015 ) . Besides interferon responses , there are intrinsic restriction factors that recognize viral components and block replication ( Yan and Chen , 2012 ) . For flaviviruses , a few intrinsic anti-viral factors have been described: YTHDF1-3 proteins that bind methylated RNA and regulate stability , were reported to interact with ZIKV RNA and inhibit infection ( Lichinchi et al . , 2016 ) ; YB-1 and QKI bind to the DENV 3′ UTR and repress DENV RNA translation ( Liao et al . , 2018; Paranjape and Harris , 2007 ) ; similarly , FBP1 blocks translation of JEV through interacting with untranslated regions of JEV RNA ( Chien et al . , 2011 ) . The association of these antiviral factors with the viral genome and their functional consequences have only been observed in cell culture and it remains to be seen whether or not these will be validated in vivo . Furthermore , there are no reports that the intrinsic anti-viral function of these proteins is counteracted by flaviviral factors . In this work we describe that FMRP is an intrinsic , direct-acting ZIKV restriction factor that interacts with ZIKV RNA and is counteracted by the sfRNA as it accumulates during infection . We provide evidence for de-repression of endogenous FMRP target mRNAs in the context of ZIKV infection of cultured cells and mouse testes . Specifically , we observed that the FMRP targets , FXR2 , TLN1 , BRD4 , and PNPLA6 were elevated at the protein level as a consequence of ZIKV infection . Moreover , infection with Δ10 ZIKV , which produces less sfRNA than WT virus , resulted in weaker or no effects on FMRP targets compared with WT ZIKV . We further observed that introduction of synthetic sfRNA into cells leads to upregulation of two FMRP targets: FXR2 and BRD4 . These results provide functional evidence indicating that the ZIKV sfRNA interferes with the activity of FMRP , although sfRNA-independent mechanism ( s ) for modulation of FMRP targets by ZIKV may also be at play . Our observations have implications for ZIKV pathogenesis in tissues with high expression of FMRP: brain , placenta and testes ( Hinds et al . , 1993 ) . Given the well-established role for FMRP in promoting neurodevelopment , it is tempting to speculate that certain aspects of ZIKV neuropathogenesis may be explained by sfRNA-mediated FMRP inhibition , leading to inappropriate expression of FMRP target mRNAs . Development of congenital Zika syndrome is likely multifactorial and symptoms such as microcephaly , which is only one disease manifestation ( Aliota et al . , 2017 ) , are unlikely to be the consequence of FMRP inhibition . Nevertheless , our findings warrant further research into how ZIKV interactions with FMRP might contribute to disease outcome .
JEG3 ( human choriocarcinoma ) , HeLa ( human cervix adenocarcinoma ) and Vero cells were maintained in DMEM ( Thermo Fisher Scientific ) supplemented with 10% fetal bovine serum ( FBS ) and 1 × penicillin/streptomycin ( pen-strep ) ( Thermo Fisher Scientific ) in a humidified incubator at 37°C with 5% CO2 . C6/36 cells ( Aedes albopictus ) were grown in RPMI medium ( Thermo Fisher Scientific ) supplemented with 10% FBS ( GenDEPOT ) and 1 × pen strep and incubated at 28°C with 5% CO2 . Cells were tested for mycoplasma contamination every three months . HeLa and JEG3 cells were authenticated by STR analysis at the UTMB Molecular Genomics Core . Dakar 41525 ( ZIKV-Dakar ) , Cambodia FSS13025 ( ZIKV-Cambodia ) and PRVABC59 ( ZIKV-Puerto Rico ) strains were kindly provided by Scott Weaver ( UTMB ) and Nikos Vasilakis ( UTMB ) and propagated in C6/36 cells . DENV-2-NGC ( New Guinea C strain ) was propagated as previously described ( Sessions et al . , 2009 ) . Nano-luciferase reporter ZIKV was engineered following the same strategy reported in Xie et al . ( Xie et al . , 2016 ) . Δ10 ZIKV and its WT version ( ZIKV-Cambodia ) were derived from infectious clones and propagated in Vero cells as described in ( Shan et al . , 2017b ) . Viral infections were performed in DMEM supplemented with 1% FBS . After 1 hr of incubation , media was replaced with complete DMEM media . MOIs and post-infection times are specified in the figures . ZIKV stocks and supernatants from Figure 3 were analyzed by plaque assay in Vero cells ( Agbulos et al . , 2016 ) . Titers of DENV-2 stock , Δ10 ZIKV and WT ZIKV from infectious clones were performed in Vero cells using foci-forming assay as previously described ( Shan et al . , 2017b; Sessions et al . , 2009 ) . The following antibodies were used for IP , western blotting and/or immunofluorescence analysis: anti-envelope protein 4G2 Henchal et al . , 1982 , rabbit IgG ( 2729S , Cell Signaling Technologies ) , anti-FMRP ( ab17722 , ABCAM , Cambridge , UK ) , anti-FXR1 ( 12295S , Cell Signaling Technologies ) , anti-FXR2 ( 7098S , Cell Signaling Technologies ) , anti-G3BP1 ( A302-033A , Bethyl Laboratories ) , rabbit anti-DDX6 ( 9407S , Cell Signaling technologies ) , rabbit anti-PTB ( homemade ) , rabbit anti-ZIKV NS4B ( GTX133321 , Genetex ) , anti-ZIKV NS2B ( GTX133308 , Genetex ) , anti-BRD4 ( 13440 , Cell Signaling Technologies ) , anti-GAPDH ( ab9485 , ABCAM ) , anti-TLN1 ( SC-365875 , Santa Cruz Biotechnology ) , anti-PNPLA6 ( SC-271049 , Santa Cruz Biotechnology ) . Synthetic DNA fragments corresponding to the complete 3′ UTRs of African ( strain MR766 , accession number KX377335 . 1 ) and pandemic Asian-lineage ZIKV ( PRVABC59 strain , KX377337 . 1 ) were obtained from IDT ( Integrated DNA Technologies ) and cloned into pcDNA3 . 1 ( + ) plasmid containing tobramycin aptamer ( Liao et al . , 2018 ) . As a control RNA we used a sequence corresponding to the DENV NS2A sequence as previously reported ( Manokaran et al . , 2015 ) . PCR was used to generate DNA templates for T7 in vitro transcription using the MEGAscript T7 kit ( Thermo Fisher Scientific ) . RNA affinity chromatography using JEG3 cell lysate was performed following the published protocol ( Ward et al . , 2014 ) . Eluted RBPs were identified by label-free mass spectrometry at the UTMB Mass Spectrometry Core . Samples were dissolved in 30 μl denaturation buffer ( 4% SDS and 100 mM DTT in 0 . 1M TEAB pH 8 . 5 ) , heated at 65°C for 15 min , and loaded onto 30 kDa spin filters ( Merck Millipore ) . The buffer was exchanged three times with UA solution ( 8 M UREA in 0 . 1 M TEAB pH 8 . 5 ) by centrifugation at 14 , 000 g . After removal of SDS , cysteine alkylation was accomplished through the addition of alkylation buffer ( 50 mM IAA , 8 M UREA in 0 . 1 M Tris-HCl pH 8 . 5 ) for 1 hr at room temperature in the dark . UA buffer was exchanged with TEAB buffer ( 40 mM TEAB pH 8 . 5 ) . The proteins were digested with trypsin ( enzyme-to-substrate ratio [w/w] of 1:100 ) and 5% ACN at 37°C overnight . Peptides were centrifuged through the size exclusion membrane and collected into a clean microcentrifuge tube , followed by a rinse with 80 μL of 0 . 2% formic acid . The combined peptide solution was then dried in a speed vac and resuspended in 2% acetonitrile , 0 . 1% formic acid , 97 . 9% water and placed in an autosampler vial for LC/MS analysis . Raw data files from the mass spectrometer were aligned by accurate mass and time in Progenesis QI for proteomics ( version 2 . 0 . 5556 . 29015 , Nonlinear Dynamics , a Waters Company ) . The top five spectra for each feature were exported as * . mgf files and searched against a combined Uniprot-Human canonical database . Peptide identifications were imported into Progenesis QI , and all peptides with −10logP scores < 30 ( Mascot or confidence score ) were removed . Conflict resolution was performed in order to remove lower scoring peptides when multiple peptides were assigned to a single feature . Protein quantification was calculated from normalized peptide abundances using a summed abundance of unique peptides . Unique peptide abundance was calculated by the area of the corresponding peaks in the ion chromatograms . Protein abundance from RBPs interacting with PRVABC59 and MR766 ZIKV 3′ UTRs were compared with control RNA to calculate ratios of enrichment ( PRVABC59 3′ UTR/NS2A and MR766 3′ UTR/NS2A ) . Eight proteins that were identified by ≥2 unique peptides and were found ≥2 times in both the ZIKV sequences relative to the control RNA were considered high-confidence ZIKV 3′ UTR binding proteins RBPs with ratio >2 were considered as enriched . 1 . 5 × 106 HeLa cells were plated in 10 cm dishes . 24 hr later , cells were infected with ZIKV-Dakar , ZIKV-Cambodia and ZIKV-Puerto Rico at an MOI of 3 for 48 hr . Cells were washed three times with cold PBS , scraped and pelleted in 3 mL of PBS ( 5 min , 1500 rpm at 4°C ) . Cell pellets were resuspended in an equal volume of RIPA buffer ( Cell Signaling Technologies ) with protease inhibitors ( Roche ) . All steps of the IP were performed at 4°C . For each IP 100 µL of protein A/G PLUS-Agarose ( Santa Cruz Biotechnology ) was washed with 1 ml NT2 buffer ( 50 mM Tris-HCl pH 7 . 4 , 150 mM NaCl , 1 mM MgCl2 and 0 . 05% IGEPAL ) and blocked for 30 min in NT2 buffer with 0 . 5 mg/mL BSA . Beads were washed three times with NT2 buffer and incubated with either 10 µg of rabbit anti-FMRP , anti-PTB or rabbit IgG in a total volume of 600 µL of NT2 buffer for 2 hr with head to tail rotation . After coupling antibodies , beads were washed three times with NT2 buffer . One mg of protein from clarified lysate was diluted in NT2 buffer in a total volume of 600 µL and added either to IgG or FMRP beads and incubated with rotation for 1 hr . After incubation , beads were pelleted , washed four times with NT2 buffer and resuspended in 115 µL of NT2 . 15 µL of beads were used for WB and 100 µL for RNA isolation using Trizol . Equal volumes of RNA ( 5 µL ) from IP reactions were processed for Northern blotting . The NorthernMax kit ( Thermo Fisher Scientific ) was used following the manufacturer´s instructions . RNA was mixed with 3 volumes of formaldehyde loading dye , then incubated 15 min at 65°C and 2 min on ice . Electrophoresis was performed in denaturing 1% agarose gels at 95 V . After electrophoresis , the gel was incubated in alkaline buffer ( 0 . 01N NaOH , 3M NaCl ) for 20 min and transferred to a Biodyne B nylon membrane ( Thermo Fisher Scientific ) by downward transfer . RNA was crosslinked to the membrane using the UV stratalinker 2400 auto-crosslink program . The membrane was pre-hybridized for 30 min with ULTRAhyb-oligo Buffer ( Thermo Fisher Scientific ) at 42°C and hybridized overnight at 42°C with a Biotin-labeled DNA probe to detect ZIKV 3′ UTR ( 10453–10739 nt ) . The DNA probe was generated by PCR using 60% of unmodified dTTP and 40% biotin-16-dUTP ( Roche ) . Specific primers ( forward: 5′ AGGAGAAGCTGGGAAACCAAGC 3′; reverse: 5′ GATAATACGACTCACTATAGGAAACTCATGGAGTCTCTGGTC 3′ ) and apt-MR766 3′ UTR plasmid ( template ) were used to generate the biotin-labeled probe . For Figure 7 northern blots , RNA was mixed with 2X TBE-Urea loading sample and resolved in 6% acrylamide TBE , 8M Urea gels and transferred to nylon membrane in 0 . 5% TBE for 3 hr using the XCell II Blot module ( Thermo Fisher Scientific ) . The membrane was crosslinked and hybridized with a DNA probe to detect the ZIKV 3′ UTR DB and SLIII ( 10528–10807 nt ) . This probe was generated as above using specific primers ( forward: 5′ GATAATACGACTCACTATAGGGCCCCTCAGAGGACACTGAGTCAAAAAA 3′; reverse: 5′ AGAAACCATGGATTTCCCCACACCGGCCGCCGCT 3′ ) . After hybridization , the membranes were washed and incubated for 1 hr at room temperature with IRDYE 800CW streptavidin ( LI-COR Biosciences ) in Odyssey Blocking Buffer ( LI-COR Biosciences ) with 1% of SDS . After three washes with TBS containing 0 . 1% tween , the membrane was scanned using a LI-COR Odyssey instrument . Densitometry was performed using Image Studio Lite Ver 5 . 0 software . Protein knockdown was performed using individual siRNAs ( Qiagen ) or a pool of siRNAs . 3 × 104 HeLa cells were plated in a 24 well plate . The next day cells were transfected with AllStars Negative Control siRNA ( Qiagen ) , individual siRNAs , or the pool using 30 nM of siRNA and 1 . 5 µL of RNAiMAX ( Thermo Fisher Scientific ) diluted in 50 µL of Opti-MEM media ( Thermo Fisher Scientific ) . After 4 hr , transfection media was replaced by fresh media and cells were incubated for 48 hr . HeLa cells were infected as noted in the figure legends . Cells were lysed in RIPA buffer ( Cell Signaling Technologies ) with 1X protease inhibitor . 10 µg of protein samples were resolved under denaturing conditions on 4–12% acrylamide gels ( Thermo Fisher Scientific ) . ZIKV NS4B , and NS2B , BRD4 , TLN1 , PNPLA6 , Actin and GAPDH were detected using primary antibodies mentioned above . Goat anti-mouse IgG and goat anti-rabbit IgG coupled with IRDye 800 or 680 ( LI-COR Biosciences ) were used as secondary antibodies . Blots were developed in the LI-COR Odissey luminescence system and protein expression was quantified by densitometry with the Image Studio Lite Ver 5 . 0 software . Cells were harvested and fixed with 1% formaldehyde , permeabilized and blocked for 20 min ( 1X PBS , 0 . 1% tween 20 , 5% FBS ) , incubated for 1 hr at room temperature with mouse 4G2 antibody and then 1 hr with goat anti-mouse IgG Alexa fluor 647 ( Thermo Fisher Scientific ) to detect ZIKV envelope protein ( E protein ) . E protein fluorescence was measured with the Guava easyCyte system ( Millipore ) using the red laser ( 642 nm ) for Figures 3 and 5 and Figure 3—figure supplements 1–3 . Double staining was performed for Figure 6 to detect E protein and cellular FXR2 protein . E protein was detected using mouse 4G2 antibody and goat anti-mouse IgG Alexa fluor 488 . FXR2 protein was stained using rabbit anti-FXR2 and goat anti-rabbit IgG Alexa fluor 647 . Sample analysis , scatter plots of infected cells , mean fluorescence intensities and histograms were performed in FlowJo V10 software . To determine percentage of infection , 3 × 104 HeLa cells per well were plated in 24-well plates . 24 hr later , cells were transfected with control siRNA or FMR1 siRNA ( pool ) . After 48 hr of knockdown , cells were infected with ZIKV-Dakar for 48 hr , fixed with 1% paraformaldehyde , permeabilized and blocked for 20 min ( 1 × PBS , 0 . 1% tween 20 , 1% FBS ) , and incubated for 1 hr at room temperature with mouse 4G2 mouse ( E protein ) and 1 hr with Goat anti-mouse IgG Alexa fluor 647 . Nuclei were counterstained with Hoechst 33342 ( Sigma-Aldrich ) . Image acquisition and infection rates were calculated using a high-content imaging microscope ( Opera Phenix , Perkin Elmer ) . To visualize the accumulation of FXR2 protein in infected cells , glass slides with non-infected or ZIKV infected cells ( WT or Δ10 ZIKV ) for 24 hr were fixed , permeabilized and blocked as described above . Viral protein was detected using 4G2 antibody and goat anti-mouse IgG Alexa fluor 488 . FXR2 protein was stained using rabbit anti-FXR2 and goat anti-rabbit IgG Alexa fluor 647 . Nuclei were counterstained with Hoechst 33342 ( Sigma-Aldrich ) and images were acquired using an Olympus fluorescence microscope . 5 × 104 HeLa cells were plated in 24-well plates . Next day , cells were transfected with control siRNA or FMR1_2 siRNA . After 46 hr of knockdown , cells were treated with 0 . 05% DMSO or 20 µM NITD008 for 2 hr before infection with the ZIKV reporter at an MOI of 3 . NITD008 or DMSO were retained in media during the infection . After 1 hr of incubation , inoculum was retired and replaced with fresh media containing DMSO or NITD008 . 2 . 5 hr later , cells were washed five times with 1 × PBS and lysed with Renilla lysis buffer ( Promega ) . Additionally , cycloheximide ( CHX ) treatment ( 200 µM ) was used as a background control in absence or presence of NITD008 . For this condition , cells were pretreated with CHX ( in absence or presence of NITD008 ) for 2 hr before ZIKV reporter infection and CHX was retained during infection until cell lysis . Luciferase assays were performed using the High-Affinity NanoBit evaluation system ( Promega ) and the Enspire plate reader ( Perkin Elmer ) . Viral genome and GAPDH RNA were quantified by RT-qPCR . Cell-associated RNA from HeLa was extracted by the Trizol method ( Thermo Fisher Scientific ) and reverse transcribed using the MultiScribe Reverse transcriptase protocol ( Thermo Fisher Scientific ) . qPCR was performed with SYBR mix on a StepOne plus instrument ( Thermo Fisher Scientific ) . The ΔΔCT method was used to calculate relative expression levels of viral genome . Primers used for ZIKV ORF ( nucleotides 4541 to 4631 of ZIKV-Cambodia: forward 5′ CTGTGGCATGAACCCAATAG 3′; reverse 5′ ATCCCATAGAGCACCACTCC 3′ ) . Primers used for human GAPDH ( forward 5′ AGCCACATCGCTCAGACAC 3′; reverse 5′ GCCCAATACGACCAAATCC 3′ ) . To evaluate viral genome in mice , RNA from testes lysates were obtained by diluting 15 µL of cell lysate in 85 µL of NT2 buffer and 300 µL of Trizol LS . RNA transcription and qPCR were performed as mentioned above using primers for ZIKV ORF and mouse GAPDH ( forward 5′ AAGGTCATCCCAGAGCTGAA 3′; reverse 5′ AAGGTCATCCCAGAGCTGAA 3′ ) . All animal studies were done in accordance with IACUC protocols as per UTMB policy . A129 mice were obtained from colonies maintained under specific pathogen-free conditions . Male 8–9 week-old A129 mice were infected with 1 × 105 FFU of WT ( n = 4 ) or Δ10 ZIKV ( n = 4 ) mutant viruses through the intraperitoneal route . PBS was given to the mock-infected mice through the same route ( n = 3 ) . At 6 days post-infection mice were euthanized , and testes were removed immediately as previously described ( Hansen et al . , 2014 ) . Testes were flash-frozen in dry ice and stored at −80°C . Tissue was homogenized and lysed with a tissue grinder ( OmniTHQ ) in 500 µL of polysome lysis buffer ( 10 mM HEPES pH 7 . 0 , 100 mM KCl , 25 mM EDTA , 5 mM MgCl2 , 1 mM DTT , 0 . 5% NP-40 ) . RNasin 1:1000 dilution ( Promega ) and protease inhibitor ( Roche ) were freshly added to samples . Samples were rotated for 10 min at 4°C to induce lysis and then flash-frozen on dry ice . Samples were thawed and nuclei were pelleted at 3000 x g for 10 min . 10 µg of protein samples from tissue lysates were fractionated in a 4–12% acrylamide gel ( Thermo Fisher Scientific ) , transferred to a nitrocellulose membrane and blotted for FXR2 , PNPLA6 and GAPDH with antibodies mentioned above . For RIP , testis lysate obtained from a mouse infected with WT ZIKV ( ZIKV-Cambodia ) was pre-cleared by adding 50 µL of protein A/G PLUS-Agarose ( Santa Cruz Biotechnology ) and rotating for 30 min at 4°C . 50 µL of beads were blocked in 600 µL of NT2 buffer ( 50 mM Tris-HCl , pH 7 . 4 , 150 mM NaCl , 1 mM MgCl2 , 1 mM DTT , 0 . 05% NP-40 ) with 0 . 5 mg/mL BSA ( Thermo Fisher Scientific ) and 0 . 5 mg/mL yeast tRNA ( Sigma-Aldrich ) . After blocking , beads were washed and incubated overnight at 4°C with 10 µg of either FMRP or Rabbit IgG antibodies . Antibody-bound beads were washed four times with ice-cold NT2 buffer . For the IP , 2 mg of pre-cleared protein was added to the antibody-bound beads . Final volume was adjusted to 600 µL with NT2 buffer containing RNase and protease inhibitors and incubated for 1 . 5 hr at 4°C . After incubation , beads were washed four times with NT2 buffer and processed for WB and RNA isolation as in the RIP for HeLa cells . qPCR was performed using ZIKV ORF primers described above and specific primers that amplify ZIKV 3′ UTR ( nucleotides 10623 to 10722 of ZIKV-Cambodia: forward 5′ CCTGAACTGGAGATCAGCTGTG 3′; reverse 5′ GGTCTTTCCCAGCGTCAATA 3′ ) . 5′ triphosphate RNAs were generated as described above using the HiScribe T7 High Yield RNA Synthesis Kit ( New England Biolabs ) and specific primers for sfRNA ( forward: 5′ GATAATACGACTCACTATAGGGT GTTGTCAGGCCTGCTAGTCAGCCACAGC 3′; reverse: 5′ AGAAACCATGGATTTCCCCACACCGGCCGCCGCT 3′ ) and the DBSLIII mutant ( forward: 5′ GATAATACGACTCACTATAGGGCCCCTCAGAGGACACTGAGTCAAAAAA 3′ ) . 5′ monophosphate RNAs were by generated by synthesis of capped RNAs followed by treatment with Tobacco Acid Pyrophosphatase ( Thermo Fisher Scientific ) . 5′ tri- and monophosphate RNAs were treated with the Turbo DNAse I ( Thermo Fisher Scientific ) . RNAs were purified using Trizol LS and Direct Zol RNA MiniPrep ( Zymo Research ) for RNA clean up . 36 pmol of DBSLIII or sfRNA were electroporated into 2 × 106 HeLa cells in 250 µL of Ingenio electroporation solution ( Mirus ) using the Gene Pulser X cell Electroporation System ( Bio-Rad Laboratories ) following manufacturer´s instructions . 48 hr post electroporation , cells were analyzed for expression of FXR2 and BRD4 and transfection efficiency by NB as described above . The independent experiments were performed with three biological replicates except for Figure 2A and B . For Figure 2C two infected mice , each with three technical replicates , were processed separately . Differences between treatments and control groups were evaluated using the SigmaPlot/Stat package 11 . In all cases , parametric or nonparametric tests and the appropriate post-hoc test were applied . If data met with assumptions of normality ( Shapiro-Wilk test ) and equal variance test , a t-test ( parametric ) was conducted . Data that did not meet with either normality test or equal variance test were analyzed using Mann-Whitney U test . | Certain mosquitoes can carry pathogens that are able to infect humans , including Zika and dengue viruses . Most people infected with Zika virus only develop mild symptoms , or no symptoms at all . But if the virus infects a pregnant woman , it can lead to miscarriage and other pregnancy complications , or cause severe birth defects in her unborn baby . Viruses must infect the cells of a host to multiply . To do so , they hijack the cellular machinery to make proteins needed to copy their genetic material and assemble new virus particles . The genetic material of Zika virus is made of ribonucleic acid ( RNA ) . When the Zika virus infects cells , pieces of the virus RNA , known as subgenomic flavivirus RNAs ( or sfRNAs for short ) , accumulate in the cell . Cells infected with dengue virus , which is closely related to the Zika virus , also accumulate sfRNA . Dengue sfRNA is known to bind to and inhibit the activity of specific proteins in cells that would otherwise block the virus from multiplying . Nonetheless , it is not clear whether the sfRNA from Zika virus performs a similar role . Soto-Acosta et al . searched for human proteins that could bind to Zika sfRNA and may affect the ability of the virus to multiply . The experiments showed that a protein known as FMRP , which , when faulty , is linked to a genetic condition that causes a range of developmental problems , binds to Zika sfRNA in human and mouse cells infected with Zika virus . FMRP inhibits the production of virus proteins in the cells and limits the ability of the virus to multiply . However , as Zika sfRNA gradually accumulates during infection , the sfRNA binds to FMRP and interferes with its activity , allowing the virus to multiply more efficiently . Soto-Acosta et al . also found that Zika sfRNA affects the ability of FMRP to regulate the production of other proteins that are normally found in cells . These findings suggest that the interference of the virus with FMRP may contribute to Zika disease in humans . Moreover , a mutant Zika virus unable to produce sfRNA could be developed into a vaccine to potentially prevent Zika . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"microbiology",
"and",
"infectious",
"disease"
] | 2018 | Fragile X mental retardation protein is a Zika virus restriction factor that is antagonized by subgenomic flaviviral RNA |
Bacteria live in environments that are continuously fluctuating and changing . Exploiting any predictability of such fluctuations can lead to an increased fitness . On longer timescales , bacteria can ‘learn’ the structure of these fluctuations through evolution . However , on shorter timescales , inferring the statistics of the environment and acting upon this information would need to be accomplished by physiological mechanisms . Here , we use a model of metabolism to show that a simple generalization of a common regulatory motif ( end-product inhibition ) is sufficient both for learning continuous-valued features of the statistical structure of the environment and for translating this information into predictive behavior; moreover , it accomplishes these tasks near-optimally . We discuss plausible genetic circuits that could instantiate the mechanism we describe , including one similar to the architecture of two-component signaling , and argue that the key ingredients required for such predictive behavior are readily accessible to bacteria .
Organisms that live in changing environments evolve strategies to respond to the fluctuations . Many such adaptations are reactive , for example sensory systems that allow detecting changes when they occur and responding to them . However , adaptations can be not only reactive , but also predictive . For example , circadian clocks allow photosynthetic algae to reorganize their metabolism in preparation for the rising sun ( Bell-Pedersen et al . , 2005; Husain et al . , 2019 ) . Another example is the anticipatory behavior in E . coli , which allows it to prepare for the next environment under its normal cycling through the mammalian digestive tract ( Savageau , 1983 ) ; similar behaviors have been observed in many species ( Tagkopoulos et al . , 2008; Mitchell et al . , 2009 ) . All these behaviors effectively constitute predictions about a future environment: the organism improves its fitness by exploiting the regularities it ‘learns’ over the course of its evolution ( Mitchell and Lim , 2016 ) . Learning such regularities can be beneficial even if they are merely statistical in nature . A prime example is bet hedging: even if the environment changes stochastically and without warning , a population that learns the statistics of switching can improve its long-term fitness , for example , by adopting persistor phenotypes with appropriate probability ( Kussell and Leibler , 2005; Veening et al . , 2008 ) . The seemingly limitless ingenuity of evolutionary trial-and-error makes it plausible that virtually any statistical structure of the environment that remains constant over an evolutionary timescale could , in principle , be learnt by an evolving system , and harnessed to improve its fitness ( Watson and Szathmáry , 2016 ) . However , the statistical structure of the environment can itself change , and this change can be too quick to be learned by evolution ( Figure 1A ) . For example , an organism might experience a period of stability followed by a period of large fluctuations , or an environment where two resources are correlated , and then another where they are not . Note that there are two key timescales here – that of the fluctuations themselves ( which we assume to be fast ) , and the slower timescale on which the structure of those fluctuations changes . One expects such scenarios to be common in an eco-evolutionary context . As an example , consider a bacterium in a small pool of water . Its immediate environment , shaped by local interactions , is fluctuating on the timescale at which the bacterium changes neighbors . The statistical properties of these fluctuations depend on the species composition of the pool . As such , the fast fluctuations are partially predictable , and learning their structure could help inform the fitness-optimizing strategy: a neighbor encountered in a recent past is likely to be seen again in the near future . However , these statistics change on an ecological timescale , and such learning would therefore need to be accomplished by physiological , rather than evolutionary , mechanisms . On a physiological timescale , this problem is highly nontrivial: the organism would have to perform inference from prior observations , encode them in memory , and act upon this knowledge ( Figure 1B ) . It is clear that solutions to this problem do exist: such behaviors , common in neural systems , can be implemented by neural-network-like architectures; and these known architectures can be translated into biochemical networks ( Hjelmfelt et al . , 1991; Kobayashi , 2010; Fages et al . , 2017; Katz and Springer , 2016; Katz et al . , 2018 ) . But single-celled organisms operate in a severely hardware-limited regime rarely probed by neuroscience . Streamlined by evolution , bacterial genomes quickly shed any unused complexity . Whether we could expect learning-like behaviors from bacteria depends on whether useful networks could be simple enough to plausibly be beneficial . Known examples of phenotypic memory , for example , when the response is mediated by a long-lived protein , can be interpreted as a simple form of learning ( Lambert et al . , 2014; Hoffer et al . , 2001 ) ; circuits capable of adapting to the current mean of a fluctuating signal , as in bacterial chemotaxis ( Barkai and Leibler , 1997 ) , also belong in this category . Prior theory work has also proposed that simple genetic circuits could learn more subtle binary features , such as a ( transient ) presence or absence of a correlation between two signals ( Sorek et al . , 2013 ) . Here , we show that a simple generalization of a ubiquitous regulatory motif , the end-product inhibition , can learn , store , and ‘act upon’ the information on continuous-valued features such as timescales and correlations of environmental fluctuations , and moreover , can do so near-optimally . We identify the key ingredients giving rise to this behavior , and argue that their applicability is likely more general than the simple metabolically inspired example used here .
For a simple model capturing some of the challenges of surviving in a fluctuating environment , consider a situation where some internal physiological quantities P→= ( P1 , … , PN ) must track fluctuating external variables D→= ( D1 , … , DN ) . For example , the expression of a costly metabolic pathway would ideally track the availability of the relevant nutrient , or the solute concentration in the cytoplasm might track the osmolarity of the environment . In abstract terms , we describe these environmental pressures by the time-dependent D→ ( t ) , and postulate that the organism fitness is determined by the average mismatch -⟨∑i=1N ( Pi-Di ) 2⟩ , a quantity we will henceforth call ‘performance’ . Here and below , angular brackets denote averaging over time . In this simple model , a given static D→ clearly defines a unique optimal state P→; the regulatory challenge is entirely due to D→ being a fluctuating quantity . The challenges faced by real organisms are certainly vastly more rich: even in the static case , the optimal behavior may not be unique , or even well-defined ( optimal under what constraints ? ) ; and in the dynamic case , the future state of the environment could be affected by past actions of the organism . These considerations can add layers of complexity to the problem , but our minimal model is sufficient to focus on the basic issues of sensing , learning and responding to changing fluctuation statistics of external factors . If D→ changes sufficiently slowly , the organism can sense it and adapt P→ accordingly . We , instead , are interested in the regime of rapid fluctuations . When changes in D→ are too rapid for the organism to match P→ to D→ exactly , it can rely on statistical structure . At the simplest level , the organism could match the mean , setting P→≡⟨D→⟩ . However , information on higher-order statistics , for example correlations between D1 and D2 , can further inform the behavior and improve fitness . To see this , in what follows , we will consider the minimal case of such structured fluctuations , namely a N-dimensional vector D→= ( D1 , … , DN ) undergoing a random walk in a quadratic potential ( the Ornstein—Uhlenbeck process ) : ( 1 ) D→ ( t+Δt ) =D→ ( t ) -M⋅ ( D→ ( t ) -D¯→ ) Δt+2ΓΔtη→ , with mean D¯→ , fluctuation strength Γ , independent Gaussian random variables η→ with zero mean and unit variance , and the matrix M defining the potential . In this system , the relevant ‘fluctuation structure’ is determined by M and Γ . In one dimension , Equation ( 1 ) gives D a variance of Γ/M . In two dimensions , denoting the eigenvalues of M as λ1 , 2 , the stationary distribution of the fluctuating D→ is a Gaussian distribution with principal axes oriented along the eigenvectors of M , and standard deviations along these directions given by Γ/λ1 and Γ/λ2 . Intuitively , we can think of the fluctuating D→ as filling out an ellipse ( Figure 1C ) . Going forward , when we refer to learning fluctuation structure , we mean learning properties of M and Γ . If M and Γ are known , the optimal strategy minimizing ⟨ ( P→-D→ ) 2⟩ , where D→ ( t ) is set by Equation ( 1 ) , can be computed exactly , as a function of the maximum allowed rate of change ∥P˙∥2 ( Liberzon , 2011 ) . ( If we do not constrain ∥P˙∥2 , the optimal behavior is of course P→=D→ . ) Briefly , the optimal behavior is to steer P→ toward the best guess of the expected future D→ ( see Appendix 1 , section 'Control theory calculation' ) . This best guess depends on the fluctuation structure , as illustrated by the comparison between Figure 1D and E for an isotropic and an anisotropic M . However , in our problem , we will assume that M and Γ do not stay constant long enough to be learned by evolution , and thus are unknown to the system . In this regime , it is not clear that the behavior of an M- and Γ-optimized system is relevant . Nevertheless , we will describe a regulatory architecture consisting of common regulatory elements that will adapt its responsiveness to the fluctuation structure of its input ( ‘learn’ ) ; for example , in the two-dimensional case , it will indeed develop the anisotropic response shown in Figure 1E . Moreover , we will find the steady-state performance of our architecture to be near-optimal , when compared to the theoretical ceiling of a system that knows M and Γ perfectly . The section above was intentionally very general . To discuss solutions available to cells , it is convenient to restrict the scope from this general formulation to a more specific metabolically-inspired case . From here onwards , let Di be the instantaneous demand in metabolite xi ( determined by external factors ) , and Pi be the rate at which the metabolite is produced , with both defined in units of metabolite concentration per unit time . The number of components of the vector D→ now has the meaning of the number of metabolites , and we will denote it as Nx . The cell needs to match P→ to D→ ( or , equivalently , maintain the homeostasis of the internal metabolite concentrations xi ) . The simplest way to solve this problem is via feedback inhibition . Consider first the case of a single metabolite x . If an accumulation of x inhibits its own synthesis , a decreased demand will automatically translate into a decreased production . For our purposes , we will model this scenario by placing the synthesis of metabolite x under the control of a regulator a ( e . g . a transcription factor ) , which is , in turn , inhibited by x ( Figure 2A ) . For simplicity , we will measure regulator activity a directly in units of equivalent production of x . The dynamics of this system , linearized for small fluctuations of metabolite concentration x , can be written in the following form ( see Appendix 1 , section 'Simple end-product inhibition ) : ( 2a ) x˙=P−Dxx0source-sink dynamics of metabolite x ( 2b ) P=aP0 definition of regulator activity a ( 2c ) a˙=x0−xλregulator activity inhibited by x Here , we introduced P0 with dimension of production ( concentration per time ) to render a dimensionless . In Equation 2c , λ has the units of concentration × time , and setting λ≡x0τa defines a time scale for changes in regulator activity . Assuming the dynamics of metabolite concentrations x are faster than regulatory processes , and choosing the units so that x0=1 and P0=1 , we simplify the equations to: ( 3 ) x=P/DP=aτaa˙=1−x . We will refer to this architecture as simple end-product inhibition ( SEPI ) . For two metabolites x→= ( x1 , x2 ) , the straightforward generalization is to have two independent copies of this circuit , with two regulators a1 , a2 ( Figure 2B ) . Denoting the number of regulators as Na , we note that in the SEPI architecture , there are as many regulators as there are metabolites: Na=Nx . The architecture we will describe builds on this widely used regulatory motif , and relies on three added ingredients: Here and below , we use index μ for regulators ( μ=1…Na ) and index i for metabolites ( i=1…Nx ) . These three ingredients , we claim , will be sufficient for the circuit to both learn higher order statistics and to use this information appropriately when matching the production to demand . It is important to emphasize that all three are readily accessible to cells . In fact , there are multiple ways to build regulatory circuits exhibiting the proposed behavior using common regulatory elements . To focus on the general mechanism rather than any one particular implementation , we will defer describing these example circuits until later in the text ( Figure 6 ) ; here , we will consider a minimal modification of Equation ( 3 ) that contains the required ingredients: ( 4a ) xi=Pi/Di ( 4b ) Pi=Σμσμiaμ ( 4c ) τaa˙μ=aμmax ( d , ∑iσμi ( 1−xi ) ) −κaμ . This architecture bears a similarity to neural networks , and , as we will see , the familiar intuition about the value of extra ‘hidden nodes’ indeed holds . However , we caution the reader not to rely too heavily on this analogy . For example , here σμi is a constant matrix describing how the activities of regulators aμ control the synthesis of metabolites xi . For two metabolites ( Nx=2 ) as in Figure 2C , each regulator is summarized by a 2-component vector σ→μ= ( σμ1 , σμ2 ) ; its components can be of either sign ( or zero ) and specify how strongly the regulator aμ is activating or repressing the synthesis of metabolite xi . For simplicity , below , we will choose these vectors to be of unit length . Then , each regulator σ→μ is fully characterized by an angle in the ( x1 , x2 ) plane , which allows for a convenient visualization of the regulatory systems ( Figure 2D ) . The σμi defines the regulatory logic of our system and does not change with time . The parameter d≤0 allows us to tune the strength of the simple nonlinearity ( Figure 2E ) ; below we set d=0 ( strong nonlinearity ) unless explicitly stated otherwise . As we will show later , the learning behavior is also observed for more realistic functions such as the Hill function , but the simple piece-wise linear form of Equation ( 4 ) will help us relate the observed behavior to specifically nonlinearity as opposed to , for example , cooperativity ( the Hill parameter tunes both simultaneously ) . Finally , the parameter κ reflects degradation and is assumed to be small: κ≪x0 . Previously , for SEPI , it could be neglected , but here , it will matter due to the nonlinearity; for more details , see Appendix 1 , section 'Simple end-product inhibition' . The parameters used in simulations are all listed in Appendix 1 , section 'Parameters used in figures' . Just like simple end-product inhibition in Equation ( 3 ) , the modified system Equation ( 4 ) will correctly adapt production to any static demand ( see Appendix 1 , section 'Adaptation to static demand' ) . In the following , we will show that the added ingredients also enable learning the structure of fluctuating environments . For this purpose , we expose our system to demands D ( t ) with fixed means ( D¯i=1 ) but a changing fluctuation structure . To show that our system is able to adapt to different fluctuation structures , we probe it with changing environmental statistics , and show that it , first , learns these statistics , and , second , is able to make use of this information in its behavior . For simplicity , we start with the 1-dimensional case ( Figure 3A–F ) . In dimension Nx=1 , an excess of regulators means we have both an activator a+ and a repressor a- for the production of x ( Figure 3A ) . This is reminiscent of paradoxical regulation ( Hart et al . , 2012 ) . We probe our system with changing environmental statistics by exposing it to a demand D ( t ) with an increasing variance ( Figure 3B , C ) . As a reminder , here and below , the mean demand is fixed at 1 . Faced with a faster fluctuating input , our system upregulates both a+ and a- while keeping a+-a- constant ( a+-a-≈D¯=1; Figure 3D ) . In this way , the two activity levels a+ and a- encode both the mean and the variance of fluctuations . Crucially , the system makes use of the information it stores: The increased regulator activities allow future changes in P to be faster . The system’s responsiveness , which we can define as ℛ≡dP˙dD , increases as a++a- ( Figure 3E; see also Appendix 1 , section 'Defining the system's responsiveness' ) . As a result , as shown in Figure 3F , our system is able to perform approximately equally well ( after adaptation time ) in each environment , unlike a system like simple end-product inhibition , which is unable to adapt its sensitivity . In summary , Figure 3D–F show that the simple architecture of Figure 3A can not only learn the statistics of environment fluctuations , but also ‘act upon this knowledge , ’ effectively performing both computations of Figure 1B . The idea of learning the fluctuation structure is perhaps clearer in dimension Nx=2 , since the two demands can now be correlated with each other , and it seems intuitive that a system able to learn the typical direction of fluctuations ( the angle α in Figure 3H ) should be able to track the input better . Indeed , as we saw in Figure 1D–E , when environment fluctuations are anisotropic , the responsiveness of a well-adapted strategy must be anisotropic as well: the preferred direction must elicit a stronger response . Mathematically , the responsiveness ℛ is now a matrix ℛij=dP˙idDj , and for a well-adapted system we expect its eigenvectors to align with the principal directions of M . In Figure 3G–L , Figure 4 and Figure 5 , our discussion will focus on this two-dimensional case . Figure 3G–L show the behavior of our system ( Equation 4 ) with Na=5 regulators ( Figure 3G ) , exposed to an input structured as shown in Figure 3H , where we vary α . As represented pictorially in Figure 3I , we rotate the fluctuation structure matrix M in Equation ( 1 ) , keeping its eigenvalues λ1 , 2 fixed with λ1/λ2=10 ( this fixes the ratio of major to minor semi-axis lengths ) . With Na=5 regulators , matching the mean value of D→ would leave Na-2=3 degrees of freedom that can be influenced by other parameters ( such as variance in each dimension and correlation between different demands ) . And indeed , changing environment statistics induces strong changes in the regulator state adopted by the system , with regulators better aligned to the input fluctuations reaching higher expression ( Figure 3J; note the diagrams at the top , where dot size reflects the activity reached by the corresponding regulator at the end of each epoch; compare to the diagrams in Figure 3I ) . This activity pattern shapes the responsiveness matrix ℛ . Figure 3K plots the ‘learned angle’ , defined as the direction of the dominant eigenvector of ℛ; we find that it tracks the stimulus angle . Finally , Figure 3L demonstrates that our architecture is able to make use of this learning , outperforming the SEPI system , whose responsiveness is isotropic and fixed . In the previous section , we have shown by example ( Figure 3 ) that the proposed regulatory architecture can learn the statistics of the environment . We now characterize systematically the conditions under which learning improves performance and compare our system to the theoretical performance ceiling . Note that unlike the general statement that learning correlations improves performance , the optimal performance ceiling is necessarily specific to a given model of the environmental fluctuations . Nevertheless , this comparison is informative . The fluctuation structure in our model is defined by Γ and M . We first investigate the dependence of performance on Γ ( Figure 4A ) , exposing our system to a two-dimensional input structured as in Figure 3H with λ1/λ2=10 as before , α=π/4 , and a changing Γ . Although the input is two-dimensional , changing Γ scales the overall magnitude of fluctuations , and the behavior is analogous to the simpler one-dimensional example shown in the first column of Figure 3 . At Γ=0 ( static input ) , and by extension , for Γ finite but small , examining the steady state of Equation ( 4 ) shows that only Nx=2 out of Na regulators can be active . In this regime , our system is essentially identical to SEPI—the extra regulators , though available , are inactive—and in fact performs slightly worse . This is because at nonzero κ , the steady state of Equation ( 4 ) is slightly offset from the ideal state ⟨xi⟩=1 . ( While this effect can be corrected , it is only relevant in the parameter regime where no learning occurs , so we chose to keep Equation ( 4 ) as simple as possible; for additional discussion , see Appendix 1 , section 'Performance penalty from the degradation term' ) . When Γ becomes sufficiently large , the first term in Equation ( 4 ) ( proportional to the fluctuation size Γ ) for one of the inactive regulators finally exceeds , on average , the degradation term . At this point , the system enters the regime where the number of active regulators exceeds Nx , and its performance deviates from the SEPI curve . Beyond this point , further changes to the stimulus no longer affect performance , as our system is able to adapt its responsiveness to the changing fluctuation magnitude ( compare to Figure 3F ) . The threshold value of Γ satisfies Γ∝κ; the proportionality coefficient of order 1 depends on the specific arrangement of regulators but can be estimated analytically ( see Appendix 1 , section 'The minimal Γ needed to initiate adaptation' ) . The theoretically predicted deviation points are indicated with arrows , and are in agreement with the simulation results . When a regulator in the system is particularly well-aligned with the dominant direction of fluctuations , the deviation occurs sooner , explaining the better performance of our system when the regulators are more numerous . To better assess the performance of our system , we compare it to the theoretical optimum derived from control theory , which we represent with dotted lines in Figure 4A . For given M and Γ , the family of optimal behaviors is parameterized by Control Input Power ( CIP ) , defined as ∫∥P˙∥2𝑑t . If P→ could react infinitely fast , it would track D→ perfectly , but increasing response speed necessarily comes at a cost ( of making more sensors , or more enzymes for faster synthesis / degradation of xi ) ; constraining the CIP is thus a proxy for specifying the maximum tolerable cost . In order to compare our system with the optimal family of solutions , we compute 1T∫0T∥P˙∥2𝑑t of our system at each Γ ( T is the simulation time ) , and compare to the performance of the optimally steered solution with a matched CIP; details of the calculation can be found in Appendix 1 , section 'Control theory calculation' . Figure 4A demonstrates that the simple architecture we described not only benefits from matching its responsiveness to its input , but is in fact near-optimal when compared to any system of equivalent responsiveness . It is important to note that for low Γ , the performance of the SEPI architecture also tracks the optimal curve . Throughout this work , our intention is not to demonstrate that SEPI is a ‘poor’ architecture . To the contrary , the surprising efficiency of SEPI has been noted before ( Goyal et al . , 2010; Pavlov and Ehrenberg , 2013 ) , and Figure 4 similarly shows that at its own CIP , its performance is virtually optimal . The advantage of our learning-capable architecture derives from its ability to increase responsiveness when necessary , in the correct direction . Our simplified treatment of the SEPI architecture is not a strawman we seek to dismiss , but an example of a system that exhibits no learning . Having investigated the effect of fluctuation variance ( changing Γ ) , we turn to the effect of their correlation . Up to now , we subjected our system to a strongly correlated two-dimensional input with anisotropy λ1/λ2=10 ( illustrated , to scale , in Figure 1E ) . We will now consider a range of anisotropy values , down to anisotropy 1 ( uncorrelated fluctuations , Figure 1D ) , keeping the variances of D1 and D2 constant , α=π/4 as before , and Γ=0 . 05 . The result is presented in Figure 4B . With Na=5 or larger , our system is able to take advantage of the correlation , assuming it is strong enough to activate the learning mechanism . ( In fact , its performance can reach values that exceed the theoretical ceiling achievable by any system that assumes the two dimensions of D→ to be independent , and thus must be exploiting the correlation in its inputs; see Appendix 1 , section 'The system makes use of correlations in the input' and Appendix 1—figure 1 ) . For Na=4 , the performance curve remains flat . This is because the four regulators are arranged as two independent copies of the system shown in Figure 3A ( one {a+ , a-} pair for each of the two inputs ) ; this architecture can take advantage of the learned variance , but not the correlation . Finally , the SEPI architecture can adapt to neither variance nor correlation; its performance curve is also flat , but is lower . As expected , the advantage of our architecture manifests itself in environments with periods of large and/or strongly correlated fluctuations . The model described above was a proof of principle , showing that simple regulatory circuits could learn the fluctuation structure of their inputs . Given the simplicity of our model , it is not to be expected that the exact dynamics of Equation ( 4 ) are replicated in real cells . However , the benefit of this simplicity is that we can now trace this behavior to its key ingredients , which we expect to be more general than the model itself: an excess of regulators , nonlinearity , and self-activation . In this section , we examine their role: first in our model ( Figure 5 ) , and then in more realistic circuits , relaxing our simplifying assumptions ( Figure 6 ) . In Figure 5 , the parameter d on the horizontal axis is the strength of nonlinearity ( see Figure 2E ) , from perfectly linear at d=-∞ , to strongly nonlinear at d=0 . The vertical axis corresponds to an increasing number of regulators Na , which we label as in Figure 2D; for completeness , we also include the simplest system with a single regulator co-activating both x1 and x2 ( bottom row ) . Panel A examines the performance of our system as defined in Equation ( 4 ) , that is , with self-activation included . In panel B , we remove self-activation by omitting the prefactor aμ in front of the max function in Equation ( 4 ) . The color scheme is chosen so that red indicates an improvement , and blue a deterioration , of the circuit performance relative to the SEPI architecture , which approximately corresponds to the point highlighted in Figure 5B . The difference between the labeled point and the SEPI architecture is that all models represented in Figure 5 include a small degradation term , which becomes important in the nonlinear regime . For the SEPI-like case , its effect on performance is negligible ( see Appendix 1 , section 'Performance penalty from the degradation term' ) . Performance is averaged over five angles α; see Appendix 1 , section 'Parameters used in figures' . Unsurprisingly , the performance of the simple SEPI-like architecture can be improved by adding extra regulators ( pink region in Figure 5B ) : each new regulator allows the system to respond more quickly in a yet another direction of perturbation , with which it is ‘aligned’ . However , such a strategy would have limited utility in a biological setting , since the marginal improvement per regulator must offset the cost of added complexity . The mechanism described here corresponds to the red area in Figure 5A . Importantly , in the presence of both nonlinearity and self-activation , even a single extra regulator ( Na=3 ) can already provide a significant benefit . Figure 5A shows that in the context of our model , the reported behavior requires Na to exceed Nx , and d to be sufficiently large . However , these ingredients are more general than the specific implementation in Equation ( 4 ) . In our model , additional regulators were required because they supplied the slow degrees of freedom to serve as memory; such degrees of freedom could be implemented in other ways , for example , as phosphorylation or methylation ( Barkai and Leibler , 1997 ) . Similarly , while nonlinearity is essential ( linear dynamics cannot couple to higher-order terms , such as fluctuation magnitude ) , its exact functional form may be changed while retaining the learning behavior ( see Appendix 1 , section 'Nonlinearity acts as a sensor of fluctuation variance' ) . Finally , the explicitly self-catalytic behavior of aμ in our model is only one possible strategy for translating the stored memory into a faster response . To demonstrate the generality of these ingredients , we constructed two circuits with very different architectures ( Figure 6A , B ) , both reproducing the results of Figure 3C–F . These are not the only ways that the logic described above can be implemented; rather , our intention is to show that as long as we keep the key elements , we can relax our simplifying assumptions , such as the form of the nonlinearity and self-activation , while still retaining the ability to learn . The first of these proposed circuits ( Figure 6A ) is based on a pair of allosteric enzymes with the toy nonlinearity of Figure 2E replaced by more realistic cooperative binding , and implements dynamics otherwise very similar to those shown above . In this circuit , the enzymes E+ and E- can be in an active or inactive state: The active form of E+ , which we denote E+* , catalyzes the production of x; similarly , E-* catalyzes degradation of x . In addition , the active enzymes can bind to molecules of the metabolite x to control the self-catalytic activity . The total concentration of E+* , bound and unbound , then plays the role of the activating regulator a+ from above ( a+=[E+*]+[xE+*] ) , while E-* plays the role of the inhibitor a- ( a-=[E-*]+[xE-*] ) . The equations defining the dynamics are then: ( 5 ) {τxx˙=γ+a+−γ−xa−−xD ( t ) , τaa˙+=a+c+nc+n+xn−a+κ+ , τaa˙−=a−xmc−m+xm−a−κ− . Despite the extensive changes relative to Figure 3A , the system is still able to learn . Figure 6C compares its performance to a non-learning version with only the activating branch a+ , which is analogous to the single-activator SEPI architecture ( compare to Figure 3F ) . For a detailed discussion of this more biologically plausible model , see Appendix 1 , section 'A pair of allosteric enzymes' . Our other proposed circuit ( Figure 6B ) differs significantly . Here , instead of seeking to match P to D , the system maintains the homeostasis of a concentration x perturbed by external factors . In this implementation , the production and degradation of x are both catalyzed by a single bifunctional enzyme; the responsiveness of this circuit scales with the overall expression of the enzyme E , and larger fluctuations of x lead to upregulation of E due to the nonlinearity , as before . ( For a detailed discussion , see e Appendix 1 , section 'An architecture based on a bifunctional enzyme' . ) Defining A=a++a-=[E]+[xE] as the total concentration of the enzyme E in both its bound and unbound states , the bound and unbound fractions are described by Hill equations , ( 6 ) a+=Acmxm+cm , a-=A-a+ . The dynamics of our system are: ( 7 ) {τxx˙=P0+γ+a+−γ−xa−−xD ( t ) τAA˙=−Aκ+f ( x ) . Despite its compactness , this circuit is also able to learn ( Figure 6D; compare to Figures 3F and 6C ) . Interestingly , this particular logic is very similar to a small modification of the standard two-component signaling architecture ( Figure 6E ) . In this architecture , the signal s determines the concentration of the phosphorylated form YP of the response regulator Y; the rapidity of the response is determined by the expression of the histidine kinase X , present at a much lower copy number . Although the signaling architecture of Figure 6E , at least in some parameter regimes , is known to be robust to the overall concentrations of X and Y ( Batchelor and Goulian , 2003 ) , this robustness property applies only to the steady-state mapping from s to YP , not the kinetics . Thus , much like in Figure 6B , a nonlinear activation of X by YP ( known as autoregulation [Goulian , 2010] or autoamplification [Hoffer et al . , 2001] , and shown as a dashed arrow in Figure 6E ) would endow this signaling system with self-tuned reactivity that learns the statistics of the input .
In this paper , we have studied a regulatory architecture which is able to infer higher-order statistics from fluctuating environments and use this information to inform behavior . For concreteness , we phrased the regulatory task as seeking to match the production P→ of one or two metabolites to a rapidly fluctuating demand D→ . Alternatively , and perhaps more generally , the circuits we constructed can be seen as maintaining the homeostasis in a quantity x→ that is continually perturbed by external factors . We demonstrated that a simple architecture was capable of learning the statistics of fluctuations of its inputs and successfully using this information to optimize its performance . We considered one-dimensional and two-dimensional examples of such behavior . In one dimension , learning the statistics of the input meant our circuit exhibited a self-tuned reactivity , learning to become more responsive during periods of larger fluctuations . Importantly , we have shown that this behavior can be achieved by circuits that are highly similar to known motifs , such as feedback inhibition ( Figure 2A–C ) or two-component signaling ( Figure 6B , E ) . The latter connection is especially interesting: There are at least a few examples of two-component systems where autoamplification , a necessary ingredient for the learning behavior discussed here , has been reported ( Shin et al . , 2006; Williams and Cotter , 2007 ) . Moreover , in the case of the PhoR/PhoB two-component system in E . coli , such autoamplification has been experimentally observed to allow cells to retain memory of a previously experienced signal ( phosphate limitation; Hoffer et al . , 2001 ) , a behavior the authors described as learning-like . As reported , this behavior constitutes a response to the signal mean and is similar to other examples of simple phenotypic memory ( e . g . Lambert et al . , 2014 ) ; however , our analysis demonstrates that a similar architecture may also be able to learn more complex features . Such a capability would be most useful in contexts where the timescale of sensing could plausibly be the performance bottleneck . Since transcriptional processes are generally slower than the two-component kinetics , we expect our discussion to be more relevant for two-component systems with non-transcriptional readout , such as those involved in chemotaxis or efflux pump regulation . In the two-dimensional case , our simple circuit was able to learn and unlearn transient correlation structures of its inputs , storing this information in expression levels of different regulators . Our argument was a proof of principle that , for example , the gut bacteria could have the means to not only sense , but also predict nutrient availability based on correlations learned from the past , including correlations that change over faster-than-evolutionary timescales , such as the life cycle ( or dietary preferences ) of the host . Importantly , we showed that this ability could come cheaply , requiring only a few ingredients beyond simple end-product inhibition . The mechanism described here could suggest new hypotheses for the functional role of systems with an excess of regulators , as well as new hypotheses for bacterial function in environments with changing structure .
All simulations performed in Python 3 . 7 . 4 . Simulation scripts reproducing all figures are included as Source code 1 . | Associations inferred from previous experience can help an organism predict what might happen the next time it faces a similar situation . For example , it could anticipate the presence of certain resources based on a correlated environmental cue . The complex neural circuitry of the brain allows such associations to be learned and unlearned quickly , certainly within the lifetime of an animal . In contrast , the sub-cellular regulatory circuits of bacteria are only capable of very simple information processing . Thus , in bacteria , the ‘learning’ of environmental patterns is believed to mostly occur by evolutionary mechanisms , over many generations . Landmann et al . used computer simulations and a simple theoretical model to show that bacteria need not be limited by the slow speed of evolutionary trial and error . A basic regulatory circuit could , theoretically , allow a bacterium to learn subtle relationships between environmental factors within its lifetime . The essential components for this simulation can all be found in bacteria – including a large number of ‘regulators’ , the molecules that control the rate of biochemical processes . And indeed , some organisms often have more of these biological actors than appears to be necessary . The results of Landmann et al . provide new hypothesis for how such seemingly ‘superfluous’ elements might actually be useful . Knowing that a learning process is theoretically possible , experimental biologists could now test if it appears in nature . Placing bacteria in more realistic , fluctuating conditions instead of a typical stable laboratory environment could demonstrate the role of the extra regulators in helping the microorganisms to adapt by ‘learning’ . | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"physics",
"of",
"living",
"systems",
"computational",
"and",
"systems",
"biology"
] | 2021 | A simple regulatory architecture allows learning the statistical structure of a changing environment |
Newly born cells either continue to proliferate or exit the cell division cycle . This decision involves delaying expression of Cyclin E that promotes DNA replication . ORC1 , the Origin Recognition Complex ( ORC ) large subunit , is inherited into newly born cells after it binds to condensing chromosomes during the preceding mitosis . We demonstrate that ORC1 represses Cyclin E gene ( CCNE1 ) transcription , an E2F1 activated gene that is also repressed by the Retinoblastoma ( RB ) protein . ORC1 binds to RB , the histone methyltransferase SUV39H1 and to its repressive histone H3K9me3 mark . ORC1 cooperates with SUV39H1 and RB protein to repress E2F1-dependent CCNE1 transcription . In contrast , the ORC1-related replication protein CDC6 binds Cyclin E-CDK2 kinase and in a feedback loop removes RB from ORC1 , thereby hyper-activating CCNE1 transcription . The opposing effects of ORC1 and CDC6 in controlling the level of Cyclin E ensures genome stability and a mechanism for linking directly DNA replication and cell division commitment .
In addition to its role in the initiation of DNA replication , ORC1 , the largest subunit of the Origin Recognition Complex ( ORC ) controls Cyclin E-dependent duplication of centrosomes and centrioles in cells by acting as an inhibitor of Cyclin E-CDK2 activity ( Hemerly et al . , 2009; Hossain and Stillman , 2012 ) . The Cyclin E-CDK2 kinase inhibitory activity is compromised by ORC1 Meier-Gorlin syndrome mutations that also alter the interaction between ORC1 and histone H4K20me2 ( Hossain and Stillman , 2012; Kuo et al . , 2012; Zhang et al . , 2015; Bicknell et al . , 2011b; Bicknell et al . , 2011a; de Munnik et al . , 2012 ) . Unlike the well-characterized yeast complex that is a stable , six subunit complex throughout the cell division cycle , ORC in human cells is a very dynamic complex ( Sasaki and Gilbert , 2007; DePamphilis , 2005 ) . ORC1 binds to mitotic chromosomes as cells enter into mitosis ( Kara et al . , 2015; Okuno et al . , 2001 ) , and in human cells , it is modified by ubiquitin and then degraded during the G1 to S phase transition ( Abdurashidova et al . , 2003; Kara et al . , 2015; Kreitz et al . , 2001; Mendez et al . , 2002; Ohta et al . , 2003; Siddiqui and Stillman , 2007; Tatsumi et al . , 2000 ) . The assembly of the full ORC occurs in mid G1 phase of the cell division cycle in preparation for its role in assembly of the pre-replicative complex ( pre-RC ) at sites across chromosomes ( Kara et al . , 2015; Siddiqui and Stillman , 2007 ) . The ORC1-related protein CDC6 is also required for pre-RC assembly , but it is targeted for proteasome degradation by the SCFCyclin F ubiquitin ligase complex late in the cell cycle and the anaphase-promoting complex/cyclosome ( APC/C ) in early G1 phase and then stabilized in mid G1 phase by Cyclin E-CDK2-mediated phosphorylation ( Mailand and Diffley , 2005; Petersen et al . , 2000; Walter et al . , 2016 ) . This phosphorylation is mediated by the direct interaction between Cyclin E and CDC6 and CDC6 and Cyclin E-CDK2 cooperate to promote the initiation of DNA replication ( Coverley et al . , 2002; Furstenthal et al . , 2001; Cook et al . , 2002 ) . As proliferating cells divide , they must make a decision whether to continue to proliferate or enter into proliferative quiescence . This decision is made by a complex regulatory process known as START in yeast and the restriction point in mammalian cells ( Johnson and Skotheim , 2013 ) . Key among these regulators are the Cyclin D-CDK4/6 kinases that mono-phosphorylate the retinoblastoma ( RB ) protein and contributes to the release of repression of E2F-transcription factors ( Narasimha et al . , 2014; Ewen et al . , 1993; Hinds et al . , 1992; Lundberg and Weinberg , 1998; Resnitzky et al . , 1994 ) . E2F1-regulated genes include genes encoding Cyclin E ( CCNE1 ) and CDC6 ( Hateboer et al . , 1998; Ohtani et al . , 1998; Yan et al . , 1998; DeGregori et al . , 1995 ) . Cyclin E-CDK2 amplifies the phosphorylation of RB , but how this is achieved is not known ( Narasimha et al . , 2014 ) . Here , we demonstrate that ORC1 is required for repression of the gene encoding Cyclin E and that CDC6 is involved in relieving the repression in cooperation with Cyclin E-CDK2 . We suggest that ORC1 establishes a period in the newly born cells during which Cyclin E is not expressed , allowing time for the cells to decide whether to proliferate again or enter into replicative quiescence .
Apart from its role in DNA replication , human ORC1 controls centriole and centrosome copy number by binding and inhibiting the kinase activity of Cyclin E-CDK2 ( Hemerly et al . , 2009 ) . That work suggested that ORC1 might also control Cyclin E by regulating its protein level during the G1 phase of the cell division cycle . To determine if this was the case , ORC1 was depleted using siRNA in U2OS cells that had been synchronized in mitosis by nocodazole treatment and released into the next cell cycle . As early as 9 hr post release , Cyclin E protein level was elevated in ORC1-depleted U2OS cells , compared to control siRNA treated cells ( Figure 1A ) . The expression of CCNE1 mRNA increased at all times following ORC1 depletion in synchronized U2OS cells ( Figure 1B ) and quantitation of multiple experiments showed significant increases from 6–24 hr post release ( Figure 1—figure supplement 1A–D ) . This data suggests that ORC1 inhibits CCNE1 gene expression . 10 . 7554/eLife . 12785 . 003Figure 1 . ORC1 represses Cyclin E gene expression and interacts with RB . ( A–B ) Nocodazole arrested U2OS cells were transfected with control or ORC1 siRNA then released into the next cycle . ( A ) protein levels were estimated by immunoblotting with antibodies against ORC1 , Cyclin E , Cyclin A and α-Tubulin . Low and high indicates different exposures . ( B ) mRNA levels of ORC1 , Cyclin E ( CCNE1 ) , Cyclin A ( CCNA2 ) and GAPDH . Quantitation of mRNA levels from multiple experiments is shown in Figure 1—figure supplement 1 . ( C–E ) Interaction between ORC1 and RB or its pocket mutants . GFP , GFP-tagged wild-type or mutant RB were co-transfected in HEK293 cells with either ORC1-Flag or empty vector . Immunoprecipitation with anti-Flag antibody ( C ) or GFP antibody ( D ) from cell lysates followed by immunoblotting with the indicated antibodies . ( E ) Cell lysate from HEK293 cells overexpressing GFP-tagged wild type or mutant RB and ORC1-Flag were immunoprecipitated with normal rabbit serum ( NRS ) or ORC2 or ORC3 antibodies , immunoblotted with the indicated antibodies . Binding of ORC1 to wild-type and RB mutants and the effect of Cyclin E-CDK2 is shown in Figure 1—figure supplement 2 . GFP , Green fluorescent protein; ORC , Origin Recognition Complex; RB , Retinoblastoma . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00310 . 7554/eLife . 12785 . 004Figure 1—figure supplement 1 . ORC1 represses Cyclin E gene expression and interacts with RB . CCNE1 ( Cyclin E ) gene transcription is up-regulated upon ORC1 depletion . ( A–D ) Quantitative PCR analysis of ORC1 , CCNE1 ( Cyclin E ) , CCNA2 ( Cyclin A ) and GAPDH transcript levels in U2OS cells transfected with either control siRNA targeting GFP ( blue bar ) or ORC1 siRNA ( grey bar ) . The siRNA treated cells were released after nocodazole arrest and mRNA levels estimated at different time points as indicated in hours . The values shown are average fold change ( mean±SEM ) from three independent experiment normalized to β-actin transcripts . Statistical analysis was performed using the Student’s t test . *p<0 . 01; **p<0 . 001; ***p<0 . 0001; NS , not significant . ORC , Origin Recognition Complex; RB , Retinoblastoma; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00410 . 7554/eLife . 12785 . 005Figure 1—figure supplement 2 . ORC1 represses Cyclin E gene expression and interacts with RB . Cyclin E-CDK2 phosphorylation controls ORC1 and RB interaction . ( A ) Interaction between purified RB and ORC1 in a MBP pull down assay . MBP-fused wild-type RB was bound to amylose resin and further incubated with in vitro translated , S35-labeled wild type ORC1 or its mutants in the presence or absence of Cyclin E-CDK2 and 1 mM ATP . Beads were isolated and bound proteins were separated by gel electrophoresis . MBP was used as a control in the assay . ( B ) Alignment of ORC1 sequences is shown with conserved LxCxE motif . The conserved residues of LxCxE motif are indicated with different colors . The alignment shows conserved residues in ORC1 from different species in vertebrate and invertebrate classes ( Invertebrates: Brugia malayi , Caenorhabditis briggsae , Caenorhabditis elegans , Strongylocentrotus purpuratus , Culex quinquefasciatus , Apis mellifera , Drosophila melanogaster , Aedes aegypti , and Pediculus humanus; Vertebrates: Danio rerio , Xenopus laevis , Xenopus tropicalis , Gallus gallus , Taeniopygia guttata , Mus musculus and Homo sapiens ) . ( C ) Interaction between ORC1 and RB in a MBP pull-down assay . GST-fused wild-type RB or its mutant proteins were incubated with wild type MBP-ORC1 in the presence or absence of Cyclin E-CDK2 and/or 1 mM ATP . Amylose-bead-bound proteins were isolated and bound proteins were separated by gel electrophoresis followed by immunoblotting with anti-GST antibody . Recombinant MBP was used as a control in the assay . ORC , Origin Recognition Complex; RB , Retinoblastoma; GST , Glutathione S transferase; MBP , Maltose binding protein . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 005 Transcription of the gene encoding Cyclin E ( CCNE1 ) is known to be regulated by the E2F1 transcription factor and to be repressed by RB protein ( Nielsen et al . , 2001; Geng et al . , 1996; DeGregori et al . , 1995 ) , a member of the so-called 'pocket protein' family ( Dick and Rubin , 2013; Dyson , 1998; Giacinti and Giordano , 2006 ) . Since ORC1 binds RB ( Mendoza-Maldonado et al . , 2010 ) , we explored a role for ORC1 in transcriptional repression of the CCNE1 gene . RB was expressed as a fusion to the maltose binding protein ( MBP ) and it bound to S35-labelled ORC1 protein ( Figure 1—figure supplement 2A ) . RB binds other binding partners dependent on a canonical LxCxE motif and although ORC1 has a conserved LPCRD/E sequence , it was not required for the interaction between ORC1 and RB ( Figure 1—figure supplement 2A and B ) . Consistent with this finding , two pocket mutants within RB ( R661W and N757F ) that are defective in binding to LxCxE containing proteins ( Chen and Wang , 2000 ) showed no defect in binding to ORC1 in vitro ( Figure 1—figure supplement 2C ) or in vivo ( Figure 1C ) . In fact , the mutant RB proteins bound ORC1 better than the wild type , perhaps due to loss of competition between ORC1 and other RB binding proteins or due to conformational changes in RB . When Green Fluorescent Protein ( GFP ) -RB fusion protein was expressed in cells with ORC1-Flag-tagged protein , the ORC1-Flag protein bound WT and pocket mutant RB , but other ORC subunits did not bind to RB ( Figure 1D ) . We confirmed this observation by showing that immunoprecipitation with anti-ORC2 or anti-ORC3 antibodies failed to precipitate GFP-RB or its mutants , but did bind ORC1-Flag ( Figure 1E ) . Given that the ORC1-RB interaction is independent of other ORC subunits , it suggests that ORC1 has additional functions that are separate from its role in DNA replication . The phosphorylation of RB is an important step that relieves repression of E2F target genes ( Rubin , 2013 ) . In the presence of purified Cyclin E-CDK2 kinase plus ATP RB no longer bound to ORC1 ( Figure 1—figure supplement 2A and C ) , suggesting that Cyclin E might feedback and relieve repression of the CCNE1 gene by disrupting the RB-ORC1 interaction . RB interacts with chromatin and histone-modifying enzymes to repress E2F1 transcription activity ( Dick and Rubin , 2013 ) . Specifically , it has been suggested that RB binds to the SUV39H1 histone methyltransferase that tri-methylates histone H3K9 and then HP1α binds to histone H3K9me3 and that these direct interactions contribute to repression of CCNE1 transcription ( Nielsen et al . , 2001 ) . It is important to note , however , that others suggested that another protein mediates the RB-SUV39H1 interaction ( Vandel et al . , 2001 ) . Since ORC1 interacts with both RB and HP1α ( Prasanth et al . , 2010 ) , we tested whether ORC1 also interacts with SUV39H1 . Using purified proteins we found that both RB and ORC1 directly bound to SUV39H1 , and ORC1 directly bound to CDC6 ( Saha et al . , 1998 ) ( Figure 2A and Figure 2—figure supplement 1 ) . Under these conditions , ORC1 bound to HP1α weakly , but the binding increased when higher levels of GST-HP1α were added , consistent with published data ( Prasanth et al . , 2010 ) . Although both RB and ORC1 directly interacted with SUV39H1 , ORC1 bound SUV39H1 much better than RB ( Figure 2—figure supplement 2A–D ) . Consistent with the in vitro data , we observed that anti-SUV39H1 antibodies precipitated ORC1 and RB in the RB-positive cells ( U2OS and MCF7 ) , with ORC1 being the predominant interacting protein in MCF7 cells ( Figure 2B ) . Furthermore , the interaction between SUV39H1 with ORC1 protein did not require RB because anti-ORC1 antibodies co-precipitated SUV39H1 from RB-negative SaOS-2 cells ( Figure 2C ) . This observation was confirmed when GFP-RB or GFP-ORC1 were transiently expressed with T7-SUV39H1 in 293 cells; ORC1 more readily bound to SUV39H1 ( Figure 2D ) , whereas higher levels of GFP-RB were required to observe an interaction with SUV39H1 ( Figure 2—figure supplement 3A ) . Domain mapping demonstrated that SUV39H1 interacted with ORC1 through its SET domain containing C-terminus , which is required for its histone methyltransferase ( HMT ) activity . Moreover , SUV39H1 interacted with ORC1 and ORC1 mutants that could not be phosphorylated by Cyclin-CDK and did not interact in a complex with other ORC subunits , similar to the ORC1-RB interaction ( Figure 2—figure supplement 3B–C ) . These data suggest that the ORC1-SUV39H1 interaction does not require other ORC subunits or CDK phosphorylation of ORC1 , and therefore ORC1 plays a role independent of its role in DNA replication . 10 . 7554/eLife . 12785 . 006Figure 2 . ORC1 binds SUV39H1 to control Cyclin E gene transcription . ( A ) Purified MBP-ORC1 and various GST-fused proteins were mixed and proteins bound in a GST-pull down were detected by immunoblotting with anti-MBP antibodies . The purified proteins are shown in Figure 2—figure supplement 1 . ( B ) U2OS and MCF7 cell lysates were immunoprecipitated with SUV39H1 antibody and immunoblotted with the indicated antibodies . Rabbit IgG served as control antibody . Asterisk indicates the cross-reacting antibody band; arrow indicates the SUV39H1 protein . ( C ) Immunoprecipitation from RB-negative SaOS-2 cell lysates with ORC1 antibody or IgG and immunoblotted with antibodies against ORC1 , SUV39H1 or ORC3 . ( D ) HEK293 cells were transiently co-transfected with GFP , GFP-ORC1 or GFP-RB plus T7-SUV39H1 plasmids ( 2 . 5 μg each ) . GFP antibody immunoprecipitates were immunoblotted with the indicated antibodies . The interaction between ORC1 and SUV39H1 and between RB and SUV39H1 is shown with purified proteins and quantitated in Figure 2—figure supplement 2 . Higher levels of RB are required to demonstrate an interaction with SUV39H1 in vivo and ORC1 interacts with the SET domain of SUV39H1 , Figure 2—figure supplement 3 . ( E ) MBP-ORC1 was incubated with bead-bound histone peptides with or without the indicated modifications and bound MBP-ORC1 was observed by immunoblotting with anti-MBP antibody ( lower box ) or silver staining ( upper box ) . ( F–G ) Wild-type CCNE1-luciferase reporter assay in U2OS cells . U2OS cells were transiently co-transfected with 500 ng of 10–4 CCNE1 promoter , 50 ng E2F1 , 50 ng DP1 and 20 ng pCMV-LacZ plasmids along with the indicated amounts ORC1 and/or SUV39H1 plasmids . ( F ) Increasing amounts of ORC1-Flag or T7-SUV39H1 repress Cyclin E gene promoter . Experiments were carried out in triplicate . Expression of proteins was confirmed by Immunoblot; α-Tubulin as loading control . Statistical analysis was performed using the Student’s t test . *p<0 . 05; **p<0 . 005; ***p<0 . 001 . ( G ) ORC1-Flag cooperates with wild type but not mutant SUV39H1 to repress CCNE1 gene expression . The experiments were carried out in triplicate . Expression of proteins was confirmed by Western blots . α-Tubulin served as a control for equal loading of each sample . Statistical analysis was performed using the Student’s t test . *p<0 . 01; **p<0 . 005; ***p<0 . 0001 . Repression of transcription by ORC1 and SUV39H1 was also demonstrated using an artificial promoter and tethering the proteins via the GAL4 DNA binding domain in Figure 2—figure supplement 4 . ORC , Origin Recognition Complex; MBP , Maltose binding protein; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00610 . 7554/eLife . 12785 . 007Figure 2—figure supplement 1 . ORC1 binds SUV39H1 to control Cyclin E gene transcription . Bacteria expressed and purified recombinant proteins . Silver stain of purified GST , GST-RB , GST-HDAC1 , GST-SUV39H1 , GST-HP1α , GST-CDC6 and MBP-ORC1 proteins . MW stands for protein molecular weight marker in kilodalton . ORC , Origin recogntion complex; MBP , Maltose binding protein; RB , Retinoblastoma; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00710 . 7554/eLife . 12785 . 008Figure 2—figure supplement 2 . ORC1 binds SUV39H1 to control Cyclin E gene transcription . SUV39H1 interaction with ORC1 and RB . ( A ) Interaction between purified MBP-ORC1 or MBP-RB with GST-SUV39H1 in a MBP pull down assay . MBP-fused proteins were bound to amylose resin and incubated with GST-SUV39H1 . Bound proteins were separated by gel electrophoresis followed by immunoblotting with either anti-GST or anti-SUV39H1 antibodies . Recombinant MBP was used as a control . ( B , C ) GST-SUV39H1 protein was bound to the resin and incubated with either MBP-ORC1 or MBP-RB proteins . Bead-bound proteins were immunoblotted with anti-MBP antibody . Recombinant GST was used as a control . ( D ) Concentration-dependent interaction with increasing levels of GST-SUV39H1 and either 100 nM of MBP or MBP-ORC1 or MBP-RB followed by pull down with amylose beads and subsequently immunoblotted with anti-GST antibody . Bands were quantified and represented in a graph after background subtraction with MBP control protein . ORC , Origin recogntion complex; MBP , Maltose binding protein; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00810 . 7554/eLife . 12785 . 009Figure 2—figure supplement 3 . ORC1 binds SUV39H1 to control Cyclin E gene transcription . ORC1 interaction with SET domain of SUV39H1 does not involve other ORC subunits . ( A ) HEK293 cells were transiently co-transfected with GFP , GFP-ORC1 or GFP-RB and T7-SUV39H1-expressing plasmids at the indicated amounts in micrograms . The whole cell lysate prepared from HEK293 cells expressing the indicated constructs were immunoprecipitated with GFP antibody followed by immunoblotting with specific antibodies . ( B ) GFP-tagged wild-type ORC1 or its mutants ( A-A: [ORC235ARA237'Cy' motif mutant] or ORC1S258A , S273A , T375A [CDK: with mutants in CDK target sites] ) were co-transfected in HEK293 cells with either Flag-SUV39H1 or its empty vector . Immunoprecipitation with anti-Flag antibody from cell lysates of HEK293 cells overexpressing the indicated constructs followed by immunoblotting with the indicated antibodies . ( C ) Schematic showing domains of human SUV39H1 protein . In the GST-pull down assay , GST-SUV39H1 or its truncation mutant proteins were incubated with MBP-ORC1 protein as indicated and immunoblotted with anti-MBP antibody . GST protein served as negative control . ORC , Origin recogntion complex; MBP , Maltose binding protein; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 00910 . 7554/eLife . 12785 . 010Figure 2—figure supplement 4 . ORC1 binds SUV39H1 to control Cyclin E gene transcription . ORC1 , but not ORC3 or ORC4 can repress gene transcription . ( A ) The U2OS cells were transfected with a Gal4-driven luciferase reporter as shown in the schematic with increasing amounts of Gal4DBD-ORC1 or Gal4DBD-SUV39H1 together with pCMV-LacZ plasmids . Relative luciferase activity was determined and normalized to lacZ activity . Experiments were carried out in triplicate . The whole cell extract was immunoblotted with anti-Gal4 antibody for expression of Gal4DBD fusion plasmids . α-Tubulin served as a loading control . Statistical analysis was performed using the Student’s t test . **p<0 . 01; ***p<0 . 005 . ( B ) Wild-type CCNE1 promoter-luciferase reporter assay in U2OS cells . The U2OS cells were transiently co-transfected with 500ng of the 10–4 CCNE1 promoter , 50 ng E2F1 , 50 ng DP1 and 20 ng pCMV-LacZ plasmids along with the indicated amounts ORC3-Flag or ORC4-Flag plasmids . The increasing amounts of ORC3-Flag or ORC4-Flag do not repress the CCNE1 gene promoter as indicated by relative light units ( RLU ) normalized to β-galactosidase activity . Experiments were carried out in triplicate . Expression of proteins was confirmed by Western blot . α-Tubulin served as a loading control . ORC , Origin Recognition Complex . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 010 Since SUV39H1 tri-methylates histone H3K9 after its de-acetylation by HDAC1 , a prerequisite step for establishing CCNE1 gene repression , we explored the ORC1 protein interactions with different histone H3 modifications ( Nicolas et al . , 2003; Stewart et al . , 2005; Vaute et al . , 2002 ) . Purified ORC1 bound to unmodified or mono- , di- or tri-methylated H3K9 or H3K14-tri-methyl modifications , while the interaction was abolished when histone H3 was acetylated at the same positions or when H3 serine-10 was phosphorylated , which normally occurs during mitosis ( Figure 2E ) . Thus , ORC1 can bind to RB , SUV39H1 and to the repressive H3K9-me3 modification on histone H3 , suggesting that it might mediate repression of E2F1-dependent CCNE1 transcription . To test if ORC1 has a role in repression of the E2F1-regulated CCNE1 gene , the CCNE1 promoter was linked to the luciferase-coding region to create a reporter for CCNE1 gene transcription ( Geng et al . , 1996 ) . Transfection of plasmids expressing E2F1 and its binding partner DP1 activated gene expression , whereas expression of ORC1 or SUV39H1 protein repressed CCNE1 expression in a dose-dependent manner ( Figure 2F ) . In a GAL4-based reporter gene assay , expression of either GAL4-ORC1 or GAL4-SUV39H1 also repressed transcription of the reporter gene in a dose-dependent manner ( Figure 2—figure supplement 4A ) . In contrast , neither expression of ORC3 nor ORC4 altered E2F1-driven CCNE1 transcription ( Figure 2—figure supplement 4B ) . Moreover , SUV39H1 co-operated with ORC1 to further repress the CCNE1 promoter , but a mutant of SUV39H1 ( H324K ) ( Li et al . , 2002; Rea et al . , 2000; Stewart et al . , 2005 ) that has lost its catalytic activity was unable to do so ( Figure 2G ) . We therefore conclude that ORC1-SUV39H1 co-operation for transcriptional repression of the CCNE1 gene is mediated through the HMT activity of SUV39H1 . Having established that ORC1 can repress CCNE1 gene transcription in cooperation with the methyltransferase activity of SUV39H1 , we investigated the positioning of ORC1 , RB and SUV39H1 proteins within the CCNE1 promoter in vivo . A previous publication ( Dellino et al . , 2013 ) reported ORC1 chromatin immuno-precipitation following crosslinking ( ChIP ) and re-analysis this whole genome Chip-Seq data revealed a weak ORC1 peak ( peak height of 7 reads ) within the CCNE1 promoter; however , a duplicate was not reported . Therefore , to test whether ORC1 associated with the CCNE1 promoter , we modified a chromatin immunoprecipitation ( ChIP ) method for ORC1 initially using asynchronously growing MCF7 cells and the immunoprecipitated DNA was analyzed by polymerase chain reaction ( PCR ) using multiple primer pairs across the CCNE1 promoter ( Figure 3A ) . ORC1 bound to the CCNE1 promoter encompassing the region from -280 to +63 base pairs ( probes E and F ) , a region that is known to bind the E2F1 transcription factor ( see Gene Expression Omnibus [GEO] transcription factor binding site accession numbers GSM935484 and GSM935477 ) and contains five E2F1 consensus binding sites ( red bars , Figure 3A ) . Our ChIP analysis also showed that RB bound to the same regions bound by ORC1 , but not to other probes ( Figure 3B ) . When a similar ChIP analysis was performed using antibodies targeted to SUV39H1 , histone H3K9me3 and CDC6 , these proteins bound to the same region ( -342 to +63 ) of the CCNE1 promoter ( Figure 3C , also see Figure 4 below ) . 10 . 7554/eLife . 12785 . 011Figure 3 . Binding of ORC1 , RB , SUV39H1 and CDC6 proteins to the CCNE1 promoter . ( A ) Schematic of the CCNE1 promoter and the regions amplified with different primer pairs used for ChIP assay were indicated as follows: A ( −1462 to −1318 ) ; B ( −683 to −575 ) ; C ( −456 to −357 ) ; D ( −342 to −195 ) ; E ( −280 to −143 ) ; F ( −159 to +63 ) ; G ( +350 to +490 ) ; H ( +816 to +944 ) . The red bars indicate five E2F1 consensus sites . The truncated box indicates the first exon of the CCNE1 gene . ( B–C ) The occupancy of ORC1 , RB , SUV39H1 and CDC6 proteins was analyzed by chromatin immunoprecipitation at the CCNE1 promoter in asynchronous growing MCF7 cells . ORC1 and RB are mouse antibodies , while SUV39H1 and CDC6 are rabbit antibodies . In the marker lanes , the two bands are 100 and 200 base pairs . The experiments were done in triplicate and two of these experiments are shown . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 01110 . 7554/eLife . 12785 . 012Figure 4 . Dynamic association of ORC1 , RB , SUV39H1 and CDC6 proteins to the CCNE1 promoter during the cell cycle . ( A–B ) Nocodazole arrested U2OS cells were released for different times ( 3 , 6 and 9 hr ) and analyzed for occupancy of ORC1 , RB , SUV39H1 and CDC6 proteins at the CCNE1 promoter by ChIP assay . The primer pairs used to analyze two different regions of the CCNE1 promoter are indicated . The experiments were done in triplicate with results similar to those shown . ( C ) Whole cell protein levels of nocodazole arrested and released U2OS cells at different time points ( as indicated in hours ) by immunoblotting with antibodies against ORC1 , RB , SUV39H1 , CDC6 , Cyclin E , ORC3 and α-Tubulin . ORC , Origin Recognition Complex; RB , Retinoblastoma . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 012 We next studied the temporal dynamics of protein binding to the CCNE1 promoter during the G1 phase of the cell cycle in synchronized U2OS cells . In this analysis , cells were blocked in mitosis with nocodazole and released into the next G1 phase and ChIP analyses for ORC1 , CDC6 , RB and SUV39H1 were performed using primer pairs B and E ( Figures 4A and B ) . ORC1 and RB bound to the probe E region , but not the probe B region , at 3 hr post mitosis , and then binding was reduced at 6 hr and eliminated by 9 hr ( Figure 4A ) , even though ORC1 protein remained in the cell ( Figure 4C ) . SUV39H1 was detected at 3 and 6 hr , but not at 9 hr . Interestingly , CDC6 transiently bound to the promoter only at 6 hr ( Figure 4B ) , precisely the time when Cyclin E proteins levels dramatically increase ( Figure 4C ) . In the following section , we investigated the physiological role of CDC6 binding to CCNE1 promoter . Cyclin E-CDK2 cooperates with CDC6 to stimulate entry from G1 phase into S-phase of the mammalian cell division cycle ( Cook et al . , 2002; Coverley et al . , 2002; Hateboer et al . , 1998 ) . Since CDC6 protein levels increase during late G1 phase ( Hateboer et al . , 1998; Mendez and Stillman , 2000 ) and CDC6 binds ORC1 ( Figure 2A ) , we hypothesized that CDC6 may help alleviate ORC1-mediated repression of the CCNE1 gene . CDC6 and Cyclin E-CDK2 were purified and increasing amounts were titrated into a mixture containing purified RB and ORC1 ( Figure 5—figure supplement 1 ) and the interaction between RB and ORC1 was monitored . MBP-ORC1 bound to CDC6 ( Figure 5A ) and increasing amounts of Cyclin E-CDK2 did not interfere with this interaction ( Figure 5B ) . Moreover , increased binding between CDC6 and ORC1 did not interfere with binding between ORC1 and RB ( Figure 5B ) . Cyclin E-CDK2 disrupted the interaction between ORC1 and RB ( Figure 5C and Figure 2—figure supplement 2A ) , but it took relatively high levels of Cyclin E-CDK2 . Importantly , the addition of CDC6 along with CyclinE-CDK2 cooperatively disrupted the interaction between ORC1 and RB , even at low concentrations of Cyclin E-CDK2 ( Figure 5C ) . As shown before with Xenopus proteins ( Furstenthal et al . , 2001 ) , human CDC6 bound to Cyclin E-CDK2 in a manner dependent on the CDC6 Cyclin 'Cy' binding motif CDC694RRL96 but not to the CDC694ARA96'Cy' mutant ( Figure 5D ) . 10 . 7554/eLife . 12785 . 013Figure 5 . CDC6 co-operates with Cyclin E-CDK2 to activate E2F1-dependent CCNE1 gene transcription . ( A–C ) Equimolar amounts of MBP-GFP-ORC1 and MBP-RB proteins were incubated with increasing amounts of GST-CDC6 and/or Cyclin E-CDK2 . MBP-GFP-ORC1 protein was immunoprecipitated with GFP antibody , then immunoblotted with the indicated antibodies . The purified proteins used in these experiments are shown in Figure 5—figure supplement 1 ( A ) MBP-GFP-ORC1 binds GST-CDC6 . MBP protein served as control . ( B ) The binding of MBP-GFP-ORC1 protein to either GST-CCD6 in the presence of Cyclin E-CDK2 ( left section ) or MBP-RB ( right section ) . ( C ) The binding of MBP-GFP-ORC1 to MBP-RB in the presence of increasing molar amounts of Cyclin E-CDK2 ( right section ) or both GST-CDC6 and Cyclin E-CDK2 ( left section ) . ( D ) GST-pull down assay using GST-CDC6 wild type or CDC694ARA96 mutant ( CDC6A-A ) with purified Cyclin E-CDK2 protein followed by Immunoblotting with Cyclin E antibody . GST protein served as control . ( E ) Nocodazole arrested U2OS cells were transfected with 500 ng of GFP , GFP-CDC6 wild type or CDC694ARA96 mutant ( CDC6A-A ) plasmids , then released into the next cell cycle . At indicated times , whole cell extracts were immunoblotted with specific antibodies against GFP and Cyclin E . α-Tubulin served as loading control . ( F ) , CCNE1 promoter-luciferase reporter assay in U2OS cells . Cells transiently co-transfected with 500 ng of 10–4 CCNE1 promoter , 50 ng E2F1 , 50 ng DP1 and 20 ng pCMV-LacZ plasmids together with increasing amounts GFP-CDC6 WT or CDC694ARA96 plasmids for 24 hr . Relative luciferase activity was normalized to co-transfected LacZ control . Experiments in triplicate . Protein expression determined by immunoblot; α-Tubulin as loading control . Statistical analysis was performed using the Student’s t test . *p<0 . 05; **p<0 . 001; ***p<0 . 0005 . GFP , Green fluorescent protein; MBP , Maltose binding protein; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 01310 . 7554/eLife . 12785 . 014Figure 5—figure supplement 1 . CDC6 co-operates with Cyclin E-CDK2 to activate E2F1-dependent CCNE1 gene transcription . Purified Proteins . Coomassie Brilliant Blue stained gel of purified MBP , MBP-GFP-ORC1 , MBP-RB and GST-CDC6 proteins . MW stands for protein molecular weight marker in kilodalton . MBP , Maltose binding protein; RB , Retinoblastoma; GST , Glutathione S transferase . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 014 Based upon the biochemical data , we hypothesized that CDC6 and Cyclin E-CDK2 cooperated to relieve ORC1-SUV39H1-RB-mediated repression of CCNE1 . If this is the case , then CDC6 overexpression should increase CCNE1 expression in vivo . To test this hypothesis , we expressed GFP-tagged CDC6 wild type or its 'Cy' mutant in nocodazole arrested U2OS cells and released them from the nocodazole block to estimate the level of endogenous Cyclin E protein at different times in G1 . Expression of wild type CDC6 but not its 'Cy' mutant led to an increase in the endogenous level of Cyclin E protein ( Figure 5E ) . In fact , the Cdc6 'Cy' mutant acted as a dominant negative to prevent normal Cyclin E expression . Consistent with these results , overexpression of CDC6 wild type but not its 'Cy' mutant further enhanced E2F1-DP1 activated transcription from the CCNE1 promoter ( Figure 5F ) . Based on these data , we suggest that CDC6 co-operates with Cyclin E-CDK2 kinase to abolish the interaction between RB and ORC1 , contributing to alleviating the repression imposed by RB on the CCNE1 gene , leading to CCNE1 gene transcription . Data presented so far suggests that ORC1-mediated CCNE1 transcriptional repression also requires SUV39H1 . We therefore transfected into RB+ U2OS cells a CCNE1-luciferase reporter plasmid with E2F1-DP1 and depleted ORC1 , SUV39H1 or CDC6 using siRNA ( Figure 6A ) . Depletion of either ORC1 or SUV39H1 with two different siRNAs led to significant increases in CCNE1 promoter activity above its basal , E2F1/DP1-dependent level . In contrast , depletion of CDC6 protein had no effect on the basal , E2F1/DP1-dependent promoter activity ( Figure 6A ) . We further investigated the binding to the CCNE1 promoter of SUV39H1 and the presence of its product , the histone H3K9me3 mark , upon depletion of ORC1 protein in asynchronously growing U2OS cells by ChIP assay . Upon depletion of ORC1 compared to control siRNA the methyltransferase activity of SUV39H1 was drastically reduced on the CCNE1 promoter as evident by a dramatic reduction in the histone H3K9m3 mark at the promoter , while the binding of SUV39H1 was only slightly reduced ( Figure 6B ) . The reduced binding of H3K9me3 was most evident within the region -280 to -143 of the CCNE1 promoter , where ORC1 binding was centered ( Figure 6B , Figure 6—figure supplement 1 ) . Depletion of ORC1 also resulted in reduced levels of the histone H3K9me3 mark as well as a slight reduction in the level of SUV39H1 protein ( Figure 6C ) . Our results support the hypothesis that ORC1 is involved in transcription repression by recruiting the SUV39H1 protein , which thereby creates the histone H3K9me3 transcriptional repressor mark on CCNE1 promoter . 10 . 7554/eLife . 12785 . 015Figure 6 . ORC1 depletion decreases association of SUV39H1 and H3K9me3 with the CCNE1 promoter and increases CCNE1 gene transcription . ( A ) CCNE1-luciferase reporter assay in U2OS cells . U2OS cells transiently co-transfected with 500 ng of 10–4 CCNE1 promoter , 50 ng E2F1 , 50 ng DP1 and 20 ng pCMV-LacZ . U2OS cells were also transiently transfected for 24 hr with two different siRNAs targeting either ORC1 , SUV39H1 or CDC6 . GFP siRNA was used as a control . Relative luciferase activity was normalized to co-transfected LacZ control . Depletion of proteins was confirmed by Immunoblot; α-Tubulin as loading control . Statistical analysis was performed using the Student’s t test . *p<0 . 005; **p<0 . 001; ***p<0 . 0005 . ( B ) The CCNE1 promoter was analyzed for SUV39H1 binding and the presence of the H3K9me3 mark by ChIP assay in U2OS cells treated with either control siRNA or ORC1 siRNA for 48 hr . The experiments were done in triplicate and one experiment is shown . ( C ) Immunoblot of protein levels following control and ORC1 siRNA treatment at different times post nocodazole release . α-Tubulin was used as loading control . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 01510 . 7554/eLife . 12785 . 016Figure 6—figure supplement 1 . The ChIP qPCR bands were quantified using ImageJ software to analyze the extent of binding of SUV39H1 and histone H3K9me3 to the CCNE1 promoter region ( -280 to -143 bp ) in ORC1 siRNA treated U2OS cells compared to control siRNA-treated cells . All values were normalized to the corresponding input DNA . Statistical analysis was performed using the Student’s t test . The p-values are indicated above the bars . The reduction in SUV39H1 was not significant due to one out of the three replicates showing little reduction in ORC1 siRNA-treated U2OS cells compared to the control . Binding of SUV39H1 to the CCNE1 promoter was also low compared to histone H3K9me3 in asynchronous U2OS cells treated with siRNAs . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 016 To separate the function of ORC1 as a transcription co-repressor from its well-established role in DNA replication , we identified C-terminus truncation mutants ( 1–700 aa and 1–768 aa ) of ORC1 ( full length ORC1 is 1–861 aa ) that were defective in binding to the other ORC subunits , but still capable of binding RB and SUV39H1 ( Figure 7A–C ) . These ORC1 mutants were fully active in repressing CCNE1 transcription ( Figure 7D ) , demonstrating that the effects of ORC1 on CCNE1 gene repression were not due to an indirect effect of the role of ORC in DNA replication . 10 . 7554/eLife . 12785 . 017Figure 7 . ORC1 mutants separate its role as a transcription co-repressor from its role in DNA replication . ( A ) HEK293 cells were transfected with ORC1-Flag or its truncation mutants and GFP-RB as indicated . Whole cell extracts were immunoprecipitated with anti-Flag antibody and immunoprecipitates were analyzed by immuoblot with the indicated antibodies . ( B and C ) In vivo interaction between SUV39H1 and ORC1 or its truncation mutants . GFP-tagged wild-type ORC1 ( 1–861 ) or its truncation mutants ( 1–700 and 1–768 ) plasmids were co-transfected into HEK293 cells with Flag-SUV39H1 plasmid and either GFP-vector or empty vector as a control plasmids . Immunoprecipitation with anti-GFP antibody ( B ) or anti-Flag antibody ( C ) from cell lysates of HEK293 cells expressing the indicated constructs , followed by immunoblotting with the indicated antibodies . ( D ) . U2OS cells transiently transfected for 24 hr with increasing amounts of wild type ORC1-Flag ( 1–861 ) or truncation mutants ( 1–700 and 1–768 ) . Relative luciferase activity normalized to co-transfected lacZ control . Experiments were in triplicate . Expression of proteins determined by Immunoblot; α-Tubulin as loading control . Statistical analysis was performed using the Student’s t test . *p<0 . 005; **p<0 . 001; ***p<0 . 0005 . ORC , Origin Recognition Complex . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 017
ORC1 in plants is known to activate transcription in via a plant homeodomain ( PHD ) that is not found in animal and fungi ORC1 ( Sasaki and Gilbert , 2007; de la Paz Sanchez and Gutierrez , 2009 ) . In contrast , ORC1 in the budding yeast S . cerevisiae binds to the Sir1 protein and is involved in repression of transcription of mating type genes ( Bell et al . , 1993; Triolo and Sternglanz , 1996 ) . In Drosophila and animal cells , ORC , including human ORC1 , binds the HP1 proteins that are often involved in transcriptional repression , but a direct link between this interaction and transcription repression has not been established ( Auth et al . , 2006; Badugu et al . , 2003; Lidonnici et al . , 2004; Pak et al . , 1997; Prasanth et al . , 2010; Sasaki and Gilbert , 2007 ) . In this report , we have established opposing roles for ORC1 and CDC6 in control of transcription of the Cyclin E gene CCNE1 . ORC1 , with RB and SUV39H1 , bind to the CCNE1 promoter adjacent to the E2F1 transcription factor binding sites and repress the E2F1-dependent promoter and this repression is relieved by CDC6 that is bound to Cyclin E-CDK2 kinase . Consistent with this model , CDC6 binds to the CCNE1 promoter transiently , just at the time during G1 phase when Cyclin E levels increase dramatically . We suggest that low levels of both Cyclin E and CDC6 , both E2F1-controlled genes , appear upon activation of Cyclin D-CDK4 ( Narasimha et al . , 2014 ) . The expression of CDC6-Cyclin E-CDK2 provides positive feedback , disrupting the repressive RB-ORC1 interaction . The disruption of RB-ORC1 interaction further weakens ORC1 association the with promoter and results in a dramatic reduction in the repressive histone H3K9me3 mark at the promoter , increasing CCNE1 gene transcription and amplifying the levels of CDC6 and Cyclin E-CDK2 ( Figure 8 , bottom panel ) . The opposing expression levels of ORC1 and CDC6 during the cell cycle ( Figure 8 , top panel ) are consistent with this model . The production of higher levels of Cyclin E-CDK2 and CDC6 can then function in pre-RC assembly , as they are known to do ( Cook et al . , 2002; Coverley et al . , 2002; Hateboer et al . , 1998 ) . 10 . 7554/eLife . 12785 . 018Figure 8 . Model showing contrasting roles of DNA replication proteins ORC1 and CDC6 in the regulation of CCNE1 transcription and commitment to pre-RC assembly and cell division . Top panel summarizes the cycle levels of ORC1 and CDC6 . The bottom panel shown the ORC1 associated complexes at different stages of the cell division cycle . Blue arrow , positive feedback inhibition of ORC1-RB interaction . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 018 Our work also shows that ORC1 binds to CCNE1 promoter and recruits SUV39H1 , forming the repressive H3K9me3 mark that can then bind the HP1 protein . Interestingly , HP1 is known to bind both SUV39H1 and ORC1 ( Prasanth et al . , 2010; Stewart et al . , 2005 ) . Thus , ORC1 participates in multiple protein-protein interactions to ensure stable repression of CCNE1 and perhaps other E2F1-regulated genes during early G1 phase . In this way ORC1 , which binds mitotic chromosomes and is then inherited into the daughter cells ( Kara et al . , 2015; Okuno et al . , 2001 ) , creates an opportunity to repress CCNE1 gene transcription and allow the newborn cells time to integrate information to decide whether to proliferate or exit the cell division cycle . If proliferation and cell division are destined to occur , Cyclin D-CDK4/6 mono-phosphorylates RB and primes it for Cyclin E-dependent activation of E2F regulated genes ( Narasimha et al . , 2014 ) . We show that low levels of CDC6-Cycin-E-CDK2 can , in a feedback loop , amplify this commitment by antagonizing ORC1-RB interactions ( Figure 8 ) . Consistent with the model , we demonstrated that over-expression of CDC6 , but not a CDC6 that cannot bind Cyclin E , enhances the levels of endogenous Cyclin E during the very earliest period of G1 phase . Cell-cycle regulated Cyclin E is essential for maintenance of genome stability since cancer cells that have de-regulated Cyclin E fail to produce sufficient pre-RCs during G1 phase and as a consequence have problems in S phase and accumulate DNA damage ( Ekholm-Reed et al . , 2004; Jones et al . , 2013; Odajima et al . , 2010 ) . Of interest is our observation that at the time CDC6 is recruited to the CCNE1 promoter , ORC1 binding to the promoter declines . In human cells , ORC1 shows a dynamic , temporally regulated nuclear localization pattern such that in early G1 phase it is distributed in a punctate pattern throughout the nucleus , but in late G1 phase , ORC1 predominantly binds to regions of the genome that replicate late in the subsequent S phase ( Kara et al . , 2015 ) . We suggest that the events that occur at the CCNE1 promoter , ORC1 binding first and recruitment of SUV39H1 and RB , followed by recruitment of Cyclin E-CDK2 and CDC6 , may occur at many ORC1 binding sites in the genome , even those sites that are destined to assemble the entire ORC protein and promote pre-RC formation . SUV39H1 binding to ORC1 may influence its temporally dynamic nuclear localization during G1 phase . Such a scenario may also explain observations that have implicated a role for RB in DNA replication ( Bosco et al . , 2001; Kennedy et al . , 2000; Sterner et al . , 1998 ) . The ORC1-CDC6 switch might temporally influence pre-RC assembly just like it does for CCNE1 gene regulation , a possibility we are investigating .
U2OS , HEK293 and MCF7 cells were obtained from the Cold Spring Harbor Laboratory cell culture collection and cultured in DMEM containing high glucose ( Gibco ) supplemented with 10% inactivated fetal calf serum and Penicillin/Streptomycin . RB defective SaOS-2 cells were obtained from ATCC ( HTB-85 ) and were cultured in McCoy’s media supplemented with 15% inactivated fetal calf serum and Penicillin/Streptomycin . All the cell lines were tested negative for the mycoplasma contamination . To synchronize U2OS cells at G2/M boundary , 100 ng/mL of nocodazole was added to fresh medium for 16 hr . After 16 hr of block the cells were washed two times with 1x Phosphate Buffered Saline ( PBS ) and subsequently , released into the fresh media . For depletion of ORC1 protein in synchronized U2OS cells , the U2OS cells transiently transfected with 100nM of siRNAs ( control GFP as well as ORC1 siRNA ) using Lipofectamine RNAiMax and at the same time were treated with nocodazole , then the cells were incubated and released as described . Plasmids expressing exogenous genes were transfected using 2 . 5 μg of DNA except where indicated using lipofectamine 2000 transfection reagents ( ThermoFisher Scientific; Waltham , MA ) . The sequences of the siRNAs used are listed in the Supplementary file 1 . Plasmids expressing GFP-RB , 10–4 CCNE1 ( Addgene: Cyclin E gene ) promoter , E2F1 , DP1 , HDAC1 and pGL2-GAL4-UAS-Luc were purchased from Addgene . pCMV-LacZ plasmid was purchased from Clontech . Plasmid Flag-SUV39H1 was a gift from Peter Zhou ( Dong et al . , 2013 ) . ORC1-Flag or its mutant derivatives , ORC3-Flag , ORC4-Flag and T7-SUV39H1 plasmids were generated with a C-terminal Flag tag or N-terminal T7 tag in the pLPC vector ( McCurrach et al . , 1997 ) ( gift from Scott Lowe , Memorial Sloan Kettering Cancer Center ) for mammalian expression . Human ORC1 or its mutants , RB or its mutants , CDC6 or its mutants and SUV39H1 or its mutants were also cloned into pEGFP-C1 ( Clontech , Mountain View , CA ) for transient overexpression in mammalian cells . GAL4-DBD , GAL4-ORC1 and GAL-SUV39H1 plasmids were made by replacing the GFP insert in pEGFP-C1 constructs with in frame GAL4 DBD sequences . MBP-GFP-ORC1 plasmid is made by cloning GFP-ORC1 insert ( amplified from GFP-ORC1 plasmid ) in pMALc2E vector , while MBP-RB plasmid is made similar to MBP-ORC1 as described previously ( Hossain and Stillman , 2012 ) for protein expression in bacterial cells . Human RB or its mutants , CDC6 or its mutant , SUV39H1 or its mutant , HP1α and HDAC1 were also cloned in pGEX-6P1 vector to produce Glutathione S transferase ( GST ) fusion proteins in bacterial cells . The mutant plasmids were generated following the site directed mutagenesis protocol ( Quickchange , Stratagene , CA ) . All the plasmid constructs were verified by sequencing . The oligonucleotide sequences used to generate the plasmids are listed in the Supplementary file 1 . Wildtype ORC1 , GFP-ORC1 and RB were fused to the maltose binding protein ( MBP ) in frame by cloning in the bacterial expression vector pMALc2E ( New England Biolab ) . To generate Glutathione S transferase ( GST ) fusion proteins , wild type RB or its mutants ( R661W and N757F ) , HDAC1 , SUV39H1 or its fragments , HP1α or CDC6 were cloned into the bacterial expression vector pGEX6P1 ( GE Healthcare Life Sciences , NJ ) . The MBP fusion recombinant proteins were expressed and purified using amylose beads according to the procedure described previously ( Hossain and Stillman , 2012 ) . For GST fusion proteins , transformed E . coli BL21 cells with their respective plasmids were induced for 12 hr with 0 . 3 mM of IPTG at 16°C . The induced cells were pelleted , washed , and further lysed with sonication in a lysis buffer A containing 25 mM Tris-HCl at pH 7 . 5 , 150 mM NaCl , 0 . 02% NP-40 , 5 mM benzamidine-HCl , 1 mM phenylmethylsulfonylfluoride , Protease cocktail inhibitor tablets [Roche] , 10% glycerol ) plus 100 mg/ml lysozyme . The lysed bacterial cells were centrifuged and the clarified supernatant is incubated with pre-washed Glutathione sepharose beads for 3 hr at 4°C . The bead-bound proteins were washed with three column volumes of buffer A plus 0 . 05% NP-40 + 500 mM NaCl and further with five column volumes of buffer A alone . Fusion protein was eluted in a stepwise manner with buffer A containing 20 mM reduced glutathione , pH7 . 5 . Fractions containing purified proteins were pooled , concentrated , and dialyzed , and protein concentration was estimated using a standard Bradford protein assay . For MBP-RB or MBP-ORC1 pull down assay , the same protocol as described previously was used ( Hossain and Stillman , 2012 ) , except that here we have used a different binding buffer with following composition; 25 mM Tris-Cl at pH 7 . 5 , 100 mM KCl , 0 . 1% Nonidet P-40 , 0 . 1 mM EDTA , 5 mM magnesium acetate , 1 mM DTT . For MBP-ORC1 pull down , Cyclin E-CDK2 protein was incubated with wild type GST-RB or its mutants in the presence or absence of 1 mM ATP . The pull down was further immunoblotted with anti-GST antibody ( 27-4577-01; GE Healthcare Life Sciences , NJ ) . The histone peptide pull down assay followed the procedure described previously ( Hossain and Stillman , 2012 ) with the minor modification of binding buffer composition ( 50 mM Tris-HCl at pH 7 . 5 , 150 mM NaCl , 0 . 05% NP-40 ) . The pull down was silver stained or immunoblotted with anti-MBP antibody ( E8038S; New England Biolab ) . The biotin-labeled histone H3 peptides were purchased from AnaSpec ( Fremont , CA ) and bound to Streptavidin beads ( Sigma-Aldrich , St . Louis , MO ) prior to pull down studies . For GST pull down assay , bead-bound GST fusion proteins ( RB , HDAC1 , SUV39H1 or its mutants , HP1α , and GST-CDC6 or its mutant ) were incubated with either MBP-ORC1 or MBP-RB or Cyclin E-CDK2 protein . The composition of binding buffer used in the assay was; 25 mM Tris-Cl at pH 7 . 5 , 150 mM KCl , 0 . 15% Nonidet P-40 , 0 . 1 mM EDTA , 5 mM magnesium acetate , 1 mM DTT . The pull down was further immunoblotted with respective anti-GST ( 27-4577-01; GE Healthcare Life Sciences , NJ ) or anti-Cyclin E antibodies ( sc-247; Santa Cruz Biotechnology , Dallas , TX ) . For titration-based binding experiments , increasing molar amounts of GST-CDC6 and/or Cyclin E-CDK2 along together with equimolar amounts of MBP-RB ( 20nM ) and MBP-GFP-ORC1 ( 20nM ) were used in different combinations . The reaction mixture was incubated for 4 hr at 4°C followed by 1 hr more incubation with bead coupled anti-GFP antibody ( ABP-NAB-GFPA025; Allele Biotech , San Diego , CA ) to precipitate MBP-GFP-ORC1 . The reaction mixture was incubated and washed with binding buffer with following composition; 25 mM Tris-Cl at pH 7 . 5 , 150 mM KCl , 0 . 15% Nonidet P-40 , 0 . 1 mM EDTA , 5 mM magnesium acetate , 1 mM DTT and 1 mM ATP . The pull down was immunoblotted with the following antibodies; anti-ORC1 antibody ( pKS1-40 ) , anti-RB antibody ( #9309; Cell Signaling , Danvers , MA ) , anti-GST ( 27-4577-01; GE Healthcare Life Sciences , NJ ) and anti-Cyclin E antibody ( sc-247; Santa Cruz Biotechnology , Dallas , TX ) . For expression of proteins , HEK293 cells were transiently transfected with the indicated plasmids with lipofectamine 2000 transfection reagents ( ThermoFisher Scientific ) . Immunoprecipitation from HEK293 cells was performed using the procedure described previously ( Hemerly et al . , 2009 ) with slight modification in the protocol . Following expression of proteins , the cells were harvested and washed in PBS and lysed in a buffer containing 20 mM Tris-HCl pH7 . 5 , 200 mM NaCl , 0 . 3% NP-40 , 5 mM MgCl2 , 0 . 1 mM EDTA , 10% Glycerol , 1 mM DTT , 1 mM CaCl2 , 20 uM MG132 and protease as well as phosphatase inhibitor tablets ( Roche ) . Benzonase ( Sigma-Aldrich , St . Louis , MO ) was added to the buffer and the suspension incubated for 30 min on ice with intermittent mixing . The concentration of NaCl and NP-40 was reduced to 100 mM and 0 . 15% , respectively with dilution buffer after 30 min incubation on ice . The extract was centrifuged at 14 , 000 rpm for 15 min at 4°C . The proteins were precipitated with specific antibodies as indicated in figure legends using FLAG , GFP , ORC2 or ORC3 antibodies . The whole cell extract was first incubated with antibodies for 4 hr and subsequently , 2 hr with pre-washed gamma bind G sepharose beads with end-to end shaking at 4°C . The beads were washed 3 times with washing buffer containing 20 mM Tris-HCl pH7 . 5 , 100 mM NaCl , 0 . 1% NP-40 , 5 mM MgCl2 , 0 . 1 mM EDTA , 10% Glycerol , 1 mM DTT and protease as well as phosphatase inhibitor tablets from Roche . Finally , the washed beads were suspended in Laemmli sample buffer and 8% SDS-PAGE gels were run and immunoblotted . For immunoprecipitation of endogenous ORC1 and SUV39H1 proteins from U2OS or MCF7 cells and SaOS-2 cells , we have used monoclonal ORC1 antibody coupled to magnetic beads as described previously ( Kara et al . , 2015 ) as well as rabbit polyclonal SUV39H1 antibody ( A302-128A; Bethyl Laboratories ) , respectively . For immunoprecipitations , the following antibodies were used: rabbit polyclonal anti-Flag antibody ( F7425; Sigma ) , GFP nAb ( ABP-NAB-GFPA025; Allele Biotech , San Diego , CA ) , polyclonal anti-SUV39H1 antibody ( A302-128A; Bethyl Laboratories ) , mouse monoclonal ORC1 78-1-172 ( Kara et al . , 2015 ) , rabbit polyclonal anti-ORC2 antibody ( CS205-5 ) ( Prasanth et al . , 2004 ) , and rabbit polyclonal anti-ORC3 ( CS1890 ) antibody ( Siddiqui and Stillman , 2007 ) . For immunoblots , monoclonal FLAG antibody ( F1804; Sigma ) , polyclonal GFP antibody ( G1544; Sigma ) , monoclonal mouse SUV39H1 antibody ( 05–615; Millipore ) , monoclonal mouse T7 antibody ( Cold Spring Harbor Laboratory antibody facility ) , mouse monoclonal anti-ORC1 antibody ( pKS1-40 ) ( Hemerly et al . , 2009 ) , monoclonal mouse ORC2 antibody ( 920-2-44 ) ( Siddiqui and Stillman , 2007 ) , mouse monoclonal ORC3 antibody ( PKS1-16 ) ( Prasanth et al . , 2004 ) , goat polyclonal anti-ORC4 antibody ( ab9641; Abcam , Cambridge , MA ) , monoclonal mouse E2F-1 antibody ( KH95; Santa Cruz Biotechnology , Dallas , TX ) and rabbit GAL4 antibody ( sc-577; Santa Cruz Biotechnology , Dallas , TX ) were used . Nocodazole arrested U2OS cells were transfected with 100 nM siRNA ( Dharmacon Inc . , Lafayette , CO ) , and RNA was prepared at the indicated time points post-drug release using the RNeasy Mini Kit ( Qiagen cat . #74104 ) including on-column DNase digestion ( Qiagen cat . # 79254 ) and eluted in the supplied RNase-free water . The cDNA used for Q-PCR was prepared from 1μg each RNA sample using TaqMan Reverse Transcription Reagents ( Applied Biosystems #N808-0234 ) with random hexamer priming in a GeneAmp PCR system 9700 thermocycler . Each Q-PCR reaction was prepared using 2 μL of 1-to-20 diluted cDNA and 13 μL LightCycler 480 SYBR Green I Master Mix ( Roche #04887352001 ) and were performed in 384-well plates using the LightCycler 480 ( Roche ) as per manufacturer’s instructions . For semi-quantitative RT-PCR , equal amounts of RNA ( 0 . 2 μg ) were used for RT-PCR using the Qiagen’s One-Step RT PCR kit following manufacturer’s instructions . PCR was performed for 22 cycles and subsequently , run on 1 . 8% agarose gel . The primer sequences used for RT-PCR analysis are listed in the Supplementary file 1 . For the luciferase reporter assay , 0 . 3 x 106 U2OS cells were seeded in six-well plates . Cells were transfected with 0 . 5 μg CCNE1 promoter luciferase plasmid ( p10-4 ) , 50 ng E2F1 plasmid , 50 ng of DP1 plasmid and 20 ng LacZ plasmid . Together with above plasmids , the cells were also transfected either with ORC1-Flag or its mutants , ORC3-Flag , ORC4-Flag , T7-SUV39H1 , GFP-SUV39H1 or its mutant , GFP-CDC6 or its mutant for overexpression at the indicated amounts in micrograms . The cells were transfected using Lipofectamine 2000 ( Life Technologies ) for 24 hr . For GAL4-based luciferase assay , the U2OS cells were transfected with 0 . 5 μg of GAL4-UAS-Luciferase promoter plasmid and 20 ng of LacZ plasmid along with GAL4-DBD fused ORC1 or SUV39H1 at the indicated amounts . In each of the experiments , the amount of plasmid was kept constant by the addition of empty vector DNA . For siRNA-mediated depletion of proteins and luciferase measurement , 100 nM of specific siRNAs targeting ORC1 , SUV39H1 or CDC6 were used together with p10-4 , E2F1/DP1 and LacZ plasmids . GFP siRNA is used as negative control . β-Galactosidase activity was measured as described previously ( Smale , 2010 ) , while luciferase activity was measured using luciferase ( Promega ) luminescent assay kit according to the manufacturer’s instructions . Luciferase activities were normalized to β-galactosidase activities and denoted as relative light units ( RLU ) . Expression or depletion of proteins was confirmed by Western blot . All the experiments were done in triplicates . MCF7 and U2OS cells were used for chromatin immunoprecipitation . The cells were harvested by trypsinization and washed with cold PBS . The cells ( 3 . 0x107 ) were fixed with 1% formaldehyde for 10 min at room temperature . The cross-linking was stopped by addition 125 mM of glycine for 5 min on ice . The fixed cells were washed with cold PBS and lysed for 10 min on ice with buffer containing 10 mM Tris pH 8 . 0 , 10 mM NaCl , 2 mM MgCl2 , 0 . 4% NP40 , 1 mM DTT , 10% Glycerol , protease and phosphatase inhibitors . Preliminary experiments demonstrated that ORC1 was most efficiently extracted from the resulting cells by digestion of the cross-linked chromatin with micrococcal nuclease ( MNase; Sigma ) and extraction with high salt ( 300–600 mM NaCl ) . Thus , the nuclei were treated with MNase with 1 mM CaCl2 at 37°C ( so that most of the extracted DNA ran on an agarose gel as ~1–4 nucleosome length ) , the reaction was quenched with 2 mM EGTA and washed with buffer containing 10 mM Tris pH 8 . 0 , 10 mM NaCl , 2 mM MgCl2 , 1 mM DTT , 10% Glycerol , protease and phosphatase inhibitors . The pelleted residual nuclei were further lysed in buffer containing 20 mM Tris pH 8 . 0 , 1 mM EDTA , 0 . 5 mM EGTA , 300 mM NaCl , 0 . 5% TritonX-100 , 0 . 05% Sodium Deoxycholate , 0 . 1% IGE-PAL , 1 mM PMSF and protease/phosphatase inhibitors for 30 min at 4°C with rotation . The lysate was further diluted to bring the salt concentration to 200 mM and sonicated very briefly . The lysate was centrifuged ( 20 , 800 X g in an Eppendorf centrifuge ) at 4°C to pellet down the debris and supernatant was used for pre-clearing with protein G Dynabeads ( Invitrogen ) for 2 hr 4°C . The antibodies were bound to protein G Dynabeads for 5 hr at 4°C in PBS containing 0 . 5% BSA . The antibodies used for chromatin immunoprecipitation were as follows: mouse monoclonal ORC1 [78-1-172; ( Kara et al . , 2015 ) ] , mouse RB ( #9309 , Cell Signaling ) , rabbit polyclonal SUV39H1 ( A302-128A; Bethyl Laboratories ) , rabbit polyclonal CDC6 ( CS1881 ) , rabbit polyclonal histone H3K9me3 ( ab8898 , Abcam ) as well as control mouse and rabbit IgG ( Invitrogen ) . Pre-cleared nuclear extracts was incubated with antibody-bound beads overnight with rotation at 4°C . 0 . 4 ml of the nuclear extract was used for each IP and 0 . 1 ml was kept aside for 'Input' in Q-PCR analysis . Beads were then washed with three times with low-salt buffer ( 20 mM Tris pH8 . 0 , 200 mM NaCl , 0 . 5% TritonX-100 , 0 . 05% Sodium Deoxycholate , 0 . 1% IGE-PAL , 1 mM PMSF ) , three times high-salt buffer ( 20 mM Tris pH8 . 0 , 500 mM NaCl , 0 . 5% TritonX-100 , 0 . 05% Sodium Deoxycholate , 0 . 1% IGE-PAL , 1 mM PMSF ) , three times with lithium chloride buffer ( 20 mM Tris pH8 . 0 , 200 mM NaCl , 250 mM LiCl , 0 . 5% TritonX-100 , 0 . 05% Sodium Deoxycholate , 0 . 1% IGE-PAL , 1 mM PMSF ) and twice with TE buffer ( 10 mM Tris pH8 . 0 , 1 mM EDTA ) . Chromatin was eluted from beads twice with 100 ul elution buffer ( 100 mM sodium bicarbonate , 1% SDS ) at room temperature . Protein-DNA crosslinks in the IP samples as well as in input samples were reversed overnight by addition of 300 mM NaCl and 2 ug RNase A at 65°C . The ChIP and input samples were then incubated with 60 ug of proteinase K for 2 hr at 42°C . The samples were further extracted with phenol:chloroform:isoamyl alchol [IAA] ( once for ChIP and twice for input ) and once with chloroform extraction . The ethanol precipitation were done by adding 10 ug of glycogen , washed with 70% ethanol , air dried and then re-suspended in 50 ul of RNase-DNase free water . The purified DNA was used as template in different PCR amplifications ( Applied Biosystems Thermocycler ) . The sequences of the various primers is listed in Supplementary file 1 . Due to very high GC content of human CCNE1 promoter , the number of PCR cycles was extensively and independently optimized with different DNA polymerases ( Supplementary file 2 ) for each primer set to maintain linear amplification in all experiments . PCR products were resolved by 2% agarose gel electrophoresis . Images of ethidium bromide-stained DNAs were acquired using an UV trans-illuminator equipped with a digital camera . Intensities of the amplified PCR bands were quantitated by ImageJ software . Data are shown as the average ± the standard deviation ( SD ) of results of at least three independent experiments . In the luciferase assay , the statistical differences between cells overexpressing ORC1 and/or SUV39H1 or CDC6 or no E2F1/DP1 were compared to E2F1/DP1 alone without overexpression , and statistically evaluated by Student’s t test analysis . The lines above the bar graph were used to indicate the statistical differences between overexpressing samples . In the luciferase assay with depletion of specific proteins , the statistical differences between siRNAs against the ORC1 , SUV39H1 or CDC6 were compared to control GFP siRNA using the Student’s t test . | Living cells must replicate their DNA before they divide so that the newly formed cells can each receive an identical copy of the genetic material . Before DNA replication can begin , a number of proteins must come together to form so-called pre-replicative complexes at many locations along the DNA molecules . These protein complexes then serve as landing pads for many other DNA replication proteins . One component of the pre-replicative complex , a protein called ORC1 , helps to recruit another protein called CDC6 that in turn acts with Cyclin E to promote the replication of the DNA . Cyclin E is a protein that is only expressed when cells commit to divide . Previous research has shown that a lack of ORC1 causes the levels of Cyclin E to rise in human cells , but it was not understood how cells regulate the levels of Cyclin E . Now , Hossain and Stillman show that the ORC1 protein switches off the gene that encodes Cyclin E early on in newly born cells , and therefore prevents the Cyclin E protein from being produced . The experiments show that ORC1 does this by binding near one end of the gene for Cyclin E and interacting with two other proteins to inactivate the gene . Thus , ORC1 establishes a period when Cyclin E is absent from a newly formed cell . This essentially gives the cell time to ‘decide’ ( based on external cues and its own signaling ) whether it will divide again or enter into a non-dividing state . When a cell does decide to divide , the levels of CDC6 rise . CDC6 is another component of the pre-replicative complex and Hossain and Stillman find that CDC6 works to counteract the effects of ORC1 and reactivate the gene for Cyclin E . This activity leads to a dramatic increase in the production of Cyclin E , which in turn allows the cells to commit to another round of DNA replication and division . The opposing effects of ORC1 and CDC6 control the levels of Cyclin E and provide a link between DNA replication and a cell’s decision to divide . Further work is now needed to see whether ORC1 inactivates other genes in addition to the one that encodes Cyclin E . DOI: http://dx . doi . org/10 . 7554/eLife . 12785 . 002 | [
"Abstract",
"Introduction",
"Results",
"Discussion",
"Materials",
"and",
"methods"
] | [
"chromosomes",
"and",
"gene",
"expression",
"biochemistry",
"and",
"chemical",
"biology"
] | 2016 | Opposing roles for DNA replication initiator proteins ORC1 and CDC6 in control of Cyclin E gene transcription |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.